Neural Pickup Modeler

Ok so this is AWESOME! You should posted on the thread that is up talking about what features the next gen of Fractal products should have 👍🏼.
 
swiss studio said:


Just saw that on TGP forum. As next gen fractal flagship could read nam, that would be killer feature! Also I tested the SH4 to split nam and it’s perfect even for my own pickup.
Maybe the dream of the Variax will raise again.
After amp modeling, after amp cloning, time for pickup cloning
Click to expand...

Hi Francois (or whatever your real name is), how are you?
Nice idea but using that method (apart from the reliability of the results using a bluetooth speaker) only measures the frequency response and eventual non-linearities of the pickup itself but tells you nothing about its magnetic aperture (i.e. the portion of string it "picks up"), the effect the instrument scale has on it and its position (neck vs bridge, etc.), these things might not matter that much in your example but would definitely matter when trying to copy the sound of a pickup in a different position or on a different guitar.

In that case, in theory, it'd probably be better to simultaneously record (somehow) the two pickups with normal playing and use the dry-wet pair method on tone3000 (I've done this myself with an acoustic piezo output and mic with great results), or maybe you could even get good results by using something like Sonulab's MagicMatch if you can't record them simultaneously.

But in practice, using different positions also means different null points in the frequency response of each string, which can't be fully compensated by nam (you can't infinitely boost a frequency) and results probably wouldn't be very accurate anyway... that's why for this kind of things it's usually best to use piezo saddles or hex pickups placed very near the bridge, cuz at the node (aka the point where the string touches the saddle) there are no nulls and your signal will be a blank canva frequency-response-wise.

EDIT: check these links for a deeper explanation of what I'm talking about and for a simulator
https://till.com/articles/PickupResponse/
https://till.com/articles/PickupResponseDemo/

PS: I think this idea could greatly improve systems like Variax, Antares ATG and Roland hex processors, cuz the major drawbacks of those have always been the pickup response simulation... Using a NAM model for pickups instead of a simple IR/filter might produce great results.
 
Last edited:
DLC86 said:
Hi Francois (or whatever your real name is), how are you?
Nice idea but using that method (apart from the reliability of the results using a bluetooth speaker) only measures the frequency response and eventual non-linearities of the pickup itself but tells you nothing about its magnetic aperture (i.e. the portion of string it "picks up"), the effect the instrument scale has on it and its position (neck vs bridge, etc.), these things might not matter that much in your example but would definitely matter when trying to copy the sound of a pickup in a different position or on a different guitar.

In that case, in theory, it'd probably be better to simultaneously record (somehow) the two pickups with normal playing and use the dry-wet pair method on tone3000 (I've done this myself with an acoustic piezo output and mic with great results), or maybe you could even get good results by using something like Sonulab's MagicMatch if you can't record them simultaneously.

But in practice, using different positions also means different null points in the frequency response of each string, which can't be fully compensated by nam (you can't infinitely boost a frequency) and results probably wouldn't be very accurate anyway... that's why for this kind of things it's usually best to use piezo saddles or hex pickups placed very near the bridge, cuz at the node (aka the point where the string touches the saddle) there are no nulls and your signal will be a blank canva frequency-response-wise.

EDIT: check these links for a deeper explanation of what I'm talking about and for a simulator
https://till.com/articles/PickupResponse/
https://till.com/articles/PickupResponseDemo/

PS: I think this idea could greatly improve systems like Variax, Antares ATG and Roland hex processors, cuz the major drawbacks of those have always been the pickup response simulation... Using a NAM model for pickups instead of a simple IR/filter might produce great results.
Click to expand...
Hi Carmelo!
Yes, I tried the piezo thing, thinking that’s what the Variax had used, and telling myself that reusing that with NAM—and my method of having the pickup “read” sine sweeps using a loudspeaker (a method I still need to refine with a coil without a magnet, which I’m working on)—could work.

But the problem is that when you actually play the guitar, the pickup will capture a string vibration that has nothing to do with what the piezo captures, since the piezo is often at the bridge (and therefore doesn’t have the harmonics and the modes “seen” by the pickup). In practice, NAM hates that and refuses to converge. So I’ve dropped that method on my side.

