Would you find a modeling pickup useful?

I've been looking at the Anteres Luther Kit, and then came upon this. Actually the pitch bend tracking isn't bad - no need for hex pickup as its only designed to receive one note at a time. Even with that in mind (one string only) this kind of analysis is normally not easy to do in a single mode. Lead mode and Riff mode basically choose between pitch tracking as a priority, vs obtaining the proper envelope. Doing both, rapidly enough to track everything and output MIDI, would seem very challenging IMHO (I'm only a non-engineer javascript programmer) without some hefty parallel processing.



Still trying to grasp the different products Anteres ONCE HAD for guitar. (I guess the guitarist market was more than they could handle and they quit...) And the ways they can be used. I came upon this:



Try to say this: "I will be using a Roland GK-2A pickup on a parts-ocaster to drive an Antares ATG-1 guitar tuning processor using a Vo Wond and the following gear...."

:D

Didn't mean to hijack your thread tho - the issue I'm having - like others - is the influence of fret position and pickup position, and understanding how much processing needs to be done and how fast. I know it can be done, but it seems like a significant investment of the mind.
 
Last edited:
I want to add something to this. The Variax was the "best" attempt at modeling pickups, I think, and it fell short of reality. Part of it, in my opinion, is the comb filtering that creates a nasal quality - just like a poor IR - that is extremely off-putting.
So in addition to a lack of immediacy and dynamic which are both essential elements of the "feel" of the guitar, there was this slight honk that told you this is not the real thing.
I rather associated the Variax honk with the piezo pickup. Piezo pickups usually have some honk, and seemed the modeling didn't get rid of it completely. Another piezo pickup limit seemed to be the palm mute sounds, as some piezo gunk gets in there and modeling can't get rid of it.

Piezo hex pickups seem to do better at acoustic sounds, and vice versa, magnetic hex pickups seem to do better at electric sounds.
Yes, the comb filtering varies with string tension and fret position. But is the comb filtering of Variax/Roland stuff really static? I thought that the main reason they use a hex pickup was to be able to apply different comb filters for each string, and I assumed that they also varied that with at least the fret position. It requires pretty fast pitch detection to avoid weird artifacts during note changes so maybe they don't?

I haven't done the math on how much the string tension from plucking the string and/or vibration amplitude affect the filtering, that is an interesting point, I'll check.
Why would they need to vary it with the fret position? Pickup position is modeled by summing a delayed inverse signal, the amount of delay dependent on the modeled pickup's distance from the bridge. The physical guitar itself then varies the comb filter depending on its string, pitch, tension, fret position, etc.
 
Why would they need to vary it with the fret position? Pickup position is modeled by summing a delayed inverse signal, the amount of delay dependent on the modeled pickup's distance from the bridge. The physical guitar itself then varies the comb filter depending on its string, pitch, tension, fret position, etc.
You answered yourself: the comb filter with a real pickup varies with the tension of the string, and the tension changes as soon as you fret, bend, make a vibrato or pluck the string with some energy.

If the hex pickup is a piezo saddle it will not have any comb filtering at all by itself since it's placed on the node of the string, so the only comb filter would be the static "pickup position simulation".

If it is a magnetic hex pickup it needs a correction for the comb filtering caused by its own position to make it "flat response", but that correction goes off as soon as you fret, bend, make a vibrato or pluck the string.
So all you'll obtain is a pickup with a weird changing response going into a static comb filter.

Both the correction and the position simulation need to be dynamic to obtain a realistic behaviour.
It probably can't be 100% precise though, cuz the algorithm can't know exactly by how much the tension changes (for example) when fretting on a specific guitar, cuz that depends on the height of the string, its gauge, the neck curvature, etc..
I don't know if it's possible to determine all that just from the pitch.
 
You answered yourself: the comb filter with a real pickup varies with the tension of the string, and the tension changes as soon as you fret, bend, make a vibrato or pluck the string with some energy.

If the hex pickup is a piezo saddle it will not have any comb filtering at all by itself since it's placed on the node of the string, so the only comb filter would be the static "pickup position simulation".

If it is a magnetic hex pickup it needs a correction for the comb filtering caused by its own position to make it "flat response", but that correction goes off as soon as you fret, bend, make a vibrato or pluck the string.
So all you'll obtain is a pickup with a weird changing response going into a static comb filter.

Both the correction and the position simulation need to be dynamic to obtain a realistic behaviour.
It probably can't be 100% precise though, cuz the algorithm can't know exactly by how much the tension changes (for example) when fretting on a specific guitar, cuz that depends on the height of the string, its gauge, the neck curvature, etc..
I don't know if it's possible to determine all that just from the pitch.
Ok, minus tension then (which I think is probably negligible). Why would the modeling need to vary it per fret position?
 
