I’ve been considering what my first Eurorack module would be if I ever decided to go ahead and attempt to mass produce one, and I have decided on the kind of module and its functionality. Oddly, it ended up being a digital module.
It is just in the drawing-board stage as of now, but it is not a module anyone else offers - to my knowledge.
It is a sample-based noise source with 64 different noise samples of varying type and length. I call it Noise 64 for now.
Each noise source can have a different amplitude and pitch. Amplitude would vary from
infinity Db to +6Db. Pitch would vary from extremely slow for use as an LFO - to C10. This gives it the widest range of possible timbres and uses. More than any other noise source has ever allowed.
The kinds of noise on board range from very smooth and natural noise loops to harsh digital signals.
The big issue with this project would be the amount of sample memory. If I used the OGG Vorbis compression library, I could fit all 64 samples (if less than 40 seconds) into only 12.5 Mb. If I used WAV files, it would take up to 900Mb. That is a crazy increase in size for not much of an increase in quality. I am re-sampling high quality WAV files into OGG for the best result.
I had considered developing some kind of a sample library for use with Kontakt or UVI, but they both have issues I don’t care for. It would also seem to be cheaper to make a piece of hardware than to buy a license from them. I looked into making an Ableton rack as well, but it also isn’t as flexible or open as I would like it to be. It would also seem to make sense to make this noise source hardware as several new synthesizers on the market either have no noise source or an uninspiring one.
I figured this would be the best place to ask for advice as everyone here has a real passion and great knowledge for hardware synthesizers.
Any advice is welcome even if you think the idea is bad/horrible. I don’t want to pour money into a DOA module.
Your idea of “pitching” noise samples is weird.
Idea 1: something very basic with a SD card. The sound-clips can be resampled offline, in advance, to various octaves (it wouldn’t take more than 8 or 16 GB in total) to deal with the faster data transfer rate at the highest notes.
Idea 2: I’d just look at each original sound and model it as a proper stochastic process; and from there it’s just code, no need for storage.
I’m just wondering which microcontroller has a OGG Vorbis decoder built in?
@pichenettes: I knew that you would suggest coding the response. I had considered that, but wouldn’t 64 simultaneously running simulations overflow/crash most microcontrollers? If I am wrong, that would be awesome. I figured that each step in the program the next value of the 64 samples would be read then played back after being attenuated. For all I know, that too would overflow. The idea was to have 64 unique noise sources in one compact device, and they would all be mixed internally.
@V`cent: That is another issue I have no idea about. OGG sounds much better than mp3, and almost sounds like WAV in most frequencies.
About the pitching, this is just a feature that I am used to from using sampling synthesizers a lot. It is what happens to the sample when the sample read head is slowed down or sped up. It basically adds some destructive aliasing in the higher frequencies and when lowered causes the sample to sound entirely different.
Oh I didn’t understand you’d want the 64 samples to run at the same time.
Are you sure people would need that?
The cost of reading 64 ogg files is greather than the cost of running 64 “coloration” filters on random sources.
Being a senior pro at ChucK after the first week of the Coursera Course, i just tried: while running 64 Oscillators at once is next to no Problem, decoding 16 MP3 Streams simultaneously is quite challenging. On a 3.1 GHZ I5 Mac. As pichenettes said, id suggest you look into coding your Noise, you then could also have more realtime control over your Noise (“Does this modulate at audio rates?”) than you ever could have on a sample and might discover that you can go with a flimsy ATMega instead of a High Powered Micro. I guess the market for people paying more than 100€ for a Noise Generator isnt quite large…
I’m crazy. If I’m going to do something, I will make damn sure that it isn’t something anyone else would do.
I am certain that people don’t “need” it, but they have also never had this option. It would be very easy to scale this down to say Noise 8/16 and make a series of specialized modules. They should only be 4HP each. I could make it less, but I am not sure pots go that small. Do they?
A coloration filter should be able to achieve several great tones, but I am not so certain that it could pull off the sample I call “brain freeze”. It is great to know that coloration costs less than I thought.
@fcd72: Sounds like I should do that course then. Are they still enrolling?
Also, did you use noise oscillators or what kind of oscillator?
But seriously, all 64 running at the same time and summed together?
I simply copy pasted a series of SinOsc as i presume the just use a Lookup table which is quite equivalent to read and play a sample from memory.
I could easily do the course, only thing to do is asking “Okay” after every sentence
could be a nice clean result with all the stuff canceling each other out
haven’t done last weeks homework yet but just acquired the ebook. can you help me with my homeworks?
Nope. As i was some time away im not at the point where i can do my homeworks…
@pichenettes: In the initial idea, yes. All 64 at the same time. Now that I know oscillators cost less, I will definitely look into those methods. Remember, I will take it down to 8/16 instead of 64 if that would work too.
@fcd72: Hmmm… if I could rip certain cycles from my samples and encode them as a “look up table” I might be able to do that. It would take 3 of these oscillators, one being a noise source, to recreate some of the more complex ones. Oh and some kind of a randomizing clock. The dwon side to this is that the HP would be considerably larger.
@rosch: That was a huge consideration. Especially when slowing down the noise samples to LFO speeds. You’d have an LFO that wouldn’t repeat for well over a YEAR with proper cancellation of two similar samples. Great for natural textures and drone music.
An overall question: What is the smallest Eurorack compatible pot? That is something I need to consider.
4 HP for the Pot. But then you will have to put the PCB rectangular to the Panel and the Hight End people willing to spend hundreds of € will start complying that it wont fit into their shallow skiff etc…
I already have an LFO that wont repeat for Months or even Years, its called Maths. BTW im selling mine (2k13 revision)
BTW, a Sample is nothing else as a laaaarge Lookup Table. For nice Noise id suggest some esoteric hybrid solution, grab one of these russian minitubes and some vintage Transistors, sample the noise that comes from them, run this thru some obscure modeling algorithms and dont forge to have some weird named Parameters under CV control.
I am well aware of the skiff issue. Is it not possible to use smaller turn pots? I could include a jeweler’s screwdriver with each module, haha. The Maths is a great module, but I am not so certain it can pull of textures like I am able to create so easily.
.the 9mm alpha you can get from smallbear. i mean the rectangular ones that are mounted like the pots in the shruthi. they have threadings and metal shaft. so you can have the board parallel to the panel.
@rosch you’ve got 46 min! …
tip! …use the Alarm clock code from the video with Ge Wang. And then just chuck a bunch of SinOscs in there running in for loops.
really? after that i’m out?
like in school, like in school
The reason im selling mine that its actually not such a great module and even less if you had the chance to testdrive Mutable Instruments Tides. Maths is 75% overhyped and 25% an imperfect double AR generator with an 4 Ch attenuverter Mixer. Not that its not usefull and versatile, i just get annoyed that it changes the output level/swing if you tweak any pot, so modulating Rise/Fall with CVs gets rather pointless and unpredictable.
Have you seen the Ataraxic Translatron?
@rosch you’ll just get 0 points…