Raspberry Pi based Synthesizer

RMW5 said:

Well, if you think you can get a 64-voice polyphonic,

Yes, 64 voices@48KHz is surely possible.

16-Part multitimbral synthesizer with 512 instrument sounds, 8 drum kits

It could even be 64-Part multitimbral, doesn't matter at all. And a 200 Mbyte Soundfont can have much more instruments (and drumkits) than the 2MB of an U-110.

, 40 built-in effects

No problem - they won't be needed all at once.

and better sounds than Roland engineers

To make better samples than those in the U-110 is absolutely no problem. That has already been done - by Roland engineers and by these ppl.

then I will be the last person to stand in your way.

Good.

Btw: The U-110 has only 31 voices and 6 Parts.

I"m an amateur electronic musician and instrument builder/experimenter going back to the days of analog 1 volt/octave modules pioneered by the likes of Bob Moog, Oberheim, Arp, etc. I haven"t tried it on the R-Pi emulator, yet, but, a well-developed, open-source, digital synthesizer library that includes pretty much every kind of module one could want and that runs on any Java virtual machine is JSyn:

http://www.softsynth.com/jsyn

There are some simple web browser example Java applets that show what can be done with the library, and some folks have even built commercial software products using JSyn. If I get time, I"ll try running it on the R-Pi emulator and, if I ever receive my R-Pi, I will certainly try running Jsyn under Fedora, Debian, and ArchLinux.

what interests me is the interface between human and machine.  sure the rpi won't produce better sound quality than a powerful x86 box and studio.  It however can be built-in to instruments easier ; -)

colincoach said:

Artales said:

I"ve got one of these: http://www.akaipro.com/lpk25

Is that the kind of thing you mean? Supposed to work with Linux without drivers.

Hmmm....between £35 and £45 in UK.....worth looking at if Linux compatible then?

I have an M-AUDIO Keystation Mini 32 - it works fine with Linux.

Seems like if you want 49/61 keys you have to pay a lot more,

Keystation seems okay where as Axiom has more features, don't know about the Akai

davidgoodenough said:

Yes, bristol is in the armel Debian sid repository.  Don't know about Fedora.

Would the Raspberry Pi be powerful enough to run the emulation properly? I would be interested if it did!

darkcity said:

what interests me is the interface between human and machine.  sure the rpi won't produce better sound quality than a powerful x86 box and studio.  It however can be built-in to instruments easier ; -)

This interests me too.  I've been wondering whether it might be fun to build one into a guitar.  I shall, therefore, be lurking...

Looks like it'll be a few months before my Pi ships, but I am downloading Russell's emulator now.  Hopefully I'll have some time to mess around with it soon.  Not sure how the performance of the emulator compares to an actual device, but it's a good starting point.

I'd also like to thank everyone for the input!

Themroc said:

rwaltman said:

colincoach said:

No great expectations of something for £25 replicating a Polymoog …

Why not? I think a modern ARM processor such as the one in the Pi is capable of emulating an eight-sounds synthesizer released 37 years ago.

The Polymoog has 71 oscillators, 71 VCFs and 71 VCAs. Thats still a bit of a challange for one little ARM (without floating point, IIRC).

Edit: http://www.vintagesynth.com/mo.....lymoog.php

The Polymoog has 71 note polyphony but only 2 oscillators (and I think 2 filters) didnt it? This is very attainable as there are plenty of vsts which are more complex than the poly moog. I am assuming if this can run debian and programs on top of that it can be used to create a complex dedicated synthesiser. I for one have absolutely no idea where to start however but I am willing to help with higher level stuff.

sacredgeometry said:

Themroc said:

The Polymoog has 71 oscillators, 71 VCFs and 71 VCAs. Thats still a bit of a challange for one little ARM (without floating point, IIRC).

Edit: http://www.vintagesynth.com/mo.....lymoog.php

The Polymoog has 71 note polyphony but only 2 oscillators (and I think 2 filters) didnt it?

Well, yes and no: It does have only 2 VCOs, but there are also top-octave-synthesizer-chips and a bunch of digital frequency dividers, so an emulator has almost as much work to do as with 142 real VCOs.

The Filters:  Once, i had to repair one of 'em and i still remember the 71 PCBs, one for each key. That quote above reflects my memory, but according to analoguediehard, there are 2 VCAs and a VCF for each voice.

This is very attainable as there are plenty of vsts which are more complex than the poly moog. I am assuming if this can run debian and programs on top of that it can be used to create a complex dedicated synthesiser.

I am certain it can. After learning that the r-pi has a vector FPU capable of performing up to 8 single-precision-ops in parallell plus a dedicated FMAC pipeline, i now think it could deliver a perfect polymoog emulation.

I for one have absolutely no idea where to start however

Get the bristol source, read the polymoog owners manual (if there are unclear bits in it, look at the service manual (only 175.66Mb)) and start coding

Interesting indeed, Thank you for the information. I would not be looking run the synth on-top of linux although it might be the most practical thing to do eventually neither would I be looking to emulate another synth directly although I would probably borrow a few bits and pieces.

I guess this is the main problem, I have no idea how to start from a low level point, if i wanted to make this the pis dedicated function. I could indeed work on a really stripped down distro of linux which would make the DSP programming side easier but I have no experience with that either so either way its a learning curve and I would rather do it the way that yields the best end result.

