in version 2019.6.9.0
Fixed a bug with legacy audio, it would get stuck in a start / stop loop, broke
that loop ‐ audio still seems to suck though in legacy mode. I've been
considering removing legacy mode completely.
As of 2019-04-10 legacy audio mode was completely removed. Great ‐ this
made patch recompilation work faster (legacy audio was started and stopped for
recompiles (adding or removing modules or cables, or loading a new patch) but
PortAudio is kept running on recompilation).
Fixed a bug for the ScaleQuantizer and the ProbSequencer modules where the initial
scale after patch start was not shown correctly when scale modulation was used. It
would not show the initially selected scale in the notes switches.
Fixed some of the most obviously wrong help items, had forgotten to update the
text for some of the more recent modules.
Changed the patch format to include the tuning parameters that were set. The patch
version did not change for this as the modification is both upwards and
downwards compatible. Old versions of Wren will skip tuning parameters
when present and new versions of Wren will not change the tuning when
the tuning parameters are not present.
Tuning parameters will only be read when a new patch is loaded from disk, they
will never be used on parameter imports and such. They will be read when a a patch
template is loaded, as this is the same as loading a new patch from disk.
An option was added to the Settings Screen F3 to never import tuning parameters
from a patch. When this is active tuning parameters are still read from the ini
file when Wren is started, and tey will then be the same as when
Wren was last closed. This option is on by default to be compatible
with previous versions.
Also fixed the hint popup for knobs to show the right frequency after tuning
changes were made. The note names are still incorrect, but I do not have ideas
about how to properly name notes in general, for instance in a tuning of say
11.21 notes per "octave" sized 2.13, say …
I've brought the tuning controls to the main toolbar, they were in the
Settings Screen F3 before.
Then had to change the patch version from 7 to 8 anyway as I did add a couple of
new LFO ranges and wanted older patches to be read with correct range values and
the variations to stay intact as well, see the next paragraph for more on this.
Added three new LFO ranges, being fast2, medium2 and slow2.
These have ranges of values mapping nicer to whole time intervals (for instance a
rate of 1 Hz can be set now) and also the mapping to standard oscillator
frequencies is better (for instance an LFO rate of 55 Hz can be set now now,
although this will only make sense in a standard 12 tone octave based tuning with
an A of 440 Hz, as LFOs are not rescaled with the current tuning). Had to change
the patch version number for this to be able to read in older patches correctly.
The new patch version is 8. There are 32 steps now for a change of a factor of
two, for the existing fast, medium and slow ranges this is 24 steps. Which also
means that the total span of the new ranges is a bit smaller, but of course the
old ranges can still be used for the more extreme settings.
Changed the A and R time modulation behaviour for the EnvARRetrig envelope module,
the modulation is now added instead of multiplied. Added A and R modulation for
the EnvAR module as well, working the same as for EnvARRetrig. Also made the
modulations to work in real time ‐ before they were sampled only when the
envelope got triggerd. This will break old patches when they use time modulation
for EnvARRetrig modules.
Added some new fuctions to the Monadic and Dyadic modules. To Monadic was added
1 - abs( x) and to Dyadic x ^ ( 1 / y). Also changed the behaviour
for the case where the function can not be calculated (like division by zero, for
instance). In the new situation the output value will not change from it's
previous value in case of an impossebility (output used to be zero for the Monadic
module and X for the Dyadic module in such cases). This may break some old patches
that were based on "undefined behaviour". Also the correct function may
not be loaded for old patches.
Added 32 row sum outputs to the LifeSeq module, just so the thing can control more
and better. This was inspiered by PHOBoS, but he is using an 8 by 8 grid where
each cell is a different note. This seemed a bit much for a 32 x 32 grid, so I
modified the idea a bit. Also made a new 12 by 12 SmallLifeSeq, more symmetrical,
with both a controllable horizontal and vertical cursor. Then also added yet
another game of life module the VitaPHoBium. This one has a 16 x 8 grid where
input values can be clocked in to turn cells on. This one also has X Y cursor
control as well as stepped cursor control, and when the cursor lands on a life
cell a note value of sorts will be output. Later added a "Skip dead"
mode which will try to find the next live cell on a next clock. skipping the dead
ones. When X and Y are connected but Step is not the cursor will be updated
asynchroneously, and when Step is connected as well the eX and Y inputs will be
sampled on an upgoing Step edge. When only Step is connected the cursor will move
in a stepped way, either forwards or backwards.
Made an unclick module, this is just an averaging filter with a fixed alpha
setting ‐ it is a bit cheaper than the average filter, and it will filter
out the most annoying clicks without giving a lot of lag.
Added 'raw' outputs to the Seq16 module, they are raw in the sense that they are
not cross faded. Marked the existing outputs with an X and the new ones with an R,
for cross-faded and raw respectively. Did the same for the LUT module.
Added a changed output on the ScaleQuantizer module, it goes active briefly when
the current output value is not equal to the previous one. Likwise for the poetry
Added inverted outputs in addition to the regular ones for the RandSigs module.
The inverted outputs are using the same output type as selected for the regular
Changed the way scale modulation works on the ScaleQuantizer module, to make it
look a bit less confusing maybe … For an unconnected or an invalid scale
selection input the pre‐selected scale will be used. When the scale
modulation is connected and it has a valid value the pre‐selected scale will
not be used but the modulation in itself will determine the scale to be used.
Before this the pre‐selected scale would be added as an offset to the scale
modulation input. So there is no longer a need to set the pre‐selection to
the first value in order to get predictable scale modulation. Also it is possible
now to set the preset to a fallback value to be used in case the modulation goes
out of range. In the old situation invalid modulation input could cause the input
value to be passed unchanged to the output, this is no longer possible, so this
change may break some older patches where this "flaw" was used as an
Added effect modulation on the AmMod module. This makes it possible to
programmatically fade between amplitude and ring modulation.
Made a Dif module, a differentiator, with a fixed gain of 1. There already was an
IntDif module as well, but the integrator bit did not seem too useful to me
… so made it a bit cheaper. The old IntDif module is still available as
Added a ChangeDetector module. It detects changes on it's input and it will send
out a trigger pulse when a change happens. Also the value of the change will put
onto an output, it can be selected whether the value of the change is taken as is
or has it's absulote value being taken. This sort of is a differentiator with a
built-in sample and hold function.
Made a voltage controlled pattern sequencer, it was modelled after
by fonik but made it a 16 step one and addded reverse and inverse modes as
well as step length modulation and random control over the slider positions. The
Gain and Offset controls work with bits, so they can be used to set the step where
the effect begins and ends. Hmm … not quite … but will leave it as
is for now.
Made an AutoQuantizer module, it samples input values clocked in on it's learn
input. It will remember the note value (but not the octave value) of the last N
ones, where N ranges from 1 to 64. Signals coming in on the (reqular) in input
will be split into an octave and a note value and the nearest note value present
in the history will then be output in the octave of the incoming value. This can
sort of improvise with notes it had seen before by putting some sort of a random
value into the in input. When a less random input is being used like a saw or a
triangle it would output arpeggios. When the new output differs from the one
before a trigger pulse will be generated on the changed output, this can be used
to trigger an envelope, for instance.
Made a NoteSplit and a NoteMerge module. The splitter splits an incoming note
value into it's ocatave and note parts. The merger does the reverse thing, it
combines a note and an octave value into a note. The octave size is obtained from
the global NotesPerOctave setting. The merger is really just an adder, but still,
being the inverse of the splitter I'll just keep it.
Made a Queue module. It remembers the latest N clocked in values (with N going
from 1 to 64). With the select controls one of the remembered values will be
selected, optionally providing cross fading between old and new values.
YASM ... this speaker uses windows SAPI voices. Right, can only get it to work in
8 bit mode and it is very slow to translate speech to text. I've made a loading
indicator which lights up while the conversion takes place in the background. This
way it at least is possible to go on patching while the speech synth compiles the
text. Funny though how it can reasonably quickly perform a voice change, text
changes take much longer. Anyways, it is more like speech than the other speech
Windows is a bit limited in the number of voices it supports out of the box, to
fix that eSpeak can be used. On installation you can select languages and speakers
to be installed, some modifiers are supported too.
Ok, got the sluggishness out a bit, turned out it was forever appending and
partially overwriting to the same audio queue instead of renewing it. So had to
destroy and recreate it instead of reuse it. Still, longish texts at high
rendering rates will take a while to render, like a couple of seconds for the text
from YASM to here.
AYASM ... this speaker uses the eSpeak voices and the espeak-sapi.dll. This one is
a bit slow to compile speech too (albeit a tad faster than the SAPI one maybe),
for short text it will do. It more or less provides the same functionality as the
older song module, except it has a better speech quality. The module will signal
speech events like phonemes, words, sentences and silence. Also markers can be
placed in the text to generate user events. The user events can be made to
generate an output pulse, or they can put some value on an output. The number of
voices is fairly large, and there are modifiers for male, female, whisper and
croak. Also … it can speak Dutch with a propperly rrrroling godverdomme :-)
Too bad tho that the eSpeak provided voices seem somewhat arbitrarily tuned.
Anyways put some tuning controls, inclusing a cents control, on the module to
cope with that.
ESpeak should be installed for this module to work, and the espeak-sapi.dll should
be somewhere in the path (it can be put into the wren executable filder, for
instance), the standard windows DLL search policy applies.
I've added the eSpeak DLL to the wren package, but forgot to include it into the
source zip I think.
Added cents values to be displayed in the popups for FM modulation on audio rate
oscillators as the concept of semitones is a bit odd for some scalings.
Changed the way the "Freeze" indictor works on the Reverb module, it no
longer shows the value of the Freeze input but the actual freeze status (which is
the xor of the input and the mode selection button).
Made the patch compiler a bit more robust. It would on bad module or cable
definitions crap out and not load the patch at all. It will now load the patch
skipping the stuff not being understood. Any skips will be logged.
Removed the splash screen animations, it was uggly in more than one aspect
in version 2019.3.5.0
Fixed a bug causing unnescessary crackling. Did this by totally removing all
automation stuff ‐ it was not working anyway.
Fixed a bug with portaudio initialization where it was not possible to have an
output only system. It is now possible to have an input-only system as well
… how usefule is that ;-) But there need to at least be inputs or outputs
for Wren to be able to do audio I/O.
Fixed a bug with reading wren.formants where the file could not be read on some
windows systems with a code page differing from 1252.
Various small changes on several modules, and a couple of new ones. I didnt keep
track of this process very well though, so … surprise.
Oh, the ones I do know are: Added a couple of new modules, mostly Chladni related,
a generator a filter bank and an oscillator bank. Oh and the Van der Pol stuff
… more info later. Also added some inputs here and there … like the
"1" input on the Sequencer16 and LUT modules, these randomize one of the
16 steps (a randomly selected one) instead of all at once (which is now the
"all" input). Also added an interpolating slow to fast mode conversion
module, helping to get less crackle on delay line modulations when the control
rate division is set high. Also made it possible to have "octaves" of a
factor different from 2, this is in the Settings Screen F3.
Trying to port stuff over to free pascal … but that will be a while. First
splitting off the engine, that would be Wrengine. It compiles and it succeeds on
some unit tests (like building a simple patch and executing that). I've found DLLs
for Windows, Linux and OSX too. Had it compile under all three OSes as well, but
didnt do any 'real time' audio testing as of yet. Also the user interface will
need to be completely redone. There may be some windows only releases still before
this wil take off. Anyway, for the curious, this is available for download (in the
downloads section at the top of this page), just note that it is not a synth
yet but an experiment. now at wrengine.2018.5.19.
A little update on the above, it can now read Wren patches version 2 to
7. Not all modules were implemented for it yet though. It does not have the
Swansong (there is an issue with the forth interpreter) and it does not have the
Van der Pol modules (which were recently added to Wren). Now at
And a lil bug fix for the variation values. Now at wrengine.2018.5.23.
Added variations, morphing and automation stuff, not finished (but consuming CPU
time). Also added support for reading Wren patches and added some
profiling code to check the performance of the synth engine ‐ not too bad
sofar, better than the original Wren it seems:
Running synth for 1000000 cycles
Duration was 9099 ms : time / sample : 9.1 us, sample rate : 110 ks/s
Which was for a patch doing about 60% for Wren at 48 ks/s (or about 80
ks/s at 100% CPU, but no actual audio i/o was performed for Wrengine where it was
for Wren), and optimization mildly set to o1 and many checks on. For o4
optimization, i/o checking, range checking, assertions and debug code off it was
Running synth for 1000000 cycles
Duration was 5672 ms : time / sample : 5.7 us, sample rate : 176 ks/s
For all cases the control rate decimation was set to 8.
Also some small bug fixes were made, should check all the p_ stuff to not use
FInputs but FParams instead … now at wrengine.2018.5.25.
in version 2018.5.18.0
Added a matrix multiplier module. It calculates Mabcd * in →
out (or a * in1 + b * in2 → out1 and c * in1 + d * in2 →
Made a new modal filter able to load presets from a text file. The frequencies
read will be fixed, but an overall tuning can be applied either exponentially or
linearly. Individual Q and amplitude settings are present. Up to eleven formants
can be specified and the overall Q factor can be set as well. The text file read
is wren.modal, and it needs to be present in the Wren executable
directory for this module to work.
Made a new MultiGate module providing four dual input gates all performing the same
selectable function. The input and output signal types can be set separately.
Made a 3 x 32 / 1 x 96 sequencer. Notes can be set per step as well as an octave
and a duration.
Interpretation of the duration depends on the mode set (either free or sync). In
sync mode step durations are fixed to 16 clock periods and the gate time will be
equal to the duration, while in free mode the step duration and the gate time are
both set to the selected duration.
Trigger mode can be set to Trig, Gate or Repeat. Trig outputting a clock for 1/16
of the input clock period (which is 16 clocks long), Gate for the set duration and
Repeat does repeating 1/16 clock pulses during the set duration.
The three button controlled modes can also be modulated.
Split the clock input into three separate ones so the three play pointers can move
at different speeds. This will also work in 1 x 96 mode, where the first output
will be the sequence proper and the other two will be canon like variations on it.
The dir input when active will reverse the playing direction, the inv input will
play the sequence upside down when active. These are global controls working on
all three play pointers the same way (just like act and res do).
The / 16 outputs output the incoming clock divided by 16.
A van der Pol oscillator module.
A van der Pol LFO module.
Added a drive input to the Attractor and AttractorLFO modules ‐ just like
the van der Pol thingies have that.
Added Sin and Cos outputs to the rotator module. These can be used
to control delay lines to get a leslie type effect. Also added a double sin mode
* where the Cos output will actually be -Sin and where
the rotation being berformed works a little different.
[*] Or maybe it's zero sin with the minus involved.
Made a stereo version of the Delay module, DelayStereo. This is just two delays
sharing the same delay length parameters.
Made the phase of the RandSig module to have a random value after reset. Also made
reset to cause an immediate 'wrap' action (starting a new random cycle). Reset
used to only set the current phase value to zero.
Retuned all of the fixed filter banks a bit in such a way that they will use 55,
110, 220, 440 Hz etc.
Added an option to the Pad mdule to have a user selectable background. Also added
three new outputs refelcting the color of the image pixel the cursor dot is on.
Added four extra multiplication units to the MultipleMult module, giving it eight
multipliers in total.
Some minor bug fixes.
I've moved from Delphi XE5 to XE10. The latter seems to be slowish .. but hey it's
the fuuture, one can't stop that.
I'm totally rewriting the automation recorder. It wil use a flat model now with a
track for each control, and no extra module level anymore. The user can select
between recording all knobs or just a set of exported ones. Each track can have
it's own graphical editor. It will also be possible to add or delete events trough
that editor. Currently it is not working at all.
in version 2017.11.19.0
The SwanTester program, for testing the contents of mazes.4th
in version 2017.7.31.0
Consider this release to be a beta version. I've changed quite a lot of internal
stuff and have added some half finished new features. But it's about time for a
new release, and some bugs were fixed … at least.
Error in Divider module, in the pulse mode it would flash the output too often.
Fixed now for the next release.
There was an issue with inserting and deleting 'non-active' modules, like blind
panels and the notes module. Such an action would cause the modules in the editor
patch and the (compiled) synth patch to get out of sync, resulting in the lights
not being updated (and possibly worse). Fixed this to always resync when modules
are being added or deleted even when no structural differences between the old
and new compiled patch are present.
Pasting text into a display editor (like in the notes module) did not work
properly, as windows on double click on a static text control will fuck the
clipboard. I've changed the way to call up the editor, must use Ctrl+Left Click now instead of a Double Click. It
is also possible to focus the display with a Left Click and
then use Enter. Will need to update some help text too …
k … diddat.
On display editing EndStateChange was not always executed when editing ended,
resulting in the patch no longer being updated correctly on wire edits. Also
resulting in strange behaviour on undo and redo operations. Fixed this. Also there
was a second mechanism locking undo/redo without performing a recompile at end,
possibly resulting in patches no longer being compiled. Removed this mechanism
Added a speed input to the RandSig module. This is a linear FM input, but for the
RandSig module the input value will be clipped to a range [0.0, 1.0] (where a
speed of zero will stop the output from changing and a value of 1.0 will make it
run twice as fast as it would with an unconnected speed input. So it is not
trough‐zero for this mdodule.
Made the LifeSeq module rate smart, so it can run at audio rates now.
Sped up the AD and DA conversions as used in the DAC, ADC and Gate modules a bit
by using lookup tables for bit shifts. This also fixed some levels, like the DAC
would output twice as much as it should have. This will change existing patches
making use of the DAC or ADC. It also changes patches using Gate modules with a
bitcount other than one.
Changed some things in the internal working of the code, handling of controls, the
way inputs and outputs are being specified, some control flow trough the program.
This may result in some stuff not working anymore, sorry about that. I will fix it
when I notice it … but having almost 250 mdule types now and various ways
different modules can interact with the patch I may have missed some things in
Added a manual compilation mode to the program. When it is on patch changes will
not result in recompilation, instead they have to be done manually. When this mode
is on you can make changes to the patch without the output going silent, then when
done a manual compile (or leaving manual compilation mode) will effectuate all
changes at once. The keyboard shortcust for this are Ctrl+F12 to toggle the mode and F12 for manual compilation. The help
desk will get cross with me for this one … but, there is some visual
feedback on this in the form of a flashing 'manual compilation' text in the top
pane and also it is made visible in the Action menu. When there are any uncompiled
changes the 'manual compilatin' will flash yellow, and white otherwise. At program
startup the feature will always be switched off (it is not saved in the ini flle).
Knob changes will still change the patch when manual compilation is on ‐ it
only affects rewiring and module insertions / deletions, but this includes loading
a new patch or performing any undo / redo operations.
Anyway, thought this might be useful for live patching, we'll see about that. The
work flow would be to hit Ctrl+F12 to turn the mode on, make
some changes, then hit Ctrl+F12 again to compile the patch and
hear the changes.
Actually this is pretty weird on structural changes, it may cause some unexpected
behaviour. Still, to do a bunch of knob edits, and have them kick in at the same
chosen moment has been useful for me.
I've changed the default behaviour for modules and controls to be enabled for
randomization. Randomization used to be all enabled by default, it's now all
disabled by default. I've also changed the feedback on this, screen items enabled
for randomization are now marked as being special, this used to be the other way
around. (Gabe you were right about this from the beginning, took me some time to
see it, thanks :-))
I've added some control over the randomization range. It is now possible to set a
minimum and a maximum value for randomization on knobs. This can be set by
changing the knob value with Ctrl or Shift being held down wile changing the value (either trough the mouse, the mouse wheel,
the keyboard or by clicking the up / down arrows under the knob). The range, when
not maximal, will be indicated by an arc being painted on the knob's edge (welp
… it doesn't look too pretty right now, but it works) and the selected
range is also visible in the hint popup (when simple hints are not enabled). Ok,
not the best image, but it's the blue hat on the topmost knob … so here is
another one, the hat can be red too, meaning the maximum
is less than the minimum, giving inverse response (not that such matters much for
random signals). It is still possible to navigate the knob outside the set range
manually. I did not implement this on selectors … as of yet … As I'm
not sure how to give visual feedback on it, and also there is an issue with some
of the selectors having dashes in their value list … separators …
those are a bit of a PITA. I did implement it for the simple knobs not having a
knob at all but jsut the up / down arrows. Here the results will be visible in the
popup hint only.
Made a multi‐sine oscillator, which generates six sines with individually
settable amplitudes and frequencies individually settable relative to the main
tuning. The amplitudes can be modulated by an external signal. The chain input is
summed with the oscillators into the output. There are mutes for each sine
individually and on the resulting output as well. The sync input when going active
resets the phase to zero for all the oscillators. The main tuning has exponential
FM modulation and a separate frequency fine tuning. Id did not find room to
include PM, but each oscillator has a separate AM modulation input.
Made a text writer module which can be used to write text into a file. I use this
to couple a radio stream title to a patch, such that when I change the patch the
stream title will be updated automatically. It can hold multiple lines of text,
and sending out the next text is controlled from a trigger input. When the module
does not get triggered it will statically output the first text line. The text
file will only be updated when a new line of text gets selected. This happens at
patch load time, patch reset or when a trigger comes in and there is more than one
line of text. When the file does not exist it will be created, and when such fails
nothing bad should happen.
Made a stereo version of the cubic clipper. It treats both channels
independently but with the same parameters, as if there are two clippers set up
with the same alpha and hysteresis values.
This module will step trough recorded automation snapshots, this can be complete
snapshots or recorded single value changes. The control can loop once it reaches
the end, and it can be made to play in two directions.
I've added value history on knobs, modules and the patch as a whole. This with the
possebility to undo an edit without a recompilation being needed. It is also
possible to make selective undos on the patch by selecting a bunch of individual
items in the list. This works on the patch level only, for the control and module
undo only one item can be selected from a popup menu. See the images below.
With this I've also split the undo history up into two separate histories. A
history for structural changes (those can be undone and redone with
Ctrl+Z and Ctrl+Y) and this new 'value
history', which can no longer be undone/redone with Ctrl+Z and
Ctrl+Y. The control history, on undo, will not silence the
patch while a structural undo / redo does need patch recompilation, which will
stop audio briefly.
Also this causes automated changes (like randomization, morphing etc.) on the
patch to not be recorded in an undo history anymore. Which also causes live
morphing to no longer mess up the undo history, it is just not being recorded
anymore. (but you can still make snapshots for the automation recorder.)
Still working on a method to let cable changes not cause silence, as those would
not really be needed (a cable change results in some pointers being changed only).
And while at it, I've also made an event recorder. In its current state it can
only be used to play back snapshots, timed playback is not working yet. With the
new AutomationStep module described above this can be controlled. I'm working on
more features, and also working on keeping things consistent when modules are
added or removed from the patch (it's all a bit messy still, trying ideas mostly).
See image below, and the mini version of it on the left.
There is no automatic Save and Load currently, but the automation control screen
has a load and a save button. Automation is saved into a separate file with the
same name the patch jas file has and an .automation extension, on load it will be
checked if it matches the patch it was generated on.
Another thing I'm working on is an abstraction editor. It is possible to make
details of a patch public, the abstraction editor will pick up all the publics and
generate a module from that with the publics at odd positions, but stuff can be
edited. This is even more just in the 'gathering ideas' phase. Anyway, I'll not
disable it for a release even whn it is not very useful currently. See image
There is no automatic load and save for abstractions, but there is a Load and a
Save button in the abstraction editor. Abstractions are saved in a file with the
same name as the patch it was generated from using .abstract for the file
extension. On load it will be checked if the abstraction matches the currently
Added a 'unique ID' (GUID) to patches to keep track of mappings between patches,
automation and abstractions. These should not be to easily tracable to the
computer which generated them … when I can believe the µSoft docs on
this. Anyways, you can always strip them out, the patch will still load.
The GUID will not change during the lifetime of a patch, not even when structural
changes are being made. This means that the sync between patches, automation and
abstraction may still get fukt. At least its not possible to load abstractions and
automations on a patch it was never based upon.
I'm trying to make it so that when a patch is being change structurally and an
abstraction or automation is loaded for it that these will be updated in a
sensible manner. (And eventually such would mean automatic save / load will be
needed on abstraction and automation files (or tht they get stored into the patch
itself).) Anyways … for now, you will loose somme work probably while
working with the new features.
Single items can be reached (and be undone) trough the control popup menu.
Single items can be reached (and be undone) trough the module popup menu.
Multiple items may be selected to set them to the value present in the 'Prev. value'
The knob on the highlighted module marked in green is the currently selected
control, the one to which the graph applies. In the patch mod 1 is the player,
allowing for snapshots to be played in a clocked manner (not respecting the time
they were made on). So his acts as a sequencer, basically. A click on the send
button will send a whole column of values back into the editor. The Clear button is
destructive, it removes all recorded events. Points in the graph can be moved
around, when moved in value only the value will change for that step, when moved in
time a new time steop will be added. I'm working on more functionality.
