Optimizing input signal for standard guitar tracker

iOS
1

I’m currently beginning a journey to try and settle on optimized signals, from my several different guitar pickups, to send variously as inputs to the following intended signal chains:

  • Electric guitar audio (“usual” guitar amp sim/FX plugins, including MG3)
  • Bass guitar audio (MG3 Transposer for 8vb, thence to bass DI and/or amp sims and FX)
  • Acoustic instrument simulator (e.g. NAM models for electric guitar sims of acoustic guitar, dobro, banjo, etc.)
  • MG3 tracking (i.e. find the optimal sound from each of my instruments’ pickups, to send to MG3 for the cleanest tracking on the standard Guitar tracker in Expressive mode)

I will always welcome input or suggestions of things to try for any of these, but I’d mostly like others’ input on the fourth–optimizing for MG tracking.

After reviewing a few things already posted here, it seems clear enough that the starting point or overall goal, is to hand MG3 the cleanest, brightest possible signal, as free of potentially confusing resonances as you can get it. This may not be a very good “human-listenable” sound but MG3 will find it a more accurate set of instructions.

Lovely. What I’m looking for, then, is maybe some more detailed guidance on how to process an initial input signal to get to this point. FTR, my guitar pickups include a Fat Strat’s single-coil (neck/middle) and humbucker (bridge) pickups, an Ovation undersaddle piezo, a Fishman Neo-D single-coil soundhole pickup for acoustic, and whatever is in the proprietary piezo-like tube of the Wright Guitars SoloEtte. (It’s unique, but has a personality I like.)

Also, please keep in mind I am almost exclusively on iOS.

Presumably, I should be able to craft the dry input sound using some combination of gate, EQ, and compression to 1) clarify the signal across the spectrum, 2) brighten the signal across the spectrum, and 3) tame any resonances. If I’m missing something obvious there, please do let me know!

I’m pretty clear on gate (recommendations are welcome but so far I’ve been very pleased with Audio Damage’s “Channel Strip” and its -60dB input gate), and I’m starting to learn about EQ in a way that I can definitely hear, in terms of cleaning up the basic Strat signal that comes in off the audio interface. I’d like to learn more about what might be the best EQ profile for MG3’s tracker. Does anyone have any specific suggestions of what EQ frequency ranges I might really pay attention to, or specific places where resonances are known to confuse the tracker?

I’d also like to learn what might be done with compression. Does anyone know if a particular type of compressor might work really well with MG3, or have other guidance for how to eke out a bit more sustain on the input signal, for MG to work with?

Any input is much appreciated; I’ll certainly forge along with trial and error, but I’m always game to learn from the experiences of others too! :nerd_face:

1 Like
2

Interesting question to ponder. I should qualify this by saying most of my recent MG3 use has been with MIDI Guitar 3 Hex, not the standard mono tracker, so this is not a direct like-for-like answer.

That said, the hex side may be a useful clue. The individual string signals don’t necessarily sound like a great guitar tone. They can be thin, direct and unglamorous. But for tracking, that is kind of the point: less inter-string blur, less pickup character, and fewer confusing resonances.

For the mono tracker, I’d probably think of the input signal as something to be clarified rather than beautified.

I’m also assuming MG3 is already doing some internal preprocessing of its own. I don’t know the technical specifics, and maybe nobody outside @JamO does, but I’d be surprised if it were just listening to the raw guitar signal with no filtering or conditioning. So I wouldn’t try to completely redesign the signal before it hits MG3. I’d just remove the obvious stuff that might confuse it.

If I were experimenting, I’d start with EQ like this:

  • Below 60–80 Hz: high-pass to remove thumps, handling noise and low-end junk.
  • 100–250 Hz: check for boom/body resonance, especially with acoustics, piezos or neck pickups. This is one of the first places I’d look if low strings feel vague.
  • 250–600 Hz: boxiness/honk. Sweep here for anything obviously sticking out.
  • 700 Hz–1.5 kHz: possible quack/nasal area, especially with some piezos or soundhole pickups.
  • 2–5 kHz: useful clarity/attack zone. A small lift here may help the note speak more clearly.
  • Above 6–8 kHz: mostly pick noise, fret noise and squeaks. I’d only boost this if it actually improves the MIDI output.