Ok, minus tension then (which I think is probably negligible). Why would the modeling need to vary it per fret position?
Cuz probably the change in tension is not the same if you push at the 1st fret vs the 12th fret
 
I'm mainly talking about the sound quality of the pitch shifting, it's literally flawless.
No perceived latency and absolutely no glitches even on extremely difficult (for the pitch detector) situations where both variax and roland struggle a bit, for example on palm mutes and pinch harmonics.

IIRC the method they used for the pitch detection is decribed on the Antares ATG manual.
EDIT: attached below, chapter 1 and 2

The manual mostly describes what the system achieves, not how it does it. Reading an old patent that seems related to this, it seems like their method is a fairly conventional autocorrelation algorithm with some performance tweaks. At that time, those algorithms were not commonly used because of how expensive they are but that's not really a problem nowadays. If that's what they use, then its latency is higher than the lowest latency algorithms available (autocorrelation needs two full audio cycles => 24ms for low E). I guess they have simply tuned the algorithm very well.
 
The manual mostly describes what the system achieves, not how it does it. Reading an old patent that seems related to this, it seems like their method is a fairly conventional autocorrelation algorithm with some performance tweaks. At that time, those algorithms were not commonly used because of how expensive they are but that's not really a problem nowadays. If that's what they use, then its latency is higher than the lowest latency algorithms available (autocorrelation needs two full audio cycles => 24ms for low E). I guess they have simply tuned the algorithm very well.
You're right, I probably confused that for another article I've read on cycfi.com about pitch detection, sorry.
 
Cuz probably the change in tension is not the same if you push at the 1st fret vs the 12th fret
No, @Per seems to differentiate the two. I was asking in response to this:
Yes, the comb filtering varies with string tension and fret position. But is the comb filtering of Variax/Roland stuff really static? I thought that the main reason they use a hex pickup was to be able to apply different comb filters for each string, and I assumed that they also varied that with at least the fret position. It requires pretty fast pitch detection to avoid weird artifacts during note changes so maybe they don't?
 
You're right, I probably confused that for another article I've read on cycfi.com about pitch detection, sorry.

No worries at all, because of your post I went and found that patent which was a good read. The patent describes an idea that is based on the same underlying technique as the one described on cycfi.com: autocorrelation, which tends to be too inefficient to use directly. Cycfi and Antares have completely different approaches to optimizing it to be fast enough for real-time computation. It's quite interesting to compare them.
 
@yeky83 I think you're right, I must have been confused/tired when I wrote that post, sorry!
Oh, no need for a sorry! Just wanted to check my understanding.

I didn't know pitch detection could be done sooner than in two audio cycles, I'm gonna have to look that up... thanks to you and @DLC86 for the fun reads ;)
 
Oh, no need for a sorry! Just wanted to check my understanding.

I didn't know pitch detection could be done sooner than in two audio cycles, I'm gonna have to look that up... thanks to you and @DLC86 for the fun reads ;)

It can be done, but it comes at the expense of accuracy so would have to be used together with conventional algorithms. Machine learning algorithms are promising in this area.
 
The closest I can remember anyone getting to this idea successfully was Antares with the ATG-1 Autotune for guitar, which used the 13 pin output of Roland type p/up guitars to model single coil, humbucker, and acoustic sounds as well as 12 strings and various different tunings all using software inside a floor unit resembling the AX8/FX8 type floor units by Fractal. I bought in to it and was sadly disappointed when Antares seemingly sold their existing stock and abruptly dropped the whole idea after actually getting quite close to the finish line. I think if perhaps Atares had a Cliff Chase among their ranks it would in fact be the class leading unit in guitar modeling that the Axe-Fx III is in the amp/efx modeling world today. I still have a VERY bitter taste in my mouth over the way Antares baited folks into buying those units they had backlogged and then dropped the whole idea as soon as they were sold "ON SALE", with no further updates or improvements.
 
Last edited:
I've been all about on the fly tuning changes and modeling for decades. Of course nothing will ever be exactly same as the original instrument but for me that has never been the point since I am not a cover artist. Creativity and the ability to travel lighter with an increasing sound arsenal that is inspiring is far more important than imitation. No judgement on those looking to get things to sound the exact same way they did 50 years ago, I love those tones but personally I am also looking for new tones.

Since the mid 90's I've owned just about every type of Hex pickup (Roland GKs, RMC, Fishman, etc) and for the last two or three years I have been using a Cycfi Nu pickups. As far as my personal experience these are the best individual string pickups if you are looking for flat response, minimal noise and minimal bleed. Joel has also made the pickup designs open source (MIT license I believe so useable for modifying, building and selling without concern).