Items marked in black are made public, the green item is the currently selected on
in the editor for which properties can be changed. This is a toy for now to gather
in version 2017.5.21.0
There is an issue with the Seq16 module in that it will sometimes not run due to
the number of steps being seen as zero, I've fixed this. Meanwhile a workaround is
to press F7 or F12 to make the sequencer run again.
There was still a race condition in the PulseSync module, solved now. A negative
edge on the sync input would clear the internal state, but the incoming pulse
might not have been processed yet (in case of a long sync pulse time). Made a
counter now which will clear the internal state and the output after a fixed
number of cycles, and in case the pulse input had gone active again within that
time it will just be armed anew.
Changed the module order on the Control tab to have all scaling modules grouped on
the left side.
Upped the maximum count values for the triggerable oscillators and LFOs. This will
change existing patches using these modules.
Added Hold mode on EnvAHD module. When on it prevents the module from going into
Release as long as the trig input is still active. When off the module works as it
Added an active output to the LfoTrig modules. This differs from the inverse of
the ready output in that ready will never be active before the first cycle ran
(and the module will not run then yet when it saw no trigger signal).
Added some swell control to the EnvAHD module. Swell when active will cause a rise
during the hold phase of the envelope. The amount of the effect can be preset and
modulated, and the shape can be selected. The time for the swell is the current
hold time. The envelope graph more or less shows the effect of this new control.
Added swell to the EnvADSR module as well. It works a bit different here, the
swell time is taken from the release time instead of the decay time (as there is
no hold time to be set). Also setting a large amount of swell will reduce the
attack level by up to one third.
With swell set to zero the modules behave as they did before. For shape
impressions see the image below.
Added a freq (unattenuated FM) input to the HrastOsc, also added shape and morph
modulation on it.
I've changed the sensitivity of all speed inputs on LFOs and oscillators. They
will run twice as fast as normal now with an input signal of 1. A zero input
signal will make them stop, and with negative values they will run backwards. Of
course the only wave shape changing on that will be the saw wave. This will change
the behavior of some existing patches.
The phase modulation on oscillators and LFOs was changed to something more
correct (for large phase modulation values), turned out to be a tad cheaper as
For the following modules I've changed the range of the slew control to make it
more useful: TypeFlip, SeqChord, Slew and SeqRandVal. This will change existing
patches making use of the slew control on these modules, however when slew was off
on them it will still be off for the new control range. Also it is still possible
to fully freeze changes with the control turned hard right.
Added 'function modulation' to the Gate module. The blue input when going from
zero to one will go trough the eight selectable logic functions (it is being used
as an offset (modulo 8) on the value from the Gate type selector).
Renamed the Pulser module into Wavolver ‐ as it was really that …
and then made it work properly. The renaming will break existing patches using
this module, the module will still be loaded but all connections to it will be
lost. But then again, the name was plainly wrong, and it did not function
properly either. The actual pulser thing can be done now with the triggered
oscillators, I've added pulse count modulation to those. These are all variations
on VOSIM of course.
Added a new display mode to the Quantizer module. It used to show fractions of
notes, it can now show fractions too. Not that this changes anything on it's
workings, but with the mode set to Fractions it is easier to understand how it
works for linear control of (low frequency) oscillators. For instance (see
illustration), when the fraction is 1/4 and the rounding type set to 'round' and
the output is used for speed control of an LFO the LFO will run at n/4 of it's set
speed (for n = 0 to 4 assuming an input ranges of 0 to 1, and assuming the linear
frequency control is being used (called speed in most places)). I've added the
same option for the built-in quantizer in the SeqRandVal module.
I've added different rounding types for the internal calculation. The module did
use the Round method before, but it can now also use Flooring and Ceiling. The
default still being Rounding. The illustration shows the quantizer set to 1/4 with
from top to bottom the input signal, the floored quantized value, the rounded one
and the one using ceiling. This option has not been added for the SeqRandVal
module, that one will always round.
Changed the Song module to have active 'end' and 'silence' outputs at patch
startup and after reset (both global and module specific), they used to be
inactive then. This makes it easier to use a PulseSync module on the 'end' and
'silence' outputs. This may break some patches that are making use of these
outputs, but not likely so.
Added a signal shaping module, it can transform a linear input signal into an
envelope shape type output signal with inversions and signal types being applied.
The module is rate smart, so it can be used for audio signals as well (but there
will be some aliasing). For shape impressions see the graph below.
Added yet another scaling module, the PropScaler. This one has a modulatable
parameter p which controls the in to out transfer as:
out = in * ( p - 1) + p.
This replaces a set of four modules that would otherwise be needed to make the
same functionality. Effectively p sets a lower limit on the output value and the
input range is proportionally distributed over the remaining output range.
Added a new sine oscillator with more modulation control than the other
oscillators have. It has amplitude modulation, phase modulation, linear FM
modulation (trough zero), exponential FM modulation and in addition the frequency
fine tune can be modulated trough the ct input. The sync input is a hard sync.
when it goes active the internal Phase accumulator is reset to zero.
This is after an idea by Ian Fritz. The analog xor makes it's output active
only when the sum of the input signals is over the set set Threshold value, but
below 0.5. The threshold can be set from -1 to +0.5, and it can be modulated
trough the T mod input.
Envelope shape impressions
Shaper shape impessions
in version 2017.4.22.0
Moogfilter could produce denormals - added some guards there. Also optimized the
filter a bit (20% off from the execution time), but it still uses an awful lot of
cycles ‐ more than a reverb does ‐ I don't really understand why that
Compressor could produce denormals - added some guards there. Pluck module idem.
Graph Image for S shaped attack on the ADSR envelope module was wrong, fixed it.
Made the S attack and decay modes working on the Seq16 module, they acted like exp
Added code for trying to get some better idea as to why PortAudio initialization
failed. It will report failures to the logs.
Added an off-beat output to the EuclideanRhythm module, it is active whenever the
regular output is not (and the clock is active). Basically it runs the reverse
pattern of the regular output.
Added an invert input to the Seq16 and the Sequencer modules. In invert mode the
sequence is put upside down.
Added inputs to the Switches4 module. When these are not connected it will behave
as it used to. And when they are connected they will be switched to the outputs.
Added range modulation to NoteQuantizer, ScaleQuantizer and ModQuantizer modules.
Added tanh( x) function to the Monadic module.
Added chain inputs to the TransInv module, these are just added to the calculated
Made the pulse delay module rate smart ‐ however ‐ the shown times are
eight times as short when the module runs at audio rates (this is being indicated
in the displays). Also made the MultiCompare module rate smart, so that now all
the logic modules can switch to audio rate computation.
Added some values to the MConst module to make it usable for controlling note
intervals. uni (0), mi 2nd (1), 2nd (2), mi 3rd (3), ma 3rd (4), 4th (5), dim 5th
(6), 5th (7), mi 6th (8), ma 6th (9), mi 7th (10), ma 7th (11) and octave (12).
These values are TET based and are fixed, even when the main tuning is changed.
Added a sent output to the MIDI CC output module to allow for easier sequence
Added a clear input to the Delay module, when it goes from inactive to active it
sets all the delay memory to zero. This is glitchy, as it was intended to be.
Added input attenuation controls for the Formant and Formant2 modules.
Added a speed input for Osc, Square, MultiOsc, OscTrig and Squaretrig
oscillators. This basically is linear thru zero frequency modulation. When speed
is set to zero these oscillators will stop and they can be used as lookup tables
by controlling them from the pm input. And while at it … also added speed
controls for Lfo, SquareLfo, LfoTrig and SquareLfoTrig modules.
Changed the maximum acceptable input level for the GVerb module to 100, in the
original code this was 1e5 or so and had that lowered to 10 before ‐ but it
could cause distortions that way … so upped it a bit. These are silly
levels, but this can happen when the input level is high and the output levels are
set low to compensate for that. Which would be silly to do probably, but I needed
some more forgiveness there.
Added a 'hold' mode for the EnvAR, EnvARRetrig and EnvConrol modules, when active
they will not enter the release phase as long as the trig input is still active.
Also removed a possible click from these modules ( on the transition from Attack
to Release or Hold mode) and removed some needless calculations … like
subtracting Ln( 1).
Added mode selectors for the FixedDiv and PrimeDiv modules, these can be set to
either Square or Pulse. In Square mode the outputs will all have a 50% duty cycle
and in pulse mode they wil be active for one positive input clock time.
Added a reversed mode for the arpeggiator, and added a reverse input for playing
the arpeggio backwards. The reversed mode plays the notes in reversed order of
reception i.e. last note first. Also added a transparent mode, when the
Transparent control is active no arpeggiator actions takes place but instead the
last sampled input value is being output when the output clock goes active (making
the module behave like a sort of a sample and hold module (except that it has two
sample clocks then).
Added a trans(parent) input on the Poetry module. When it is active no poetry
generation takes place but instead the last sampled input value is being output
when the output clock goes active (making the module behave like a sort of a
sample and hold module (except that it has two sample clocks then).
Added count modulation inputs for LfoTrig and SquareLfoTrig modules.
Added a level modulation input for the Quantize module.
Made the control decimation programmable, it can be set from the
Settings Screen F3. The default value is 8, what it always had been, but it can
be lower or higher now. Higher values will mean that less processor resources are
being used. Changing the value however also means that some modules will run
faster or slower, this affects some of the blue filter modules i.e. the slew times
may change etc. It is of no concern for the red (audio) filters though.
Working on some MANIAC features …
I've dded cross fade modulation to the Seq16 module and also added a random input.
The random input will set all the faders to random positions. Added three 'canon'
outputs, these are shifted by four steps for each one. So when the first output
outputs step1 the 2nd out sends step 5, etc. This way the sequencer can be used as
a 2 x 8, 3 x 6, 4 x 4, … sequencer, but it is also possible for instance to
set like 12 steps and use output 1 and 3 and then they will have eight sliders
overlapped and four unique ones for for themselves. The xfade works on all four
outputs, but there is still just one chain input which is being added to all four
outputs. Added a 'hold' input, which when active will hold the last sequencer
output value and it will also block the chain outputs from being updated, but it
keeps the sequencer running. Then added skip inputs for each step, when a skip
input is active that step will not be used (making the sequence a step shorter).
Added a clockskip input too, when it is active the sequencer will freeze at its
current step. This is different from the Active input in that the Active input
outputs no step values but passes on the chain outputs only. When a clock is being
skipped the current step and output values will be he held and the chain inputs
are no longer sampled. Then I went on and made 16 step outputs, going active when
the step is active and the incoming clock is high or the trigger mode is set to
(G)ate. Oh, and also added a TimeWarp (tw) output, which can be used to control
the speed of the driving clock, it is meant to drive the speed input of an LFO.
The idea is that when steps are being skipped the sequence length can be kept the
same when using the TimeWarp output in this manner.
Getting closer to the M.A.N.I.A.C.
Added the 'canon' outputs on the LUT module as well. Added a 'steps' control as
well to map a zero to one signal range on the set number of steps, step count
can also be modulated. Then added a random input for doing fader shaking too. With
a negative step count (which can be set trough the steps modulation input) the
unit will run 'backwards' ‐ that is ‐ for larger selection values it
will move to a lower step value. Note that with fewer steps and a ramp on the
input there will be less steps in the same time, so when used as a sequencer the
sequence will run slower then. Added 16 step outputs here too, there is no clock
or active input on this module, so the step outs are active for the whole time the
step is active.
Added a DC blocking module with a -3dB point that can be set from approximately
20 ~ 40 Hz to 400 ~ 800 Hz in audio rate (red, depending on sample rate) mode and
from 2.5 ~ 5 Hz to 50 ~ 100 Hz in control rate (blue, depending on sample rate)
mode. The further the knob is turned right the higher the cutoff frequency will
be. DC will always be totally blocked no matter what value the cutoff is set to.
Feedback delay lines (and there are a couple of other modules that do not like DC
very much) can lock up when there is DC present in the loop, this module can be
used to get rid of the DC, probably best before the loop and not inside it.
Added a MIDI Program Change Output module, also featuring a sent output for
easy command sequencing. It will send out a a program change message when the gate
goes active or when the program is manually changed. The actual program is the SUM
of the values read from the prg control and the progr input. A change on the progr
input will not trigger a message to be sent ‐ a gate signal is needed then.
Added a MIDI Parameter Number Output module. This module can send both RPN and
NRPN data and it can do so with the most significant septet only or it can send
both data septets. Added the chain output to this one as well. Data will only be
sent when the gate signal goes active.
Changed the compressor looks a bit. The LED scale was not very informative (as the
scale is pretty large things barely moved, and there was no feedback on actual
gain either) so … removed it, but added a dB value indicator showing the
Also added a side chain input and output. Effectively this is a 3rd compression
channel where the compression is being calculated on either the side chain channel
(side mode) or on the stereo signal channel pair (norm(al) mode). So it is also
possible to side-chain a mono signal from the stereo inputs.
Then added a DC blocker on the inputs as in RMS mode DC on the inputs can do some
weird stuff. The blocking can be switched between AC and DC. AC mode has a -3dB
cutoff point at approximately 20 ~ 40Hz (depending on the sample rate) and DC mode
does nothing, The filter will only operate in the compression calculation chain,
the signal chain is not being filtered.
Added a PulseSync module. The module gets triggered by an incoming positive edge
on the pulse input, which it will then hold internally. When a subsequent positive
edge appears on the sync input the output will be set active. Once the pulse got
processed the output and the internal state are both being cleared when a negative
edge is seen on the sync input. The incoming pulse is guaranteed to never get
This module can be used to sync up a pulse event appearing at a random time to a
syncing clock signal.
Added a Transpose module. Four separate transposition amounts can be set each
having an activation input. The output will be the input note value transposed
over the sum of the active transpositions. Transposition range is -24 to +24
Added a SquareSineLfo module, it outputs a sine along with a PWM controlled square
wave. I often end up clocking things with sine LFOs as sine are easy to multiply
up to higher frequencies (using the ChebGen module) which is easier often than
starting with a higher clock and dividing it down. Also the sine shape is more
useful to do (synced) modulations with. However its not easy to change the 'pulse
width' of the sine, so that is what the square PWM is for, that one is easy to
make changing note lengths for driving sequencer.
Made a PulseSkip module. It will pass on input trigger signals only when the input
level is larger than the compare level. Levels are computed as the sum of the
preset level (as made visible in the displays) and the modulation level. Once a
pulse got started will only be ended when the trig input goes inactive, so either
a complete pulse is being passed on or it is being skipped.
in version 2017.3.14.0
Bug in DataGraph module for the splines mode, it outputs X values instead of Y
values. Using Y works way better - so that is fixed now. And its not in the
AudioGraph module, as that one can not use splines ‐ pwew :-)
Mode and range changes through selectors were not handled when controlled with the
keyboard (only mouse clicks worked for that), fixed it.
Added an 'act' input for the EuclideanRhythm, SeqRandom, SeqRandVal, Pattern and
Added inv outputs on EnvAHD and EnvADSR modules, they output 1.0 minus the value
that is on the env output.
Made a controller module for the DataGraph module to allow it to do envelopes. The
help blurp says : A retriggerable envelope controller. The envelope will start
when the trig input goes from zero, or below zero, to above zero. The output will
then in the set Attack time go to the level set with the level control. When the
trig input falls off the output will start getting larger again for the set
Release time, till a value of 1 is reached. At every new trig start the output
value is reset to 0, and the above cycle will restart. Both the Attack and Decay
phase are linear, any non linearity is to be set in the controlled graph. This
module is meant to control the GraphControl module, the actual envelope shape is
then drawn in that module. The set level will then correspond to the vertical
distance in the graph where the hold value will be.
Made a pattern sequencer module to be used for drum patterns. When set at low
speed it can also be used to control the time structure for a self playing patch.
A full pattern is made from eight parts, where each part is composed of eight
sections and each section has eight elements. This makes for a total of 512 steps,
where each step controls 24 bit outputs simultaneously. The steps can be set to
off, to pass the incoming clock, or to be on.
For detail information please click on the module, or for some background
information have a look at the forum
Sequences can be saved to disk and reloaded just like any module (or any group of
modules) can - as a preset. Right click on the module and select load / save
preset from the context.
in version 2017.2.25.0
Range control for the pulses module was not working, fixed that.
Output signal type selection for the pulse delay module was not working, fixed
that. Also tried to clarify the help text a bit.
Constant module when put into 'note' mode is off. And this seems to be wrong in
more module types. Indeed it did affect the Constant, LUT and Knobs4 modules and
it is fixed now. Also made it so that when 'notes' mode is selected the note
controls have 128 steps now, and possibly some other number in the other modes.
Also the LUT module was having an offset in the 'notes' mode, this got fixed too.
These changes will affect old patches using any of these modules in 'notes' mode.
Made a Moog style 24 dB low pass filter.
Inspiration came from a github
site, the Huovilainen variation on the theme. I've found it to be a bit
explosive … the volume that is, so made it to oversample a bit more and
also added a saturation function (tanh) on the output, as it had a tendency to go
like very loud when resonating. This is not a cheap filter CPU wise, but I think
it to be a nice one with the built-in distortions. On the referenced web site
there are some more variations, I may try some other and make a selector maybe for
Made a fixed divider module.
The module has a series of outputs with fixed division rates. After
reset all the outputs will be off, to be toggled each time the number of input
clock periods matches the division count for that output. The output will stay
high for one clock period then.
Made a multiple waveform oscillator.
This module simultaneously makes a sine-, a
trianglar-, a saw- and a square-waveform. The PWM input provides linear Pulse
Width Modulation (on the square output only).
in version 2017.2.19.0
Delaytime range on stereo looper was not working, fixed now. Also changed the
start and length control modulation amount knobs to be displayed as linear (they
were actually working as linear controls already but not displayed as such).
Some screen painting took place on the desktop's device context, changed that to
happen on the editor's context only. This draws better selection rectangles and
better lines (on cable drawing). This may as well solve an issue with the screen
not being repainted properly (which was reported by someone using Windows 10). Also
thickened up the lines a bit for better visibility.
Added inv outputs on EnvAR and EnvARRetrig modules, they output 1.0 minus the
value that is on the env output.
MIDI transmit bug. The command was never sent only the data bytes … Fixed
now. And of course I had forgotten to test some code …
Added f(x) = - x and f(x) = 1 - x functions to the Monadic module.
Added a a modulation depth control module
. The graph illustrates it's workings
(where the red signal is the wanted modulation depth (mod), the green signal is
the modulation (in) and the blue signal represents the resulting output).
For the inputs and the output the signal type can be set, for the graph they were
set as in the module image. The mod signal and the output being unipolar and the
modulation (in) signal being bipolar. I've made this module primarily for AM
modulation (tremolo), but it would work for vibrato as well when applied to a PM
input of an oscillator.
The internal calculation used is : out = mod * ( in - 1) + 1, so that when mod = 0
the output will be 1 and when mod = 1 the output will be a copy of the in signal.
This is being calculated on signals assumed to be positive only (unipolar), the
input output type selectors are to be used to ensure such scaling.
in version 2017.1.29.0
When windows screen magnification is set to something else than 100% everything on
screen gets messed up. Turned off screen scaling, which paradoxically results in
better scaling - I guess the scaling property on forms is a promise to windows ...
anyways, windows will now scale all stuff properly - or so it seems.
Bug in LifeSeq module, the chain input does not work ‐ fixed now.
Obscure bug in the wave file reader code, it would read more samples than present
in the audio file which would cause glitches on playback. Fixed now.
Some of the 6 dB filters (the high pass parts) could produce denormals, fixed
Bug in multiplier module where the leftmost input would not work due to having a
wrong name. Fixed that, but all patches using that input were not working properly
and on reload will not restore the connection. When needed this can be fixed
manually. Open the patch in a text editor, search for all occurences of the text
'multiplier_in47' and replace all of them by 'multiplier_in7', the patch would
then load again as intended.
Bug on pasting modules. It looks like when a selection larger than what is
present in the current patch can not be pasted. This has to do with an internal
naming conflict for modules. This issue must have been present for years I think,
looking into it now. Is fixed now, it made the code somewhat simpler too :-)
Added a reverse input to the LifeSeq module, when active it plays from right to
left. Also added stops to the module. When one stop is set it will play from the
first column up to the row where the stop is set and it will then jump back. When
two stops are set it will start in the column after the leftmost and it ends in
the column just before the rightmost. When the reverse input is active this gets
reversed. When no stops are selected the full range will be used. On reset the
cursor will jump to the column after the leftmost stop, unless only one stop is
set ‐ in that case it will go to the first column. This also gets reversed
when reverse mode is active. I thinks the stops to be interesting mainly for the
Changed MIDI NoteOn handling a bit. When a NoteOn with velocity 0 is received it
gets translated into a NoteOff event. The velocity for the generated NoteOff was
set to zero, it is now set to 64 instead. Changed that for the MIDI file player
Implemented running status for the MIDI transmitter, this was already present for
Updated the MIDI decoder and encoder to be able to handle RPN and NRPN data, so
could make modules handling those MIDI messages now. Implemented a mode for this.
When 7 bits mode is on the callback will be called when the MSB comes in, in 14
bit mode the callback will be delayed until the LSB comes in. When MSB and LSB get
swapped by the sender it will still work, provided 7 bit mode is selected, as in
7 bit mode LSB reception will not trigger the callback (and the internal value
gets updated anyway on both MSB and LSB reception.
Data increment and decrement are implemented as well, in 7 bit mode they will
change with steps of size 128 and in 14 bit mode with steps of size 1. Increments
and decrements are clipped to the range [0 .. 16383] (which may get interpreted as
a range [-8192 .. 8191] depending on the actual receiver), no wrapping takes
place ‐ could make a mode for that too though I guess.
All this is in preparation for MPE (Multidimensional Polyphonic Expression)
… going to be some messy fun that for what's basically a mono synth
… sorry paraphonic :-)
But first MIDI will need little more attention.
For PortAudio I've changed the way buffering is handled. It is now possible to set
buffer size of zero, whech means that PortAudio decides on it. This works for ASIO
[As of 2019-04-10 legacy audio mode was completely removed.]
For legacy mode this is not possible, a 4096 byte buffer will be used in legacy
mode when 0 was set. I've noticed that the WASAPI interface gives me best results
for latency, but I've also noticed that I'll have to shave off some of the peak
load for some of the modules to be able to use that for larger patches. I've added
a buffersize display under the CPU load display to be able to see what gets
actually being used (it shows the number of samples delivered and requested by the
PortAudio callback). Setting the buffer size to zero may cause the VU meters of
the main window to be updated a tad irregular, but didn't actually really see
that. Also, in the Debug Screen F5 I've added latency indicators, they show what
is being reported by PortAudio.
Processing for the LifeSeq module is spread a bit better over time now. After a
change request was seen a column counter is set up which gets adjusted on every
control rate tick. On each tick only one column of the automaton gets updated. As
a consequence the updates should not occur any faster than once every 32 control
rate ticks (SampleRate / ( 8 * 32) Hz). Also the display will now get updated only
when there was a change.
Made it so that all MIDI sends from a patch on a channel > 16 are sent back into
the patch itself (and the channel is decremented by 16 then). This makes it
possible for a patch to send MIDI data to itself by allowing it to send on a
channel > 16. For the MIDI send modules these channels can be selected as ic 1 to
ic 16 (ic standing for Internal Channel). The MidiNoteOut and MidiCCOut modules
now allow for such channels to be selected, so a patch can now send notes and CC
messages to itself. Such MIDI messages will be handled internally in the synth
patch and will not be visible outside it (like in the user interface, or in the
outgoing MIDI stream).
Two new modules, a Phasor
and a painted waveform thingie, the AudioGraph module
The Phasor was created to control the AudioGraph, but it can also do phase control
The AudioGraph is better suitable than the DataGraph module for audio rates as it
internally uses Differentiated Parabolic Waves to reduce the aliasing a bit.
I've added an auto scaling feature which after every data alteration rescales the
graph to use the full Y value range. This feature can be turned on or off.
Also added an input to randomize the graph, randomization will occur when the
random input goes from zero or lower to something greater than zero. The actual
momentary value at that stage is used to control the amount of random to be
applied. Small amounts (near zero) of random cause the graph to be changed just a
bit, a value of one would create a totally uncorrelated new graph. Unfortunately
the rnd input acts sluggish … maybe something to fix later on.
Added a tilt filter from the music dsp site.
A center frequency can be set for it and a tilt factor for the
frequency response line. With tilt in the mid position there will be no filter
action, turn it up for more highs or down for more lows in the output signal. The
center frequency is the point the response line rotates around. The maximum tilt
is ± 6dB / Octave.
Added a txt end output to the Song module, it will fire when all phonemes are
spoken. It will not fire initially so it can not be used to make the module self
looping. However, the module does have a loop mode for that purpose. I've added it
to be able to only start the module when it was not already running (with an RS
Added hold mode to the Demux module. When active XFade will no longer work and
outputs will keep their value when they get muxed out (integrated sample and
hold). The new button toggles between cross fade and hold modes.