The goal here is mainly to transform one pickup into another, and I have to admit the humbucker-to-single-coil conversion surprises me: it sounds the same to me, and I feel it the same in blind tests. The reverse is less true—first because turning a single coil into an HB won’t eliminate noise from the source (50-cycle hum) and other interference that a humbucker cancels out, but also for a reason I don’t know—probably impedance and system loading—maybe it’s easier for NAM to emulate in the HB→SC direction than the other way around.

As for cloning the pickup and the guitar’s construction, you’d need something much heavier-duty:

you’d have to find a way to excite the strings perfectly and in a perfectly reproducible way… maybe with a speaker resonating with each string… but I find variation of the magnetic field more reliable and reproducible, especially with large coils to avoid edge effects.

I’m also pretty blown away by the transformation of noiseless pickups into vintage single coils. The big advantage here is getting the HB benefit (silence) with the snappy SC sound. Otherwise, nothing beats acute gear-collecting syndrome ;)
 
I do hope someone does something similar to what Slate is doing with Mics (and now headphones).
Slate mics are a 'clean' - his software adds the 'color' for the mic of your choice.
Someone could do the exact same thing with pick ups.
 
swiss studio said:
But the problem is that when you actually play the guitar, the pickup will capture a string vibration that has nothing to do with what the piezo captures, since the piezo is often at the bridge (and therefore doesn’t have the harmonics and the modes “seen” by the pickup). In practice, NAM hates that and refuses to converge. So I’ve dropped that method on my side.
Click to expand...
Actually it's the opposite, a piezo at the bridge captures all harmonics the string produces and thus you can easily create the same nulls of a given pickup position simply by using a comb filter... but that comb filter needs to be different for each string and that's probably the main reason NAM can't converge, it's really not made to perform this task and can't know which string is being played in the provided sample.
 
DLC86 said:
This might be the closest thing currently available: https://www.cycfi.com/xr-spectra/
Click to expand...
Interesting - combine this with a 'shaping/coloring' software and it seems you could emulate other pickups.

Their pickup, picks up a lot of extra high end content (almost sounds like an acoustic piezo). These would be great for using with acoustic Tone Match and IRs, to get a decent acoustic sound from your electric.

*edit - just noticed cycfi sells a small unit to change the tone of the pickup. Pretty cool.
 
BBN said:
Interesting - combine this with a 'shaping/coloring' software and it seems you could emulate other pickups.

Their pickup, picks up a lot of extra high end content (almost sounds like an acoustic piezo). These would be great for using with acoustic Tone Match and IRs, to get a decent acoustic sound from your electric.

*edit - just noticed cycfi sells a small unit to change the tone of the pickup. Pretty cool.
Click to expand...
That's a somewhat simplified version of the crazy-younger-me pickups and tone controls that i just took out of my old 335 (see huge thread here). Low impedance pickups, filters with variable frequency and resonance (Q).

Cyfi clearly means for people to set that module's controls and forget them, not change them during a live set. Processing outside the guitar with presets etc, seems like a much better idea.
 
swiss studio said:
I imagine that playing with those would wake up the anti aliasing topic ;)
FRFR pickups is full of 5kHz-20kHz energy, perfect to « feel » aliasing ;p
Click to expand...
They are meant to be filtered before hitting any drive or amp, so no more than normal pickups
 
Dave Merrill said:
Cyfi clearly means for people to set that module's controls and forget them, not change them during a live set. Processing outside the guitar with presets etc, seems like a much better idea.
Click to expand...
Agreed.

Line6 was going after that with the Variax (I used to have one, and loved it for drop D tuning on the fly)....but the tone left much to be desired.
 
Joshua_L said:
Isn’t that just refined eq ?
Click to expand...
In theory it’s more of that. The compression feels very different. But I struggle finding more reliable way to convert on pu to another, using the strings themselves. Recording both pickups (with proper electronics and each own jack) with same strings playing on a bench could do the trick.
But is it really worth the try…?
I don’t make a lot of diffence in my style of playing using mid to high gain, humbucker are unbeatable.
And I play mostly in studio, with 13 guitars available.
better spend time having fun with Fractal gear and practicing guitar.
We all saw what happened to Variax. We all want a reason to buy more guitars and gear :)
 
You must log in or register to reply here.
Back
Top Bottom