posted 11 August 2013 18:45            

What I am trying to do is to create a sort of mini-replica of the sort of "pin" matrix used on the EMS VCS3. Instead of pins, I want to use toggles. I have the basic idea down, or so I thought. I have thus far created two "oscillator sections", each with two different waveforms, ala the VCS3. and like that instrument, I want everything to be able to modulate everything. The output of each oscillator should be able to modulate the frequency of each oscillator, filter and like that whenever the appropriate toggle is ticked off. The amount of modulation will be dependent on the output level of each section. I have Oscillator 2 of my synth modulating Oscillator 1 as it should, but I'm damned if I can figure out how to make Osc 1 modulate the frequency of Osc 2, or how I can make each oscillator modulate themselves. I know I must be misunderstanding how, for instance, the Constant can be used here. I simply referenced the mod matrix in the Multiwave Graincloud Sound in the factory Sound bank.

What am I missing here? I've uploaded the sound here (It's the one called "VCS3WIP" in the zipped "Synthi" Sound):
http://www.theelectronicgarden.com/Scot/Synthi.zip

posted 13 August 2013 15:49            

Yes you've sussed the fact that feedback is going to come into it. What seems like a simple thing to do with a soldering iron and patch pannel on the VCS3 becomes quite involved in DSP as you have to account for every eventuality. In a pre wired synth (non patchable) the order of modules tend to be close to fixed and the chances are that when you patch your synth most of the time the order of modules will be similar. If you can add a few restrictions that will not detract from the flexibility it could help a lot.

Other things to consider before you start. What signals are control signals that could be at capy talk rate (1ms) and what needs to be at sample rate. i.e. gates and triggers and may be LFOs and ADRSs (smoothed at sample rate if needed) could be at capy talk rate. capy talk signals can feed sound to global controllers which can be picked up any where before or after the transmitting module. Sample rate signals will need a feedback pair each, to be sure you can send the signals where every you choose.

An output can feed many inputs but an input can only ever be feed from one output. This is the clue on where to start.

Each input will need one multiway switch. This could be made up of say 3 (4 mix) modules into a mix module if you had 12 outputs to choose from. There are newer kyma modules that work as 8 way selectors that you may want to look into that could help. There may even be a matrix switch that will help no end. Each of these input switchers will feed either a sound to global controler or a feed back module depending on the type of input signal required. These are all put in a mixer with the output turned down to act purely as a place holder for all these modules. This hangs off a mixer just before the output as they have to live somewhere. Only one input selector will go directly to the output so you can hear the sound. All your processing modules will be to the left of these selectors and their outputs will feed all of the selectors (one for each leg of each input). This is a lot of knitting here so try to be systematic.

The other end of your feedback pairs will go to the left of and feed the one processing module that needs its signal, but the capy talk type inputs will be written into the fields (as hot paramittors) of the processing modules that need them. Any of these hot parramitors that need to be smoothed (to feed a sample rate input) could be pasted into a constant which feeds a single sample delay with feedback (acting as an SR smoother) which then feeds the processor sound input.

It's a big job you are taking on here and the more planning you do before hand the easier it will be, and I wish you luck.

I hope this helps.

posted 14 August 2013 16:25            

As you may tell I haven't downloaded your patch (rushed for time at mo). just a few points if it helps.

Trapezoidal envelope would be an ADSR but with the sustain always set to 1 and tick linear (not exp). When the sustain is 1 the decay has no relevance so using A and R and Scale gives you the shapes you need.

Another important thing with analogue synths is that they normally used a 1V per octave CV. This means pitch spacing not frequency spacing. nn is pitch spacing (hz is not) so make sure you convert your control signals into nn (or nn/128 so that it doesn't clip and *128 to put it back in range in the frequency fields). If you don't do this you will find that things like LFOs will work on LF (bassy) stuff but will need to be turned up loads for HF stuff. Portamento won't be linear, switching octaves will screw up mod levels. Envelopes won't have a consistent effect on VCFs. VCFs won't track Keyboards. Nothing will be stable and every time you turn one knob you will have to turn another three to compesate.

Hope this makes sense

posted 14 August 2013 16:45            

The synth I linked isn't the one I am now working on. I abandoned that one and started over with a different approach. My head is reeling with all of the things I would need to do to make a true emulation... But I'm not too concerned that it be accurate. It's just a learning exercise. I will attempt to put into play your many good tips once I have my basic structure finished... Only one row and one column left for the pin matrix now.

On the trapezoid... Yes, I knew that would work, but there are some slightly more difficult arreas, like the unusual behavior of the ON knob (holds full volume even after the note is released for a time determined by the setting) and the OFF knob (the length of time between the ATTACK begins again when in "looped" play... The DECAY always finishes its cycle before that silence... That bit is stymying me, too...

posted 15 August 2013 09:16            

Synthi.zip

Nevertheless, this sounds cool already and works pretty well as is. It isn't exactly like the real thing... I added a delay and omitted the joystick patch points in the matrix because it's just as easy to add those in the Virtual Control Surface as needed. Still, this version has pretty much all of the ridiculous routing potential of the original... everything can modulate everything else. I decided against free-running oscillators, as they were giving me too much trouble (and, frankly, I was so upset about the accident with my own VCS3 that I couldn't work up the energy).

Also, I never did get the envelope set up the way I wanted it, so instead used a selectable sound to switch between the plain ol' ADSR and a looping multisegment envelope. This provides many of the same sorts of sounds that I make with the real thing's envelopes. A compromise to be sure, and hopefully I can come up with something more authentic.