Added a sync output to the RandomWalkLfo module. The sync output goes active when
the output values change and it goes inactive halfway the LFO period. This so I
can get triggers at the LFO rate. Also added an output smoothing mode to this
module. When smoothing is off a stepped wave is produced, when it is on the wave
glides from old to new values. And also added a distance modulation input.
Added a maximum value display to the Value module, it will display the maximum of
the absolute value seen. A reset button was added too to be able to reset the
maximum to zero. Also a patch reset will set the maximum value to zero.
Changed ini file loading a bit, at first Wren will try to load an
inifile based on the name of the actual executable file name. When such does not
succeed it will try to read a file 'wren.ini' from the same folder the executable
is in. This assumes Wren originally will be called 'wren.exe', but
makes it possible to have a copy of the executable under a different name which
will then try to read the original settings (but they will be saved into an ini
file named after the actual executable fiel name). Anyway, this should make it a
bit easier to have two (differently named) Wren executables in the same
folder, and to have the new one to initially inherit the settings of the default
one (they can be set to differ later on). The looks will always be shared.
in version 2016.10.23.0
Oops ... a bug in the AM modulator causing an addressing error, fixed now.
I've changed the implementation for the intdif module a bit. For the integration
function the output value is divided now by the execution rate of the module. This
can be scaled with the gain control slider, but it means the output will be way
lower than it used to be. Likewise for the differentiation function the
output value is multiplied by the execution rate, and this is also scaled with the
gain slider. This results in way more output signal for the
differentiator. Also the output values for both integrator and differentiator are
clipped to a range [-10 .. 10] (this used to be the case for the integrator only).
The execution rate depends on the set sample rate and it can be either control
rate or audio rate depending on the rate of the input signals.
Trying my hands on some Markov stuff. Not sure if I can get the learning mode fast
enough for real time use, playback should not be a problem.
This is a first order implementation.
The internal state consists of a number of nodes each holding some value (for
which a maximum must be set with the max size control), a State variable holding
the current node index (which is not under direct user control) and a matrix of
transition chances (which is not under direct user control).
When the learn input goes active the current input value is looked up in the
nodes, and when found the transition probability from the previous input value to
this new one will be increased with the current strength value (which can be
negative). When the new value was not found it will first be appended to the node
list. When the node list becomes full and the mode is set to 'once' an overflow
condition occurs and the new transition will not be learned (also the Ovfl light
will go on then). When the mode is set to 'cont' no overflows will occur but the
oldest transitions will be discarded instead.
When the 'new output' input goes active the current State variable will move to a
new one, this is a random process controlled by the transition chance matrix.
The state of the module can be saved to and loaded from a file (must still
implement some user interface for that), so … training could be done in
non-real time mode when the need arises. Anyways .. let's test things first
‐ actually it does not seem to perform so bad, great :-) It will depends on
the input material, when there are many different values things may be worse, but
for discrete events (like note values) it seems all fine.
Now it needs to be made a bit leaky … as currently when the input patterns
change it will take longer and longer before those new patterns will be learned.
Although setting random strength values seems to help here ‐ just like
setting not too many nodes and using 'cont' mode.
Now working on some code to do do the Markov thing a bit more efficient so I can
hopefully implement a second order implementation which is fast enough for
Ok, that seems to work. It also seems to be not too hard to generalize it a bit to
arbitrary orders. I've implemented something, but found no time yet the check the
workings or the performance.
After a little break and some bug fixes … arbitrary order seems to
basically work, it needs more testing though.
Made a variation of he Average filter with direct frequency control (instead of
just exposing alpha). The set frequency references the - 3dB point. Filter falloff
is 6 dB / Octave, just as for the Average filter. The filter should work on
control and audio rate signals (Rate Smart).
Made a bandpass variation on this and then made a stereo version of that one too.
I've used this one on the final output of a patch to trim off excessive lows and
highs a bit.
When in control rate mode the frequencies are off by the control rate decimation
factor (currently 8), the displayed value being too high. I should fix that by
displaying the lower frequencies, but as of yet there is no proper mechanism for
Added a new tab for generators (gen) and moved some of the 'sequencers' there.
This tab is meant to hold modules to be primarily used for generative music. So
state machines, song generators, cellular automata etc. Moved some modules from
the control tab to the new tab as well.
Automata … well … just one automaton for now, but will likely make a
couple more based on the same underlying engine. This one is a rhythm generator
This module implements an 'elementary cellular automaton'. Each cell can have a
value of either zero or one. Three cells are evaluated according to the set rule
and this will yield a new value for a next generation cell. The rule number [0 ..
255] is being interpreted as a bit pattern which is indexed by the three bit
pattern generated from the three evaluated cells. This is all according to
Wolfram's theory about elementary cellular automata. The start state of the
automaton consists of one single cell having a value of one.
With the steps parameter you can set the number generations to be calculated, this
results in a bit string. On every step a next bit will be sent to the output and
when the end of the string is reached the sequence will wrap back to the beginning.
The Faults parameter sets the average number of cell faults generated for each new
cell calculated. The set number is divided by 10,000.
The total length of the generated pattern increases with two on each new evaluated
step, which results in length = 2 * steps + 1, which is odd always. The reset
input can be used to wrap back sooner.
The blue output has a weighted sum of the last three bits ( Bit[ anIndex] + ( Bit[
anIndex - 1] shl 1) + ( Bit[ anIndex - 2] shl 2)) / 8.0. The dir input when active
reverses the play direction.
Some rules will not generate any output at all (like rule 0 or 255), some will
generate simple repeating patterns and others will be almost random. For some
rules the steps setting may not change the sequence much, for others a next step
may differ wildly from the previous one.
For some ideas about possibly interesting rules have a look at my rules page. At
some stage I may build some of the visuals into Wren , but not at this
stage. That page, at the very end, also shows an implementation of the algorithm
And another cellular automaton based module. This one uses Conway's game of life.
the current help text says (subject to change though):
A Life based sequencer.
Based on Conway's game of Life, but there are some extra modes selectable from the
mode selector. The step input moves the cursor, which will put the cell values
from the current column on the green outputs. The change input triggers a new
generation of live cells, according to the currently selected rule. The blue sum
output has the column sum of the cells under the cursor, this can be a binary sum
or a linear one. The grid can be set to be wrapped or unwrapped. In wrapped mode
the top connects to the bottom and the left side to the right. The fault rate
parameter determines the amount of faults made in applying the evolution rule.
When the number of cells drops below the low count the grid will be filed with
random values, the chance of a cell becoming alive is set with the rnd amt
parameter. The rnd button will fill the grid with a random pattern, obeying the
rnd amt setting. The clr button will clear the grid, this usually will then
immediatly set random pattern as the live cell count drops below the minimum. It
is possible to draw on the grid, a single click will toggle the cell state, with
shift cells can be turned on and using ctrl instead will turn cells off.
Using a right click on the grid the current state can be saved to or loaded from a
This a a pretty big module, and 32 rows seems a bit much maybe, but I've found the
game of life to be more interesting on larger grids. The current 32 by 32 size
seemed a minimum for interesting patterns to occur. And at the same time 32 logic
outputs seems like overkill as well, for the sum output it's a nice amount though
(2.5 octaves in the note modes).
Made an inhibit control on the change input to be able to stop rule evaluation,
and also made a 'Freeze' mode (in addition to the various ;ife modes) in which the
current pattern will not change ‐ this allows for painting sequences.
Now if only I could get the module to load itself from a patch file … it
would then double as a 32 x 32 (event) sequencer (the sum output can be set to
make note values though) … it also reinitializes on every module insertion
or deletion … which it should not do. But otherwise it seems to work well.
Ok, the save and load stuff works too now, but the mode has to be set to 'Freeze'
for this to work properly (or else the internal Life machine will have it's own
ideas about the pattern).
The new grid control thingie could be nice for other modules as well :-)
When Wren is closed the current patch will be saved in the
application's home directory (where wren.ex resides) as
'__l__a__t__e__s__t__.wren'. This file can be reloaded from the file menu trough
the 'Load last patch' item. For just in case … This is not saved in the
most recently used patch list (unless you manually open it; it will not be opened
automatically on Wren startup).
in version 2016.10.2.0
Added the Markov module, but it has not been fully tested yet, see remarks in the
'working on' section above. Also I did not implement the code for saving and
loading models yet ‐ or rather ‐ I did, and then the code changed, so
it needs to be done anew.
Delay module, in 'short' mode, does not get correctly initialized on patch load,
it seems fine on patch recompilation though.
Or actually … several parameters were not updated in the synth after for
instance a range change on a mixer or when clicking an rnd button on a sequencer,
etc. etc. Fixed that, it was introduced with the randomization features in the
last couple of previous releases.
Made some new noise modules with Perlin like noise, in which the roughness can be
controlled and the pattern length. A clocked one and a free running one. New noise
values can be inserted to change the pattern. The maximum pattern length is 4096,
the length controls set a fraction of that. The number of layers can be set, more
layers give more smoothness, at the expense of processing time. The new input when
unconnected is active, giving noise like behavior, when it is set inactive the
behavior is tone like. When made active less frequently it will change the timbre.
I'll probably make LFO versions for this too, welp, would like to at this time, it
seems like a slowly clocked one can make nice percussion patterns. Also I'd like
modulation of the length ‐ ok, added that. Not sure how strictly Perlin this
is, I've tweaked the interpolation function from one layer to the next to more or
less give the same amplitude for smooth and rough settinging (but it is random of
With the above … there now are 200 module types.
And then when on the Perlin you do not set new values very often it makes sense to
have frequency control as well. The maximum sequence length was set a bit shorter
for this one (256 vs 4096 for the Perlin noise modules). This can make some nice
metallic or tone like noise sounds. I should add a button to the Perlins to invert
the new input. As it is now the new input is set by default, this is not always
I've added the invert option ( i(nvert) vs n(ormal), the latter being the default
value on the noise modules, the former for the tone module).
And another new module, Random scales … the help blurp says something like:
This module implements a scale randomizer. Each time the rnd input goes from
inactive to active, or when the rnd button is clicked, a new random scale mapping
will be computed. A scale will always have 12 notes per octave, and all octave
mappings will be the same (with an octave shift, of course). Within an octave the
12 notes will be randomly mapped to 12 (probably, but not guaranteed so, it's
random after all) different notes.
A simple arpeggiator
On each clock on the 'new in' input a new input sample is is taken and added to a
list of samples, the oldest value present will be dropped from the list then.
This list can be left unsorted, it can be randomized or be set to play the lowest
or highest note first .
The list length is set with the length control.
On each output clock on the 'new out' input the next stored value from the sorted
list will be output.
In random mode the order in the sorted list will be randomized when the playout
wraps back to the first value.
In unsorted mode the playout will be the recording order.
On the input side an acceptance filter can be applied. All incoming values can be
accepted, values differing from the previous value or values not present yet in
the list of samples.
The toggle input when inactive causes the selected sort order to be in effect and
when active the next sort order will be used. This can be used to toggle between
low to high and high to low modes, for instance.
When there are not enough input values available (yet) arbitrary values will be
clocked out until the playout counter wraps back to zero.
The reset input (or a global reset) will cause the playout pointer to be set back
to zero without changing any stored values.
And another new one, a probabilistic sequencer.
For each note the probability with which it will be played can be set with a
slider (and this value can be multiplied by the value on the associated input).
Notes have to be be enabled before they can sound, and note enables can be
selected from a scale preset.
The presets will be dynamically loaded from a file wren.scale that must be in the
same folder as the Wren executable file. When this file does not exist
Wren will create it. After that you can edit it with a text editor
(like Notepad for instance) and Wren will not touch it anymore.
A selected scale can be altered with the note buttons, and such a change will be
saved with the patch (making the shown scale preset name meaningless, of course).
The rnd button will set a random probability distribution, the clr button will set
a flat distribution. The distribution can be altered with the tilt control. A zero
value means that there is no tilt, a positive value makes higher notes more likely
to be selected and negative values will produce more low notes.
A transposition can be applied to the selected scale with the trans(pose) control.
C is the 'neutral' base note (not leading to any changes). Transpositions are not
reflected in the shown note pattern. The low and high note controls limit the
range of possible notes.
The sm(oo)th controls determine the likelihood that the next random value is
about the same as the current one. So with a larger smoothness the results will be
Added noise type selection to various noise modules (OSC and LFO). The default
type is Linear (which was the only available mode before), the added types are
Gaussian (with a fixed sigma of 1.0) and exponential (with a fixed lambda of 3.0).
This affects the Noise, NoiseTrig, NoiseLfo and NoiseLfoTrig modules, but
backwards compatibility was maintained.
Added mode control to the Pattern module. A mode can be set on each step, it can
be normal (nrm), random (rn1), weighted random (rn2, or rn3) or skip. Normal
meaning that it will do it's full count. The random modes will generate a random
number [0,Count] and use that as the step count. The first random mode considers
all numbers in the interval, the second random mode only divisions by 2, and
likewise rn3 uses divisions by 3. Skip will just skip the step. Also added
individual step outputs which will go active when the associated step causes an
output pulse to be made. The step outs will produce a 'per step' copy of the main
Added clear inputs to arpeggiator and poetry modules to allow them to forget
stuff. These are latching inputs capturing a low to high transition. When they are
in the set state a next value to be clocked in will not be the input value but
zero instead. After having been processed the set state will be cleared.
Added an Amplitude / Ring Modulator module, which is a multiplier with control
over modulation depth and modulation offset. This makes it a bit easier to apply
less than full depth AM or RM to a signal. The input signal type can be set, and
it can do a fade between amplitude modulation (AM) and ring modulation (RM).
And made a simple pattern rewrite module for which a couple of rewrite rules can
be specified, and an expansion level. There is a small alphabet of the letters A,
B, C, D and E. For each letter a substitution is specified in terms of other
letters from the same alphabet. For instance, given the rules:
Then a number of expansion steps is specified resulting in patterns being
generated from the start symbol 'A'. With the rules above the successive patterns
- A → ABC
- B → C
- C → D
- D → AC
- E → B
etc. The maximum depth to be set is 8, which would result in a pretty long pattern
already with the rules above.
- A → ABC
- ABC → ABCCDEDEACB
- ABCCDEDEACB → ABCCDEDEACBACBABCDEC
At each positive input step a next symbol from the generated pattern is selected
and the corresponding output will be made active. At the end of the pattern it
will start again from the beginning. Some rules (or all) can be left blank for
simpler patterns. All letters must be from the set [A..E], everything else will be
ignored for pattern generation.
This is after an idea found at
RobotPlanet, which in turn references this article.
Decided to generate all 8 expansions whenever the rules change, which made it
possible to have depth modulation ‐ which I've added.
Made a variation on the previous module generating values instead of events. It
also has a couple more rules to compose with. It adds a lookup system so that for
each active step a value is put out. The value knobs are mode controlled, one of
the modes being 'note value'. There are the usual dB and linear ranges too.
A common chain input is available which is simply added to the current rule based
value, and then there are individual rule chain inputs which are added to the
output only when the corresponding rule is active ‐ with this the module can
be used as a rule based multiplexer too (when all values are set to a neutral C4
or to zero, depending on the selected mode).
A TuneSmithy module. This is a melody generator after an idea by Robert
Inventor and it is explained at the website of the author. The sequence can be arbitrary long in
principle, but decided to cap it at 1024 steps. The pattern is a comma separated
list of integer values. The second link above explains the inner workings.
in version 2016.8.21.0
When during live morphing new modules are inserted some odd stuff happens. Ok,
only when the module does odd stuff, see next item ‐ the GVerb module would
crash. Ah, and modules get set to all zero values when inserted while live morph
is active. Aha optical mostly … the first variation gets set from the
default knobs and all other variations get set to all zeros, which looks confusing
in a live morph situation. Still it would be better that on insertion of a new
module all the variations would be be set the same ‐ according to the
default knob positions. Ok, fixed that.
Live morphing when the GVerb module is present sounds bad (even when it is not
being morphed), for FreeVerb it is fine tho. Ah, it was being recreated all the
time even when the parameters did not change, and then it would crash. Ok, fixed
it. But there also is the issue that automation will change parameters that it
should not change. Ok, stopped that too.
Bug in DataGraph where the rightmost point would always end up top right (1,1)
even when it would not be drawn there. Fixed.
The active variation randomizer got mixed up a bit with the randomize function
from the morpher, unmixed that. The active variation randomizr is selected from
the action menu or by pressing 'R' on the keyboard and it sets random values for the
current variation only. The randomize function for the morpher sets random values
for variations 2 to 6 (for quickly generating a handful of new genes).
Bug in ScaleQuantizer, the selected scale selection is not properly read from a
saved patch, the buttons are OK though. Was an issue for all dynamic count
selectors and got introduced with the introduction of variations. Fixed now.
Oopsie .. the filter inversions used '-' for 'off' (which can not work anymore
after some changes on how dashes are handled for popup menus in relation to
automation) changed them to '.'.
in version 2016.8.20.0
Made a Time Of Day (Tod) module which is a saw tooth LFO with a period of 24 hours
synced to the local PC clock. So at 00:00 it's output value will be zero and at
01:00 it will be 1 / 24. The clock has millisecond resolution. The output value
may drift a bit as windows does time syncs when it feels a need ‐ but it
will be the best estimate of the current time. A couple of level shifters and
ChebGen modules could be used to obtain signals that run faster. Like first make
the signal bipolar, then multiply by 24 (first by 12 then by 2 using two ChebGens
in saw mode) and then level shift it again to unipolar positive to get a signal
with a rate of one hour. Then the same trick with a factor of 60 to get minutes,
60 again for seconds.
Hmm … instead, I've just added a couple of outs for hour, minute and second.
Made a new option for the morpher, it can do a live morph over the breeder too
now. Normal morphing is between the left and the right morph and morphing over the
breeder uses all eight variations present in the breeder instead (when the breeder
is morphed with the Morph button this is equivalent to the non breeder morph, but
when mutate or mate is applied variations will be introduced). This new option
tends to (depending on settings and actions) give more lively morph results (like
for the XY controls the dot can then move in a rough curve instead of in a
straight line only).
And made a new 'soapy' worm shape ‐ how cruel it now seems to hold them in a
cage like this …
I've turned the morpher code inside out a bit, that is, I've made variations for
patches and based the morpher on that. Variations are stored with the patch, so
they will still be there after patch reload. As the variation values are stored in
the controls (knobs and such) now, this also means that module additions and
deletions do not mess up the worms for other modules anymore. All this
needs some work though still; the worms are not always updated correctly, and
there is an issue with XY controls loosing data or something ‐ but it
stopped crashing at least :-)
Changed the randomizer as well, it now works through the same mechanisms as used
by the morpher. It will set random variation values for the active variation now
(which then in turn moves the knobs, the randomizer used to work on the knobs
directly (which would then change the current variation)). It is all a little
easier this way. Anyway … some more work to do on this.
Most stuff is working now, but drag-drop of worms is messy still … some
more work …
But it's not working well; turning some more stuff inside out ‐ the
mutations are all carried out on the patch now and then update the worms, instead
of working on the worms and updating the patch. Let's see what surprises will
arise from that …
Ok, that goes better, it seems to all work now as it should ‐ except for it
being very slow all. Not sure if its the editor updates or the worm updates
… there are a lot of params that change on a live morph … more work
Improved the speed somewhat, still not very fast, but it will do I guess.
Livemorphs are controlled trough the lights update routines, that is not ideal
(as the tempo goes all over the place) and this should be changed. Ok, changed
that, it is currently updated at a rate of 10 Hz triggered from the audio
generation rate (and so independently of the lights update rate). So this should
work properly for recording to disk as well. The exact rate may change later from
10 Hz, not sure yet, or make it a user pref maybe.
Live morphing works over the variations now, and I've removed the left/right morph
worms. When a left/right morph is wanted you can do a morph operation on the
breeder, then copy the worms to the variations and then perform the live morph. So
with less options the same functionality is still available.
Added a couple of buttons to the randomizer / morpher screen to make copying
easier. Drag / drop now works properly as well. Moved the live morph options from
the randomizer to the main toolbar ‐ turn auto mode on or off, and the morph
control knob. Morphing no longer changes any variations, instead there now is a
temporary variation which is being used.
Changed some keyboard shortcuts related to the randomizer / morpher and added some
(Ctrl+R still is recent patch reload).
- R - randomizes the patch (was Ctrl+Shift+R).
- Shift + R - toggles the exclusion visibility. (was Ctrl+Shift+Alt+R)
- Alt + R - shows or hides the randomizer window.
For changing variations you can use the keys 1 … 8, or they can be selected
by clicking on a worm in the toolbar or from the View menu. Variations can be
dragged from one spot to another, or even from or to the randomizer window.
Made it so that it is still possible to see all of the toolbar in the space needed
for four horizontal modules … which is like 1024 pixels. This would be a
practical minimum value for the horizontal screen resolution needed for
Wren (see image above). And also added the mutator commands in there
(as well as into the action menu).
Had to change the patch format again … not really for the variations, but
could not store the active variation with the patch in the current format. Made
the format a little more flexible again, so that future patch parameters will not
need a new patch format (which already was the case for control and module
parameters). Anyway, it's backwards compatible, but not forwards, as in: old
patches can still be read.
Made an Auto module which can perform Auto-randomize, -mute, -mate or -morph in
the same way the buttons in the randomizer / Mutator window do. Making this a
synth that can truly auto-mate!
The '-' avoidance code has a bug it seems … causing Wren to
hang in an endless loop on some patches when the randomizer is used ‐
investigating. Ok, solved that, but still saw dashes on morphing … fixed
Which … of course … introduced an issue for selectors having only
two values and no dashes at all, fixed that too.
module can not support randomization on the delay parameter. I disabled that. The
problem is that when delay and shift are changed at the same time invalid internal
values can occur. All automation has to be and was set disabled for that knob.
When a control was set to be non-randomizable, or non-automatable actually, and in
the patch it was stored to allow for randomization the red exclusion rectangle
would not show. Fixed that, and when such a patch is saved the control will be
saved as non randomizable too.
PitchShift module would produce denormals under some conditions. Probably fixed it
with normalizing the linear interpolation code's output.
Gabe made some new bitmaps for the looks. I've made some new ones too – had
a lag of about 40. Looks file got updated to 2016.8.2.
in version 2016.7.27.0
Fixed the patch randomizer for selectors having '-' values (these are meant to
generate separators in the associated popup menu). Trough the menu interface such
items can not be selected, but the randomizer found 'm allright …
Ah, and clicking on a selector or using the cursor keys or the space bar hits them
as well. Fixed that too.
Made the Pad, Spiral and DataGraph modules randomizable.
I'm working on a patch mutator, it's very much like what the G2 has. Still some
quirks to iron out on that, like what to do when the patch gets modules inserted
or deleted. The current implementation does not handle that as it should. Also the
images seem to not always be drawn correctly. Ok, found the issue, I'm misusing,
sort of, some selectors as indicators (on the ScaleQuantizer). This means that the
chord selector has to be disabled for randomization. It should maybe also indicate
that the key selection does not map to the selected chord anymore. But then again
… I could just consider the selector to be a preset selector only, not
reflecting the actual state. On a patch reload the actual keys will win anyway.
I may change this when i get ispiration.
I've made a FAQ entry for the morpher / randomizer.
The Wren forum thread may have some info not listed
Made the mutator work for Pad, Spiral and DataGraph too. There are some issues
with the DataGraph, as points can be added to or removed from it. This means that
the number of data points present in the mutator may change. What I did for now is
that such a situation is detected and the mutator will be re-initialized (which
looses all the current worms in the mutator, making a random set of new ones, and
the scratch area will be cleared). Module insertions and deletions are handled in
the same way.
Added two extra worms to the mutator, a left and a right morph, and a knob to fade
between these two. Also added a modulation option to the knob, and a new module
which can do the fade from the patch. This works at the lights rate, which is user
controllable, but it can not be very fast nor can the exact timing be very
predictable, as a lot of screen updates need to be done. Still for slow morphs
between two variations it would be nice. It is possible to have more than one
LiveMorph module in the patch, but things may get a bit jumpy with that.
The randomization exclusion works for both the randomizer and the mutator in the
same way. There is a keyboard shortcut to visualize the exclusions, and also when
the randomuzer / mutator window is active exclusions will be made visible in the
patch (with red rectangles around the excluded controls or modules).
Made it so that when patch randomization is requested (as opposed to single
control or single module randomization) and there are no modules selected, that
all (randomizable) modules are randomized.
Changed some code for the components introducing a couple of interfaces for them:
IAutomatable (for MIDI control), IRandomizable (for plain randomization) and
ICollectable (for controls that can contribute to the genome used for patch
mutation). This makes things a bit easier to program, I hope … it was quite
some work. Please note that non automatable controls can not be randomized either.