So my practical starting EQ would be: high-pass, one or two cuts for ugly resonances, then maybe a small presence lift around 2–5 kHz. I would not aim for a finished guitar sound before MG3.

On gating, I’d use it mainly to stop idle noise and handling noise. I’d avoid setting it so aggressively that it clips note starts or tails.

For compression, I’d try something clean and transparent rather than characterful. Not an opto/tube/vintage “make it nice” compressor — more a boring utility compressor. The goal would be slightly more stable level, not obvious compression.

Possible starting point:

  • Ratio around 2:1
  • Medium attack, so the pick transient survives
  • Medium release
  • Only 2–4 dB of gain reduction on normal playing
  • Careful level matching after compression

I would avoid heavy squash, fast limiting or obvious pumping. If compression is useful here, I suspect it would be subtle.

The most useful test would be to record a dry input phrase and loop the exact same performance through different EQ/compression settings. Then judge the result by the MIDI, not the audio. False notes, missed notes and wobbly held notes probably point to different causes.

My rough guess would be:

  • False notes: too much noise, sympathetic ringing, fret/pick noise, or a strong resonance.
  • Missed notes: level too low, gate too aggressive, transient softened too much, or not enough useful attack.
  • Wobbly held notes: unstable sustain, pickup resonance, heavy compression movement, or too much ambiguous pitch content.

Based on the hex experience, I’d say the target is not a beautiful guitar sound. It is a boring, stable, unambiguous signal that gives MG3 the least possible guessing to do. Not very rock and roll, perhaps, but probably the right direction. :slight_smile:

These are just my thoughts. Others may have more direct experience with the mono tracker.

One extra thought: I should probably separate “tracking input” from “tone input” here.

My EQ/compression thoughts above are mostly about feeding MG3 the clearest possible tracking signal. I would not assume the same chain is ideal for NAM, amp sims, bass 8vb, or acoustic/dobro/banjo-style sounds.

For those, I’d probably save separate input chains. The boring, clinical signal may win for MG3 tracking, but the more colourful pickup sound may be exactly what you want for tone.

1 Like
3

Boy, thank you for that really quick and thoughtful response! I’ll take a bit to digest all of it over the next few days, but actually I’m happy to say that some of this looks a lot like what I’ve been picking up so far from my own research and WAGing.

Also, your extra attention to some generalizations and nuances is very much appreciated, as I am keen not just for boilerplate settings but to really understand the -fu of what I’m doing.

4

Oh, and I’m totally with you on sending optimized input signals to each of the separate destinations. AUM makes it so easy for me to do that–take the originating signal off the audio interface and do the universal things first (HPF, gate at -60dB, and tuner) and then send to a mix bus for use by other Channels (guitar, bass, acoustic sim, and MG tracker) to refine that as they need.

5

@craftygrass I love your insightful thoughts—thank you for sharing them with us here time and again.

The guitar’s input signal (and even more so with the bass) is particularly important. Different pickup configurations alone probably aren’t the problem, but the signal must remain clean without any effects processing.

Thanks also to @Vaultnaemsae for the helpful comments. I’d just like to add that guitar cables and their capacitance (long cables often result in a duller tone) can play a role, as can the guitar’s potentiometers. It’s best to ask an experienced guitar technician.

Going forward, I expect that the guitar tone at the input won’t just be detected by neural networks, etc., but will also be reconstructed via a second neural network for a “virtual” guitar with ideal tones,

By comparing it to the ideal sound of a guitar string, the real sound could then be separated from its “superfluous” components, which hinder accurate detection.

This approach could also play an important role in the future for single-string detection in MidiGuitar3 Standard.