So come up to the lab / and see what's on the slab / I see you shiver with antici...
  • Just a couple of speculations and questions for the new module.

    When it is a new audio signal processor where are the audio inputs on it?
    Audio Inputs on the right does not seem to fit the other of Oliviers Designs!?
    Could it be an expander/modulator for a Mixer Module? That works also without the Mixer?
    Are three of the outputs uni and 3 outputs bipolar?
    Do you remember Oliviers doodle for module 1 with the pentagon?(see page1 of this thread)
    The Pentagon and the other drawings seems to me like a modulator!
    Bias could be a control to change the direction in which rings modulate?
    Jitter could add randomness to the modulation?
    Step could control from which corner of the pentagon the modulation begins?

  • 6 Outputs > Pentagon!?

  • Maybe the poti on the top controls the corner of the pentagon and the one below is unipolar cv-control over the position? These two seems connected.
    The switches next to the Bias could switch between different slopes?

  • Well, after looking at the video again, and from what the PCB tells us, it is clear that the four right hand outputs in the bottom row output bipolar control voltages (hence the bipolar LEDs), and the two left hand outputs in the bottom row output gates or triggers (and their LEDs are probably just monochrome – it looks that way in the video). And despite the fact that it uses an STM32F4 processor, it isn’t an audio module – as Olivier points out, Cortex M4 processors are now actually cheaper than the much less powerful Cortex M3 processors used in Frames, Peaks, Braids, Tides etc.

    I’m sure it will do some very novel and interesting things. And it will be a great platform for hacking in all manner of other interesting capabilities…

  • can’t help but suggest some manner of vcf/vco in the line of the dr. octature/mankato – potentially with allowing independently morph-able wave shapes/phase thus the x/y/z in lieu of traditional “degree” outputs (thinking of how grids interpolates patterns from a map, only applied to wave shape). the “jitter” may skew the phase relative to the clock input. also, the bottom left jack seems to my blind eyes to read “q”... perhaps freq and q for the larger pt3-ps knobs? if the module doesn’t process audio, the filter could be a similar digital implementation to that on tides for modifying wave shape… or maybe i’m just completely wrong!

  • What about a random generator Olivier talked about recently in a french forum
    And with some sequencing quantising env lfo addings

  • I’m thinking multi-channel CV pattern generator, too.


  • No, that’s Oracles, scheduled for 2017, with drum pads to input patterns :)

  • I only mentioned it, in jest, because someone suggested suffix trees as a possible solution for higher-order HMMs learnt from large volumes of health care data, which led me to factor oracles, which reminded me that you had mentioned them at some stage, which led me to read the 2004 Assayag and Dubnov paper about musical applications of them.

    An Oracles module might even be feasible, although with a MIDI input so you can play it canned MIDI music, and then it riffs endlessly on the theme. Although of course it makes no sense to train the oracles in the module, but modular synthesists hate general-purpose computers, so that might be unavoidable. And besides, firmware for your Oracles module should be written in Montparnasse, in Python. Or in an object-oriented dialect of Pascal, if you insist.

  • @BennelongBicyclist : do you have any reference(s) for “a possible solution” wrt. suffix trees & higher-order HMMs? I am currently working on something related, I am just curious :-)

  • My colleague, who is a computer scientist, just said “consider suffix trees” in a terse txt message, so first I looked at the wikipedia page for them and then googled and picked the first paper that was readily accessible, which was this one: – it is mostly comprehensible to a non-computer scientist like me. We work mostly in R, and conveniently there is an R package that implements probabilistic suffix trees: We haven’t investigated how it scales yet, but it seems to work well with small test data. There’s also a Python library that wraps a C++ PST implementation. One of the papers, I forget which, mentioned factor oracles as an alternative approach, and that rang an Olivierian bell. Wikipedia to the rescue again, and it pointed to the Assayag and Dubnov paper from IRCAM. That’s as far as I have gotten so far with factor oracles, because the musical applications suddenly seemed a lot more interesting than trying to predict where a patient would pop up next in the health system (actually, that is a very interesting question). I haven’t looked at their OpenMusic implementation yet, but I have played with first-order HMMs in OM, so will have a look.