In general anything that needs patch recompilation can not be automated or
in version 2016.7.23.0
Made a scale control for the Reverb module. It sets the internal length of the
delay lines to some factor of the original length. This alters the color and
buzziness of the reverb. As the lengths of the internal delays are modified a
change of scale needs patch recompilation( e.g. a change causes a brief silence).
Added an End Of Cycle (EOC) output for the AHD and ADSR envelopes. When fed back
into the trigger input turns the envelope into an LFO, but can it be used to chain
envelopes as well such that one triggers after another.
Moved the Constant, Amplifier and Ratio modules to the Math tab. This distributes
the modules a bit more evenly over the tabs.
Working on a patch randomizer. Randomization is possible on Knobs, Sliders and
Selectors. It is possible to randomize a single control, all controls on a module,
or al controls on a selected set of modules. There will be options to disalow
randomization on selected modules and on selected controls. The action menu will
get an option to make visible the modules excluded from randomization.
Made a randomizer control wimdow to make it possible to exclude module types and
control types, partly by category.
For the randomizer I had to change the patch format to allow for a new module
parameter type. Old Wren patches can still be read by the new
Wren but new Wren patches can not be read by old
Wren versions. I did make the change in such a way that in the future
it should be possible to add new module parameter types without altering the patch
version. This already was the case for new detail (knob etc.) parameters. Old
Wren's should then just skip the parameters they don't know about.
in version 2016.7.4.0
Made a popup menu for the DataGraph, or for it's DataMaker component rather. Some
wave shapes can be selected and some basic operations on the data can be selected.
With the symmetries there are some duplicates in the operations, like a Bell
shape is the same as a vertically flipped cosine, and for the squares there are
some duplicates as well. Added load from file and save to file to the menu. Added
a left – right adjustment feature as 'make loopable' to the context menu.
MixMS8to1 and MixM16to1 were not working without modulation signals being
connected, fixed that.
The timer controlling the lights and VU meter updates is started in running mode.
When the startup is slow for some reason this may cause the VU meters to be
updated from non‐existing objects. Fixed that by disabling the timer at
startup, after each patch compile it was turned on already, so things should be
all fine once an actual patch is present.
Decided to make all timers inactive during startup to enable them only in
FormShow or even later – when needed. Also typed some extra checks around
the VU meter calculator objects to see if they were allocated before being used.
MIDI TX messages were logged from the audio thread directly (when logging was
enabled), which was a bug. Fixed that.
Error in DataGraph module. It was meant to react to double left clicks or
ctrl+left clicks by adding a new data point, but ctrl+double left click would add
two points and would then mess up the order of the points. Fixed that and checked
at other places if such a problem could exist elsewhere ‐ did not find such
There is some debug stuff left in the looper module (like the light), this should
be removed once the thing gets settled … which currently it is not.
Delayed help file creation a bit so that the looks are completely set up before
the images are being created ‐ image previews on the module selector look
Made the trigger LEDs for the envelope modules light up 125 ms always, even for
very short trigger pulses. Some more will follow as some modules output very short
trigger pulses that could not been seen on all input lights. This also means that
for a non changing input signal the light will go off after a while, it will only
show the trigger moment. made an error in this making the sample and hold turn
into a track and and hold … and fixed that again.
It may be the case that there is a denormals issue for the Formant2 module,
running some debug tracking on that, but lost focus … focusing again
… After 10 or 11 hours of runtime on a patch, yup, got a denormal report,
looking into the why now, it is on the output signal, should not be possible
inspecting the code, odd.
Restyled the module documentation a bit and added a link to it in the main menu of this page.
Also for newer modules I started to add links from the module images in this page
to the documentation page (meaning that module images here are clickable now).
Also fixed some inaccurate help text.
I've started using a mark‐down processor for easier markup of the generated
help file. See the Acknowledgements section for
more on this. This will allow for clickable links in the help, better looking
lists, etc., Expect more help restyling.
Changed some things for popup menus. They should now popup near where the cursor
is instead of near where the top‐left of the control happens to be. Also in
the patch editor I've made the menu key work for the element the mouse cursor is
Moved wire toggling and wiggling from the edit to the action menu, keeping the
same shortcuts (Ctrl+space bar and Ctrl+W)
Made three new mixer types MixQS2to1, MixQS1to1 and QuadPan (which actually is a
panner, but it's on the mix tab …). Although these do have blue connectors
they are intended to be used for audio mainly and when audio signals get connected
the connectors will change to the audio rate color. The panner uses an equal power
distribution only. The mixers allow for 24 dB of amplification.
Made an audio to control rate converter, which can be used to keep large areas of
a patch from becoming red.
Added a 'retrig'(ger) option to the LfoTrig module. When set into retriggerable
mode and when the trigger input goes inactive and active again during the current
pulse train the module will be reset to start the train anew. The default setting
is 'normal' for compatibility with older patches. Together with the DataGraph
module this could be used to make retriggerable arbitrary envelopes, for instance.
Hmm … not quite … anyways …it seems useful … maybe.
Added a Polar
module, it translates rectangular coordinates into polar coordinates.
Added a Rectangular
module, it translates polar coordinates into rectangular
Made a stereo version of the looper. It still has the drawbacks of the mono one,
but really wanted stereo.
Added a Spiral module, an experiment really. Per spiral winding of mouse movement
a settable amount of note steps is added. Along the positive X axis the white
places correspond with the steps. As this internally works with polar coordinates
I made outputs for the angle (phi) and distance / radius (norm) as well. The len
represents the spiral length for the current (X,Y) location (from which the note
output is calculated). So at least it doubles as a Polar sort of Pad module. (But
the latter can also be done by adding a Polar module after a Pad.)
Added an unattenuated frequency control input for the Lfo, SquareLfo and RandSig
modules. This makes it easier to add overall speed control to a patch.
Added an MConst module for selecting some more or less often used mathematical
constants, like 1, 2, e and 𝜋 etc.
Added a 'wrap' mode to the DataGraph module, the standard behaviour was 'clip',
which still is the default. In clipped mode an out of range input value will clip
on one of the graph edges and in wrapped mode it will wrap around and come in on
the other side again (modulo operation).
Added f(x)=x·𝜋, f(x)=x·2·𝜋,
f(x)=x∕𝜋 and f(x)=x∕(2·𝜋) functions to the
Monadic function module. Also added f(x)=floor(x), f(x)=round(x), f(x)=ceil(x) and
The Pad and Spiral module did not set their output values after a module load or
patch change, it needed a movement first ‐ I've changed that. I may have
thought this to be a feature for the Pad module, but for the Spiral it was just a
Removed all the all-too-obvious email addresses from the web page and
Removed some unused stuff here and there from the code, making the executable a
Added some volume damping around patch recompile operations, the worst of the
recompilation click is out now.
I've made an option to disable automatic doc file generation. This option is off
by default (so docs will be generated out of the box), and this feature will only
work when you use the old wren.ini file, e.g. when you copy the new executable
over the old one (or copy the old ini file to the new location). You can always
manually generate documentation from the Settings Screen F3 by using the
'generate documentation ...' button there. Note however that when no docs were
generated the popups from the module selector will not show a module, or an old
version of the image.
I've added command line passing of a patch file name. You can now associate
Wren with '.wren' patches in windows such that double clicking on a
patch will open it in the Wren instance that got associated. It is
still possible to run multiple Wren instances, but they will have to
run from different directories now, and patches opened through the explorer will
always open in the associated instance. When starting Wren directly it
will still open the last patch it had loaded as well. BTW it is possible (it has
always been) to copy patches, or parts of it, from one Wren instance
to paste them into another) or into a text editor, or whatever program that will
in version 2016.6.5.0
Changes in text controls and data graphs marked the patch for needing
recompilation which was overkill. It should have been marked for change only.
Changed that, and internally renamed some 'PatchChanged' stuff into 'Recompile'.
While all modules seem to be properly execution locked now during heap operations
the patch as a whole was not locked. Changed that.
MIDI CC assignments to knobs are not working properly ‐ CCs move the knobs
alright but the synth value is not changed. Ok fixed that, but now on very fast
MIDI I'm getting all kinds of memory issues. Ok, that actually is due to some
errors in MIDI transmission ‐ it was doing memory allocations from the
audio thread. Ok, four bugs fixed in this … messy, but fixed now.
Added exponential FM control to looper. Oh that it was no longer clicky was a
mistaken observation, with long loop times or record being off it is not too
noticeable maybe. And a remark about the interpolation mode, all pass mode is not
to suitable when the interpolation point moves relatively fast, that is for
in version 2016.5.29.0
in version 2016.4.27.0
There was a bug in how the selectors worked, setting the value to zero by clicking
on the selector instead of using the left mouse click popup menu would not set
dependent modes correctly. Fixed that.
There was an error in the stereo RMS calculation used for the compressor. It
calculated the RMS of the sum of the left and right signals where it should have
calculated the root mean of sum of the squares. Changed that, it works better this
Fixed the range setting for the MixMM16to1 module.
Changed the looks a bit for this web page, checked the spelling for it (and for
the module help) a bit better, updated the online module docs and updated some of the module help
I've made the IntDif module rate smart, it would not otherwise differentiate
control rate signals the preferred way. Also added a gain control for the
integrator to make it possible to avoid clipping. Also made a horizontal slider
control type for this. Transparency acts on the differential output only and gain
on the integral output only.
Adding RMS mode to the envelope follower module. Note that in RMS mode the RMS
averaging filter time is set from the Attack time, which is a bit odd, but setting
it from the Release time seemed even less even.
Added a 'saw' mode to the ChebGen module. The image shows this mode for the order
five output for some standard input wave forms. Basically this is:
out := FloatMod( N * ( in + 1), 2) - 1;
This would 'sort of' work for non integer N as well. Will make a separate module
for that, the SawMill module.
(Also changed the graph colors back to something a little less psychedelic.)
Added the SawMill module. This module is rate smart and will switch to audio rate
when such a signal is present at one of it's inputs. The frequency multiplication
proper only works for integer multiplication factors, for non integer factors it
is more like a PWM effect on a sawtooth wave (that is the 'sort of' disclaimer
mentioned in the lemma above). This is quite a nice effect on sine or triangle
waves as well. This is a naive implementation … up to like 200 Hz this is
sort of OK.
Example output for the SawMill with modulation in the range [1.0, 5.0], controlled
by the red triangle wave. The blue signal is the input saw and the green signal
the frequency multiplied saw:
Added pulse modes to the divider modules. They can now give a square output, a
pulse output (a copy of the current binary value of the trigger signal) or pass
the analog trigger signal's current value when it is larger than zero and false
when it is not. The output level to be used for falseness can now be set on the
divider modules too – see next lemma.
Added output level selection to some logic modules. The default mode will always
be bipolar and non inverting to be compatible with the old situation. Any lights
will still show the internal logic state independent of the selected output type.
Added this for the divider modules as well now.
Added a 5 to 1
and an 8 to 1
stereo mixer to save some screen space in patches with many stereo mixers.
Made a phase distortion module, after an idea found here. The input signal is fed into a (short) delay line, and
the delay time is being modulated with a low pass filtered copy of the input
signal. The modulation depth can be set and modulated by an external signal. In
addition to the modulation from the filtered input signal The delay time can be
modulated by an external as well. The filter Q can be set and it's cutoff
frequency can be set and modulated.
Added Hysteresis to the CubicClip module, and moved it from the Control page to
the FX page (but it still can run at control rate as well). The default Hysteresis
value is zero for compatibility with old patches. It is very possible to let this
module clip now into one of it's two extreme states due to the positive feedback
used for Hysteresis control. I'll consider this to be a feature.
Added an option to choose between two modes for the fractional delay calculation
in the KS module. It used to have linear interpolation only, but can now do
all-pass interpolation alternatively. The difference is a bit more subtle than I
had thought it to be, but still may add this to more delay based modules.
Added an AllPass filter, which is a delay line with feedback and feed forward
around it. The feed amount can be modulated, which moves the frequency for the
'halfway phase shift point'. As it is a delay based algorithm it ended up on the
'delay' page. The delay time is specified in (whole) samples, so the actual delay
time depends on the currently set sample rate. I put this in mainly to do some
more experiments with phase distortion on arbitrary signals, it may be useful too
for reverb experiments … or to help make stuff LAWDy like (by smearing out
peak energies a bit for impulse like sounds).
Added a Rungler module. This is after idea from Rob Hordijk. It is to be found
amongst the logic modules. The idea is to let the Rungler control the frequency of
an oscillator where the output of that same oscillator goes to the data input of
the Rungler. The clock input is then clocked by an LFO, or faster for weird
effects, and from this little semi random patterns will emerge. For more info see
http://hordijk-synths.info/ (which is not Rob's own website). It is nice to
put some note quantization into the frequency control feedback loop (for tuned
patterns). This module is rate smart, it will switch to audio rates when needed.
I've made the number of taps used settable as well as the shift register length.
It is possible to set the number of taps larger than the length of the shifter,
this will result in non proper working. Default settings are 8 shift steps and 3
taps, like the original circuit has them. The output value is scaled with the
number of taps to always have a range [0.0, 1.0], and after that a final level
scaling is applied.
Added some measurement ranges for the Value display module. Possible settings
currently are : 'units', 'dB', 'Freq', 'Note', 'Lfo fast', 'Lfo med', 'Lfo Slow',
'Lfo BPM' and 'Lfo Lin'. All ranges assume the associated module parameter to be
set to it's minimum value and the modulation input to be fully opened for a
bipolar control signal. I.e. for the LFO fast range it is assumed that the speed
is set to it's minimum value (0.254 Hz, knob position 110) and that the 'fm'
control knob is set to the max (1.000).
In the image shown the current cycle time for the bottom LFO will be 10 minutes
and 36 s. (The top LFO provides the modulation for it.)
Made a logic pulse delay module which can either delay an incoming pulse or
stretch the length of it. The delay/stretch time can be set to the same ranges as
available for envelope times, and it can be modulated.
Added positive and negative Q modulation to the SVF filter.
Made a tap control … as in tap twice and it gets the tap rate from that to
control LFOs with, or delay times. There is a rate multiplier which can set a
factor of the tap rate to be applied to the control output. Also added a slave
input and output to allow for chaining of Tapper modules (so one can be in LFO
mode and another in DELAY mode,or a different rate multiplier can be used). Tap
rate can be set by clicking the mouse on the tap area or when the keyboard focus
is on the tap area (when it has a red colored border) the space bar can be used as
well. It is also possible to set the rate with a knob, or by typing it in
There is a sync output which goes active on every tap action, and also the sync
button can send a sync signal independent of the tap control (this is meant for
resetting LFO's, for instance).
An example patch for the tap control:
in version 2016.3.5.0
2016-02-20 : A user reported that when the windows theme is selected the text
used on the module selector tabs is too large (not fitting the buttons). The
styled font drawing for the module selector is a bit tricky, made a change to
TKnobsModuleSelectorStyleHook.DrawTab which might help, at least the font is
controllable then from the form designer … and no longer defined by the
component's owner, so should be fixed to 8 point Tahoma now.
Fixed the memory leak in the TextSequencer module, it was present after all.
A new module, the Pulser. This is after an idea from Ian Fritz. This
module only implements the pulsing bit though, and not the wave wrapper. The
pulsing works by letting trough the 'signal' input when it's absolute value is
larger than the 'shape' input value. When this is not the case the output is
zeroed. To get the full implementation of the Wavolver the 'inact' output could be
used to control mixing in some other signal, for instance from a wave wrapper
module. This output goes active when the main output (out) is inactive (zeroed).
And another new module, the Formant2. This is a formant type of filter, or a vowel
filter, for which presets can be loaded from a file.
Sofar I've typed in formant sets named swedish, usa, french, csound_alto,
csound_bass, csound_countertenor, csound_soprano and csound_tenor. See the
supplied wren.formants file for details. Not all formant sets are equally well
specified, only the csound ones did specify levels and bandwidths.
The filters used are second order State Variable Filters used in bandpass mode,
one for each formant. Bandwidth is being set by calculating a Q value and applying
that. When no bandwidth was specified a Q of 10 will be used.
in version 2016.2.14.0
Tested Wren with the JACK Audio
Connection Kit, looks like it is working fine here, but note that you'll need
to select some channels to be used from the audio setup. I've added a FAQ item for this.
There is an issue with MIDI connections, when the server can not be found there is
an endless amount of popups telling so. Ok, fixed that, and a couple of other
possible ways for the NetMidi connection to go haywire. When something fails
unexpectedly the auto connect feature will be turned off now, and error messages
will go into the log instead of into a popup window.
There is an instability in the Formant filter in that it makes huge signal levels
when modulated at audio rates. Ok, can fix this by setting the update rate lower,
at a tenth of the control rate it seems to behave better – set it to 1 / 20
to be sure and at 44k1 this still seems smooth enough without filtering (update
rate is then about 275 Hz).
An idea came in to make double click on a cable work like ctrl+normal mouse click, i.e. move or delete a connection.
This was supposed to be working, but the double click was tested at the wrong
point in the code, fixed it.
Added an average (avg) output to the Median module, it outputs the average value
of all samples present in the sample bin. This provides a linear moving average,
as opposed to the exponential type the Average filter and the Slew module use.
Added a simplified version of the Karplus - Strong oscillator. This one uses less
CPU, the noise stuff went out and modulation update frequency was lowered to
control rate. Also the brightness modulation was removed.
Changed the data buffering for the scope module a bit, seems to work better that
in version 2016.1.24.0
Thanks to Gabe I found a bug in the sample rate switching mechanism. Or two
actually. First one would freeze Wren when using legacy audio and
switching the sample rate. Second one caused general protection faults in the
PitchShift module on a sample rate change. I've fixed those.
Looks like I fukt the math modules … if they ever worked … something
to look into. Ok, it was only the monadic function module, fixed now.
The Third filter had some trouble with denormals, changed the code a bit, this
affects all the multi-filters.
Solved the no-drag issue with module imports. Also made a distinct module import /
export folder to be remembered over Wren sessions. This introduced a
problem with regular module insertions … modules sometimes jump to silly
places now … will need to check that. Found a quick and dirty patch for it,
this will need some more attention at some stage (as the underlying issue is one
that has been present for a long time already), but it's fine for now.
Module import and export:
Made regular patch folder and template folder to be remembered over
Added an Alternates item to the module pop menu, this can be used to change the
type of the selected module. After the change it will try to reconnect the module
(based on connector names, where they exist in both the old and the new module)
and it will try to copy over parameters from the old module to the new one (based
on control names, where they exist in both modules). Wire and parameter copying is
not always possible however as there are some naming inconsistencies between
modules (and to solve those would mean old patches can not be read anymore, not
naively at least).
And while at it … implemented non strict parameter paste as well, where you
can copy over values from one module type to another type. Only the matching
parameters (name based) will be copied in that case and the rest will be skipped.
This fuzzy parameter pasting can only work from one module to another.
This does not really work like I'd want it to work yet … it needs some more
Added peak / RMS mode selection to the compressor module. It used to do peak mode
2015-09-24 : Different distribution laws for panner, say power law or volume law.
Added some different fading curves for the XFade and Pan modules: linear (equal
volume), square root, sine (equal power), sine squared (s-shaped) and squared.
Added a mixing level scaler, an "attenuverter'" like thing. The
MixingScaler is on the Control page.
Added a a stereo widener module. For signals that are stereo already the base can
be widened by raising the inversion amount, this adds a bit of -L to R and a bit
of -R to L. For signals that are mono some delay can be added, amount and delay
time can be set. Modulating the delay time will result into a flanging effect.
This module can be found on the FX page.
Added a smaller mono mixer, taking less patch space. It has one input less than
the other mono mixer, but is still modulatable, mutable, chainable and rate smart.
The MixM3to1 module can be found on the Mixer page.
Added a volume control, with a preset, a mute and modulation. The Volume module
can be found on the Mixer page.
Wren is now four voice paraphonic or limited polyphonic! Added a couple
of note stealing algorithms 'none', 'cyclic', 'softest', 'oldest', 'lowest' and
'highest'. I've added level inputs too, note stealing can use that information
(for the 'softest' mode), these could be connected to voice envelope outputs for
instance. You'd still need to patch out the voices manually. The MidiMultiNoteIn
module can be found on the I/O page.
Added an RMS module, calculating the RMS value of it's input signal. The averaging
time can be set with the time control. The RMS module can be found on the Math
Was asked for a utility module just displaying the value on it's input, made it
and it can either show units or dB values. The Value module can be found on the
Added a soft clipper with a cubic transfer function.
in version 2015.12.15.0
The Hrast filter had a denormalization issue for the 24 dB section, added
normalizing code there.
LevelConverter was not working properly, it took the input mode setting for both
the input and the output mode (and so it acted like a .. wire).
Removed some the InternalName stuff, there were two mechanisms with one not being
Cleaned up the source code a bit by removing unused pascal units.
The previous distribution had an old version of the executable in it – fixed
New module type, a ScaleQuantizer, which is a NoteQuantizer with user definable
The presets will be dynamically loaded from a file wren.scale which must be in the
same folder as the Wren executable file. When this file does not exist
Wren will create it, after that you can edit it with a text editor
(like Notepad for instance) and Wren will not touch it anymore.
And yet another quantizer, a modulatable one. That is, the quantization steps can
be switched on or off by both a selector and a logic input signal (the logic xor
of those two signals). This one is a lot of fun actually, it can do trills for
Changed the input level control for all note quantizers to be an input range
control with a value to be set in semitones. This will not affect existing
patches, just the displayed value changed to something more intuitive. The default
range did change from 128 semitones to 48. The range can still be negative
(meaning that an up going ramp for an input would result in a down going note
sequence to be generated).
Made a chainable chord sequencer. This one outputs four note chords for which a
degree, a type an inversion and a base octave can be set.
Which chord is to be sent to the output is selected by the active inputs, the
leftmost active active input wins. Four gate outputs are available to control
envelopes, for this to work properly a gate input (which determines the length of
the logic output activation) has to be provided.
For chaining the logic inputs and the signal inputs are to be connected to the
respective outputs of a previous module. The last module in a chain having an
activated active input will be the one sounding. The activation inputs could be
controlled from a binary counter set to seq. mode, or probably a gate sequencer
would work as well. The logic in-outs are the ones closest to the indicator
The chords can be transposed with the transp. control, this only works on the
local chords and not on any incoming ones.
A slew control is present to get some gliding effect on the note changes, this
works on the local notes as well as on the incoming ones.
I'm not to sure that the chords played are theoretically
correct … if there even is such a thing, please inform me about any
mistakes I made – I'm not too good at this and basically just googled up
And a slew rate limiter, or lag, or control rate low pass filter. This actually
does the same thing as the Average module, except that the knob action is
reversed and that this is a control rate only module. And with that … this
text became longer than the code used to implement the module.
in version 2015.11.4.0
This version does no longer make use of the 'bitmaps' folder that the previous
version used. Instead there now is a folder 'looks' which has a sub-folder for
each selectable 'look'. This version comes with two pre defined sets, the default
one and the gabe one. The gabe one was made by … gabe, also known as
LfoHead on the electro-music.com forum. Thanks Gabe, great job!
As almost all coloring is user configurable now … all color references in
this document will be according to the default coloring scheme. Except that I have
substituted the color green by lime (bright green) in that scheme (which it used
to a long time ago as well … on a darker scheme Lime seems better visible
10-03 00:09 Octangula -> Blue_Hell: is it just me, or does the knob control mode
setting not make any difference?
Ok, the modes do work on modules but not on the other knobs, like settings and
main volume control strips. Fixed it for the next release – the code to do
it was never present … forgotten.
2015-10-09 : When a module with a MIDI assigned control is removed from a patch
the MIDI connection is not deleted, resulting in GPFs … Likewise, after
that got fixed … when undo is used after a module with a MIDI assigned
control was deleted … that CC does not work, although it is still shown to
be present with a blue border. Ok, fixed that too. Apart from that there was yet
another bug, CC's were never properly unregistered, also fixed now.
OSC sync was not working properly anymore, it sent a message /sync where it should
send /*/sync to match any listening synth, fixed that. Had not tested this
apparently after I incorporated the synth name into the OSC system. Likewise, the
the OscMessage module did not send a synth name (while expecting it in listen
I was not going to make any new modules for this one … but thought a
fourfold multiplier to be handy … so here it is, taking half the space of
And another one crept in … a 16 to one mono mixer with modulation controls.
Trying to make a more general look and feel settings control.
Look and feel is stored into a 'looks' sub directory of the Wren application directory, each theme goes into it's own sub folder there. These
folders can contain a set of bitmap files and a configuration file 'looks.ini'. On
loading a a look and feel preset the bitmaps will be loaded from the selected
folder and set as the application bitmaps. The file 'looks.ini' is then read and
any settings in there will override the settings as read from the normal ini file,
including the visual theme specified there. You can still override some settings
after loading a preset, and these will be stored in the normal ini file. The
'looks.ini' file will be overwritten by Wren only when explicitly
requested (by using the 'Save as preset ...' button). For example:
The following items can be customized (and can be stored in look and feel
Each module has a bitmap associated with it, when a bitmap file is found in
the looks folder that bitmap will be used. Bitmaps have to be named like
<ModuleName>.bmp for this to work. The bitmaps should have a size
of 21 by 21 pixels. I will supply example files as part of a default looks
Some bitmaps used by the modules can be overriden, this concerns the following
items (I will supply example files as part of the default looks set):
These bitmaps will have to be 'approriately sized', except for skin_module
which will be stretched as needed.
