Divided pickup ideas

[Pixel]

When I was thinking about your monophonic project I always thought it was unfortunate that the signals were not available on a per string basis as that would have made processing simpler. At that time I thought that using one SMD inductor as a pickup per string just to provide "hints" as to what string was struck could help with processing. I also thought that multiple processors could be used on the problem because in some instances if I recall correctly the code blocks. I would be careful to understand where the ESP32 might block, especially if it has so many ADCs.

You could get additional ADCs on RP2040 by using more of them. If you use one RP2040 per string aren't you dealing with a problem you have solved before?

I haven't tried it yet, but I wanted to do some work with state machines to see if they would apply to your project.

[Mick Bailey]

My last (current) version has no code blocking issues and runs really well. I'm not sure how processing would be simpler on a per-string basis - it multiplies everything by six and adds in the need for additional output mixing as well as new considerations as to how the hold function could work. I would however maybe strip out the harmonizer functionality, but retain the multiple voices per note (possibly thickening the sound by detuning the oscillators slightly) and expand the modulation. This is particularly where a single processor has an advantage where there are features that affect all strings/oscillators simultaneously - ADSR, VCF, VGA. Presently these are analogue but I want to shift these to digital. Also expand the LFO and modulation side. If split across multiple processors these become more difficult to control.

I also want to get to the stage where I have a single encoder that encompasses most of the functionality, along with an OLED screen. Maybe a few live performance controls and patch selection. The RP2040 cannot save patches without adding external memory.

Looking at the RP2040, this has a single ADC multiplexed to 4x channels. In most, one channel is assigned to a temperature sensor and isn't available, but the Seeed has all 4 available. Three Seeeds would be needed to build a polyphonic synth on the basis of my existing code - 12 ADC channels, so 6x assigned to string inputs, 3x assigned to potentiometers and 3x left over which could be used to expand the modulation or shift the envelope function into the processor. One processor would act as the supervisory control and pass the state of the buttons/pots to the others.

It would work fine, but limit greater expansion and would give me a polyphonic version of what I already have. I wanted to simplify the design down to a single processor and all the outboard stuff incorporated. I'm also looking at how the inputs could go direct into the ADC without any preamp conditioning to eliminate the need for any input opamps.

[Pixel]

Anytime you can limit the problem domain you get to eliminate some of the options you need to consider. If you were processing on a per string basis then there are fewer valid choices of notes. For example there were some issues in prior builds which only emerged when processing low notes.

In terms of reuse, code or hardware, we typically try to reduce the problem to one we have solved before.

As usual you seem to have things well in hand. It is useful to see where you are trying to end up. Do you think putting a LOT of turns on the windings for the pickups would help? Would that introduce other problems?

Anecdotally problems often occur at or near the endpoints. When you have false trigger problems is that the case? I wouldn't be surprised if the ADCs have problems at the ends of their range.

[Pixel]

Sorry I wasn't clear when I asked about ESP32 and blocking. I wasn't as much concerned with code blocking, but rather the processor itself having problems due to the large number of ADCs. Those ADCs could be sharing something inside the processor that could cause you issues.

[Mick Bailey]

ESP32 has just two ADCs, multiplexed to 18 channels. The fewer channels used, the higher the available sample rate per channel. It's not clear whether channels requiring a lower sample rate (those used for pots) can be de-prioritised to increase the sample rate for the string channels. You're right in the ADCs being a possible problem. The spec is 2M samples/second at 12 bits.

I currently run the RP2040 at 1M/s but I can easily drop this down to find out what the lowest practical rate will still give accurate pitch detection.

Back to pickups.......

It looks like the way forward for this is low-impedance using off the shelf inductors. One thing that's on my mind is wanting a complete synth that anyone who can put together a stompbox can build using readily available parts. I'm thinking, start with the pickup and make the rest to suit. I also want to be able to fit this easily, so standard pickup size. Maybe a humbucker format which will take any necessary support electronics. I assembled x6 miniature transformers and they take up a fair bit of space along with the rest of the synth and may be too bulky.

[The Dude]

Just throwing this out there. Why not use piezo saddle pickups for this? I would think you could more easily get string to string isolation and separate signals.

[Mick Bailey]

That should work. I built a replacement synth interface for a Godin a while back and it used individual piezo transducers and the signal output from the individual onboard preamps was excellent. The downside maybe is that if someone wanted to build the synth using a Tune-o-matic equipped guitar it would be more difficult to find one with individual outputs.

Another thought was to use Hall-effect devices and place the magnets underneath each one.

[bbsailor]

[QUOTE=Mick Bailey;n1011027]ESP32 has just two ADCs, multiplexed to 18 channels. The fewer channels used, the higher the available sample rate per channel. It's not clear whether channels requiring a lower sample rate (those used for pots) can be de-prioritised to increase the sample rate for the string channels. You're right in the ADCs being a possible problem. The spec is 2M samples/second at 12 bits.

I currently run the RP2040 at 1M/s but I can easily drop this down to find out what the lowest practical rate will still give accurate pitch detection.

Back to pickups.......

It looks like the way forward for this is low-impedance using off the shelf inductors. One thing that's on my mind is wanting a complete synth that anyone who can put together a stompbox can build using readily available parts. I'm thinking, start with the pickup and make the rest to suit. I also want to be able to fit this easily, so standard pickup size. Maybe a humbucker format which will take any necessary support electronics. I assembled x6 miniature transformers and they take up a fair bit of space along with the rest of the synth and may be too bulky.

Could you please post the technical details of the x6 miniature transformer you used.

Thanks

Joseph J. Rogowski

[Mick Bailey]

The transformers are just to get some overall idea of sizing and layout. I have quite a few Eagle (Xicon) miniature transformers to hand, though the impedance is not correct for this application. They're the TL/TM series. I haven't done anything else on the pickup side as I'm still working out how to implement the signal processing. It looks like a single ESP32 may not be fast enough to perform analogue reads across six inputs.