  • FYI: Olivier updated the Modular Grid page of Warps, incl. price info:

  • Warps is officially on the main MI Eurorack product page as well.

  • And what is trails module in the foto next to warps then ?


  • @Charlie: well spotted!

    So we now know that this module is Trails:

    So what does it do? Well, its’s analogue. And the blue side looks like it is for the attack, and the red side looks like it is for the decay, and both are adjustable from linear to exponential. Is it an envelope follower? One follows a trail?

  • The clue Olivier has left is a Google StreetView shot of a road ascending Le Truc de Grèzes in Languedoc-Rousillon:

  • It’s quite big for an envelope follower. Not sure it would need so many controls. Could be a combined envelope follower and lag/slew, though.


  • Trails = analog lag/slew/slope thing that I won’t release. I have it in my rack and am having fun with it, but there are enough reasons for me to let it rest in peace.

  • Interesting. Were there technical issues, or marketing ones, I wonder? I have a feeling it might not be possible to squeeze the multi-function versatility MI users have come to expect into such a module. Maybe that was part of the reason.


  • There was a technical issue with a certain type of input signal that could not be fixed without blatantly ripping off the original Serge schematics (something everybody else is doing of course, but the point of the module was to do something free of any Serge heritage). And there are enough good modules covering this territory.

    > I have a feeling it might not be possible to squeeze the multi-function versatility MI users have come to expect into such a module

    “Multi-function versatility” is a hole I badly want to get out of.

  • On a side note: trails has a very nice icon and panel …

  • @pichenettes I can see why the multi-function thing might be a bit of a millstone (especially as it’s much easier to do with MCU-based modules than all-analogue ones), but it’s also a USP for many of your modules.


  • @pichenettes I’m still holding out for the Mutable Instruments ‘external input processor’ module, with envelope and pitch extraction, that was mentioned a while back, since Trails clearly isn’t it ;)

    I do think there’s a gap in the market for a module like that. The only module I know of that attempts pitch-tracking and pitch CV output in the digital domain is the Disting.

    I definitely think a dedicated module for extracting pitch and envelope CVs from a monophonic instrument or voice is something lots of people would buy.


  • @BennelongBicyclist: nice! thanks alot! I’ll take a look at your bioinformatics related paper. I had a short skim through the Assayang paper, which is also a nice pointer for alternative approaches to HMM construction methods, thanks! The current ISMIR conference proceedings have some related work for HMMs as well, e.g. (depending on if you are interested in generative models or pure analysis).

  • I’m mainly interested in generative stuff, but that ISMIR paper on score-following is nifty – so in hybrid human-computer performances, the human can improvise, ad lib and embellish the score, but the computer works out what’s going on and doesn’t just plough on. Thus human-computer jam sessions become possible.

  • Yes, exactly! It would be interesting to experiment with applying such a model for other musical forms where you dont have a one-to-one mapping from score to instrument performance, e.g. for consistent (wrt. score) automation/parameter control with low dimensional performance input.

  • @toneburst: If someone could point me in the direction of a pitch detection algo that is compatible with an Arduino Leonardo or an ATmega32u4 microcontroller, I could add that to my simple envelope follower and gate out code. I use it to trigger drum sounds mostly, but pitch detection would be very welcome!

    I’d need to tidy it up a bit, but I would have no issues making that code opensource as a lot of people want that kind of interface for their modular, and it is very useful even without the pitch detection!

  • > pitch detection algo that is compatible with an Arduino Leonardo or an ATmega32u4 microcontroller


  • @audiohoarder that sounds cool. I have a feeling doing pitch detection well might require something a bit more powerful than Atmega32. Maybe Olivier could comment on that. Would be cool if you could get it to work, though..


  • @pichenettes random()?


  • A pitch detection algorithm simple enough to work on an atmega32u4 won’t be much more useful than spitting random voltages between 1V and 3V :)

  • I thought that was probably what you meant.
    You could probably do something worthwhile with a 32-bit MCU though.


  • @pichenettes: Exactly, haha! :)
    The 32u4 is very powerful for a lot of audio and midi applications, but accurate pitch detection is not one of them unfortunately.

    @toneburst: Look into the Ardcore module.