The following patch related colors can be changed from Wren directly:
- Editor background color
- Module color
- Connector border color
- Selector color
- Selector border color
- Selector border color single
- Display color
- Display border color
- Knob MIDI color
- Knob focus color
- Knob MIDI and focus color
- Viewer color
- Viewer border color
- Viewer line color
- Viewer fill color
- Indicatorbar peak color
- Indicatorbar valey color
- Audiorate color
- Controlrate color
- ControlrateLogic color
- Logic color
- Unconnected wire color
- Wire highlight color
- Module group I/O color
- Module group OSC color
- Module group LFO color
- Module group Env color
- Module group Filter color
- Module group Switch color
- Module group Mix color
- Module group Control color
- Module group Note color
- Module group Math color
- Module group Logic color
- Module group Seq color
- Module group FX color
- Module group Delay color
- Module group Utility color
A visual theme can be selected, this mostly is for the non editor related
visuals. Currently user made visual themes are not supported, it seems to be
possible to arrange for that though. Anyway, several themes are built in.
Guess it will survive the coming winter with all this fresh paint :-)
in version 2015.10.2.0
A multi compare module, after an idea I found here. Added individual outputs in addition to the bus outs,
added random and clear buttons as well. Moved the module from the Sequencer to the
Logic page … not sure … but hey, there is a module search function
Moved the Pad module to the Utility page and added a new Knobs4 module there too
– featuring larger knobs and memo fields to be able to give some information
about what the knobs do. Added a switches module as well, same idea. And buttons,
for momentary switches, with indicators lighting up when a pulse starts. The
button out's are active for 1 ms, the lights for 125 ms.
Added a simple level conversion module, direct translation of input range to
Made a new Math tab for the module selector and moved some modules over from the
control tab to the new one. Added some new math modules there as well.
Current monadic functions are x2, x3,
x1/2, x1/3, sin(x), cos(x),
tan(x), asin(x), acos(x), atan(x),
2x, ex, 10x,
log2(x), ln(x), log10(x),
2x/12 and 12 * log2(x). Invalid domain values
will output zero.
The current dyadic ones are x mod y, xy,
logx(y) and atan(y/x). Invalid domain values will pass on
More may follow.
Made the Space bar work on focused selectors (red border). It
will select the next value or when out of values the first one. This also works
for momentary switches (buttons, each hit on the Space bar will
pulse it then).
Fixed the lockout mechanism which must prevent module's real time execution during
memory allocation and de-allocation. This should cure some issues like access
violations or Wren stopping audio generation or knobs stopping to be
responsive, stuff like that, which would still happen under some conditions with
modules like the filter banks. The mechanism used only gives just a smidgen of
performance loss (full fledged synchronisation degraded performance by about 20%,
which made me prefer some instabilities) … I reasoned that the dirty trick
I used should be ok, may it be so :-)
As a consequence: Starting a second instance of Wren does not stop
audio for the first instance anymore, as it occasionally would do.
As a consequence: It should no longer be needed to restart Wren after a
sample rate change was made in the Settings Screen F3.
As a consequence : It should no longer be needed to stop audio generation before a
sample rate change is being made in the Settings Screen F3, the change
restrictions that were present have been removed. Wren needs to still
stop and restart audio to make the new sample rate work, but it can automate that
As a consequence : Removed the option to not use PortAudio when it is available
from the Settings Screen F3. This option can still be set from the
Wren ini file though (set 'UsePortAudio' to 0 (zero) there).
Changed the way output to input links are followed in a compiled patch. These are
now traversed from the modules themselves instead of from the synth patch. Not
only is this a tad more efficient, but once the compiled modules get properly
sorted as well, signals will have less delay trough the patch. Right the modules
were properly sorted already – great :-)
Module execution order is top to bottom and left to right. Which means that signal
flows from top to bottom and from left to right should have zero samples of delay,
and reverse direction links should have one sample of delay.
The lost feature is that in debug mode wire performance is no longer profiled
separately, it has become part of the module execution time. But then again, there
is no profiling support for release versions of the code.
Also feedback behavior in existing patches may have changed.
Changed the drawing order for the graphs, the leftmost output (channel 1 L) is now
Made help file generation automatic at program startup. The code will check the
version information and new files will only be generated when needed. This also
means that the image cache used for module previews in the module selector should
always be up to date now. The popup asking for confirmation on writing out the
help has been removed. It also means that starting up a new Wren version for the first time will generate some 'nervous window stuff'. I have no
solution for that yet – in order to generate a proper module image the
module has to be made visible, or some controls will not be painted correctly.
Forgot to actually update the patch version to 4 in both source code and for this
page (should have done so for version 2015.8.21.0), fixed it now.
Moved some controls and indicators from the I/O level pane to the top control
pane, and made the module selector resizable with a splitter bar.
When the buttons do not fit the width of the page in the module selector a left
and a right scroll button are added.
Also made a toggle on the visibility of the top pane as a whole, accessible
through the view menu and using Ctrl+T as a keyboard shortcut.
Currently the module finder has to be disabled when the top pane is not visible
– could maybe add a modal dialog for that later on.
So can do this now:
Made the CPU overload indicator larger.
Fixed some minor issues with displaying the file render status. Audio generation
will be restarted now as well when rendering was stopped (when it was on before
rendering was started). Changed the render indicator to be a slow green flashing
Added a context menu for the patch editor. It can be used to insert new modules
cut, copy and paste modules etc.
Changed the default sample rate to 88k2, that's just the 'out of the box setting',
an ini file present will override this.
Fixed the file history (menu File/Reopen) such that the files are actually listed
according to when they were last opened now.
Fixed the issues with menu shortcut keys, the original issue text was:
2014-12-13 : Ctrl+A in the Debug Screen F5 selects all modules
in the editor … and more keyboard shortcuts go wrong there … like
Ctrl+C copies a patch, not handy.
The solution also fixes the issue with the Del key not working
in the popup editor for displays. The module title editor had the same issues,
fixed those as
Changed the way the Enter key works in display editors. Some
editors are multi-line and those will need the Enter key as
part of the text to be edited but most are single-line, and will accept a plain
Enter now for confirming the edit and closing the popup (as it
already did for the display title editor).
Re-grouped the items in the file menu, changed some other menu's as well.
Added a toggle for sound generation, Ctrl+D. This is
accessible from the action menu too. When active all sound outputs are muted but
the patch will stay in a running state.
Added a toggle for lights updating, Ctrl+L. This is
accessible from the action menu too. When active the indicator lights on modules
are disabled. This can save a little processing time, like 15% or so. Note that
the lights update rate can be changed too from the Settings Screen F3. This is
not very useful when PortAudio is being used, as in that case the lights update
process runs not on the same thread as audio processing. It would help a bit for
legacy mode though.
Also improved the speed at which the lights are handled – this used to work
with formatted strings that needed to be made and parsed, now just passing around
pointers to records with t the detail information. Also pre parsed the names of
the lights for faster comparison during the update process. Did the same for all
other data gathered for the editor from the synth, but lights are processed most
often. This could maybe be sped up more by letting the synth broadcast changes,
but it would need to maintain some state then and a synchronisation queue would be
needed. Prolly not worth the trouble.
Added 'panic' control, Ctrl+P. This will stop all pending notes
from MIDI inputs and from the MIDI player. This is accessible from the action menu
too. MIDI input modules will listen to a MIDI AllNotesOff command as well.
Made it so that when a knob or slider is locked that a rnd or clr button will
leave it as is.
There was a kludge used for the GVerb module to make the spread control work more
or less properly. That kludge was based on the lights timer, so when the lights
are disabled the kludge does not work anymore. As a kludge is still needed I
replaced it with another one, added a timer to the TModGVerb code … at
least stuff will not be broken by accident anymore that way … The thing is
that the RoomSize can not be properly modulated in the GVerb code, and I still
want to be able to do that. But the room size must be set before the spread can be
correctly calculated. The timer will cause a little delay after patch reset to
give some time for the room size to be set up, it will then recreate the reverb.
Also for this module, I've removed some needless calculations for parameters that
were not changed.
Fixed menu shortcuts for automatically generated menus, these would jump from '9'
to 'K' instead of 'A'.
Added shortcuts Alt+1 .. Alt+4 to show 1 ..
4 I/O level strips.
Some selectors have only one value to choose from, and thes effectively act as
momentary buttons (for reset or randomization mostly) I've given those a white
border instead of the selectors' default yellow one.
For PortAudio mode started using the PortAudio CPU usage indication, which is more
accurate than what was used before (and still is in Legacy mode). The CPU usage
shown will be lower, but it will run out of real–time now at 100%.
Added support for loading bitmaps from files, this needs a bitmaps sub folder of
the folder the Wren executable file is in. It also needs properly
named and sized bitmap (bmp) files – going for 21 x 21 pixels.
Changed the default for the bread and butter mixers to dB mode instead of linear.
Fixed the spacing in the module selector.
Toying a bit with skinning, for instance, with a couple of custom bitmaps in the
module selector (some more examples) :
Changed the default looks to be 'themeless' so it will look mostly like previous
versions, a user can select a theme then to fit her needs. I did have to change
some things in order to get things compatible with theming support. I will add a
zip file to this web page with a couple of bitmaps to enhance the looks a bit. For
now I'll not add any new module glyphs, that needs a bit more work, but these can
easily be added later on. The bitmaps will be included in the main zip download
It works reliable now, so theming will be available for the next release. When
needed it can be disabled from the ini file by setting UseThemes to zero –
by default theming is on though, and there is no way to change this from the
software itself. When theming is disabled the default Windows theme is used, just
like in previous versions of Wren. When theming is disabled it is still
possible to load your own user glyphs for the module selector.
Added a file opener for Scala files which shows the contents of the file selected
to ease the selection a bit (the first couple of lines having some useful comment
usually). You can enter a wildcard mask to zoom in on files as well. This is all a
bit Windows 3 like, but that was the plan with windows 8 eh :-) Anyway, it works.
The actual looks of this dialog are determined by the selected theme.
Changed the code a bit such that file open dialogs from the patch will try to go
to the directory that the currently used file is in.
in version 2015.9.2.0
And the filter bank is finally finished :-)
I had planned to realize this filter using Fourier transforms, but it ended up
using State Variable Filters … up to 102 of them when the steepness is set
to 6. It takes some amount of calculation time then, but steepness can be set
lower and it is not too bad actually. Steepness controls the number of filters put
in series 1 to 6 can be selected, where each step is a 12 dB filter and so amounts
to 6 dB / Octave of slope for the bandpass ones.
I've scaled the range for the Q control a bit so it's more or less usable over the
full range – it depends a little still on the selected steepness. Also
divided the effective internal filter Q by the number of filters cascaded, which
tames the silly amplitudes a bit, not perfect but it's workable. At high Q
settings and a large steepness it will do some ugly stuff still.
The low side filter can be set to either low pass (shelving) or to bandpass and
the high side filter can be set to high pass (shelving) or bandpass. The in
between ones are bandpass always. The Q control can be used to add in some
resonance. Added level modulation for each filter chain, and added FM control to
be able to sweep the filter as a whole, but note that there is an upper limit for
the frequency that can be used for a state variable filter. So for high
modulations the high end frequencies may turn out to be lower than expected, and
you might want to set some faders to zero there.
Decided to make two filter types, one having a half octave (tritone) filter
spacing, and the other a third-octave (major third) spacing. Both have the same
amount of filter bands and so the second type has a reduced frequency range.
And while we are at it … let's make a couple of splitters too. Added clip
indicators to these.
And … a simple modal thingie would be cool too :-) Added some random
buttons to this one to help search for timbres, it should probably have presets
that can be saved and reloaded. Come to think of it, I could make a general
purpose presets mechanism for modules making use of the existing parameter paste
mechanism. And then control it trough the module popup menu.
Just did that, added a presets mechanism for any selected set of modules or for
the patch as a whole. Module presets can be used by selecting the modules to save
as a preset, then right clicking on any of the selected
modules, and then use 'Save as preset'. Patch presets can be saved from the file
menu. On loading from a preset the selection will have to match the selection that
was used on saving the preset, or else there will be a structure mismatch. So this
probably is most useful for the patch as a whole or for single modules.
The file format used for presets is exactly the same as the one for patches, but
they will by default be saved with a '.preset' file extension. So they can always
be loaded 'as a patch' (for seeing the structure of it for instance).
Loading from a preset will only change the knob settings of the selection and
there will be no patch recompilation, and so the patch will just keep playing
using the new settings.
When there is a structural mismatch on pasting of parameters there will be a
warning generated now. These warnings can be disabled from the
Settings Screen F3.
Made the search function to operate on the module selector as well.
Non–matching modules will be hidden from the selector. and when a page
becomes empty this way the page will be hidden too.
This feature can be disabled from the Settings Screen F3, and also made it
possible to disable patch searches from there. When both options are disabled the
'find' control and the menu item for it will be disabled.
Pressing Esc while the keyboard focus is on the search box will
clear the search phrase.
Rate Smartness (where modules can change execution speed based upon the actually
connected signals) is now indicated with [RS] in the module popup description
shown from the module selector.
in version 2015.8.26.0
Added a new module type, a 16 step clocked delay (analog shift register). It is
rate smart and it has the option to clock it at audio or control rate with an
internal clock (fast mode) instead of from the clock input (clocked mode). When
the module is operating at audio rates the fast mode will be audio rate, control
rate otherwise. Just connect an audio rate signal to any input to make the
module operate at audio rates. This one could be used not only as a bucket
brigade like delay, but in fast mode you could implement your own FIR filters
with it when used in combination with the 16 input mixer for instance.
Made a NoteQuantizer module. It quantize it's input value to the note nearest to
one selected by the note selectors. When the highest and lowest nearest are
equidistant it will use the higher one, and when no notes are activated it will
pass the input value unchanged. Added an input conditioning knob to it to
attenuate the incoming signal, or to invert it (attenuverter).
Made a lookup table module with 16 sliders for the values. A select input selects
which fader will be output, and there is a cross fade option to be able to output
in-between values. This one is to be found on the sequencer tab, as it can be used
for sequencing when driven with a saw shaped input signal (range [ 0, 1]).
Made the Scala module working, copied the details about it over to a separate page.
It translates note information into scaled note information according to the Scala
keyboard mapping and scale files loaded. The input is supposed to be be connected
to anything that generates note information, like a sequencer or a MIDI note in
module, etc. This information is first translated into MIDI key numbers
internally, then the Scala algorithm is applied, then the output value is
translated back to an internal note value again (128 notes / unit).
You can get scales and keyboard mappings from http://www.huygens-fokker.org/scala/ but also there is the Scala program
itself on that site which can make such files to your own liking .. it will also
tell you how to hand-craft such files.
Made an inconvenient choice for the signal levels of the 'rec' buttons in the
granulators and had to invert them. This means that on patch reload granulator
recording will be inverted from how it was aved with the previous Wren version. Sorry about that …
I seem to have over 150,000 lines of code now
or 150 KLOC, sooo … there should be about 1500 – 7500 bugs …
or 'code defects' as they call them … when I perform as 'good' as the
'code industry'. Or around 75 when I'm as good as µsoft … or ZERO
when it is as good as space shuttle code. (see here for a rationale of this).
Fixed a bug causing Wren to refuse to stop under certain conditions. It
would also not allow the Delphi environment to be shut down after Wren had run in
debug mode. Forgot to call the inherited handler in a Notification function ... or
forgot ... it was omitted in some old code that used to work long ago without
that call (TKnobsWormPanel.Notification). I should have disabled the Friemels
module as it does nothing usuful anyway.
The lookup tables were about twice as large as I had meant them to be, halved
Poetry generator, XYScope, Vocoder, Convoder and Filterbank had a memory leaks,
The WaveWiper had its dezipper spec in the constructor instead of in SetDefaults.
in version 2015.8.21.0
Added a 'Poetry' generator, modeled after an idea found in a Phd thesis by
Johnathan Mackenzie found here. It finds seed patterns in a buffer starting at some random buffer
position, and it will get it's next output from the value following the found
seed. Then it will repeat this strategy, producing variations on a read in
pattern as it moves along. New values can be clocked into the buffer and the
search algorithm is triggered from a separate input. This is very similar to a
hidden Markov model.
This module is primarily meant to generate variations on control signal sequences,
but the module is rate smart and will switch to audio speed when needed. As is,
the history may be too small though for audio processing (and the search algorithm
is probably too slow to allow for substantially more history). Anyway, I like it
for control signals.
Added an eight tap delay where each delay output can be individually modulated.
Just like the regular delay module there is a fixed output having the maximum
delay with no time modulation, and the delay time can be set from 10 µs to 5
minutes - in three separate ranges. The delay modulation inputs modulate between
zero delay and max delay for input values from 0 to 1.
Added OSC support. Wren can be either an OSC transmitter or an OSC
receiver. See the FAQ for more info on this and about how to
set up OSC and about what OSC messages are implemented.
In the current OSC implementation there can only be one listener in a patch for
any specific address. The first listening entity seeing the message will disable
it for other potential listeners. There can however be multiple transmitters for
the same address. This is a feature of the OSC module I've used, and I may change
this for a future version – it does not seem to be a serious limitation
currently as further signal propagation can be done on the wire level in the
Added a sync module which can send sync signals from one Wren instance
to another. Or alternatively external OSC enabled software could make use of this.
This module sends and listens to OSC /<synthname>/sync messages. See
the FAQ for more info on OSC messages.
The sync rate must be lower than a quarter of the control rate (currently
amounting to the sample rate divided by 32), or else it will not work properly. A
practical maximum working for all sample rates would be 1 kHz.
Added an OscMessage module which can send or receive arbitrary float values over
OSC using a user defined address. When the current Wren instance is an
OSC transmitter it will send the value present on the blue output when the trigger
input goes from low to high, this value will be set as the output value too then.
When Wren is set to be an OSC receiver incoming OSC messages matching
the set address and having a float value as their first argument will set the
module output value (it will still copy the input value to the output as well on a
Messages may be sent with more than one parameter, only the first parameter will
be used, and only when it's type is 'f' (float).
Working on Granulator modules. These are all based upon sample storage in a delay
line (hence, the module can be found in the delay section). The length of the line
can be set up to five minutes. A maximum of 128 grains has been set currently,
which seems to be about what the system can do (although the maximum actually
obtainable in real time may be lower, depending on settings).
There is control for the grain playback speed (dly mod), the grain inter-onset
time, the grain duration, the grain volume and grains can be panned as well. There
is a record control to be able to stop recording into the delay line, when it is
off the delay line is being looped. The input left of the record control morphs
between recording and looping for dynamic record control.
For the first type the inter-onset time, the duration, the volume and the panning
can be randomized. Not quite sure yet on how parameter modulation should work.
Once the grain starts all the parameters are set, and they will not change during
the grain's lifetime.
Added a second type of granulator which does not have the parameter randomization,
instead this should be controlled trough the modulation inputs. It looks exactly
the same but the modulations work in a different way.
Added an input on both versions to control the grain start point within the
storage loop (this was just a random point before). Also upped the maximum grain
count to 256. This can easily get the CPU on it's knees for certain settings, but
for other settings it's nice to be able to have a higher grain count than
Renamed Granulator into RndGranulator and SimpleGranulator into Granulator, and
then added a new SimpleGranulator. For this one new grains are started from a
trigger input. This one has no pan or volume control. It should be a tad more
efficient CPU wise and the trigger input allows for tuned inter-grain delays.
Current envelope shapes are Bell, Exp, RevExp, Pulse, RndBell and RndExp.
Added a new mixer type, mono, 16 inputs, a chain input and one output. Control
type can be linear or dB with a couple of ranges.
Added S-shaped envelopes for all modules having envelope slope shape selection
(including the interpolation curves on the Seq16 module). Also sped up the
envelope shape calculation a tad by working out an unnecessary multiplication.
Added an 'env' out to the compressor module, an envelope of the sum of the
Changed the timing for the net MIDI and OSC lights from 250 ms to 125 ms and added
an OSC enabled indicator light.
Added a SynthName property which can be used to distinguish Wren incarnations for the purpose of OSC addressing. The synth name can optionally be
shown in the title bar and it can be changed as part of the OSC settings.
Also added an OSCAddress property to compiled modules in the synth patch, it's
value is set from the module title in the patch editor. A module title starting
with '@' will enable the module for OSC activities.
Furthermore I've changed the value popups for knobs to show the OSC address
fragment to use for addressing them as the first line in the popup. When the
control is not automatable an exclamation mark will be shown on the first line
The OSC specific modules such as Sync and OscMessage are OSC enabled always
regardless of their title.
Working on OSC enabling all modules, and maybe the controls of the main program as
well, the above is in preparation of that.
OSC patterns starting with /*/ should be accepted by all running
incarnation that do have OSC enabled.
Optimized Pan, XFade, MultiMult and Multiplier a bit by getting some
Changed the patch format to version 4 to allow for MIDI CC values to be bound to
ValuedControls. Old patch format versions 2 and 3 should still be readable, but
older Wren versions will not be able to read the new version 4 patches.
Also allowed for more options to be added in the future with better backward
compatibility. This change was needed to allow for CC to knob associations to be
stored into the patch.
Added an item on the context menu for knobs, sliders and selectors for associating
a MIDI CC with them. Most knobs, sliders and selectors have support for this, not
all. When a a CC is associated with a control it will be painted with a blue
border (when not focused, otherwise the indicator color will be red still). The
associated CC will also be shown in the hint popup when it appears, and the popup
menu will show the currently associated CC as well.
Also added a 'learn' feature to quickly assign the last CC received.
Also added a CC un-assign item to the popup menu, and and un-assign button in the
MIDI CC assignment window.
Added a MIDI channel selector to the Settings Screen F3 to select what channel
to use for the non MIDI modules (that is: passing on CC values). This can be set
to listen to all MIDI channels. The MIDI modules have an independent channel
MIDI CCs on knobs are working now, the CC directly controls the knob, which in
turn will change the value in the synth – dezippering is not used on MIDI
Added a couple of quick select buttons to set the the amount of visible I/O
channels more easily.
Some more work will be needed before general OSC will actually work, but the
editor supports it now. Made code for general OSC support too now but did not test
it yet. Controls are addressed by the OSC pattern
/<synthname>/<modulename>/<controlname> and they want to
see a float value in the range 0.0 to 1.0 for full travel. Every control that
allows for MIDI automation can be OSC controlled as well. When a control is not
automatable this will be shown on the control's popup menu by having an
exclamation mark in the first line.
Made selectors focusable, they will get a red border to indicate focus (and a blue
border to indicate that a MIDI CC was associated, or fuchsia when both focused and
having CC control). Selectors also listen to the arrow up/down keys now when they have focus. Knobs will get a fuchsia border too now under the
latter condition. Focusable controls will get focus now when the mouse cursor
hovers over them regardless of the mouse wheel mode selected.
Output 10 of the Counter and the ChebGen module were not working, fixed now.
Demux module was not working, fixed.
Knob controls for Pan and Fade modules were incorporating the input type selected,
not handy, changed that. It is a level conversion on the control input only now.
This will probably break some patches … but it was just not good.
Fixed memory leaks.
Made the 'unfocus' feature work better with a 'sort of' hidden control that will
get focus when knob focus gets lost. The 'use unfocus' feature is on by default,
this can be changed from the Settings Screen F3. The unfocus option causes
controls to loose focus when the mouse moves out.
Fixed sample and hold (which stopped working properly after making it rate smart).
Seem to have weeded out the access violations on the granulators.
in version 2015.7.12.0
in version 2015.6.27.0
in version 2015.6.21.0
Another error in displaying the envelope curves … the range settings were
all taken from the attack range – fixed now.
Added some glyphs for a bunch of new modules, it got all a bit confusing with the
Updated some help text, some new modules still had text copied from the modules
they were copied from.
Fixed the tab order for some new modules.
Decreased the spacing between module buttons on the module selector a bit to get
room for more module types. Resized all module glyphs to 20×20 px. to make
Changed the default font for the module title to Corbel 8 pt bold, seems lighter
but more readable than the previous font used. Only thing is that it has
'oldstyle' numerals, that is lower case. When this font is not available (as in
Windows XP apparently) the module titles will use the old font (Small Fonts 7 pt
bold). But read on …
Made user selectable fonts for module titles and module text. It is probably best
to keep using Small Fonts for the module text, as some of the text was measured
out with that font, but one could still change the color.
in version 2015.6.19.0
An instant crash was reported by a user, no clues yet.
