Divided pickup ideas

[Mick Bailey]

I'm planning my next guitar synth build which will be polyphonic and most likely be built into a custom guitar and output to a regular 1/4" jack. I already have a really successful mono version, so will adapt the code to use one of the multiplexed A-D converters in an ESP32 processor.

That will be six pickup outputs into six opamps and then into six processor channels.

I visualise the pickup being a regular format S/C or humbucker size. One thought was to wind six blade-type single coils and arrange them side by side (lengthways) on a humbucker size baseplate to give a decent output. Or maybe a staggered 3+3.. Another idea is to just wind six bobbins, which would be easier but result in a lower output.

I don't want to buy a Roland diveded pickup as their integrated preamp and 13 pin output isn't suitable.

Any ideas?

[Helmholtz]

Another idea is to just wind six bobbins, which would be easier but result in a lower output.

Not necessarily.
Coil area has little influence on PU output.
Output is mainly determined by turns number, core material and magnet strength.
So you would need 8k to 10k turns per coil for a "normal" output.
But as you're using opamps, why does output even matter?

Something like this:
https://lawingmusicalproducts.com/zexcoil-design

[Mick Bailey]

The Zexcoil is the type of construction I was thinking about, though not angled.

The output question is interesting; my earlier synth prototypes used an Arduino Nano and could detect the fundamental frequency of a string very accurately with pickup output as little as 10mV. However, there was a latency on the lowest few notes (standard tuning) that was resolved when I moved to a much faster dual-core RP2040 processor. This also allowed considerable expansion of features, but an unresolved aspect of the new processor is that I need a much larger signal input. A further complication is that increasing the gain of the opamp input increases the number of false triggers - something that was absent from the previous device. This is something I've been unable to resolve and maybe is something to do with the architecture of the A-D converter, as the code in both cases is the same. The only significant difference is the Arduino has an input scaled to 5v, where the RP2040 is 3.3v. I can't figure out what the difference is between a low output pickup signal that's amplified more to give the same output as a high output pickup that's amplified less.

Now, the intended ESP32 processor is different again, so it may be when the code is ported over it can be made to work like the Arduino (though the input will be 3.3v).

It may help by giving some insight into how the string frequency is detected;

1. The guitar signal is sampled for 14ms and the the positive and negative-going peak levels recorded.
2. Calculate midline of waveform between the peak values, which gives a notional 0v axis.
3. Calculate low trigger level (midpoint between midline and peak low).
4. Calculate high trigger level (midpoint between midline and peak high).

The calculated trigger points are then used to analyse the waveform; it can be seen that on the positive and negative peaks respectively there will be two high and two low trigger levels (one on the rising edge of the positive waveform, and one on the trailing edge of the positive waveform - vice-versa for the negative half of the waveform). These points are used to calculate the period, and therefore the frequency by starting and stopping a timer. This process is continuous in 14ms chunks.

The process should be independent of overall signal level, though a minimum peak level is used to signify note end. This is also useful in keeping the whole process above the noise floor.

There's a balance to be had though with the input opamp gain; if too high the signal clips on low notes and this means that accurate calculations cannot be made for the trigger points as the true peaks cannot be identified. If the gain is too low, the top octave of the fretboard can drop out, or the note length be too short because the level quickly drops below the note-end threshold. I didn't want to compress the signal, though frequency-dependent gain could be beneficial and easier to implement than six compressors.

I do have the luxury of not having to be concerned with tone, as the synth output is entirely synthesized and a regular pickup in the bridge position will also be available. While I was writing this it just occurred that I could increase the turns on a spool bobbin by making each much larger and arranging them in two offset rows, but skipping a string on each row. This would fit on a custom humbucker baseplate using Tele Alnico 5 slugs and pretty much give the required number of turns for each bobbin.

[Mick Bailey]

I've been giving this some more thought and been looking at sizing for the bobbins. It all looks to be straightforward to get around the same number of turns in 42 gauge as a regular pickup. The bobbin design though needs thinking through as to how the leads are connected. Maybe similar to a humbucker coil.

Then there's the question of baseplate material and attaching the bobbins. I'm anticipating either fibre board ends pressed onto a magnet slug, or a thin-wall machined Delrin bobbin with a pressed-in magnet. Maybe the latter as a preference.

I'd probably want to make just one bobbin and use the output to develop the firmware. That way I could easily change the pickup coil design to suit.

[Helmholtz]

I 'd expect some "crosstalk" between neighbouring strings.
Could that be a problem?