    It can do anything a vanilla Arduino Leonardo can do, but in Eurorack form. Another version to consider is the NS1 Nanosynth which has a built in Leonardo for usb midi using the Ardcore library. Of course, you can easily add extra CVs to the NS1 while still running midi.

    Also, 8 bits is fine for coarse CV tuning. I have a usb midi to v/oct code that runs just fine after all. The low kHz and the tiny amount of ram for the buffer is what makes a pitch detection algo pretty much useless. As usual, bit depth gets blamed when it shouldn’t due to the Nintendo/Sega/Atari console wars.

    Side note: The icon for Trails is crossing the streams, you can’t release a module that actively crosses the streams. ;)

  • Warps is now well-and-truly out of the lab – it’s in the stores – but to close out discussion of it in this thread (which is dedicated to what’s still on the slab), here are the four SoundCloud demos of Warps which Olivier recently uploaded:

  • Warps sounds amazing! Yet another winner. :)
    I can’t wait to hear what people do with it!

  • While we’re listening to SoundCloud, here’s another clue from Olivier from about a month ago. The description is:

    How many polyblep oscillators will you have with that STM32F4? 72. And Elements’ reverb with whatever CPU is left.

    This may just be an experiment, but it may be a clue that one of the new modules will focus on harmony, in interesting ways.

  • If I had to code a pitch detector running on an 8-bit AVR, here’s how I would tackle the problem:

    • On the hardware side, a steep analog LPF filtering everything above 4kHz. You want the input signal to look as much as possible like a sine.
    • Sample the data at 8kHz – so that you have about 2000 CPU cycles of processing / sample.
    • Store only one bit per sample (sign bit) – so one byte in memory will store 8 adjacent samples.
    • Use XOR and a bit-counting LUT to evaluate the cross-correlation – 8 pairs of samples compared in 3 or 4 CPU cycles.
    • Get a coarse estimate of the period (say use candidate periods which are multiples of 8), then do an exhaustive search in a limited interval.
    • To deal with the low sample rate, use parabolic interpolation around the 3 maxima of the autocorrelation.

    You could probably run this 10-20 times per second, and interpolate between values.

    I think I could get it to work, the same way people make 3D demos on C64s :)

  • @pichenettes: Fantastic advice as usual. I do appreciate it. :) Since 8 samples would only take up one byte, the minimal RAM is no longer an issue.

    I do think that there would be some issues to run into when running the gate/trigger out and envelope follower at the same time. At least the way I have the envelope follower now, as it needs more bits than one per sample. Either way, I will try this in my spare time. No guarantee that I can get it all to jive.

    >I think I could get it to work, the same way people make 3D demos on C64s :)

    With more powerful microcontrollers available for a few dollars more, it is relatively pointless to accomplish. However, for a single use, not a bad cost/function trade off.

  • One outstanding thing with Warps: what is the Easter egg? Both Olivier and some beta testers have confirmed its existence on MW.

  • Olivier provided this postscript to Trails, which is destined to never got off the slab and won’t ever leave the lab, on MW:

    The module had a few unusual things:

    • Voltage control on the expo / linear / reverse-expo thing. It had a relatively simple compensation scheme that kept the time/frequency within 0.6x and 1.5x of its central value (unlike Maths/Function where it goes between 0.25x and 4x of the central frequency) – but not good enough for stable tracking as a waveshaping VCO. I suspect the WMD Minislew does the compensation in software.
    • Another voltage controlled shape control. It smoothed curve corners (like a LPF on the output, but tracking the A/D time – this one had absolutely no influence on frequency in LFO/VCO mode).
    • The one-shot/loop setting was not on/off but was a duration, from 0 to infinity (permanent loop). So it could make short woopwoops.
    • There was a 1:4 switch controlled by the stage (take off, flight, landing, landed) – which was very odd but costed nothing to add because the core of the module was a dual 1:4 switch IC. It was normalized so you could use it in one direction to get gates for each segment, and in the other direction to inject stuff into each segment of the envelope (say I want to replace the stable stage by an external LFO)... Or of course like any 1:4 switch timed by the A/D envelope.