Inspecting the code I did find a bug in the audio handling. When it ever happens
that there is no compiled patch and the audio stream runs, the audio I/O
processing will write out of the array bounds of the audio out buffer …
seems likely to be that … on a slower system audio I/O may run before the
patch got compiled. Also it would be worse with largish audio buffers (seeing that
user's buffer size is 4096, using 512 or so here). Fixed that issue.
It fixed the problem too.
in version 2015.6.18.0
Made a WaveWiper module after a schematic made by PHOBoS, which he in
turn based on a Voltage Splitter idea published by Jim Patchell. Not sure if it
does exactly the same thing, but it is quite a nice wave shaper I think. This
module is rate smart, it runs at control rate when all inputs are control rate
signals, at audio rate otherwise.
InSig1 := FInputs[ i_in1 ];
InSig2 := FInputs[ i_in2 ];
BreakSig := FInputs[ i_break];
Mute := SignalToMute( FInputs[ i_mute]);
FOutPuts[ o_out1] := ( Max( Insig1, BreakSig) + Min( InSig2, BreakSig) - BreakSig) * Mute;
FOutPuts[ o_out2] := ( Max( Insig2, BreakSig) + Min( Insig1, BreakSig) - BreakSig) * Mute;
With a sqare and a triangle input modulated (on the break input) by a sine wave:
And while wave warping .. made a WaveWrapper module too. This one will mirror the
input signal in a user set low and high bound limit. The limits can be modulated
by the mod input. This module is rate smart.
Modulation := Normalize( FInputs[ i_mod] * FInputs[ i_modamt]); // Modulation amount
LoMirror := FInputs[ i_lowlevel ] - Modulation; // Get low mirror and subtract modulation
MirrorDelta := 2.0 * ( FInputs[ i_Highlevel] + Modulation - LoMirror); // High mirror gets modulation added. calc mirror distance, times 2
Mute := SignalToMute( FInputs[ i_mute]); // Get Mute multiplier
if Abs( MirrorDelta) < 1e-6 // When mirrors are too close to each other
then FOutputs[ o_out] := 0 // make an exception ...
Amp := 4.0 / MirrorDelta; // Amplitude reduction factor due to wrapping - times two
Signal := MathFloatMod( FInputs[ i_in] + LoMirror, MirrorDelta); // Lift signal by LoMirror, and take signal modulo 2 * Delta
if Signal > 0.5 * MirrorDelta // Reflect bit above Delta back
then Signal := MirrorDelta - Signal;
FOutputs[ o_out] := Normalize(( Amp * Signal - 1.0) * Mute); // Bring signal into -1, 1 range and apply Mute signal
With a sine wave input (range -1, 1), the low level set to about -0.6 and the high
level to about 0.4:
Made a logistic map module. This one is rate smart, driven by an input clock signal.
Not sure yet what the definitive form will be. R values below three and different
start values seem not to interesting, so I may remove those. And the cross fade
works good only when the clock rate does not change .. so I may change this into
an oscillator module.
A new LFO module based on strange attractors, a Lorenz type and a Rössler
type. Both the Lorenz and Rössler systems are 3D, so I made three outputs for both types.
Made an audio rate version too and had some trouble to get the calculations stable
for that. It needed quite a bit of over sampling to get it right (or actually the
time step must be smaller than some critical value). For a sample rate of 44k1 12
times over sampling for Rössler (time step ⪅ 0.08), 15 times for Lorenz
(time step ⪅ 0.02) seem to do the trick). For higher sample rates the amount
of over sampling could be set lower but it betters the sound quality too. This may
have to be adjusted later.
Maybe these should get some means to set initial conditions, but otoh they seem to
visit a lot of places already.
» Decoupled the connector color from it's signal type. In the designer the
signal type should be set now and the default wire color will follow that. Also
made a separate color lookup for painting connectors and wires. User selected
cable colors are possible now; connector colors still reflect the signal type.
» Note : The patch version got changed to 3. The patch reader will still
read version 2 patches, and previous Wren versions can read type 3
patches as well, just ignoring the wire colors in it.
Changed the default wire color Lime (for control rate logic signals) into Green,
Lime was not very visible on most module colors.
Moved Quantizer to the Note tab and made it a control rate module. Also the Scala
and the TransInv modules are control rate now.
» Made rate smart modules. These execute at control rate normally unless one
of their inputs gets connected to an audio rate output, they switch to audio rate
» rate smart modules are: Inverter, Adder, Multiplier, Scaler, Ratio,
Rotator, Rectifier, MinMax, MultiMult, Median, HardClip, Not, Gate, Divider,
DFlipFlop, RSFlipFlop, Counter, ADC, DAC, Compare, LogicSelector, ProgDivider,
XFade, Pan, MixM4to1, MixS2to1, MixTto1, MixQ1to1, SampleAnsHold, TrackAndHold,
Switch2to1, MultiSandH, Mux, Demux, DigiMux, DigiDemux, TerraGen, Average,
GateSeq, Seq16, SeqSeq, SeqStep, SeqClockStep, EuclideanRhythm, PhaseDetect and
Changed the built–in module help to indicate rate smart Modules and what
rate smart–ness does.
Removed a tab control level, all screens are on the same tab control now, giving a
bit more screen space for the settings, graphs and debug screens.
Made a couple of new checkboxes in the Settings Screen F3 to control what appears
in the title bar.
Changed the PortAudio handler in FrmMain to combine the loops for input and for
output into one.
Sped up the Normalize function considerably by not using the Delphi internal
TDoubleHelper methods. Requested it to be inlined, but Delphi seems to mostly not
do that. The SignalToMute function basically uses the same code as Normalize,
changed that one too.
FreeVerb internally works with Singles, not with Doubles (for floats) as the rest
of the synth does. The standard Normalization routine would not always properly
deal with that – changed the value for the smallest acceptable number to
1e-20 (-400 dB).
Local copy of Normalize routine in GVerb module fixed too now.
Error in drawing the shapes for EnvAHD (and likely EnvADSR as well). I had changed
the way the knobs work and how their values are calculated, but this was not
reflected in the drawing of the envelope shapes, fixed now.
in version 2015.5.31.0
NOTE: The memory management was changed, the borlndmm.dll is no longer needed
– and it is not available for download anymore. This needed some code
changes here and there and I may not have touched all affected places, please
report bugs on the forum, will fix ASAP.
Added a debugging aid — Find small values — to help find possible
calculation bottlenecks in a patch. This can be accessed from the action menu or
from the Debug Screen F5. This feature will search for values that are either
denormal or are in the range (-1e-50 … 0) or (0 … 1e-50), small
values being not zero. Hit F10 to highlight the modules having such values on
their inputs / outputs or as internal signals. In the Debug Screen F5 a listing
will be made of the found modules and the pin / knob names found (Name [Title]
[inputs] [outputs]), knobs are listed as inputs. For compiled versions that have
profiling support (which the regular releases do not have) the small value finder
is part of the profile action.
Changed ranges a bit for knobs in dB modes and for Mute buttons – these are
off now when their value is <= -140dB. Changed the dezipper algorithm in
the same way, values <= -140 dB result in zero.
Made an option to allow for undo after a patch was saved, it is off by default.
Likewise added an option for undo after load.
Tab order of form and module elements sorted – some elements do not properly
show focus, that needs a fix too.
Cleaned out the source code a bit, removed unused units from the project, unused
code from units, etc.
Experimented with allocating memory in large chunks. This was not really a good
idea; having many smallish chunks performs better. Anyway, that idea is tested
Added user definable scales for the SeqRandom module, these will be loaded from a
file wren.scale which must be in the same folder as the Wren executable
file. When this file does not exist Wren will create it, after that you
can edit it with a text editor (like Notepad for instance) and Wren will not touch it anymore.
A Line like
(which is a dash for the scale name and no notes in the scale) adds a separator to
the selector's popup menu.
This unfortunately will likely break existing patches, in that for the SeqRandom a
different scale will be loaded than was saved into the patch. Also this will make
patch sharing a bit fuzzy as there may be a different set of scales in effect
— but then again, such already was the case for wave players and Talkie
files as well.
Some designer changes: added a 'DynamicStepCount' property for TKnobsValuedControl
and for TKnobsXYControl. When this field is set to true (by default it is false)
and the value for the control goes over StepCount - 1 StepCount is fixed to value
+ 1. This was needed to make the user definable scales work properly. This is also
used for ModTalkie to allow it to have more than 32 voice banks. Also added a
design time field '_DesignerRemarks' on TKnobsValuedControl and TKnobsXYControl,
this is just a comment field, it is not used at run–time.
Added Dyads and Triads to the default scales for SeqRandom. Should make it easier
to program scale modulations using chained SeqRandom modules.
Made Picture and Docs properties for modules, so that information is no longer in
TImage / TMemo components in the designer, which cleared stuff up a bit there.
Added a help item to the module popup menu showing the help text in a message box
instead of in the web browser.
Edited help texts a bit to get the greatest nonsense out.
Had some inputs exchanged on the audio rate square OSCs – PWM was not
working – fixed now.
Changed the memory manager – this revealed some allocation and de-allocation
issues, Like freeing something in the Median module that was also freed by the
system. Some more things like that … anyway, trying to get rid of the
memory manager DLL, reasoned that it would not be needed. Ah, and in allocating a
new synth patch as well.
Changed the knob ranges for the SeqRandom module slightly, to be able to better
set fractional values.
Freq input was not found on Pluck module, stopping execution of the user
interface. Added the forgotten freq input. Also added some exception trapping to
make connection errors issue a log item only and to not prevent the system from
running the patch. The wires will still be there in the editor (and they will be
saved on a patch save), it is the compiled patch that does not know of the input,
and so the connection will not be made.
Made the program write logs to a log file (wren.log in the application directory).
This feature will be always on at program start, but it can be set disabled from
the Settings Screen F3, which will cause the feature to be turned off once the
settings are loaded from the ini file. If the file exists it will be erased at
program startup, so there will only be a log of the most recent program run.
00:00 000 probing PortAudio DLL presence, looking for "portaudio.dll"
00:00 319 PortAudio DLL found
00:00 321 probing FFTW DLL presence, looking for "libfftw3-3.dll"
2015-05-30 12:36:43:622 compiled synth patch
2015-05-30 12:36:43:656 starting PA [api: "Windows DirectSound", in: "ADAT 1 (1+2) (R ...
2015-05-30 12:36:43:684 PortAudio was started
Made graphs to be invisible by default, it can be enabled in the
Settings Screen F3. Changed the drawing of the main PageControl a bit so it does
not show the nearly invisible tabs anymore (there were a couple of pixels in the
Removed the About Screen – program information can now be found on the
Settings Screen F3. The About button there and the About item in the Help menu
now go to this site.
Bugfix on Pluck module, on FM modulation it could end up with a negative array
index, causing a program crash.
in version 2015.5.12.0
Bugfix for crash at 00801B75 ... code tried to read data from a synth patch that
did not exist.
in version 2015.5.10.0
Changed the knob ranges for a lot of modulation attenuators … old patches
will need some work … especially FM control for oscillators. The new ranges
make it easier to set fractional values, like 0.5 etc. I'd probably need to change
the patch format so that it stores values and not knob positions, so that on an
internal change it could find the closest value. I had that before, but it was
messy, so don't know.
Made a RangeConverter module, or a linear mapper. For each input the selected
input range is mapped onto the range set by the low and high values knobs. The
input ranges are chosen to map the standard LFO output ranges.
Added a four channel mixer MixQ1to1.
Changed the layout a bit for MixM4to1 to make it clearer what are mix inpits and
what are level control inputs — hopefully.
Added a new tab on the module selector – 'Note' – and moved some
modules into that from the Control tab, it was getting a bit crowdy there.
Fixed an issue with knobs: with a high opacity setting the red rectangle
indicating selection would not be visible.
Changed knob painting a little to remove the faint rectangle it would always show.
Changed VU meter peak and valey indicator fall-off rates to be more consistent
over various sample rates, buffer sizes and channel counts.
User set lights update rate was not restored after program restart – so it
used a fixed value of 43 ms … until the setting would be manually changed
— fixed now.
Added a new connector and wire color, Lime (bright green), to indicate control
rate logic signals. Changed several modules for that.
New wave types for LFO's — 'Random Square' and 'Random'. The 'Random' type
is white noise.
Fixed bug in template support – loading or saving a template would change
the internally stored filename of the patch so that a later save action would
overwrite the template and not the patch. Also made the filename visible again in
the window title. Filename and patchname can be different …
Changed the help for the Pulses module to explain why it does not always listen
to the frequency setting, also changed the default mode for it to not be 'Async'.
PinkFilter had denormals – fixed now.
The denormal preventer itself would create denormal exceptions during debugging
when the value passed to it was a denormal already, changed the code to first
check that condition (in debug mode denormal checking is on to be able to find
such issues) – fixed now.
Added signal normalization to all mixer outputs.
Changed sensitivity of PM input for ModResonator to be 32 times less sensitive.
With the previous settings subtle modulations were impossible.
Added patch changed warning on loading a new patch or a template. This can be
disabled in the Settings Screen F3 – 'on load warn for patch changed'.
Starting to add un-attenuated frequency control inputs to modules, sofar Osc,
Square, Pluck, OscTrig, SquareTrig, SVF and HrastSVF. These will just be added to
the value set by the frequency control knob(s).
Changed the detection for unconnected modules. When a module has no connectors it
is regarded to be connected. When it has no outputs it is regarded to be connected
when an input is connected. Finally when it has outputs it is regarded to be
connected only when an an output is connected.
Added pass-trough mode to the mute buttons of the FreeVerb and GVerb modules.
Changed the order of items in the module popup-menu to make it less likely to
delete selected modules by accident.
Added input level control for Resonator module, as it is way too loud for
non impulse like inputs. For single sample impulses it's fine at 0 dB.
Added a debug like thingie that can be turned on in the Settings Screen F3 – 'show lights time' which for each update cycle shows the amount of time
spent in updating the lights in the user interface. This can help to set the
'update rate for lights' parameter – it would not make sense to set the
update rate higher than what the system can handle.
Fixed bug in 'Recent Files' mechanism, menu items were not freed but deleted only
– this led to memory corruption on patch load / save actions.
Added input type selector for Pan and XFade modules. Default type = [-1…1]
to keep existing patches compatible.
Added a new pitch shifter module which can be accurately tuned in semitones and
cents. It has exponential pitch modulation and the delay length can be set to make
a trade–off between quality and delay. This is a delay line based shifter,
not an FFT based one, meaning it will have some 'clasical artifacts'.
Changed some memory allocations to make sure they are atomic and non blocking on
the real time thread. Eventually all should be made like that, but some were a bit
easier to do …
in version 2015.4.25.0
When I googled for Delphi gverb I got this page back pretty high … which
prolly means I should actually make that gverb thingie (as there seems to not be a
Delphi translation for it yet) … in fact I just did a code translate on it
… a bit more work needed to make a module and then some testing of course.
Would love to hear an FDN based reverb …
Ok … here is a proto … its not 100% stable … when I change
the max room size from 1000 m2 to 0.01 m2 in an instant it
can produce invalid floats. Changing the relative (actual) room size is OK though,
and added a modulation input for that for some nice effects. Also some oddities
with the spread control still, on a patch recompile spread goes to zero it seems.
Spread and Max room size are non-realtime controls, and will give zippery side
The issue with sudden max room size changes got fixed, there was some out of
array memory addressing. The issue with with the spread control stay, and when the
effect runs for a long time the signal levels will explode.
Ok, spread control is fixed, there was an issue with initialization order. Also
found a way to control the 'explosiveness', this was related to room size
modulation, by changing the amplification factor in the FDN matrix thate is
control, setting it to 0.45 instead of 0.5 seems to work without altering the
sound too much. Also reduced the zipper effect a bit on changing the max roomsie
and the spread by flushing all internal data on changes.
Changed the maximum room size to be 6400 m2 and the maximum reverb
time to 30 minutes with an amp factor of 0.46 currently — seems stable,
0.475 was unstable.
Changed the spread control to multiply by 100 before it gets into the actual gverb
code, the control still goes from 0 — 1. with 180 it would try to make
negative length delay lines, I calculated 160 to be the limit, Super Collider
seems to allow for a maximum of 100 … the web is not clear on this …
I thought one to be pretty wide already … oh well 100 it will be.
Added modulation inputs for damping, reverb time, tone, early level, tail level
and dry level. These are all clipped into a [0,1] range and multiplied by the knob
setting. So the knob sets the maximum value, the modulation can lower that only.
Added a mode parameter to select between rounded or relatively prime length delays
amd between the original and an alternate FDN matrix. See the module help for
details. Differences are quite subtle, mode 1 is most edfficient computationally.
The GVerb port is in the released Wren now and in the published source
Settled on the following ranges:
- MaxRoomSize from 1 to 6400 m2
- RoomSize internal minimum is 0.01 m2, Roomsize is a fraction 0 to
1 of the maximum size, but it will never go under the internal minumum. It
is possible to get some glitchy results with this still, but that seems nice
- Spread ranges from 0 to 1 but is internally multiplied by 100.
- Reverb time is from 5 ms to 30 minutes.
- FDN matrix multiplication factor is set to 0.43, which has been stable here
for a couple of days now under wild modulations. The alternate matrix uses 0.73.
Made the Average module clear it's internal state on a system reset, so that it's
outputs will always start at zero after reset. This way an averager after a
constant module will always make a smooth start from zero on it's LP output after
a reset, and the HP output will always make an initial spike under those
Removed the dezipper action fom the Constant module's mute control so it can be
used as an on/off switch.
Found some addressing bugs, where indexing constants from one module type were
used in another … changed the scoping of the indexing constants so that
such a thing can not happen anymore.
in version 2015.4.12.0
I seem to have fukt the Vocoder, it is way too loud … hmm … I had
changed some math to a complex data type … innocently enough it seems
… build 103 seems to be ok still. Ok it was in the Convoder instead where
I forgot to divide by the Attenuation factor needed (16777216 so there is the
reason to be LOUD) … got fixed for this release.
Made a version check in the help menu, and also added button for it on the about
tab. Using your default browser it will bring you to the Wren web site
which will then do a version check based on the Wren version passed to
Added a light left of the CPU load indicator that lights up red for a couple of
seconds when any internal audio buffer over or underflows occur. The CPU load can
go above 100% before this will happen. And conversly it can go red when the CPU
load is below 100%. Not that you'd not have heard it … but still …
it may show the actual cause of odd things happening … my algorithms being
too slow …
Changed the FreeVerb parameters (for the Reverb module) to their original
settings, but as they are recalculated when the sample rate is not 44k1 the
selected parameters may not be the best (i.e. not relatively prime) — this
may need some more coding.
For example see the JCRev implementation from the Synthesis Toolkit
(this one sets the next available prime value for delay lengths recalculated for
new sample rates). Might also want to implement JCRev as a 2nd reverb algorithm.
Still need to do some tests on this … I did not like the reverb very much
as it was … too much high frequeency tone like stuff is/was going on to my
Another one to look into could be gverb.
Gverb also enforces the delay lengths to be prime. There are some other similar
implementations around, some simpler (so less CPU cycles needed). For some
examples see the Artificial Reverberation section at ccrma.stanford.edu.
It seems better now, ensured all delay lengths to be prime, but no check was
made to see if they are relatively prime, so under some conditions two might end
up having the same length.
Changed the 2to1 switch, it actuallly is a 2to2 switch now connecting (in1,in2) to
either (out1,out2) or (out2,out1) depending on the selection input.
Added two new modules, DigiMux and DigiDemux, for which the routing is sleected by
a couple of digital selection inputs. A bit like the CMOS analog switches, except
they are one-way only in Wren of course.
Added a zero / hold mode selector to the DigiDemux which in hold mode causes the
last output value the be held when an output gets deactivated. In zero mode the
output value will go to zero.
Added a programmable frequency divider. A low division and a high division factor
can be set and with the select input it is possible to change the division rate
between these two values. A step value can be set as well, the amount to add to
on each increment on the division factor.
in version 2015.4.6.0
The global tuning parameters were not working properly, fixed that.
in version 2015.4.4.0
Removed all Jittik scripting stuff — took too long to get it stable —
must be a bad idea.
Made peak (black) and valey (white) indicators on main VU meters. The peak
indicators are dynamic, the valey indicators show the lowest value seen since last
reset (to reset click on the textual representation for the min / max values). The
peak and valey indicators currently are mono, i.e. the same for left and right,
taking the extremes of left and right values. Changed the colors for peaks and
valeys, peaks are white no and valeys black.
Sanitized the VU code a bit, and made the peak and valey level indicators to work
per channel instead of Left / Right combined. Also added a valey value label for
each pair of channels. The labels still show the extremess of the Left / Right
Recolored the Seq16 module chain input and the out output to red make it show
those run at audio rate. With the xfade feature set to one these can be used as
user programmable wave forms.
Renamed parameter pasting into value pasting.
Changed the bitmaps and painting for knobs so that they will be painted better on
non gray-ish backgrounds too now.
The red box to indicate the knob being focused (and processing keyboard input) is
now drawn around the knob as well, instead hiding behind it on the right and
bottom sides. The up/down clickers were resized a bit to make rtoom for that.
MIDI modules were not working anymore, forgot to make them control mode modules
after the slow / fast mode changes in the previous release — fixed now.
Changed internal scaling for envelope times to not use a time in seconds anymore
but a a range [0,1] which makes it easier to have properly working time
modulations. Some existing patches probably will need envelope parameters to be
changed (only when the use time modulation).
Added a pattern module, meant to be used for rhythmic figures in a more compact
way than can be done with step sequencers. Basically it is a set of eight
dividers, when a divider is started it will copy the trig input to the output on
the first count only. Then when the divider counted down to zero a next divider
will be searched, skipping all dividers with a count set to zero. When all
dividers have their count set to zero there will be no output pulses.
Added an EuclideanScale module, a bit after what Kyma has in version 7, can set
the number of divisions for the octave, and then a pattern selection can be
applied to that. An ocatave still spans a factor of two with this.
Added an Interpreter module, it executes forth code at control rate. This is
experimental, undocumented and subject to change. For now eight inputs are put
onto the forth stack. Then a deferred forth word is called from Wren.
This word can be hooked by custom forth code and that code should pop the stack
values to do something with them. Outputs can be set from code by calling a
predefinedootput word whenever such is needed.
Added a Convoder module. This module performs multiplication in the frequency
domain which is equivalent to circular convolution. Nothing very speciall but it
can give some nice reverb like sounds based on its input signals. It can sound a
bit vocoder like too when used on speech signals.
Added a logic selector module, that when the select input is inactive passes its
inputs A resp. B to its outputs A resp. B. When the select input is active it
passes inputs A resp. B to outputs B resp. A (i.e. it swaps the signals then).
in version 2015.2.26.0
Had an integer overflow in the Seq16 module when the module did not get clocked.
This would stop sound generation in a particular patch after about 6 hours. The
Integer is a counter used to measure the time between clocks and is used for
interpolation. I've put a cap on it's value
Implemented value pasting, but did not test it yet. The idea is to select some
modules, then do a copy to clipboard for those. Then select the same module
structure elsewhere in the patch, or in anoter Wren incarnation's
patch, and then paste the values over to that selection to get identical knob
settings there. One could copy over the steps from a set of sewuencers to another
set of sequencers for instance.
It seems to not work yet, more later. Ok, there was an error for checking
It works now.
All in all that got about 10% off from the current patch's CPU load.
Made ModCompressor about 5 times as fast by getting some calculations out of
the loop and optimizing the remaining ones a bit.
Sped up LFOs by about 10%.
ModReverb was sped up by about 15% — also changed the execution order
of the allpass filters to be the same as in the original code, not sure if
that would change anything (sound wise).
Sped up envelopes a bit by using table lookups for the power function.
Re-implemented the FIsSlow mode request flag. A while ago I had settled on
executing all modules in slow mode and only the ones needing fast execution in
fast mode as well (with the FIsFast flag). With the reintroduction of the slow
mode flag it is possible again to have fast mode only execution. I'm getting more
detailed profiling information this way and patches seem to perform marginally
better too. Module constructors need to now set Fast and / or Slow mode request
flags always or their DoTick and / or DoSlowTick method will not be called (the
TMod base module is neither fast nor slow). [Slow being control rate / blue
signals, fast being audio rate / red signals].
Found hat denormal numbers could occur in SVF (again) and in FreqShifter modules
resulting in stuttering for some settings. Added some Normalize calls to fix that.
CSV player would play 11 channels only instead of the 16 it was designed for
— fixed now.
SVF Q calculation was wrong, changed it. It is still wrong but at leeast it is the
same as on the Hrast Filter now (I should change both to use 1 / Q internally
instead of 1 — Q … or maybe not … to not break patches).
Added a clocked delay module.
Added a bottom label to he graph to show timing to be in milli seconds (for some
reason I always think it to be micro seconds), also added checkboxes to enable or
disable individual traces.
Changed Pad module looks.
