Trying To Get That Pat Metheny Guitar Synth Sound - How To Do?

[Rob Strand]

There's quite a lot of interesting things in that schematic. Thanks for bringing it up. I'd never have thought to look at it otherwise.

You gotta love the electronics in those things.  Someone spent a lot of time getting that front-end to work.

I remember when they came out.  IIRC, there was a bass version as well.  A friend of mine bought one of those guitars for $50 in a garage sale without the synth box.

Anyone remember these from the mid 80's.   Used by Alan Holdsworth, Lee Ritenour and some others.  It's like the 80's started to engulf the guitar itself.
http://www.hollis.co.uk/john/synthaxe.html

[Paul Marossy]

Anyone remember these from the mid 80's.   Used by Alan Holdsworth, Lee Ritenour and some others.  It's like the 80's started to engulf the guitar itself.
http://www.hollis.co.uk/john/synthaxe.html

I do. I remember seeing it in Guitar Player magazine and I have several Allan Holdsworth albums where they make an appearance. Distinctive sound for sure.

[Paul Marossy]

The Youtube that you originally linked to was all single-note stuff, which is why I thought "Why get into all that divided pickup stuff when that modded Boss-Tone seems to approximate it well enough for single-note runs?"....

I love guitar synth, but if a little modded 2-transistor distortion gets you to where you want to go, I say vaya con Jordan.

Thanks for the suggestion. So I modded my mothballed Boss Tone last night with these "Trombetta" changes. That sounded cool, so I thought I'd see what it would sound like adding a tilt EQ with an "edge" control before the level pot. That sounds even better because you can really shape the sound - I guess by controlling the harmonic content at the output? Anyway, if I put that before my modded Orange Squeezer (attack, release and tone added) it evens it all out level-wise. Lots of fun to play with for sure.

Not sure how to describe the sound.... it's kind of like a synthy buzzy thing, in a good way, not the annoying overly fuzzy kind of way. There must be a lot of odd order harmonics going on in there, almost like a square wave I would think. Tonight when I get done with this annoying work thing I have to do I will look at what is happening to the signal with a scope. Should be interesting...

[Mark Hammer]

Glad to see there was some useful purpose to my suggestion.

[Paul Marossy]

Glad to see there was some useful purpose to my suggestion.

There usually is. 

[pinkjimiphoton]

About 15 years ago I attempted to design something that would give me something similar to that Pat Metheny guitar synth sound (kind of like a horn). I originally tried to use an LM2917, etc and I tried revisiting the project a couple of weeks ago. I found some ways to make it respond better to all notes on the fret board but I just end up with something that is like a ring modulator.... which is NOT the sound I am after. 

Anyone have a vague idea of how Roland does it?

Original thread from 2003 is here: https://www.diystompboxes.com/smfforum/index.php?topic=11267.msg69185#msg69185

The sound I want to emulate (without tons of delay)

try biasing some fuzz into unlinear regions.
check a couple of my circuits for that weird almost blown or bowed attack.

dirty bad model 1 driver
photon phuzz
weetie pi
double hit deux all have ranges that do kinda similar. sorta a muted almost gated fuzztone with decent sustain that kinda swells in, and swells when ya play legato.
there's videos blah blah blah
youtube.com/666pinkster

[Paul Marossy]

try biasing some fuzz into unlinear regions.
check a couple of my circuits for that weird almost blown or bowed attack.

dirty bad model 1 driver
photon phuzz
weetie pi
double hit deux all have ranges that do kinda similar. sorta a muted almost gated fuzztone with decent sustain that kinda swells in, and swells when ya play legato.
there's videos blah blah blah
youtube.com/666pinkster

Thanks will check them out too

[Paul Marossy]

So looking at those Roland Service Notes, I see that everything starts with the Fundamental Detector which strips all of the harmonics off of the fundamental note coming into it. I can see that this is one of the main problems with my attempt, too many harmonics confuses the system.

That being said, is there a relatively simple way to do accomplish that with easy to find parts? Would be interesting to mess around with just for my own self-edification.

[pinkjimiphoton]

square wave the @#$% out of the guitar signal, and use a big cap to ground to wipe out everything above the fundamental. then fuzz that with misbiased or adjustable biased fuzz may get ya in the ballpark. a diode clipper in series with the audio signal instead of to ground can help smooth it out some by compressing the signal.  use big caps in the power supply or in the emitter/ground  part of transistors, maybe a 5k resistor and a 47k cap to bump up the gain and give a longer sustain with a bit of envelope.
all the rules ya ever learned? break 'em. you WILL get this bro. i have total faith in you. lotta ways to skin a cat.

whats nice about the photon phuzz circuit guitarpcb was selling was the voice pot, it puts it thru a few different ranges of useable bias from reasonably clear to almost ringmodulated. the 'weetie pi does it even more.

i totally live to make weird sounding fuzzes, but must warn ya, sometime the things i do may not make much sense, til ya listen to them

[Rob Strand]

I can see that this is one of the main problems with my attempt, too many harmonics confuses the system.

