I've some idea, but I'd really appreciate some guidance. I'd like to amplify the weak signal from a guitar pickup to make the best of the A/D resolution of a Teensy4.0 which is powered by 3.3V. Ideally, a low noise single supply op amp which is rail to rail, good for audio, can operate from a single 3.3V supply, but is not too expensive.

Since I probably will use a PP3 9V battery, or 5V supply, I'm thinking, as it will be AC coupled (both in and out), I could probably use a less expensive audio op amp as long as I protect the input to the teensy with schottky diodes. To adjust the gain, I'd use a potentiometer, but by setting high and low thresholds for the A/D I'm thinking of using a digital potentiometer to adjust the gain (with a suitable time lag to make adjustments). Will a digital pot add noticable noise?

Whilst here, what should I do about anti alias filtering?

 

Not a bad idea. The tendency today is to go all digital and eliminate the op amp.
Put a 24-bit ADC and the MCU in the guitar. You don't need an op amp nor the digital pot. Use +/- buttons and control the amplitude in software.

 

what should I do about anti alias filtering?

What is the A/D sampling rate?

ak

 

IMHO: Not a good idea.
Most guitar players like a pure analog sound in/out of the guitar (unaltered) that can be externally altered/distorted if desired.

 

Electric guitar magnetic pickups are often followed by a bit of circuitry on the guitar. BUT there is not much room for the batteries on most guitars.
You are not likely to experience higher frequencies from a magnetic pickup, so that will not be an issue.
There are many published circuits in the "SCHEMATICS FORFREE" website. And the download is free!

 

What is the A/D sampling rate?

ak

For 12bit, it can be up to 1.3M samples a second, it’s a blisteringly fast processor, but I will be playing tricks with the signal to create distortion, chorus, echo, reverb effects so maybe I’ll go for 96 kHz but faster should be possible

 

IMHO: Not a good idea.
Most guitar players like a pure analog sound in/out of the guitar (unaltered) that can be externally altered/distorted if desired.

I probably didn’t explain very well, this will be a Pedal which will be a box external to the guitar. Typically analog pedals are strung together to combine effects. The teensy should be fast enough to combine effects in one box where the sequence of effects can be programmed. I’m aware that guitarists are fussy about the difference, for example, between a distortion pedal which uses germanium diodes rather than silicon in the feedback loop, but a fast processor with sufficient resolution should be able to mimic either

 

I would consider something like an OPA310 or OPA314 for this application, and I would set the input impedance to probably 1MOhm (pretty standard for this job). Non-inverting configuration is a must.
I doubt you'll need much more than unity gain going into a 3.3V-powered A/D. Guitar pickups (esp. humbuckers) can produce peaks as high as 2Vpp when played hard. I saw this with my own eyes on a scope.

 

The normal output of a guitar is about 10mV, so you will need some sort of a preamp.
It needs to be high impedance (at least 220k), so a JFET op-amp will be ideal.
There's not much above 8kHz on a guitar, so sampling at 22.05kHz or 24kHz would be plenty good enough.
The bandwidth is constrained by the pickpup inductance (about 6H) and the cable capacitance (200pF or so), which gives a 12dB/octave roll-off starting at about 4.5kHz. That can be your anti-aliasing filter!

 

I really don't know where 10mV came from. Maybe as an average when playing softly with a low output pickup. If you lean a bit into it and you have a fairly hot humbucker, the peaks are way larger (huge crest factor), and one must accommodate them without clipping (unless you just want to make a distortion pedal). I just took a scope video, guitar straight into scope, set at 1V/div, to illustrate this (you'll have to unzip the file since I can't attach an mp4):

 

I really don't know where 10mV came from. Maybe as an average when playing softly with a low output pickup. If you lean a bit into it and you have a fairly hot humbucker, the peaks are way larger (huge crest factor), and one must accommodate them without clipping (unless you just want distortion). I just took a scope video, guitar straight into scope, set at 1V/div, to illustrate this (you'll have to unzip the file since I can't attach an mp4):

10mV from my measurements.
It will vary from pickup to pickup. Some method of adjustment will be required, otherwise the small signals will disappear into the noise floor, especially as the TS was thinking of a 12-bit A/D. I'd go with @MrChips 's suggestion of a 24-bit audio DAC.
You would probably want an op-amp that can do better than the OPA310/314 for 1/f noise. Down at 80Hz they are both getting rather noisy.

 

1/f noise is not normally a concern with guitar signals. The ear is very insensitive to noise at low frequencies. I make guitar preamps with CMOS opamps all the time (often OPAx196), and if anything, people report that they're dead quiet.

 

10mV from my measurements.

How do you properly measure that? I also found around 2V to 3V p-p with a multimeter or scope (can't remember which) on an otherwise unloaded guitar output, striking a full chord (6 strings), output from the bridge pick-up. This is for passive humbuckers in 3 different guitars all outputting about the same. Presumably at the neck the output will be higher but I didn't measure that.

I'm considering an OPA1678 (to OP: it requires 4.5V minimum). That looks like a nice opamp except 1/f noise is not great. Can 1/f noise be rolled off with an RC or RCRC filter?

 

How do you properly measure that?

You can't - accurately. You can get an idea of the maximum at the point the string is plucked, or an average midway through the decay.
It depends on: The mass of the string per unit length, the amount of ferromagnetic material in the string, the distance between the string and the pickup (both vertically and laterally), the sensitivity of the pickup, and the transverse velocity of the string (which in turn depends on how hard it was plucked, and the length and tightness of the string, and the length of time since the string was plucked.

With limited headroom (±2V for an audio ADC) and limited SNR (72dB for 12 bit) it is going to have to be optimised. But without adjustment available it can only ever be optimised for one particular guitar and one particular guitarist.

With ±15V supplies on a preamp (or 325V if it's valve) there is plenty of headroom to prevent any particular stage from clipping, unless that is what is required.

 

people report that they're dead quiet.

That's most likely because a lot of pedals use an LM358, and most op-amps are quiet by comparison.

 

Thanks for all your suggestions. My current plan is to use an MCP602 op amp for the input, gain adjusted with a pot, 12bit A/D, LED indicator to show when the input is close to the rails for gain adjustment and an MCP4921 DAC for a 12 bit output.