[Mick Bailey]

If the unwanted signal was higher than the note-off threshold (which can be adjusted) it would trigger a note-on event for that string. That would be a problem. The note-off threshold is easily adjusted to prevent triggering from string handling noise or EMI but ideally needs keeping as low as possible to avoid prematurely cutting a note off. It will be separately adjustable for each string, though.

If the coil diameter was kept within the confine of each string perhaps any crosstalk could be minimised. Winding a pair of bobbins would give me a better idea and allow some experimentation regarding positioning and size.

[Helmholtz]

If the coil diameter was kept within the confine of each string perhaps any crosstalk could be minimised. Winding a pair of bobbins would give me a better idea and allow some experimentation regarding positioning and size.

Crosstalk can be reduced with pole pieces close to the strings.

[bbsailor]

I'm planning my next guitar synth build which will be polyphonic and most likely be built into a custom guitar and output to a regular 1/4" jack. I already have a really successful mono version, so will adapt the code to use one of the multiplexed A-D converters in an ESP32 processor.

That will be six pickup outputs into six opamps and then into six processor channels.

I visualise the pickup being a regular format S/C or humbucker size. One thought was to wind six blade-type single coils and arrange them side by side (lengthways) on a humbucker size baseplate to give a decent output. Or maybe a staggered 3+3.. Another idea is to just wind six bobbins, which would be easier but result in a lower output.

I don't want to buy a Roland diveded pickup as their integrated preamp and 13 pin output isn't suitable.

Any ideas?

Mick,

If you want a new idea about picking up a guitar string vibration and converting it’s movement into an electrical signal, try this.

Find some small audio transformers rated 8 Ohms to 20k or higher. The turns ratio can be easily calculated. 20,000 divided by 8 is 2,500. Take the square root of 2,500 which is 50.

Now comes the creative part. The common ground is the guitar nut connection that needs to be taken to the area near the tailpiece. Since the string is near 1 ohm measured from nut to tailpiece, you need an input impedance about 10 times higher to provide efficient loading. This also means that any wire resistance between the nut common ground and the input to the 8 Ohm side of the transformer needs to the low enough to not add any input losses.

Any strings that share a metal bridge and the metal nut will short out the desired test string described next.
Remove 5 strings from the guitar, leaving only one test string. Attach the wire from the metal nut and the tail piece to the 8 Ohm side of the step up transformer. Tune the string to its normal tuning and attach the high impedance side of the transformer to the amp input. Now hand hold a magnet over the string and listen. If you are using a thin rectangular magnet align it parallel to the string and the output will increase due to more string area being magnetized.

While doing this experiment, I found out that the wire from the nut to the string transformers needs to be very thick to minimize any losses on the input. Just measure the signal from the nut common ground and the 6 transformer inputs common ground connection near the tailpiece.. The tailpiece needs to be customized allowing all six strings to being secured to the tailpiece but being individually insulated from the tailpiece metal and the other strings. The last thing I experimented with was removing the fretboard and installing a copper strip down the center of the underside of the fingerboard. Then I considered using the truss rod as a common ground return because it is very thick metal.

My final design was to use an 8 pin connector set. One pin set is a common ground for all the transformer secondaries and the other pins were the hot pins from each string transformer. Then I broke out the after 10 feet to individual strands of single wire, 1 foot long, to plug into my 6 channel mixer using common 1/4 inch plugs. Thus I could control each string level electronically rather than mechanically moving magnet height to adjust individual string level.

The guitar pickup technology has not changed much since it was invented. What I offer above is just my way to look into what the future of guitar pickup technology might evolve into and share it with MEF members.

Joseph J. Rogowski

[Mick Bailey]

Thanks. I had considered the low impedance route, based on your previous posts, as well as optical sensing using an infrared pair for each string to get away from traditional methods.

Is the noise floor lower with a low impedance setup than a regular S/C pickup? Also, what kind if signal level are you getting?

[bbsailor]

[QUOTE=Mick Bailey;n1010740]Thanks. I had considered the low impedance route, based on your previous posts, as well as optical sensing using an infrared pair for each string to get away from traditional methods.

Is the noise floor lower with a low impedance setup than a regular S/C pickup? Also, what kind if signal level are you getting?



Mick,

The typical impedance loading is about 10 times the source impedance. The typical low impedance microphone input impedance load is about 200 to 250 ohms now. By keeping the very low string source impedance very low in the 1 Ohm range, minimizing the wire resistance losses will ultimately affect the SNR.