I've also owned many, many processors starting with Digitech 2112 & 2120, the Roland VG-8, VG-88 and for a long time the VG-99 which for a couple of years has been going into an Axe FX II XL+. As of a couple days ago I am using an Axe FX III (WOOHOO! :D). I never tried the Antares although it certainly looks interesting and I might pick one up just to check it out.

So, all that said, I found this thread because I really want to replace the VG-99. It is a bit long in the tooth to say the least but it seems like nothing has really exceeded it's capabilities. Bottom line, I am definitely interested but for me something foot switchable is critical. I like making fine adjustments on the guitar with the usual volume, tone and pickup selector knobs but I use too many combinations to remember them during a show.

As for technical issues, the biggest problem that I have noticed with emulation is noise while the biggest problem with retuning is string bleed. Latency isn't a problem with the VG-99 (although it is with the Axe FX pitch effect). The noise problems vary depending on the pickup type. There is the over sensitivity issue with unwanted finger, pick and hand noise particularly effecting Piezo pickups. Interference, crackling and hum from electronics is an issue with all pickups. There is also a problem with noise as a by-product of emulation, which in my experience seems to be more prominent as an additive effect of processing multiple pickups. Basically that is to say that a minimal noise floor is essential as you are getting six times the noise.

For the most part I have personally overcome most of the issues with noise and latency at this point, leaving the issue of bleed when it comes to retuning. If excessive bleed can cause weird glitches with electronic retuning but the main problem is the unwanted dissonance or doubling effect of a string crossing into the retuning of another. For me the Cycfi pickups have eliminated all of the glitches and I only find occasional issue with hearing a cross over when using higher gain. Hopefully this gives you some insight into the many issues. I think you will find similar comments experiences on the vguitarforums site and those guys would definitely be your target market for any hex pickup.

One last thing, I was going to add the links to Joel De Guzman's 4 blog posts on the subject of Virtual pickups, except I can't do to the forum rules. You can find them on the Cycfi site if you search for them by name "Virtual Pickup Placement" (parts 1, 2, 3, and Revisited). Funny was that I found that he just came out with a followup today after three years on software for pickup emulation that he has been working on which is titled "In the Meantime". As an aside, having worked with on my personal project I can tell you he is very open to collaboration. Something to consider. Cheers and best of luck.
 
Thanks JAmatulli for your thoughts!

What do you use your VG-99 for? I haven't used one but it seems like it has lots of features, with quite a bit of overlap with Fractal units. A complete multi-fx and amp sim system is quite a bit more broad than I currently have in mind to do.

When you say you need foot switchability, would being midi controllable be sufficient for you? I'm a big fan of compact and light effect units, so I'm aiming for more of a Polytune Mini footprint than a big lots-of-footswitches thing.

Also, would you be interested in using the knobs on your guitar for sending midi control messages to control presets on your other gear? The Cycfi Nexus can kind-of do that but to me it seems like a dented rotary encoder would be nicer for that.

Do you have any issues with noise on the Cycfi Nus or are they good enough for you? I'm hoping to achieve a noise floor similar to that pickup, even though the inner workings will be a bit different (and hopefully with even less string crosstalk, but I don't know how much yet).

I'll definitely head over and ask the people at vguitarforums at some point, but I have some things I want to prepare and think about first. It's a different crowd and I want to ask questions specifically about their needs.

I'm well aware of Joel's work with Cycfi, I've been in touch with him a bit also.
 
What do you use your VG-99 for?
I use it for alternate tuning and pickup/instrument emulation only. No amp sims, no effects. Not since getting an Axe FX.

When you say you need foot switchability, would being midi controllable be sufficient for you?
Yes, midi controllable. (There is also OSC but midi obviously has far better compatibility.)

Also, would you be interested in using the knobs on your guitar for sending midi control messages to control presets on your other gear? The Cycfi Nexus can kind-of do that but to me it seems like a dented rotary encoder would be nicer for that.
I have the Nexus and use a midi controller as a volume knob for just that purpose.

Do you have any issues with noise on the Cycfi Nus or are they good enough for you? I'm hoping to achieve a noise floor similar to that pickup...and hopefully with even less string crosstalk.
Noise is always an issue when you are dealing with multiple pieces of gear like this. That said, the Nu's are very good. Any reduction in crosstalk is always welcome.
 
I don't even understand how one would play the guitar after an "on the fly tuning change" unless it's something simple like dropping everything down a couple of semitones, or dropping the low E to D. Do people legit do this during performances?
 
Back
Top Bottom