That being said, is there a relatively simple way to do accomplish that with easy to find parts? Would be interesting to mess around with just for my own self-edification.

Similar detection problems comes up in other fields as well.  It's an open problem, you start with an idea which is kind of there.  Then you spend ages trying to refine that idea - this can make or break it.   When you are finished you are never quite happy with it.   So you try another idea, perhaps a completely different concept, and go though it all again.   Then one day you come across a funky simple circuit that works better than all of yours!  That's why I admire circuits like this.   In the DSP world there's a lot of tools at our disposal these days but for analog circuits of reasonable complexity it's a different ball-game.

Cleaning up signals with filtering helps a lot.    If that doesn't work then the multple filter idea like the Roland one might help.

On the other end of the scale you could try the fundamental detectors in the Boss OC-2.   They work most of the time.  The question is is that better than yours or worse?   The next question is it much worse than the Roland circuit?    Recently there was a thread on the Mutron octave circuit.  I have no idea how this compares to the OC-2 circuit.

[Mark Hammer]

So looking at those Roland Service Notes, I see that everything starts with the Fundamental Detector which strips all of the harmonics off of the fundamental note coming into it. I can see that this is one of the main problems with my attempt, too many harmonics confuses the system.

That being said, is there a relatively simple way to do accomplish that with easy to find parts? Would be interesting to mess around with just for my own self-edification.

There's simple, and there's reliable and problem-free.  The problem with guitar is that we move our fingers along the strings, and don't just press a key.  So, unlike keyboards, where there is no irrelevant signal until you press a key, guitar generates plenty of irrelevant signal.  And squaring the daylights out of it only gets you something useful if you have already managed to remove the irrelevant.

So, since you wanted easy and available parts, I would suggest using a back-to-back pair of silicon diodes in series with the signal path as a Q&D noise gate.  That way, nothing under 500-600mv gets through.  Once you've blocked out the string glisses and other irrelevancies, THEN you can boost what's left of the signal, and use a comparator, or similar, to trigger square waves.  The old EPFM "Ultra-Fuzz" is a good example.

[Paul Marossy]

So looking at those Roland Service Notes, I see that everything starts with the Fundamental Detector which strips all of the harmonics off of the fundamental note coming into it. I can see that this is one of the main problems with my attempt, too many harmonics confuses the system.

That being said, is there a relatively simple way to do accomplish that with easy to find parts? Would be interesting to mess around with just for my own self-edification.

There's simple, and there's reliable and problem-free.  The problem with guitar is that we move our fingers along the strings, and don't just press a key.  So, unlike keyboards, where there is no irrelevant signal until you press a key, guitar generates plenty of irrelevant signal.  And squaring the daylights out of it only gets you something useful if you have already managed to remove the irrelevant.

So, since you wanted easy and available parts, I would suggest using a back-to-back pair of silicon diodes in series with the signal path as a Q&D noise gate.  That way, nothing under 500-600mv gets through.  Once you've blocked out the string glisses and other irrelevancies, THEN you can boost what's left of the signal, and use a comparator, or similar, to trigger square waves.  The old EPFM "Ultra-Fuzz" is a good example.

Interesting, will give this a try.

[pinkjimiphoton]

a diode clipper in series with the audio signal instead of to ground can help smooth it out some by compressing the signal. 

hahah mark agrees!

[Rob Strand]

There's simple, and there's reliable and problem-free.

It could well work.   That's the problem with these ill defined problems.

If you look at the OC-2 circuit it captures the peaks and has an adaptive "slice" point to find the fundamental.  The compressor tries to regularize the peaks then use a fixed slice point.   So to some degree they are achieving the same thing.  But not quite ...

[Mark Hammer]

a diode clipper in series with the audio signal instead of to ground can help smooth it out some by compressing the signal. 

hahah mark agrees!

Not so fast Sir James.  I'm suggesting the diodes as a means to block low-level signals, not as a means to compress anything.

If diodes are placed in series with the signal, then the signal has to be above the forward voltage to pass.  A back-to-back pair of Si diodes would mean that any subsequent stages see nothing until what's coming in is above that threshold.  We are mostly familiar with circuits that use such diode pairs (and most often germanium) to produce crossover distortion, like in the venerable Boss HM-2.  In those instances, the signal is boosted by an earlier stage, and the series diodes clip the little bits of the nose and tail of the waveform, just above and below the midpoint.  I'm suggesting no clipping whatsoever, or at least no intention to clip.