The Filter bank does not work yet.
Could use a SeqRandVals analog to SeqRandom, but then for random non-note values
— no need for that, it is easily patchable.
in version 2015.1.31.0
Got some 15% off from the execution time of the FreqShifter module, still not
understanding really why it takes so long to compute. Well … 70 to 80%
goes into the calculation of the Hilbert transform … but then again that's
just about 100 multiply / add instructions. The vocoder does a lot more than that
in about the same time. Hmm … maybe the vocoder does not do that much more
… also looked at the compiled code for the FreqShifter, it is not that bad
really, so this is it I guess.
Mute on Square and SquareTrig modules was not working, fixed.
Triangle amplitude was a factor four low after the anti-alias modifications, fixed
now (for Osc and TrigOsc modules).
Pad module, with an X and an Y output, both simultaneously controlled with a
in version 2015.1.23.0
When the FFTW DLL is not present some patches will not work. They will load though
but they can not be compiled. When an old compiled patch was in memory that one
will keep playing, and when there was no patch there will be silence …
will need to issue a message for this situation. For now it will be the vocoder
module only causing this issue.
I've changed things so that an affected patch will load and compile but any FFTW
dependent modules in it will just do nothing — a one time warning will be
shown for this situation.
When words.lpc is missing and a Talkie module is used there will be a NULL pointer
issue. I've changed the code a bit such that the Talkie module will show
'words.lpc missing?' in it's bank selector control when that file could not be
found, or when it could not be interpreted. The file will be looked for in the
Wren executable directory. Will try to get some better diagnostics on
this at a later stage (for the case of parse errors).
Bugfix for SeqRandom, when 128 notes were in the scale there would be an integer
Added an exponential FM input for Talkie, and put a frequency range clip on it
from 0 — System_Rate / 2. With no FM connected it will behave like the
previous version. Also changed the layout a bit to create room for more controls.
Added a bank select input, which when it changes value will overrule the manual
bank selector control. Also made it work a bit different, Talkie will now only say
something when it sees an upgoing edge on the start input (and it is no longer
possible to change the phrase (or bank) while Talkie sqeaks).
Added trigger mode to SR FlipFlop, so it can now be edge or level triggered. The
default mode is edge triggered for compatibility with old patches, However I also
changed the workings a bit such that when a reset condition is present that will
always overrule the set condition.
Implemented anti-aliasing for all audio rate oscillators. Algorithm based on this article about polyBLEP and the implementation of the
HrastOsc, which uses a form of polyBLEP as well. I've added a band limiting level
control, with a default level of about 1.5 — this will change the sound for
existting patches — for better usually. For some special cases like
modulating a delay line or a nicely whistling patch you may want to set the
B-Limit control to zero. For high pitched voices you may want to nudge it up a bit
from the default value.
Some stability issues with the anti aliasing code, had to mod the code a bit for
that — copied from HrastOsc. The BLEP stuff did not like phase accu bound
over/under-flows at all.
Added a 'friendly' label field in the designer for value controls, so the text in
popups can be a bit more descriptive — filled in some, not all yet.
in version 2015.1.18.0
Moved TerraGen module from Osc tab to FX tab. Reorganized LFO tab a bit. Made a
couple of modules a tad smaller.
Changed the Debug Screen F5 and Settings Screen F3 a bit so they only have the
buttons needed locally in the right pane.
Clipped all PWM to be in an interval of [ 0.01, 0.99]
I renamed a 'reset' input pin to 'res' on one of the sequencers, on reloading a
patch with that you may loose connections — I will not fix that, and will
keep the patch version at 2 as well — sorry about that.
Experimenting a bit with Talkie. And … we have liftoff — Wren speaks! Now I need to think about how to select what to speak from the 1200 or so
letters / numbers / words / phrases available. When Talkie grows up it may look
I settled on that FTTB. Phrase selection is messy, but for random phrases its ok
now. I like this module :-) Frame speed is liniarly modulatable between 0 anf 80
Hz, sample rate is liniarly modulatable between 0 and 16 kHz.
Would like xfade on seq16 module … it would need to calculate a time guess
between steps … hmm … it would act one step late then on tempo
changes. Ok, it can actually calcultate a period estimate on each clock edge, and
it can be made such that when the xfade factor is zero it would not delay. It can
however produce large over and undershoots then when the actual period changes
fast. This can be clipped though, such that the final value is just reached early
(or late …).
With that I think I found a reasonable implementation for the single sequencer
case (it works on direction reversal too). Should check it to work for chained
sequencers as well, which probably it does currently not, but this may be
Experimenting with a Sequencer Sequencer now, to make chained seuencers more
reliable and easier to set up. Especially with interpolated sequencing.
This seems to work well for chained interpolated sequencing too (although there
still is a reset related issue). The clock speed is very high, to not have to wait
too long for a trace to be made.
Top trace is clock in (trig), then clock out (trig, with bit of delay). The orange
track is reset. Then the enable for the top sequencer (for 16 steps), purple is
the enable for the bottom sequencers (8 steps). Green is the output of the top
sequencer, blue the output of the bottom left one (not cross faded) and the the
red trace is the 100% cross faded output of the bottom right sequencer. (The
bottom two sequencers have the same step settings to be able to compare crossfaded
and normal signals.)
Only the last sequencer in chain should have the xfade turned up, it will handle
the upstream sequeencers too then.
An asside: It is interesting to see how close this is to multi-stage envelopes,
only the sustain phase is not there. With a small mdification on the trigger LFOs
this could be done too, they would need a hold input only. The sequencer
inerpolation could easily be made exp or log as well, might do that anyway.
Ok, that works … now the backwards stuff … seems to be working too
Except reset is not perfect yet when running backwards … when the last
sequencer has less than 16 steps the sequencers before will not start on the last
There seems to be no such thing as a perfect reset, I've added a couple of reset
modes for the SeqSeq module each having a different effect on how on how patterns
will be played.
in version 2015.1.4.0
Removed vocoder startup glitch.
Made sliders, a 16 step sequencer would fit now.
Made an audio mixer with two busses, 8 mono inputs with pan control and two bus
level output controls. Also added mute and solo buttons on the inputs. It is
chainable for when more inputs are needed. The first solo button seen will be the
effective one, and an active solo will suppress solo actions in a chained module.
Varous issues with clocks. Square oscillators and LFOs started with a low output
instead of a high, divider module would start on a down going edge and the Counter
had a wrong start state. Sequencer would not always reset on short reset pulses.
Also there were some race conditions in the Divider and the Counter making reset
give unexpected results. Also the Sequencer output was one step off, at least
… after the other fixes … GateSeq needed some mods too to make
sequencer sequencing working, and its reset did not work nice either.
Added random and clear buttons to sequencer, added an active input as well to
allow for sequencer sequencing.
Added active input to GateSeq to allow for sequencer sequencing.
Added reset input to Divider module to be able to use it for sequencer chaining.
The example patch here shows how to make a 16 step sequencer out of 11 step
sequencers — this is scalable to 16 sequencers, and to different sequencer
lengths as well.
Added lights for GateSeq steps, added Clear and Random buttons as well.
Added a button left of the control panel with the main volume controls to quickly
hide and get back those controls (near the top).
Added a direction input for the GateSeq.
And there it is … added a seq16 module. This one can sequence notes or
values, it can be used for sequenced sequencing. It has a direction input and a
random and clear buttons.
Directional sequencing is not direcly compatible with sequencer sequencing as
shown in the image to the right. You'd need more logic for that, try to use a
GateSeq instead of the Counter.
Source code release.
so anyway … your typical 41 step sequencer with gates, trigs, notes,
volumes and filter cutoffs would look like:
in version 2014.12.30.0
Added a VU meter clear on audio stop option, the option is on by default. Also can
double click on any VU to clear them all.
Some code refactoring for file selectors and display editors into a
Images for mute buttons
Increased maximal compresssor gain to 64 dB.
Color modulation for most noise modules was not working, fixed that.
Added a basic vocoder. Note: this needs the The FFTW DLL to be present.
Support for two template files, a save function in the file menu and two load
buttons in the editor. Also added export to file for selected modules and import
from file to addth modules in tha tfile to the current patch … however
that still sucks as the added modules will not go into drag mode …
in version 2014.12.22.1
in version 2014.12.21.0
MidiPlayer needs reset input and done output, the system reset does not work on it
Changed channel offsets, depending on midi file type, ch 1 will be the first
usable channel, skipping the meta channel for type 1 MIDI files. Added reset and
done, fixed reset issues.
Issue with channels on MIDI modules, fixed now.
Program crashes on exit with incoming MIDI it seems, put a gard on that.
Issue with MIDI command processing on other channels than ch 1 — thank you
Steve for testing :-)
in version 2014.12.20.1
Oh fuck … and now it does not start properly when an ini file is present,
selected devices are forgotten. Ok, it will in Legacy mode, but not for PortAudio
Ok, seems to start now properly with and without ini file and it seems to remember
the device selections for both legacy and PortAudio modes properly noww.
in version 2014.12.20.0
Added 'lowest' and 'highest' value outputs to Median module. Also allowing now for
even valued settings for the bin size — a binary or trinary sorter can be
made this way. Note however that an even setting for the bin size results
in the median being the average of the middle two values (after sorting)
and thus that this needs not be a value that was actually present in the input
stream. The lowest and highest values will always be values that were present in
Working on a simple MIDI player thingie — implementation will likely change.
There was a memory allocation still left in the Median module's DoTick method,
removed that one. It caused occasional runtime errors.
TransInv module could do an occasional division by zero, fixed.
SeqRandom does not recalculate the scale after a scale change, it needs the range
to be changed or the transposition. A patch recompile also works. Fixed that.
The MIDI CC output module had no synth binding, making it unusable, fixed that.
2014-12-19 : Need to review Reset behaviour, the response to the Reset button. It
may not actually cause a change in output values for some modules?
The envelope modules did not set their envelope values to zero after a global
reset indeed, fixed now.
This may not not be completely fixed yet … reset should last for
'system_rate / control_rate' periods I think.
When wren.ini is not present things will go wrong terribly … made it
survive that situation.
in version 2014.12.17.0
I've installed Asio4all … ever since the RME ASIO driver works too, and
I have 8 in and 8 out channels available :-)
Not all channel dB labels were updated, fixed that.
Added an ASIO control panel button on the device selection screen.
Added an audio test button on the device selection screen.
Added ASIO channel selection on the device selection screen.
Some bug fixes in device selection, and a better reset after device changes
… such that the device test function could actually work without
in version 2014.12.12.0
Bug in TimeToSeconds routine, the conversion produced an invalid index. Fixed it.
in version 2014.12.11.0
Added a couple of new modes for the Constant module, one of them being Note mode,
for easier note selection.
Changed LfoTrig Ready ouput to only fire after a cycle was ran, and not just when
Added a TransInv module for note Transposition or Inversion, it can do so with a
modulo operation being applied.
Found a better way to color modules.
In the Settings Screen F3 you can set a default color and an opacity value
which determines the color intensity.
in version 2014.12.10.0
Added a GateSeq module, an event sequencer with per step 'off', 'trig' or 'gate'
Added a Pulses module, generates pulses with random intervals, either synced to
an internal clock or free running. The 100th module type!
Added Exponential random distribution and Lambda control to RandomWalkLFO
module, as a sort of jumpiness control.
Added Exponential random distribution and Lambda control to RandSig module, as a
sort of jumpiness control.
Added a search option for un-connected modules (in the Edit menu, or use
Frequency range selector for RandomWalk LFO was not working, FM was not working
Put a guard on negative module top offsets to avoid ghost modules to be in the
patch. This might solve bugs 2014-11-28 and 2014-12-08. Some old patches may
look a bit strange after loading, as any ghost modules will be made visible now.
Changed the patch writer such that it will no longer write out modules with
negative X or Y positions, it will instead shift all the modules by the largest
offset it found.
Some bug fixes for sequencers and for the morse generator for division by zero
and an undefined string index respectively, these were related to Port Audio and
thread issues. Same sort of issue with XYScope, fixed too by making history an
internal signal instead of a patch signal.
in version 2014.12.6.0
Error in Delay module. When using PortAudio the setting of the delay length is
from a different thread than the module execution. It did happen that way that the
memory would be invalid at execution time. Fixed this.
Some issues with memory allocation for delay based modules and the EuclidianRhythm
module. Memory was allocated and de-allocated form the audio thread. I changed
that. Ah, and also the Under/Overflow debugging was a violation … moved
those to the Lights updater (out of the audio thread). That way the normal memory
manager can be used it seems — except maybe for debug messages — hmm
… nope it can not.
Changed the implementation for multi channel such that it will open up to 8 input
and output channels but never more than the Port Audio interface reports to be
present. In theory this should give multi channel IO under ASIO (at least with my
RME fireface UFX — except that ASIO refuses to work at all with it).
Currently there is no way to specify which channels will be used though. I did some
measurements on audio round trip times, see here for some results on my computer — not breath taking, but
better than it used to be.
Device selection sucks … it is needed now to restart Wren when
other Port Audio devices are selected — there is no need for that really, it
is a bug. Changed things a bit, this seems to help. Also moved the buffer size
selection to the wave device selection dialog.
in version 2014.12.3.0
Improved XY scope a bit, it's not super but it found some bugs for me already.
Added a Compare module for signal comparisons, > = and < outputs. Equality
holds when the inputs differ no more than the value set by the equality range
Added a 12 to one cross fading multiplexer.
Added a HardClip module for hard signal clipping, not really meant for audio
distortion but more for signal conditioning.
Bug in delay module where the longest delay setting produces no signal output at
all (on the variable delay time output). Fixed it, the interpolation code was
wrong, every exact integral delay time (in terms of sample counts) would produce
silence as the output on the variable delay time output.
Bug in the RandSig module, it should produce smooth glides from one random value
to the next, but it has jumps in the output value. Ok, solved, it was smooth for
the special case of non inverted full output swing signals only.
Bug in cents calculations, for oscillators a change of 25 cents amounted to a
semitone … and for the Resonator module it was not working at all. Fixed
Error in the naming for the displays for cents, they were called 'fine' instead of
'cents' and could thus not be found back by the program — fixed.
I've made patch reset on render start optional, by default it will be on. When it
is on the patch will be reset when rendering starts, when it is off rendering will
continue from where the patch was. Setting can be changed from the
Settings Screen F3.
Changed the Rotator to map an input swing from zero to one to a rotation of 2 *
PI radians, this was just PI. Old patches will rotate twice as fast now.
Changed the Median module to have a larger maximum on the number of bins, 99 now.
2014-05-17 : The terraGen module has some issues to be solved. Seems fine now
after adding input clipping, have been using it in couple of patches.
Added info popups to more valued control types, like mute buttons and mode
Added mouse wheel support for knobs, this is off by default which can be changed
and configured from the Settings Screen F3. On my laptop's mousepad two finger
click drag will result into mouse wheel actions, and with that one a sensitivity
of 50 seems about right. For fine control use the arrow up down keys while the
mouse is still in the control.
Added a 'use unfocus' option that is effective only when mouse wheel support is
on. When active and the mouse moves out of a mouse wheel controlled control the
focus will be set to the channel one output volume control. This option is on by
Changed the default width of the panel with the main volume controls to only show
channel 1. This can be changed by using the splitter to the left of it. Still
… for the time being only channel 1 will be working for audioio. Also made
the volume controls and device selectors for the non-working channels invisible
for now and hid the unusable I/O connectors on input and output module as well.
Also all the non working multi channel code was removed, this will have to be done
differently once using Port Audio. The device selector was changed to allow for
only one input and one output device to be selected, for multi channel operation
the devices will have to support multi channel operation.
Added experimental support for Port Audio, this needs a portaudio.dll file to be
present in the application directory or somewhere on the Windows path. It will not
work properly for multi channel I/O yet (i.e. ASIO). For Port Audio the buffer
size needs to be set smaller than for legacy mode, large buffers may perform
Sped up calculations a bit by getting rid of some FModulo operations, and
optimizing the remaining ones with some assembly code.
Removed the global Scala related settings, any future Scala support will be
through the Scala module.
Added a Scala module, which can read in a Scala scale file (.scl) and a Scala
keyboard mapping (.kbm). It translates incoming note values according to the Scala
files loaded into a signal suitable to control FM inputs. For this to work as
intended oscillator frequenciy should be set to 440 Hz, for filters likewise.
Sofar I've tested this only for a trivial mapping using all keys 1:1 with a 12 TET
scaling. This works, but there probably are some subtleties that I missed
And indeed some other scales I tested give output differing from what Scala
computes. Anyways …still experimental.
Text sequencer bug, it does not work at all, and causes a division by zero. Ah no
… it wants commas, and it does not like spaces … fixed the spaces
issue, the commas are needed still. And the help was wrong, it said spaces where
commas were meant. It still needs a new parser … as mentioned elsewhere.
A couple of changes on it: values can be just values or notes, a modiefie like Hz
or s can be part of the value when it is numeric. Values need be separated by
commas and after a value a repeat count can be set separated by a colon (:) from
the value. When there is no repeat count the implicit repeat count is one.
Anything not understood will be interpreted as a rest. A rest or a repeat will not
trigger the gate ouput (new to).
An example string 'A1:2, 333 Hz:3, _, C1' first A1 will be output for two clock
periods, then 333 Hz for 3 clock periods, then a rest for one clock period (
extending the 333 Hz signal for yet another clock time) and finally C1 for one
2013-11-20 : Make durations for the text sequencer, and a more forgiving parser.
2014-11-30 : For the text sequencer add a way to play rests, that is add a gate
output and some special character to indicate that a rest should be played, e.g. a
note time slot is to be skipped.
Added atan mode to TerraGen module, to be used as a soft clipping function. I've
been complaining about the TerraGen a bit in the past, but it does not seem too
bad now. I did add some input clipping to it, maybe that changed things. It did
not get too much thought.
Had forgotten to turn on the MustSave flag for the display of the Notes module, so
notes were not saved. Fixed now. Also text values were not quote escaped, so when
a quote was added to the text the patch would not load anymore after a save. Fixed
Instability issue with SVF when Q = 0, fixed. In older versions set Q no lower
than 0.01 (1e-2).
Start of rendering should first issue a reset. The sequence F9 to stop the patch,
then hit reset, then start rendering does not seem to work. Will have to look into
Two things, FixZippers was not called on render start, meaning knob values would
be stuck low for a while when a patch had not run before render start, and Reset
was not issued either, meaning rendering would start off from where it was. Fixed
Bugfix xfade module ... and the same one in Pan module - had mixed up the knob
and the control input
Bugfix pluck module, the mute button was not working.
Added random distribution controls to SeqRandom module, and removed the change all
input from it. Also added random distribution to RandomWalk module.
Added color control modulation to Noise, NoiseTrig, NoiseLFO, NoiseLFOTrig and
RandSig and modules.
Added modulation for the alpha parameter on the Average module (untested).
Added a phase detector / filter module, as a building block for PLL circuits.
Fixed the chart mode to work for non debug/profile builds again.
Pluck oscillator suffered from denormals .. fixed that.
Added a Morse generator.
Bugfix for the RandomWalk module, the random distribution stuff made the code
essentially non working.
XYScope module, this is not functioning properly yet.
SeqRandom, a random notes module. A scale can be set with a base note and a lowest
note and a highest note. A step count can be set and on each trigger a next random
note fitting the set conditions will be played. When the last step is passed the
sequence will repeat from the start. A pulse on the random input will select a
new set of notes for the sequence. Sequences will not be saved with the patch.
Made the reverb module to scale with the sample rate, it should sound better at
higher sample rates now.
Made new images for signal polarity selectors on LFO and Envelope modules.
Made a multi sample and hold module, one trigger input controls multiple sample
and hold units in one module.
Added some extra in / out pairs for the multi multiplier module.
PluckOsc added, a Karplus-Strong model.
New mixer, T1to1, a one input chainable three channel mixer.
Logic divider forgot to clear the ResetFlag, so after one reset it would stay in
reset mode - fixed that.
Fixed a bug in most of the sequencers .. had 'optimized' some code, they were not
working at all anymore.
Added a new FX tab and moved some effect modules there. Also moved the wave
player module to the OSC tab.
Added a zero crossing based frequency detector.
Added a resonating filter with decay instead of Q control.
Added a Bode frequency shifter module. The Bode shifter shifts all frequencies by
an equal amount, meaning that harmonic relations between overtones willl be
Added a running median filter. The median of a set of values is calculated as the
middle value of that set after the set was ordered small to large (only odd sized
sets are supported to a maximum size of 25). On each clock a new value is added to
the set and the oldest value drops out, then the median is calculated and sent to
Changed some initialization stuff, had to modify the HrastUnit a bit for that.
A counter module, can count binary or sequentialy, up or down and it has a reset
Fixed bug in pan module, the knob would go center to right only, not left to right.
Added a multmult module, where four inputs are multiplied by one common
multiplier into four seperate outputs.
For BitCount = 1 I've changed the gate module back to it's old functionality. The
boundary conditions were a bit odd with the new code.
Added level modulation to mixers MixM4to1 and MixS2to1.
Added position modulation level to Pan and XFade modules. The controls are fully
open by default to be compatible with older patches.
Added control depth knob to formant filter. The control is fully opened by default
to be compatible with older patches.
Made glyphs for output mode selector on some LFO and Envelope modules.
Made glyphs for wave shape selector on some OScillator and LFO modules.
Made a compressor, or actually it's an AGC more. It is feedback based in that it
tries to maintain a set reference level at the output. In doing that it can both
apply gain (for weak input signals, with a maximum of about 30 dB) or reduction
(for strong input signals, also about 30 dB). Reaction speed can be set with an
Attack control and fallback is controlled with a Release control. A pre-amp is
included to be able to shift the set point easily.
Made a KleeBit module, which is 8 steps of a Klee sequencer. The bits can be
chained to make a 16 or more steps Klee sequencer .. with some external glue.
Added a 3dB/octave (10db/decade) pink filter, to pinken up white noise.
Instability issue in Flanger/Chorus module solved.
Output types for all LFOs: 'Normal', 'Inverted', 'Positive Only', 'Negative Only',
'Positive Only Inverted' and 'Negative Only Inverted'.
Activity light on NoiseLfo was not working.
Output types for AHD and ADSR envelopes: 'Normal', 'Inverted', 'Normal Negative'
and 'Inverted Negative'.
Looks like the knob popups can be hanging at times ... made a breakout for it F8
Colored border on modules, user selectable.
Slow fall off on the display module (VU meter) .. instead of drop to zero
immediately ... which sorta worked with the old lights update regime but not at
all with the current one.
Finally found a nice compact form for an Euclidean rhythm generator (a variation
on Bresenham's line drawing algorithm, read more about that here), and turned that into a new module. [On the sequencer tab, 1205]
Added a fast mode to the display module to have the meters react faster.
This sometimes better fits the data to be displayed.
Added a CSV data module, it can read a Comma Separated Values file with up to 11
columns of data. The data is being clocked out with a trigger input, it can play
forwards or backwards and there is a reset input. There is an option for playing
the file data once or looped. When the mode is set to once the data sequence will
start again after a reset signal on the reset input (or the global reset button
being clicked on). Comments in the CSV file start with a # character, when a
comment is on a non-data line it is ignored when the file is read, when it is at
the end of a data line it will be displayed in the module for that step in the
sequence. [On the sequencer tab, 1206]
Added digitizing to the logic modules, they used to convert the input into a
single bit value (which is still an option) but they can now digitize to longer
bit vectors as well. The logic function is then executed on the vectors instead of
the single input value. Also changed the XOR and XNOR functions to perform odd and
even parity respectively. Added two new functions ONE and NONE that behave like
the old XOR and XNOR functions did, the output going active / inactive
respectively when precisely one input is active. The bit vector stuff will not
work on these two new functions.
2014-04-08 : Floating point overflows in places ... some input clipping is needed,
but not sure why signals get so large in the first place.
Okay .. it was SVF instability ... probably fixed now.
There was still debug info present in the release build, removed that, shrinking
the exe size quite a bit.
Changed the way rendering works .. it does not generate live audio anymore. The
thread's priority level is set to 'idle' so it wont hog a complete CPU. A limit
can be set for the maximum recording duration, or it can be set to zero (or blank,
or any string value that can not be translated into a float value) to make it
record till the file reaches a size of 2GB. At any time the rendering can be
stopped which will close the file. Likewise, leaving the program will close the
file too .. or when an exception was raised .. the file may be shorter then than
Set all floating point exceptions to be masked for release version. Also clearing
any exceptions after Tick executed (so errors wont propagate).
Added a property selector utility to set the value for the ControlType property
on knobs more easily. This is a module design time feature only, it changes
nothing for the synth itself ... although ... for some odd reason this halved the
size of the executble tho o_O .. for the curious, it's in the knobs2013.pas unit.
There seem to be some numerical instability issues in one or more of the modules
... the program crashed on me last night. This may have to do with FPU exception
masking and how windows deals with those, when the masking is set different from
the windows masking even something simple like Abs( reasonable_value) my trigger
a floating point exception.