I will probably have to wait until there is more documentation on it and hopefully some examples, I am assuming noone has really got their hands on these yet either which doesnt make researching the easiest thing in the world. Still interesting times for hardware developers.

This looks really interesting!

Where do people see the trade off between number of voices and quality lying?

I'd be keen to get some decent physical models running but think that may cost a bit in terms of resources.

@sacredgeometry

I had exactly the same idea when the Raspberry Pi game about, cheap synth, fun project.

My idea would be to get the most stripped back version of the operating system as possible, as little over head as possible.

Then use something similair to STK to program a synth, using a MIDI to usb converter (if can find one with Linux driver), then can connect any MIDI device as a keyboard/controller.

https://ccrma.stanford.edu/software/stk/

Split the stereo output so you can have maybe a drum track on the left and a synth on the right, connect to a cheap Behringer MultiFX unit circa £30 on eBay.

Seems to be lots of outboard type stuff coming along as well, so would like to have 16 LEDS, and then build some kind of simple step sequencer, with banks etc.

Wouldn't be anything fancy, but with the FX unit attached as well, should be able to make some interesting noises.

I believe ReBirth is now open source as well, somebody could port that maybe. Seem to remember it ran just fine on my 133Mhz Pentium

http://www.rebirthmuseum.com/

I'm thinking of trying a simple midi-usb -> sequencer -> synth -> audio. I have an Akai EWI USB (midi controller with saxophone key layout). Looking at using R-pi to provide limited live practice device rather than carrying a netbook around. I'm looking for command line controller, synth with one chanel and single voice output. All would be run by script at power on. Any suggestions for software sources welcome. C/C++, Python.

all QQ aside ... its still a super awesome fun rainbow happy idea. yes you can buy some purpose built hardware to get the best possible sounds. But some of the greatest inventions including sounds have come for failures.

The 303 failed at what it was intended for and was only in production for 18 months. Just think if Roland didn’t bother and told guitarists to suck it up and get out there and practice with real people.

Wait… now I am ranting.

NM just anyone who plays around with this idea pls keep the forums posted and share the fun !

I think there is real potential here. My dream is:

- Build a raspberry pi into a tiny midi controller (e.g. korg nano key / custom build)

- Script to boot straight into a VST host (difficult) or linux alternative

- Cheap low res lcd screen which flips out for preset / soft synth selection and basic (albeit fiddly) sequencing

- Laptop audio internals stuffed in for sound on the go & stereo output

- Battery power

- XY touchpad controller for fiddling around and live performance


Obviously CPU cycles and RAM are a limitation, but even "basic" sounds and effect would be not only useful - but bloody fun. Another problem seems to be the drivers for rpi audio in linux DAWs or in Jack - and finding a compact sound card with working drivers for low latency (has to be below 20 ms to be useful).

Has anyone successfully managed to get any DAW running on this board?

I've had some success with fluidsynth which uses old Soundblaster soundfonts and an Alesis Q49 keyboard

you need to run Fluidsynth using alsa only, as jack can't run on the existing sound device, and have a soundfont that's probably not more than 40MB but it's pretty straightforward. once you've apt-get installed fluidsynth you need to run

Code: Select all

fluidsynth -a alsa -m alsa_seq soundfontfile.sf2

then connect the midi controler to the synth with something like

Code: Select all

aconnect 20:0 128:0

you can use aconnect -i -l and aconnect -o -l to find out the numbers you may need to change

on my rpi the Q49 appears as device 20:0 and fluidsynth as 128:0

Bristol dosn't work in alsa mode unfortunately, I think that's because the same reason jackd dosn't, it can't open plughw:0,0 which I think is to do with ,missing mmap in the alsa driver.

video of above

sound is very quiet

https://www.youtube.com/watch?v=_U3OL_s1dLA

To add to the excitement I installed and ran amsynth (in the repos). Good news! it works, Bad News! it's incredibly laggy. Also it will max out the CPU quickly and with hilarious consequences. You might get away with just using pad sounds if you keep the polyphony down to about 4 or 5. This is on a stock rpi. I'm going to try overclocking it at some point (but not overvolting). I'll see about uploading a video in a few days.

amsymth test on youtube

https://www.youtube.com/watch?v=lRILm5n ... ture=g-upl

My polysynth on Raspberry Pi came alive last night - http://raspberrypisynthesizer.blogspot.co.uk/

My polysynth on Raspberry Pi came alive last night - http://raspberrypisynthesizer.blogspot.co.uk/

Looks brilliant, looking forwards to having a go!

It's playable by MIDI keyboard now - only old-fashioned 5-pin MIDI, not USB, USB was for some reaosn killing both performance and audio quality. But there is really grim latency in the audio path somewhere between me handing packets of rendered audio over the OpenMax and the sound coming out of the speaker. Once that's solved (and I can't solve it, it's deep Pi firmware stuff) then this will be a truly great synth.

http://raspberrypisynthesizer.blogspot. ... is-go.html

Hmmm - it's deep firmware stuff - well, it *was* deep firmware stuff, but when the idiot user doesn't drive the deep firmware stuff as directed thing will go wrong.

Panic over, latency gone, it sounds *astonishing*. Astonishing at any price, but for £30 it's insane.

More videos and audio to follow.