    Precisely because it works with a switch IC, the module has a quirk that the Serge 4-transistor “asymmetric multiplier” arrangement doesn’t have. The quirk, (the solution of which completely goes above my knowledge of electronics), prevents the module from correctly slew-limiting signals which are already slew-limited (like slow triangle or sine LFOs above a few Hz). This is a deal-breaker for me. And I’m certainly not going to steal from the Serge circuit to fix it, so please forget Trails and get the WMD Minislew. Or the new Befaco Rampage, it looks awesome!

    Hmmm, I always wanted a Befaco module (I’ve bought components from them, and have their sticker on my suitcase)....

  • Interesting reason, it sounds a bit like a non-issue to me, but with so many slew/waveshaping combo modules already available it isn’t disappointing to know that Trails will not be seeing a release.

    I’m more interested to find out what is behind that “Peace” demo. :) I have no idea how 72 voices of polyphony would be easy to access with CVs, but that is what midi over 1/8 in minijack is for, right?

    I would also guess that the Warps Easter Egg is some kind of extended FM where the big knob controls the algorithm, and timber controls the feedback.

  • Bored with working on a Sunday afternoon, I started to look for cool DIY modules to add to my ever-growing pile of unbuilt PCBs. I came across a re-released version of Jürgen Haile’s Tau Pipe Flanger/Phaser, with 20 poles.

    It then occurred to me that the Easter egg in Warps is probably a flanger/phaser. Anyway, that’s my bet.

  • On the topic of slew / function generators, anyone have any strong feelings on the WMD/SSF Minislew vs. the Make Noise Function? The Minislew has a few extra features the Function doesn’t have, like the aforementioned time compensation, while the function has the ‘hold’ feature and is a little cheaper (especially on the used market since it’s been out awhile relative to the Minislew).

  • You can emulate the “hold” feature by sending a very large CV in the fall/rise time CV input.

  • @DMR

    I have both and prefer Function’s response to triggers when in envelope duty. MS makes a better LFO, in my opinion, with the additional CV control over shape, polarity, and amplitude.

  • Thanks for the tips pichenettes and striatalda. It makes sense that the hold functionality would just be an extreme case of slew limiting.

  • So Rings is a Resonator, according to a video someone posted on MW?

  • I´m wondering when the new modules (at least Rings and the “new utility module”) will officially be announced? Last year we had this great livestream of the event at Mudular Square.

  • Yeah, it becomes long now ;-) gimme tht rings, out and seq modules for christmas

  • Rings: next week (was supposed to be this week – hence the video “leak”).

    Utility module: february.

    Sequencer: end of 2016????

  • @pichenettes can we safely say the sequencer is the module in the previous video you posted?


  • Thanks for answering Olivier. Mutable is so exiting, congrats ☃

  • > @pichenettes can we safely say the sequencer is the module in the previous video you posted?

    No, you can’t safely say that.

  • yes – Thanks! But damn, I´m really curious what the utility module will do, especially because I missed the chance for beta testing.

  • @pichenettes worth a try.. ;)


  • rings sounds really interesting. can’t wait to see what it does compared to elements.

  • Rings from top to bottom:

    Two three-position mode selectors: left one is annotated with one dot, two dots, and four dots. Can’t make out the labelling on the right one clearly, but they look like waveforms or response shapes?

    Uppermost row of pots: Frequency, Structure
    2nd row of pots: Brightness, Damping, Position
    3rd row of pots: attenuvertors for the five parameters above
    1st row of sockets: CV inputs for the five parameters above
    2nd row of sockets: Strum (input), v/oct (input), In (input), Odd (output), Even (output)

    Cf Elements, it seems that the coarse and fine tune controls have been condensed into one Frequency control, the Geometry control equates to or is cognate with the Structure control, Brightness, damping and Position remain the same, and the Space (stereo stage and reverb) control has been dispensed with, but the stereo output has been retained, with alternating bands of the modal filters sent left or right.

    The most interesting part is Strum, I think. It suggests that Rings may be able to play chords, in the same way that the PLUK model on Braids can play chords if it is “strummed” in the right way (for example). Elements can probably do that too, but arranging to send three or four pitch CV and triggers in quick succession to simulate strumming can be tricky. The Strum input may make that easier (no idea how)?