Added sum output for scaler module, it gets the sum of the other four outputs.
Made 'inverted' output for envelop module, this one get the difference between the
peak value of the normal output and the current value of the normal output. The
idea being that with this output it is easy to make something get louder when
something else gets softer. Using the peak value will prevent the inverted ouput
from going negative, but it will take some time to settle. Also added a peak decay
option on this with selectable time.
2014-06-02 : Render to file option. Rolled my own RIFF writer for this .. may do a
RIFF reader too now .. it seems not too hard ... anyway, did some tests and the
WAV played in VLC, WMP and CoolEdit here, so guess they are fine ... no file
length checks tho ... so be careful when it goes over 2 or 4 GB, the code could
do 4 I think.
Had to split the knobs2013 componenents into a runtime and a designtime package.
Turns out that in order to be able to make design time property editors on has to
specify an extra compiler option -LUDesignIDE (or elese Delphi will not find it's
design interface ... ), but this causes the program to be linked with runtime
packages ... which then in turn causes users to complain ... Anyways, when you
want to compile your own you need to uninstall the knobs2013 knobspackage and then
install the new Knobs2013Design package. Also added a couple of missing icons to
the DCR file. This explains the file size halving too.
Made popup menu's for selectors, so you can now right click on say an Lfo speed
selector or a mute button to select a value directly instead of clicking to cycle
through the values. Selectors currently are indicated by a yellow border.
Working to port over a FlangeChures and a PitchChanger I had laying around since
the Turbo Pascal days .. or was that Borland Pascal ... old anyway .. from the
days one could directly make interrupt code in Pascal. There seem to be some
overflow issues to be sorted out. Some issues still with negative rates for the
Removed profiling code for releases, this makes the program run a bit faster.
Fixed hint / popup window for knobs to keep displaying when the value is changed
with the mouse. F8 will show all knob popups .. not too useful actually.
After patch recompile dezippers are set to the final value immediately now to stop
the slow glide-in that happened.
Changed the module designer a bit, such that the module's PageName property will
be automatic if the module happens to be on a TTabSheet with a non-empty Caption.
Changing some signals for AHD and ADSR modules to control rate to make them more
Changed the trigger lights on AHD and ADSR modules, they are on now when the
envelope is in some active state, i.e. they now indicate envelope activity and no
longer the presence of a trigger signal. This works better for very short trigger
pulses taht would not be visible otherwise.
Removed the denormalized exceptions from the floating point exception mask to be
able to find cases of denormalized numbers. As a result of that I've added
Normalize() calls in a couple of places (Including the HrastUnit).
This code will only be active in debug compiles of the program, to help find
execution speed issues.
All of this seems to have sped things up a bit.
Made update rate for lights user settable (settings page - press F3).
Interpolating random signal generator. Moving from current value to random value
at a rate determined from the frequency it runs at - smoothly.
Mute was working the wrong way around on Noise and TrigNoise modules. And for LFOs
Made a built in rectifier for the intdif dif output, so it can be used for trigger
on signal change operations more easily.
Square thingies not working unless PWM modulation is being opened up - fixed.
Envelopes still act strange ... seems better now, removed some code
2014-05-01 : HrastOsc vintage knob must not be in the knob map - fixed that.
User selectable sample rate: 44k1, 48k, 88k2 or 96k. Also changed the control rate
from being 'sample rate' / 4 to 'sample rate' / 8.
Moved intdif module from control page to filter page and changed it's type ID from
312 to 1003. When an old patch will not load anymore please change the type ID
for all intdif modules (type = 312) to 1003 - this can be done with any text
editor (like notepad).
Removed lights handling from audio loop and made it timer driven instead.
Removed some synchronization code that did nothing.
Some more LFO range issues fixed, forgot range for SquareLfoTrig.
Added Lin range for LFOs, the FM will work linear then too. For trig LFOs this
means that they may never finish their sequence.
2014-04-30 : AHD envelope has clicks it should not have. Had A and D shapes
reversed, looks like a click on the attack phase is more acceptable than one at
the end of the D phase. This may need a bit more work later.
Added exception handlers for synth engine crashes. This will stop audio
processsing and issues a message about that .. instead of the usual four ...
Fixed square LFO speed range selection, I had a typo.
Made the quantizer working again, also it displays note fractions now instead of
arbitrary numbers. To do that I moved the rationals source module from the project
itself to the knobs2013 package.
Fixed LFO out lights, another typo.
Major overhaul in how knobs are handled internally. The patch version has to be
changed as a result, and current patches will not be compatible with this.
Made a popup menu for connectors - this sort of works .. break is not quite what
I'd like yet - it should break the downstream connections, but currently also
breaks the upstream one. Delete deletes the whole connection tree where the
connector is on. Disconnect removes this one connector from the tree.
Made a module popup menu too, with cut copy and delete options.
Made a knob popup menu as well, added default values for knobs and the default
can be set from the popup. Knobs can be locked and unlocked from the popup too.
When an editable display is associated with a knob editing the value through the
display will lock the kob to prevent the edited value to be changed by knob
movements. The lock can be undone by unlocking the knob or by setting it to it's
For 'File Save As ...' cut out the path in the save dialog to only show the file
Changed the connector coloring a tad.
Had messed up oscillator frequency calculations, they were two octaves too low,
2014-03-03 : It happens on module insertions that the lights will not be updated
(properly) anymore. It can also happen that a knob on a module controls something
on another module. So it looks like the patch compiler has some issues. This seems
2014-03-28 : When an envelope of type AHD or ADSR is in fast mode and the patch is
then saved the values will not always be properly set after reloading the patch.
The same thing happens when such a module is copied and pasted. This seems fixed
2014-04-01 : There is an issue with Knobs and setting values ... like an
oscillator frequency by default is 440.00 Hz but after some messing with the mouse
on that knob it is impossible to get that value set again. Only way to get it back
is to drag the knob all the way to one of it's ends, then use the cursor up/down
keys to try to get past that end, then use the cursor keys to step all th way back
to the value ... yuck. This seems fixed now.
Defaults, stepcounts and ranges for several modules were changed.
Sequencer chaining would crash things on a floating point overflow, fixed.
Adding lookup tables, my earlier conclusion that they were faster only marginally
was wrong. This will limit modulations to foreseen ranges though, and some value
clipping will be needed. See previous point tho .. this would be needed anyway.
Cleaning up code, there are duplicates all over the place and bits of unused code
Adding speed ranges for LFOs, Slow, Medium, Fast and BPM.
Selectable waveforms for LFOs and oscillators to reduce the amount of modules a
bit. Sine, Saw, Tri and Square.
Changed peak level indication for main in and output displays to slowly fall off,
also added clip indication with a highlight color on those.
Reversed knob direction for Average filter. So now the cutoff frequency get higher
when the knob is turned right, like how a filter works.
Removed the logic gates And .. XNor and replaced those by one Gate module with a
selectable logic function. The same functions are available still.
Removed variable input count stuff, it had been disabled anyway.
Fixed bug in rectifier module, full wave mode did not work.
removed linear LFO and OSC stuff .. need to rethink that.
Repeats on the up / down knobs under knobs work just once and then never again.
Ah .. forgot MouseDown := False in MouseUp handler, that fixed it.
Added HrastSVF 12 / 24 dB resonating filter.
Added cents control to HrastOsc
Frequency fine control for oscillators added, in cents. Range -50 to +50.
Added dB / Linear control selection to mixers, changed some attenuaters to dB
mode here and there, added some level controls here and there. Renamed some
stuff now that the patch versions are incompatible anyway.
De-zippering added for knobs on modules, inputs to be de-zippered can be added to
a knobmap (at compile time) and the process will then be automatic.
This will not work on mute buttons that use the SignalToLogic function to decide
what to do, best to use SignalToMute( FInCache[ x]) * OutLevel or somthing alike
instead of an if then else thingie.
Added exponential FM control for wave player. To use this the speed input must be
connected to a signal source unequal to zero.
Could have made the speed input to default to 1 to solve this , but then the
position input would need the speed input to be set to zero ... compiled patches
ideally would know if an input is connected to solve this issue.
Added color to random walk module
Added attenuators to square LFO and OSC modulation inputs.
Added time modulation for AHD and ADSR envelopes.
Fixed bug in AR and ARretrig envelopes that became visible after adding the
dezipper code, it would try to dividde by zero.
Fixed bug in triggered LFO and OSC modules, they did not stop running always.
2014-04-02 : Module paste can still mess up the module layout, and also can get
modules at negative offsets still it seems.
Looks like I finally nailed the the problem where the module lights stop working
after a new module got inserted, or rather it would stop after a module was moved
in the editor ... this was partly treated as a patch change and partly as a no
change, resulting in the editor modules to be renamed and the patch not being
recompiled - I've deleted the editor name changing line.
Repeats on the up / down knobs under knobs work just once and then never again.
Ah .. forgot MouseDown := False in MouseUp handler, that fixed it.
2014-04-01 : Unit interpretation in displays for knobs is not correct.
I think this is not relevant anymore.
Fixed some graphics stuff that got broken over the months.
Window state saved in ini file, when the window was maximized on exit it will be
set maximized again when the program is restarted. Minimized state is forgotten
Added a reset button which resets some module state on click, there are likely
issues with this as reset is not as trivial as it may sound.
Added mutes per input on MixM4to1 and MixS2to1 modules.
Solved issues with output module, it does not need to be connected anymore and
there can be as many as you like. The AudioOut module now gets compiled as any
other module, removing the need for special stuff in the main program. TAudioOut
is now treated as special in the synth engine itself. All inputs from aydio outs
will be added there and the result is then copied over to the output terminals of
the synth patch. The output mapping stuffthat was used is no longer needed.
Solved issues with input module, it does not need to be connected anymore and
there can be as many as you like. The AudioIn module now gets compiled as any
other module, removing the need for special handling in the main program. It is
now treated as special in the synth engine itself where values from the engine's
input queue are copied over to the outputs of all the AudioIn modules present in
the patch. The input mapping stuff that was used is no longer needed.
Solved issues with knob default values of zero to not be used.
HrastProgrammer contributed a HrastOsc bandlimited oscillator module, thanks
I Changed the module layout a bit for this one to make it fit better into the
scheme, and also added a little code to it to make sure that disconnected cables
on control inputs will be reset to zero.
Fixed the situation where knob values zero could not be a default, changed some
default settings for some modules.
Added the program icon to the source tree.
Fixed the project's DPR file, Delphi had put some crap in it.
A little speed up for filters and oscillators, changed some signals to control
rate. FM stuff may sound a bit different now, however I did not change the linear
A patch changed warning on program exit. This is on by default but can be turned
off in the settings screen.
Added a debug indication in the title bar for the debug version. A debug compile
is less efficient and should not be released by me, but I might do so by accident.
Removed dependency on TLed component. New Source code (2014.3.29.0, build 57)
Stereo wave players. Wave players still need to be 44k1 samples/second 16 bit wav
files, but now both mono and stereo files can be loaded. For a mono file the L
and R outputs will have the same signal, for a stereo wave the mode can be set to
stereo or mono. When set to mono L and R will both have the sum (divided by 2) of
the left and right channels from the wave. In stereo mode the L output will have
the wave's left samples and the R output will have the wave's right samples. Added
an indicator to show wheter the wave itself is stereo or mono.
Chainable 6x6 mono matrix mixer module.
Source code for 2014.3.28.0 (build 56) is made public, see the Downloads section. The customif components have
changed, so they will need to be recompiled too (before recompiling the new
Search shortcut changed to CTRL+F
Bugfix - some patches could not be read, and sometimes modules could not be copied
because the patch reader would not accept values in scientific notation (like
Search function on module title, highlights found modules, F7 shortcut.
Support for netMIDI / tangle, MIDI over TCP.
This got implemented by using the Overbyte ICS TWSocket component. When you want
to compile WREN yourself you will need to install that component. It can be
downloaded from here, look
A zipped up netMIDI executable is available for download from here - see the
Downloads section (netmidi.zip). Will make
source code available at some later stage - it is old Delphi code and it needs to
be ported to Delphi XE5 first. The old code is available on request though.
Note that MidiTangle can probably be used as well http://jcfmusic.com/miditangle.htm
Rectifier module, MinMax module, Integrator/Differentiator module, MIDI note
receiver module, MIDI CC receiver module, MIDI note transmitter module, MIDI CC
Horizontal and vertical knob change modes, in addition to the circular one used
now. Built in a treshold for movements as well, such that after the click a
minimum distance has to be moved with the mouse before the value will change.
Ok .. it sorta works .. except for knobs near the edge of the screen ... would
have to make the mouse cursor invisible .. jump to screen center then handle
the moves, then jump back to start location and make the pointer visible again ..
that is for later.
Hit F3 to go to the settings.
Made a 2D vector rotator.
Made a ready output on the triggered LFOs, and a sync output on the other LFOs.
Made a file open dialog on the wave player and now display only the file name and
not the full path anymore. Old patches will still load OK.
Added control knobs to xfade and pan modules.
Added sequencer step outputs that go active when the step is active.
Added sequencer chain input.
Added some mixers, 4 to 1 chainable mono and 2 to 1 chainable stereo.
Fixed the problem where loading a new wave player into a running patch would crash
Made oscillators and LFOs with linear frequency control - not quite happy with
Ratio control module to multiply a signal by some ratio, like by 5/8 for example.
Added D and RS flip flop types.
Combined WavePlayer and WaveStorage modules into a new WavePlayer module that both
has speed and position control. When a patch with a WaveStorage module is loaded
it will be replaced by a WavePlayer module.
Added duration ranges (speed) to AHD and ADSR envelope modules.
Logo and some looks changed.
Help system changed so that it talks about the modules as they appear in a patch
rather than as how they are seen by the synth execution engine. The help text for
the modules is now set from the module designer rather than from code.
When F1 is pressed and a module is selected in the patch editor the help will
try to directly move to the help for that module.
Added images to the hint / tooltip windows popping up from the module selector.
Some scrolling / panning issues .. it still sucks, but less so.
VU meter looks changed.
2013-11-23 : Look at notes / octave setting and the 440 Hz reference setting, does
it really have to affect LFOs too?
No ... made LFOs independent of that mechanism and based on 12 notes / oct and a
reference A of 440 Hz.
A wave player with variable speed playback and made a wave storage unit too, that
one has a position input meant to be controlled from a sawtooth wave. The storage
module seems more flexible, as it is easier to set begin and end points on the
wave by constraining the saw amplitude and offset, and also by FM modulating the
saw there is a form of pitch control. When the saw is inverted it can still play
the wave backwards.
Changed the looks a bit again, new knobs mainly, shifted around knobs, texts and
connectors here and there, fill color for envelope graphs added, clearer
Fixed the issue with moving selected modules being way too slow. However when a
bunch of modules are dragged too high up module restacking may change the order of
some modules still.
2013-11-21 : in TConnector.MouseMove make the wire panel scroll when the mouse
goes out of its visible area, but do not scroll beyond the current size (as
defined by its module content).
2013-11-22 : Autoscroll needed for when modules get dragged out of view and when
wires are dragged out of view. Ths scrollbox should keep the point where the mouse
is in view in those cases.
Implemented "follow mouse" for cable drawing and module placement such that when
the patch editor has scroll bars it is still possible to patch with it.
Semi fixed the arrow icons on the 'noknobs' (as used for instance for note
slection on sequencers). This will need a better soluition at some stage.
Started implementing mousewheel messages to perform scrolling.
Vertical scrolling works OK now, horizontal not quite yet.
Fixed crashing popup editors.
Ported application from Delphi 2009 to Delphi XE5.
Fixed delay control knobs
2013-12-17 : S&H bug, it acts like T&H. Fixed now.
Made the lights update a factor two slower, gives room for about 10% more modules
it is a bit more sluggish now tho, light wise. was about 86 ms, now about 12 ms
or just a tad slower than 6 Hz.
2013-12-16 : Locale issue - patches saved in a locale that uses a comma for the
decimal point will save patches with that comma, but on read back the program
insists on a dot. Ah .. and that went wrong on internal representations too. Looks
like fixed ... global locale settings needed to be set OK. Reported as fixed too
Added focus visibilty and keyboard arrow handling for TKnobsValuedButton(s)
Sped up sine oscs by 25% or so.
Made the bar indicators display dB's instead of arbitraty units, the Display
module also works now.
Got rid of the TKnobsValuedButton)s), made a new TKnobsSelector component which is
smaller and looks better.
Added mute buttons to a lot of modules, about 30 in total, tested most of 'm. The
envelopes have a three mode mute function, it can set the envelopes to pass
through the signal evenwhen no trigger comes in ( pass mode, the other two modes
being unmuted and muted).
Profiler gives some more detail, like wire performance and overhead. The lights
on the modules are nice, but it seems to be like 25% overhead comes from them.
Changed some colors.
Streamlined text to value conversions, so they work the same everywhere now. It
also accept both note notation styles A#4 and A4# now, and even A#4# .. oh well.
Lil fuckup with text sequencer and optimization of initialization code - it
crashed the text sequencer with a division by zero ... anyway, fixed now. It needs
a small modification to the TKnobsDisplay component .. didn't reflect that in the
published source code yet.
Fixed tab order on modules and added tab stops on some controller widgets that
didn't have it yet.
Made all value displays editable (so values can be set with better resolution than
is possible with the knobs). It should also be possible this way to enter
different modifiers from the default one's - like use note names for an
Value displays can be 'tabbed to' as well now.
A new way added to open an editor window - tab to the value display or click on it
and then hit enter.
Several small optimizations in the drawing routines, and less unnecessary screen
and synth updates.
'Move final' behaviour on knobs changed - this mostly affects the delay lines
where new memory will only be allocated when the total time had not been changed
in the last 739 ms.
Added offset parameters for the editor, so modules can have a fixed small border
on the left and top, making it easier to select modules near the left or top.
Changed contact details in the about box, and in this file, to point people to the
Added a RandomWalk module - or brownian noise as some also call it. In this one
a virtual particle performs a random walk in a confined D space, when it hits a
wall it'll bounce off. X, Y and Z outputs are available as outputs.
Made the lock mechanism on the scaler modules work, in lock and antilock modes
the knobs movements are synchronized. This way it is easier to to scale a value
and offset tit at the same time. See it as a modulation depth control.
Some visual details changed, like LEDs a bit bigger, symbols on modules lined up
Envelope shapes in documentation are better now, some preset display types added
for the TKnobsDataViewer component. Mainly a 'design time' item.
2013-11-23 : DEL key not always working for module deletion, and then ctrl+DEL
does not seem to work either. As if the patch thinks deletions are not possible.
This seems fixed now.
2013-11-19 : Get knob changes into undo history. Did it.
2013-11-19 : Compile less often. Did that too.
2013-11-18 : on edits of the text in the text sequencer the synth is reported
crashable, prolly related to the one below. Sofar I've not been able to reproduce
this on various 7, Vista and XP machines. Unreproducible.
2013-11-18 : someone reported an OSC module disappearing - leaving some wires to
go nowhere, prolly related to the one above. Sofar I;ve not been able to
reproduce this on various 7, Vista and XP machines. Unreproducible.
2013-11-18 : there is a bit much latency from input to output. Maybe start using
directx instead of legacy audio., such would prolly improve device name displaying
as well in the audio device selector. DirectX will not solve this for Vista or
later ... maybe look at ASIO later, or the underlying mechanism used by ASIO.
There was a problem when pasting multiple modules with the mouse capture keeping
stuck on all but one module, fixed this.
Some new modules, some bug fixes, renamed all controls, old patches will not load
anymore (but forgot to change the patch version). Old patches can be fixed in a
text editor BTW.
Ran into some float denormalization issues making some patches about 15 times as
slow as needed. Added undenormalization on many outputs and also reconsidered the
antirunaway code a bit. This now prevents absolute signal levels from both getting
too high and too low (too close to zero). erm .. this could be formulated better
... the anti runaway code now also prevents float denormalization to ripple
Added some profiling code, so that one can see now how many processor clocks are
needed for the tick and slowtick module methods. Run a patch for a while, then hit
the profile button. This will make a file profile.html in a data subfolder of the
application folder. So one can see the computational bottlenecks.
The keyboard shortcuts got some more attention, seems ok now.
Added a fixed delay output on delay modules - this one gives the maximum delay as
set by the delay knob, it is not modulated.
Changed the output name for Avreage filter to LP and added an HP out. Also changed
knob response for this module to give better resolution at the high end.
2013-11-25 : New envelopes have clicks that don't belong there, ah no .. it's a
gain issue ... the published version has non working gain, and on trying to fix
that I got the clicks. So ... gain on the new envelopes does not work.
Fixed now, gain works, no clicks.
Added two new envelope modules, AHD and ADSR with them nice envelope displays that
respond to the knobs :-0
2013-11-23 : Divider indicator not working, and some more. Fixed various little
issues for this release, enough to be able to actually do my first noodle with it
Source code for 2013.11.23.3 is made public, see the Downloads section. The customif components have
changed, so they will need to be recompiled too (and before recompiling the new
2013-11-23 : LFOs run a factor 2 too slow. Fixed.
2013-11-23 : Look at filter tuning, needs rescaling like the OSCs had it and it
seemed off before. Filter tuning did not work at all, fixed now and tuning seems
to be about right.
Changed the Looks a bit
Display editors can handle ENTER now inside text, use CTRL+ENTER to close a
Had forgotten to convert the trigger OSC and LFO stuff for expo control. Also
trigger lights on those modules did not work. also had forgotten the SVF.
2013-11-18 : Exponential frequency control. That, and rescaling of knobs for
frequency control on oscillators and LFOs. Also removed the quantizing to knob
steps, so that once editors are available for values all arbitrary values could
be set (within the resolution of an IEEE double float).
Implemeted a setting for notes / octave, not sure how useful that is as it is.
Various changes in the user interface, added an about page, recent file history,
patch name editor, added some keyboard shortcuts (F2 will show the editor, F3
Name changed from 'modules' into 'WREN'.
2013-11-21 : When a new version runs and F1 is pressed the help files should be
regenerated - did it.
2013-11-20 : it was reported that when scrollbars appear displaying the visible
area starts to behave odd, sofar I've not been able to reproduce this on various
7, Vista and XP machines. Ok, I see what was meant now, working on it. Seems fixed.
2013-11-18 : SVF seems to lock up for (very?) high or low FM values. Added
clipping for the sum of frequency and fm to a range 0, 1.
2013-11-18 : Feedback delay loop seems to lock up under positive feedback
conditions. Added cliping to range -16, 16 to input.
2013-11-20 : trig lights on AR modules not working, the names are wrong - changed
2013-11-20 : Note number to signal and signal to note number do not work correctly
for 'negative' octaves - (i.e. A-1) this needs a mathematical modulo calculation
and floored division to make it work correctly. Introduced two new functions
Remainder and FlooredDiv to perform modulo and division for this case.
2013-11-20 : Make a logic NOT module - did it.
2013-11-18 : Pressing DEL in an editor :: Changed the DEL behavior a bit, removed
it as a menu shortcut and instead let the application handle it. the application
will first ask the editor if modules can be deleted, and if so it will do that and
mark the key as being handled (so it will not be passed on to anything else). When
the editor is not in a state where it can delete modules the key will not be
handled by the application and be passed on to the control that has keyboard focus
(i.e. a LabelEditor (for module labels) or a DisplayEditor (for
the notes editor in the TextSequencer). This way the DEL key can still be used for
both functions, delete modules when possible and delete text from an editor
For another way to solve this see http://embarcadero.newsgroups.archived.at/public.delphi.vcl.components.using/201009/1009246070.html
2013-11-18 : pressing DEL in text sequencer editor will delete the module while
the editor is still open, crashy stuff from that
'Solved' with a bit of a kludge - module deletion is not possible when there is an
editor open - either on a text sequencer or on a module title (which has the
2013-11-19 : The resource files customif.dcr and bitmaps.res are missing in the
source zip - made them seperate downloads for now, they go into the customif
2013-11-19 : Get some unused stuff out - diddit.
2013-11-19 : Show input buffer over and under flows in debug screen - diddit.
2013-11-19 : Fix error with quotes on version number - diddit.
2013-11-19 : Make version numbers visible in the program - on the settings page.
2013-11-18 : A shorter delay - made it.
2013-11-19 : Random picks from text sequencer, as a random lookup table - added a
2013-11-19 : Introduce patch versioning, the proposed change from lin to exp
frequency control will make patches incompatible - current patch version is 1.
2013-11-19 : Globals.pas in the customif stuff should be renamed, two times the
same filename is troublesome (there is a globals.pas in the main project too) - it
is glowballs.pas now.
2013-11-18 : color on SVF filter outs is wrong - blue instead of red - changed
2013-11-18 : audio in does not work - it works, but the level display is wrong -
2013-11-18 : audio in does not work, the input values are garbage, this needed
buffering - added buffering