Want to build an organ/polysynth - what hardware?

[Taylor]

I'd like to build a polyphonic synth. I would be ok with just on/off, no envelopes, filters, etc., basically an organ. I'd like sine waves as a minimum, other waveforms would be nice but not required. Don't want to go the top-octave generator route as these are hard to find and expensive.

I have no qualms about going digital, so if such a project exists already, I'd love if someone could point me to it.

I've considered doing individual sine oscillators for each key, but this is obviously pretty intense, with a dual opamp plus some passives for each note.

Can anyone point me in a good direction?

[slabman]

http://www.electro-music.com/forum/topic-30139-0.html&postdays=0&postorder=asc&highlight=

[Taylor]

Er, thanks but:

[slabman]

sorry - meant to say more but was half asleep when I posted. Instead of a TOG, just use 12 tuneable CMS oscillators and divide down. The rest of the info on that thread is applicable. To add richness, you could build two duplicate oscillator/divider chains and slightly detune. Even if you don't want attack/decay, routing audio through the keyboard contacts will make for clicks and noise, so you need keying circuits, in which case you may as well consider envelope shaping. Transistor Electronic Organs by Alan Douglas is a useful sourcebook.

[slabman]

That should be CMOS oscillators, of course!

[droffset]

Hi, how are you planning to do the keyboard, as a key matrix or a resistor chain, or another way?

If you choose key matrix and discover good info about how to make a scanner please share.

[Taylor]

Well, since the whole tone-generation aspect is so open-ended at this point, I'm not sure about keying. At one extreme, I had considered individual opamp-based sine oscillators for each note, which would be turned on by SPST switch. From there, I could go to a transistor per key controlled again by SPST switches. I'm not too bothered by a click at note onset so this isn't much of a problem for me, however parts count would of course be monstrous.

I like that idea because it would be neat to experiment with strange tunings. I'm a big fan of Harry Partch and other microtonal composers, so having a microtonal organ would be great.

At the other extreme, if there was a digital synth based on FPGA or similar, I'd probably be forced to just use a MIDI keyboard with some MIDI decoder.

I had considered doing the plan with individual opamps per note, doing a SMT PCB design and having it fabbed up for me so I don't have to do all that soldering. Would be nice if I could get a group buy together for this, but something tells me it's a little too esoteric to get much interest.

[Rykhaard]

My tests that I had done last Summer, for building the oscillators around CD40106 (hex schmidt inverters), were successful I figured. The tuning of the polyfilm capacitor test oscillators, held their frequencies steady, for just under 48 hours. (The temperatures in my shop / studio, vary greatly in Summer when the voltages were sampled, over time.)

That was in my own idea, for building my own 'string machine'.

A CD40106 could give you up to 6 oscillators per chip. 3 dual voices? (Though - my tests were only on recording the voltages from 3 oscillators, from a single chip, over the 48 hour period.)

[cloudscapes]

I've been using an AVR interfaced to a PT8211 16-bit DAC interfaced to a 2206 function generator to get several octaves in sine, triangle and square. I use a word table (DAC values).

[Taylor]

[cloudscapes]

[Taylor]

Interesting, that's a bit over my head, but I'll be very interested to see what you come up with. Are you going to offer schems/code so others can build it?

[cloudscapes]

[Taylor]

Really cool, I looked at your thread as well as your blog and I really like the concept. If you want to go in on getting a couple of PCBs fabbed up when you're done, I'd be in.

How will your keying work? I'd like to build my own keyboard using the Wicki-Hayden isomorphic key mapping, so I'm curious how the keys will need to be set up.

[JovianPyx]

[Taylor]

Thanks, great info. I think I have looked at your site before, but I will delve a little deeper into it. I do some digital effects design, but am clueless about FPGAs, so the real-world implementation still is a bit mysterious to me.

Let's say I burn your code to an FPGA, then what do I do with it? Are there PCBs available that I can put this chip into and have a functioning hardware unit? This would be MIDI-controlled, right?

[JovianPyx]

What I've done is taken commercially available FPGA development boards and with small hardware additions make a synthesizer. Not just one synthesizer really, but several. These boards range in price from $50 on up. I have several polysynths that run on a $50 board.

The hardware additions are a MIDI electrical interface involving a 6N138 opto, 2 resistors and a diode. (so yes, these are all MIDI synths I've designed). For boards that do not have a DAC, I've included a DAC (which I bought from a friend who made up the simple boards).

An FPGA is built on static RAM. If power is removed from the FPGA, it does not remember what was in it. FPGA development boards usually have a Flash RAM that programs it at power up. There is usually a way to reprogram the Flash, so one can export a design into the board such that when it powers up - your design instead of the demo is running. But because an FPGA is RAM based, it can be reprogrammed an infinite number of times (unlike Flash based devices like a PIC). This is both a blessing and a curse. Blessing: you can rewrite the RAM as often as you like. Curse: it takes a second or two for the device to be reprogrammed at power on.

One of my designs could be used to program the Flash on the dev board and then the board would always power up configured as that synth. As long as the required hardware is attached, it will be a synthesizer. Depending on the synth design, it may be required to send patch data to it via MIDI sysex before it will make any sound. The procedure to program a given board's Flash is not standard, so specific instructions for a particular board must be performed.

At this point, I've not done my own PCB with an FPGA on it. That may come in the future, but it's not a priority for me.

There's definately a learning curve involved, but as far as getting a well featured polysynth of your own design working with the least expense, I can't think of a better deal than an FPGA.