I've actually used this little Q&D gate trick to reduce the audible carrier whine on my EPFM Ring Modulator.  That circuit has a nulling trimmer, which can certainly reduce the carrier whine, but doesn't completely eliminate it.  Using the series diode-pair robs the signal of a half-volt of amplitude, but one can easily make up for that.

The basic idea is that you don't want the comparator to be squaring anything that isn't a note.  You could use a conventional noise-gate, with envelope follower, and stick that somewhere before or after the squaring, but it is much easier to declare with a pair of diodes that "I'm not interested.  If you're not above threshold, talk to the hand!".

The Anderton EPFM circuit is a variable threshold comparator.  What I'm suggesting is a kind of pre-comparator comparator, albeit with a fixed threshold, set by the Vf of the diodes.  How much input signal is rejected/blocked could be determined by choice of diodes, or by introducing a bit of variable gain just ahead of them to provide more and less signal falling below the fixed threshold.

In some respects, one could extract more reliable behaviour from the final comparator by using an op-amp stage before the diode pair to filter out at least some of the irrelevant stuff, including both harmonic content and low-frequency string ripple.

[Paul Marossy]

I've actually used this little Q&D gate trick to reduce the audible carrier whine on my EPFM Ring Modulator.  That circuit has a nulling trimmer, which can certainly reduce the carrier whine, but doesn't completely eliminate it.  Using the series diode-pair robs the signal of a half-volt of amplitude, but one can easily make up for that.

I wish you could do something like that with the ICL7660CPA chip. I have several of those and I can't really use them in anything. 

[Mark Hammer]

Doesn't using larger value caps help?  I don't know that.  I'm just wondering if it does.

[Paul Marossy]

Doesn't using larger value caps help?  I don't know that.  I'm just wondering if it does.

Don't know... a few days ago was the first time I've ever tried using a charge pump IC to double the voltage. Got the whining sound, was expecting it with that chip. I think I need the ones with the "S" suffix that operate on a higher frequency.

[garcho]

what if you have 40-odd super narrow band pass filters, one for each note? i know that's not really practical to make with op amps or what have you, but i feel like that's somehow heading somewhere interesting...

if you squared the ink out of your guitar signal, each filter would cut out non-tonal garbage, leaving a small band around the tonic, making that weird string signal simpler, like a flute's, except distorted, so more like a trumpet's. for a digital project maybe?

[pinkjimiphoton]

a diode clipper in series with the audio signal instead of to ground can help smooth it out some by compressing the signal. 

hahah mark agrees!

Not so fast Sir James.  I'm suggesting the diodes as a means to block low-level signals, not as a means to compress anything.

If diodes are placed in series with the signal, then the signal has to be above the forward voltage to pass.  A back-to-back pair of Si diodes would mean that any subsequent stages see nothing until what's coming in is above that threshold.  We are mostly familiar with circuits that use such diode pairs (and most often germanium) to produce crossover distortion, like in the venerable Boss HM-2.  In those instances, the signal is boosted by an earlier stage, and the series diodes clip the little bits of the nose and tail of the waveform, just above and below the midpoint.  I'm suggesting no clipping whatsoever, or at least no intention to clip.

I've actually used this little Q&D gate trick to reduce the audible carrier whine on my EPFM Ring Modulator.  That circuit has a nulling trimmer, which can certainly reduce the carrier whine, but doesn't completely eliminate it.  Using the series diode-pair robs the signal of a half-volt of amplitude, but one can easily make up for that.

The basic idea is that you don't want the comparator to be squaring anything that isn't a note.  You could use a conventional noise-gate, with envelope follower, and stick that somewhere before or after the squaring, but it is much easier to declare with a pair of diodes that "I'm not interested.  If you're not above threshold, talk to the hand!".

The Anderton EPFM circuit is a variable threshold comparator.  What I'm suggesting is a kind of pre-comparator comparator, albeit with a fixed threshold, set by the Vf of the diodes.  How much input signal is rejected/blocked could be determined by choice of diodes, or by introducing a bit of variable gain just ahead of them to provide more and less signal falling below the fixed threshold.

In some respects, one could extract more reliable behaviour from the final comparator by using an op-amp stage before the diode pair to filter out at least some of the irrelevant stuff, including both harmonic content and low-frequency string ripple.

duly noted. i use them in series sometimes in things i mess with, it seemed to me with my limited understanding it was affording some compression as well as gating. my bud ethan winer uses them in some of his stuff like that.