  • Here is a screenshot from the video.

    2048 x 1536 - 328K
  • Here’s how it sounds (but not how it looks):

  • Some details of Poles, a module that never got off the slab, and which has now been consigned to the big bit bucket in the sky, rendered redundant by the release of the very similar Intellijel Polaris filter module:

    Screen Shot 2015-12-07 at 9.13.59 AM.png
    1242 x 1630 - 590K
  • Mmm.. MI Sequencer!! The only problem? My modest rig has so few HP’s left already.. How am I going keep them free for an entire year??
    Im gonna have to bribe Olivier with Capybara picts or LM13700’s or something just to get a proto :p

    (Psst, Olivier, you know its all your fault im even in this Eurocrack trap, don’t you??)

  • This is the problem about Euro for me. I spend literally 3 months contemplating on what module to get, save up money, is about to pull the trigger on a module, then, along comes the harvestman seq.
    Damn, fuck. Ok, I’ll wait and save the money for that one. And now the mutable seq. The decisions.

  • Which mutable seq????

  • The one that everyone is whispering rumors about, and gullible people like me believe in.

  • See my post:

    Sequencer: end of 2016???

    Not even sure it’ll be ready by that time.

  • My launch has been completely screwed up by an injury that forced me to postpone everything by one week (the scanner video was supposed to be published after the launch) and a shop in the Netherlands who wanted their 5 mins of fame and did not follow my instructions. Please don’t rub it in. The module will be revealed tomorrow, with explanations about what it does.

  • Get well soon …

  • Get well; I hope it’s not too bad.

  • @pichenettes I hope it wasn’t a serious injury, and you recover quickly.


  • Sorry, I honestly thought it had already been launched, or at least announced, at ModularSquare last weekend, but I now realise that was Warps, and it was a few weekends ago.

  • Get well soon Olivier.

  • Mon dieu, get well soon – plenty of rest, some calvados, and maybe some more rest. Then more calvados. Works for me.

  • Well, don’t trust module shops to be quiet about your new thing…

    So here it is…

  • Ahh. Cool!

  • This is really, really nice.

  • Kinda stupid for a shop to break the embargo. Don’t bite the hand that feeds you, and all. Rings looks awesome though. I can see it selling like fire.

  • Eh, no reason to send them future shipments. It’s better to cut the fat out of the middle men.

    Distribution BS aside, I love the placement of Microphonie next to Rings. They were made for each other. Rings is a lot cooler than a ring modulator, haha.

    I also wish you have a speedy recovery from your injury. :)

  • Very, very nice, and very smart use of “intelli-normalling” inside the module (i.e. it’s behaviour changes intelligently depending on what’s plugged into its inputs).

    And the demos! Well, to quote Alex in Anthony Burgess’ “A Clockwork Orange”:

    Oh bliss, bliss and heaven… Oh, it was gorgeousness and gorgeousity made flesh… And then, a bird of rarest-spun heaven metal, or like silvery wine flowing in a spaceship, gravity all nonsense now… I knew such lovely pictures.

  • With the release of Rings, there’s only one module lacking from the MI range, needed to implement the eurorack recreation of something like this (oh, and a lifetime of study of ragas…)

  • > “intelli-normalling”

    That’s one new feature of my 2015 digital modules (Warps does it too to get pristine digital constant values when nothing is patched in the LEVEL CV inputs, rather than some 8LSB noise from the shitty ADCs on the STM32F4). As you’ll see, my implementation is a bit over the top (and requires calibration), but I haven’t found a simpler way of doing it.

  • Design question why use only white knobs on this one ? is this what will happens to the next mutables modules ? or specificly to this one ?

  • I use color for:

    • Distinguishing groups of parameters (Braids: white = frequency, turquoise = timbre, fuchsia = color. Edges: white = frequency, turquoise, gain. Tides: turquoise = frequency, white = waveshape).
    • Separating voices/channels (eg: Shelves, Branches, Streams, Shades).
    • Separating sections of the module (eg: Elements). An interesting case is Clouds who got this particular color scheme from the time it was a FFT spectral processor – the fuchsia knobs were related to recording, the turquoise to playback, the white ones to mixing/post-processing. And it stuck :)

    For Rings or Warps, there was no obvious way of making use of color. There’s no obvious grouping of parameters. This “all-white” look might become the norm as I no longer use knobs for attenuverters, and tend to design modules in which there are more dimensions of control than just 2 or 3 parameters.