A simple experiment is to choose one guitar string as a test string. Just alligator clip the 8 Ohm side of the transformer to the string ends and listen to the signal at the output of the transformer connected to your amplifier. Whatever voltage was induced into the string will be raised by the transformer turns ratio. That is why I included how to measure the approximate turns ratio from the impedance rating. If you install a long set of magnets under the strings from the heel of the neck to the bridge, approximately 6 inches long. you will have a higher voltage than if the magnet under the string were 1 inch long.

To get a better understanding of this technique to pick up the signal from a vibrating piece of metal, look up how a ribbon microphone works with the ribbon acting like the guitar string vibrating in a magnetic field.

This is a very good experiment to see the possible evolution of guitar pickups with individual control of each strings volume and tone.

Give it a try!

Joseph J. Rogowski






]

[Mick Bailey]

I'll dig out my test 'log' - a length oak with a simple nut, bridge, single tuner for one string and a slide-in Tele neck pickup. I've used this quite a bit for developing my synth, as it takes up a lot less space on the bench than a guitar and sits flat. Should be a good candidate for experiment and I can use my current synth build.

[Helmholtz]

https://www.cycfi.com/2014/02/neo-se...honic-pickups/
https://www.cycfi.com/projects/six-pack/

[Mick Bailey]

Very interesting, and some familiar and we'll-respected names contributing to the development. Looks to be a ready-made pickup for my application, though I would always favour making something myself. The coil arrangement is very compact.

[Pixel]

The Zexcoil is the type of construction I was thinking about, though not angled.

The output question is interesting; my earlier synth prototypes used an Arduino Nano and could detect the fundamental frequency of a string very accurately with pickup output as little as 10mV. However, there was a latency on the lowest few notes (standard tuning) that was resolved when I moved to a much faster dual-core RP2040 processor. This also allowed considerable expansion of features, but an unresolved aspect of the new processor is that I need a much larger signal input. A further complication is that increasing the gain of the opamp input increases the number of false triggers - something that was absent from the previous device. This is something I've been unable to resolve and maybe is something to do with the architecture of the A-D converter, as the code in both cases is the same. The only significant difference is the Arduino has an input scaled to 5v, where the RP2040 is 3.3v. I can't figure out what the difference is between a low output pickup signal that's amplified more to give the same output as a high output pickup that's amplified less.

Now, the intended ESP32 processor is different again, so it may be when the code is ported over it can be made to work like the Arduino (though the input will be 3.3v).

How much larger signal input is required with RP2040 vs Arduino Nano?

Is there any pattern to the false triggers? Could it be the case that they are not false but are simply things you couldn't detect before?

What has led you to the ESP32 processor instead of Arduino or Raspberry Pi?

[Mick Bailey]

The RP2040 is a 3.3v processor and the input is centred on 1.65v, the Arduino is 5v, centred on 2.5v. In both cases though the voltage range is scaled to the same bit depth and I've experimented with 9, 10 and 12 bits with the same results. The bit depth is low because I only need to resolve the peaks and increasing the resolution just encodes noise and unwanted harmonics. Because the bit depth is the deliberate limitation the origin of the false triggering is unclear. I've not been able to rationalise this at all. It could be that the signal conditioning needs a different approach with the RP2040.

With the IDE I can monitor variables in real time. Using identical code and input circuits, but different supply voltages results in quite different behaviour (note that the preamp gain is adjustable). The main observation is that with the Arduino I can roll off the guitar volume considerably and it still tracks. With the RP2040, rolling off the volume to any degree causes dropouts and mistracking. Then there's the gain/pickup output relationship already mentioned.

When I had the circuit breadboarded I did a direct swap between the two processors running identical code (just the guitar capture, not the full synth) and that's where the difference showed up. There's a design error in the RP2040 ADC, but with 8 bits the values are not affected.

The reason for moving to the ESP32 is the number or ADCs is far greater than the RP2040 and the speed vastly increased. There's also the capability to store patches. All for the same price. The reason I need more ADCs is I want to do everything that's currently analogue (VCA, VCF, envelope follower) digitally. This means an entire synth using a single processor, three dual opamps and a few discreet components. The controls will still be pots, but their value encoded via ADCs. Then add in six ADC channels for polyphonic inputs.

The only way to get the additional ADCs with the RP2040 is to use an outboard multiplexer, but then the analogue read slows down and it adds complexity. I couldn't say for sure, but I don't think it would work for audio processing.