So come up to the lab / and see what's on the slab / I see you shiver with antici


#1

…pation

From Olivier’s Facebook page, we know that he is working on at least three new modules. Exciting stuff, but it is his clues about “module 3” that make me wonder if Dr Gillet is turning into Dr Frank N. Furter:

  • Started schematics/layout for new module 3. The board is so dense there are parts sneaked in under pots, and there’s a little mezzanine board above the processor!
  • Finished assembly of module 3 prototype. Yet another JTAG mistake O_O. Wrote firmware covering all I/O. Board bring-up status: 100%. Analog section tested and validated. Will start writing code next week…
  • Wrote view/controller code for new module 3.
  • Wrote more DSP code for new module 3.
  • Recorded tons of training data for new module 3. I love leaving machine learning/system identification algorithms crunching data and figuring out all the good stuff by themselves
  • Finished building a model used in a feature of module 3. It is still stuck a bit in the uncanny valley, but I have plenty of time before the production to improve things.
  • Wrote a significant chunk of new module 3’s firmware.
  • Implemented new module 3’s firmware updater.
  • Ordered new iteration of PCB prototypes for new module 3.
  • Added objectionable secret functions to new module 3.

So, Olivier has been busy analysing data - presumably musical or audio data - using machine learning algorithms to build mathematical models of something. He previously used machine learning to build the Kohonen SOMs used in Grid, and prior to that, for feature extraction in his PhD thesis.

But it is the mention of the uncanny valley that has me intrigued. Could it be that module 3 will be the first modular, voltage-controlled version of Vocaloid Are the “objectionable secret functions” an imitation of coprolalic Tourette’s Syndrome


#2

Apparently is was a mistake that it appeared for pre-order.

a|x


#3

> Already under too much stress

Sorry to hear that. I hope things slow down soon.
I got a cancellation mail from Schneider, it was a wrong posting on their shop site.


#4

drool


#5

I hope the long awaited Mutable Instruments sequencer will be coming up…
The mentioning of the board being really dense makes me think it has a large analog section… Could that be a sequencer? Or maybe some kind of digitally controlled very complex analog vco?


#6

Either he trained it on lotsa gangsta rap, or it was a combination of Frank Zappa, Blowfly and DJ Assault…

It sounds both intriguing and fun, yet somehow a bit ominous. It will be interesting to see what develops as always.


#7

You’re looking at this from the wrong angle…

I could have very well thrown computational resources at figuring out the quirks of the Moog CP-3 mixer…


#8

@BennelongBicyclist: One glaring error with this thread: the title should end with “antici” and the first word of your post should be “pation”.

If it is some sort of vocal synthesizer, a style of analog formant filter would be a smart addition. Usually a serial bandpass, bandpass, lopass set up. That takes a lot of analogue components.
Also, if you do make a vocal synth, add a USB keyboard input, please.


#9

Man, now I want a coltage controlled version of Vocaloid… :frowning:


#10

@audiohoarder: I had Riff Raff fix it.


#11

Ahh, so you modeled old Neve/SSL/Fairchild units then? Just being hopeful. Whatever it will be, Polymoog or PS-3100 resonators, Sennheiser Vocoder or something I’m sure it will turn out great.


#12

Does this look granular to you? From this week’s instalment on Olivier’s Facebook page, in which he further discusses the new modes, particularly module 3. At first I thought it was a wave-folded sawtooth wave, but I think it is more likely to be an amplitude envelope, in which case, it seems to be pretty grainy to me.

Mungo g0, watch out! (As much as I want to promote Australian technology start-ups like Mungo Enterprises, there’s really no contest with the inventiveness, design and engineering elegance, and open-source internationalism of the Mutable Instruments approach. But gosh, this thing does look awesome, or ambitious, at least.)


#13

Here are Olivier’s latest Facebook clues on the new modules:

* Finished the DSP code for the first half of new module 2. This is a totally self-contained module with two major digital building blocks. One of them done!

  • Spent 10 days with the mrs, away from code, oscilloscopes and circuit boards.
  • Started sorting, annotating, and collecting data for the second DSP half of new module 2.

My speculation regarding this, based on Olivier’s spectrograms shown above, was:

Hmmm, data mining spectrograms, and that’s only for half of module 2… could it be that you are creating a Kohonen self-organisating map ( SOM) of spectrograms based on features of each that you are annotating, and then navigate this map, as you did in Grids? But for what purpose? Maybe to drive a granular synthesis engine in some clever way that I can’t imagine, yet?

Does anyone else (who is not operating under a formal or informal non-disclosure agreement with Olivier) have any ideas of what he is up to? I’m just curious, and I love a puzzle, and Olivier clearly loves dropping clues. Which reminds me, did anyone crack the peaks Easter Egg, with the Edgar Allen Poe allusions and the Esperanto nine?


#14

hum … , no idea. You know these are spectrograms? They seem very abruptly sliced. If it’s something melodic one might expect a bit more order and maybe a dominant frequency.

Maybe something like Frames only that you process audio? So that one can jump between different filter/eq sections, a bit like vocoder snapshots, either on a signal that can be put in, or on some internal sound source (noise, whatever). If that makes any sense, probably not, i do not code any such stuff and it’s probably better that i leave that to others …


#15

@morcego Yes, that’s why I referred to them above as…spectrograms.


#16

That spectrogram looks like the harmonic wave table view in audio term. If this is a custom wavetable/additive oscillator from MI, count me in!

However, the additive slices/grains all look like Shepard tones to me. This could get interesting if it were used as a CV source. Endless cutoff sweeps, haha. For all we know Oliver could have a S&H triggering the different modes.


#17

@audiohoarder Ah, you’re interpreting the spectrogram(s) as output from the new module. I had assumed they were slices for input to a data mining or clustering process, but you’re probably right. Maybe this new module is a granular synthesis engine in which you can navigate around an n- dimensional volume of grain types, clustered or ordered by their harmonic and envelope similarity? A bit like the timbre and colour parameters for Braids, but far more generalised. That would require only a limited amount of memory to store the grains in the map, which would be consistent with the use of a Cortex M4 processor without exertnal memory. Who knows, it might even be polyphonic - an M4 probably has enough horsepower to permit that.


#18

There’s two thumbwheels bottom right. Are they a clue?


#19

No, it’s part of the Sonic Visualizer interface.


#20

Now that is a very cool tool!