  • Does the Rings logo remind me of something?

    Yes, it does.

  • From Tom Whitwell’s Instagram:

    Any ideas?

  • Intriguing…

    Other interesting-looking stuff on Tom’s Github pages, too:
    as well as

    I think I’ll be picking all those up, if they ever turn up on Thonk.


  • The Whitwell / Gillet pcb marking is certainly interesting. Maybe it’s an update / tweak to the mikrophonie, since so many of the recent M.I. modules make excellent dance partners with the mikrophonie?

    I also noticed the projects on Tom’s github a while back. The pickup is definitely a module I want to build and work into the system.

  • I’m sure that is not the right answer:

  • I’m sure it isn’t, too, but it’s good fun..


  • Spotted by REVIVER, posted on MW, an integrated Shelves plus expander. Probably just a unified panel, with the maim Shelves PCB plus expander PCB behind, linked by cables? Same price as original Shelves.

    Edit: on the Shelves page on the main MI site, it says:

    Shelves’ two parametric sections are, behind the scenes, full-blown 12-dB multimode filters. The expander panel gives you access to their low-pass, band-pass, and high-pass outputs.

    New! Modules made after september 2015 combine the main module and the expander panel into a single 18-HP module.

    (OCD zone: missing cap on September)

  • You silly english speaking people with your silly caps on day and month names. ;)

  • Just for the record, there is a very informative thread on MW in which Olivier shows off some designs that never made it off the slab.

    The Outs module would have been very useful, but by far the most interesting, to me, is the 2013 Warps, which is unrelated to the current Warps, and instead is (or was) a Markovian sequencer.

    It is described as a “CV/Gate generator from a 4-state HMM“. In answer to a question “_Did you casually drop a hidden Markov model and left us dangling on what and why? :-)
    Was this supposed to provide adaptive behavior or were the transitions fixed?”, Olivier replied:

    You programmed them with the 4x4 pad. It was very Doepfer-ish in that I did not wonder why or why not it would be relevant musically. At each clock tick it did its thing… picked a new state, generated a new voltage from the mean and standard deviation associated with the active state.

    OK, so the state-to-state transition probabilities and the emission probabilities for each state I understand, and it seems that these were intended to be explicitly set with the encoder and maybe via the Set input VCs? The State outputs were gate/triggers indicating the current emitted state at each clock tick. But the actually output voltages for each state were randomly chosen from a probability distribution (presumably Gaussian) with mu and sigma as set for each state by the two columns of knobs on the right. But I don’t understand the mu, sum and sigma outputs on the right, nor the CV ins on the right. Surely it outputs a single realised CV drawn from a probability distribution? How could it be (or with what utility) outputting the mean and standard deviation, and how can they be summed (or rather, to what purpose)?

    433 x 388 - 141K
  • SUM was Hannes misunderstanding my pitch, but that was expected from such an obscure concept!

    Each state had a MU (mean) pot with a CV input offsetting it, and a SIGMA (standard deviation) pot. There was a global transposition CV input too (at the bottom row) which adds the same value to all channels’ CV inputs.

    The output jack labelled MU outputs the value of MU for the active state (+ transposition CV).
    The output jack labelled SIGMA outputs the value of SIGMA for the active state.
    The output jack (incorrectly) labelled SUM outputs a random value of mean MU and standard deviation SIGMA. Only this output is truly random, but if gaussian values are not your thing, you can still use the MU and SIGMA outputs as two-channels of deterministic CVs.

  • Ah, OK, now it makes perfect sense!

  • 2013 Warps looks kind of like a mini MPC. ;)

  • The Warps That Never Was does look intriguing. I wonder if some of the concepts will pop up in some other MI product in the future.


  • i have to say, this module, possibly a generative cv function is the essence of what drew me to mutable. bloom was always something intrigued me and so the idea of generative, global control over a system is really exciting to me.

    2870 x 1594 - 556K

Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

In this Discussion