[sdiy] Patchable polyphonic synth with FM or AM transmission idea

[Gordonjcp]

On Thu, Jan 03, 2019 at 09:24:11AM +0100, cheater00 cheater00 wrote:
> 
> You don't need racks of equipment to make a single modular voice of a
> poly synth sound good over FM transmission. It's a single synth voice,
> not a piece of classical music performed by a philharmonic orchestra.
> 50 dB headroom is more than plenty. If you don't count "off", many
> synths don't have more than 45 dB dynamic range, and they're used to
> make music by people who win the grammy, so I'm not worried.
> 

Okay, but you're not passing the audio over the patch cable, you're
passing DC control signals.

> One of the reasons to explore this idea is that I've explored all
> others and this actually seems like it would be possible to build. But
> there are also many pro's:
> - compared to separate 1/4" leads, or multi-conductor cables: most of
> the complexity can be taken care of by a pick and place fabrication
> service, rather than hand-soldering front panel jacks, especially
> esoteric jacks; cables can be inexpensive;
> 
> - compared to digital: FM transmission has no delay (not even 1ms),
> ability to perform feedback, repeated A/D-D/A conversion with
> affordable ADC and DAC chips will result in a lot of aliasing
> artifacts as well as in cross talk (due to using codec chips)

The delay in conversion to and from digital would be in the order of
microseconds.

If you need to shave off one millisecond from your latency, sit 30cm
closer to the speaker.

> - compared to crossbar switches: crossbar switches just cannot handle
> a flexible modular synthesizer with a lot of modules, have poor
> crosstalk and headroom, and need a lot of massaging to work in the
> first place, plus they're pretty expensive and difficult to source.
> but if anyone has notes to the opposite i'd love to hear them. bear in
> mind the design goal here is 50 inputs and 50 outputs, times 16
> voices.

Every analogue polysynth you've ever played uses analogue switches to
route the control voltages.

> - compared to CAN, LAN, WiFi: all of those are either too expensive or
> won't be able to handle runs (distances) within a single moderate size
> modular or won't be able to handle all the channels due to bandwidth
> limitations at which point you need involved topology. Delays are also
> an issue

CAN bus runs over something the size of a car, or lorry, or bus.

How big a modular are you planning?  I'd love to see a modular bigger
than a city bus.

> Handling near-DC component transmission is an interesting question.

All your control signals in a practical synth are slowly-varying DC.

> A few ideas on implementation that i would love to hear (constructive)
> opinions on:
> 
> - patch cables could be normal BNC cables. this is inexpensive. 1/4"
> cables might not be good enough.

Huge, expensive and fragile, but doable

> - outputs of each transmitter should have balancing resistors of ~ 1K
> (series to output) and terminating resistors (~50 Ohms across TX pair,
> before 1K balancing resistor). This among other means the ability to
> tie together multiple outputs, and distribute them to multiple inputs,
> just by using T junctions.

This will give you intermod problems.

> - you need one FM carrier oscillator per voice. Those should be
> central and distributed to every synth module. This means 16 good
> sinewave oscillators, they can be synthesized with DSS.
> - each module would need some sort of connector for each of the 16
> carrier oscillators, possibly BNC as well, but could also be an edge
> connector.

You'd actually need an FM carrier per module per voice, if you're going
to do frequency channels.  You'll need a different frequency for every
single control signal, and they will all need to be chosen so they don't
interfere with each other.

> - given you don't need local oscillators, the FM transmitter becomes
> in essence just a few transistors and an op amp, which might be
> possible to drop. It could be easier to use a pick and place service
> rather than source a more expensive FM IC that contains a running
> oscillator just the presence of which will degrade quality. A
> transmitter IC could be $2, op amp and transistors could be quite a
> bit less, which adds up quickly.

You will need local oscillators for the transmitter to work correctly.
A transmitter consisting of "a few transistors and an opamp" will not be
clean and stable enough for this to work.

> - using your own oscillators also means you can lower the carrier
> frequency to outside the commercial band, making it easier to create
> non-radiating patches, i.e. less cross talk between modules.
> - the receiver can be a bit more involved to create, but still
> attainable, especially with modern, stable passives.
> - you might want the modules in separate metal enclosures, just due to
> the weight of 16 PCBs.

If you want to avoid leakage between modules, you'd actually be better
going up into the many-GHz region where screening will be more
effective.  This will be (more) expensive.

There are a lot of considerations which make this less than entirely
practical.

I'm not entirely sure why it would be useful, since it won't be able to
interoperate with any "normal" modules.  Why would this approach be
better than something like the Oberheim Matrix 1000, where you just
generate all the modulators digitally including the mod matrix, and
emit control voltages to the appropriate analogue circuitry?

-- 
Gordonjcp