Q for L6 developers: How did amp emulation detail improve from Pod>HD>Helix ?

[d0stenning]

In virtually every promo blurb or description of why an amp sim plugin, hardware effect or iOS amp sim is "better than the one before.. or others..." you will hear that in "THIS ONE" its "BETTER BECAUSE" the engineers modelled "in detail" every component in the amp circuitry - from valves to transistor. 

 

Then a new version comes along, HD or V2.0 of Amplitude etc - and the manufacturers will say - NOW we model "everything in the amp" ? then a "Helix" will come along and they day - "oh now we model down to the "component level".

 

Clearly compromises were made in earlier versions that could be removed in later versions due to DSP ( or power draw ) performance improvements.  So heres the question to L6. 

 

Explain how the modelling for - say a Marshall Plexy  improved in detail - and where - in which component areas - from the

 

Amp Farm Pro-Tools plugin > Pod > HD > Helix. 

 

This is a question from an electronic engineer and programmer. Not to steal ideas - but just to get a sense for just what _was_ accurately modelled as a component - versus some approximation using EQ curves, input-output transfer curves , "tricks" etc.. 

 

For example the team behind Revalver IV (  a highly respected amp-sim VST/AU plugin ) say in their promo blurb that in the LATEST version - IV - they moved to using an open-source commonly used electronic circuit emulation-design program called SPICE  ( which has been around for a long time - and used to model circuitry - originally not in real-time ) because NOW CPU power allows them to use SPICE to model everything in the circuit.... which means prior to this--- not everything WAS being emulated at component level. Presumably certain circuit groups ( tone-stack for example ) were implemented through some other way.. 

 

 

Just curious ! 

[cruisinon2]

I wouldn't expect much of a response. The powers that be are not big on discussing anything even vaguely technical. They still refuse to divulge specs as innocuous as the recommended height for the mag pickups on the Variax guitars.

[duncann]

I wouldn't expect anyone from Line 6 to start discussing their trade secrets. But Digital_Igloo might chime in with some vague, generalized concepts about this. Who knows?

[silverhead]

I expect any answer you would get would be in terms of the evolution of DSP capacity, rather than components. You could compare it to pixelation in the photo/video world. The evolution of processing power in cameras has consistently improved the depth and granularity of the images, and you can speak of this generation of camera having ten times the number of pixels as the previous generation. Similarly, the amp modelling in the HD series allowed ten times the sample rate of the X3 series, resulting in much higher granularity and realism of sound reproduction.

 

Number of frames/second and pixelation depth defines image realism. Similarly, audio sample rate/depth defines audio image realism. Same story.

[PeterHamm]

I suspect that the whole "modeled every single component in the circuit" deal is different from what they did before, where they probably measured and modeled parts and pieces together.

 

But I'm just guessin'...

[Digital_Igloo]

I wouldn't expect anyone from Line 6 to start discussing their trade secrets. But Digital_Igloo might chime in with some vague, generalized concepts about this. Who knows?

Indeed, and if Helix had four SHARC DSPs, or next-generation SHARCs, I'm sure we'd have no problem finding more granular elements of amps and pedals to model. I'll betcha $100 Ben and co. already have a list of things they'd love to do that Helix won't currently let them. No snake oil here.

 

It's not just about measuring stuff, tho'. You need a DSP and sound design team that understands why an amp or pedal component sounds and feels the way it does. Behavior modeling is much, much more complicated than measuring points in a circuit. Spreadsheets are easy; blending math and art is hard.

 

My posts are vague because I'm a product designer, not an engineer.

 

Slightly helpful insight might be had on Helix's blog page.

[phil_m]

My posts are vague because I'm a product designer, not an engineer.

 

[Digital_Igloo]

Digital Igloo: You know, I had an idea like that once. A long time ago. 

 

Phil: (nonplussed) Oh really? What was it, DI?

 

Digital Igloo: Well all right. It was a... Jump to Conclusions processor. You see, there would be this guitar processor... that you would put on the floor. And it would have different... conclusions written on its scribble strips.... that you could jump on!

 

Phil: That is the worst idea I have ever heard in my life, DI.

[PeterHamm]

Digital Igloo: You know, I had an idea like that once. A long time ago. 

 

Phil: (nonplussed) Oh really? What was it, DI?

 

Digital Igloo: Well all right. It was a... Jump to Conclusions processor. You see, there would be this guitar processor... that you would put on the floor. And it would have different... conclusions written on its scribble strips.... that you could jump on!

 

Phil: That is the worst idea I have ever heard in my life, DI.

 

[HonestOpinion]

.... Behavior modeling is much, much more complicated than measuring points in a circuit. Spreadsheets are easy; blending math and art is hard.

 

...

 

 

This comment, "blending math and art is hard" from DI truly resonates with me and is something I have thought quite a bit about.  Although I am a tireless and probably sometimes tiresome advocate for steady improvement and enhancement of the Helix I do want to take a moment to acknowledge the incredible talent at Line6.  You guys have achieved something extraordinary!  Mediocre programmers are a dime a dozen but putting together a team of gifted programmers and musicians, and if you are really lucky some who are both, is an incredible challenge. Some of the best visionary musical/technical talent in the world is currently gathered at Line6 and I can't wait to see what you guys do next. Thank you for all your hard work!

[cruisinon2]

Digital Igloo: You know, I had an idea like that once. A long time ago.

 

Phil: (nonplussed) Oh really? What was it, DI?

 

Digital Igloo: Well all right. It was a... Jump to Conclusions processor. You see, there would be this guitar processor... that you would put on the floor. And it would have different... conclusions written on its scribble strips.... that you could jump on!

 

Phil: That is the worst idea I have ever heard in my life, DI.

You think the pet rock was a good idea? ;)

[lrschmid]

Digital Igloo: You know, I had an idea like that once. A long time ago. 

 

Phil: (nonplussed) Oh really? What was it, DI?

 

Digital Igloo: Well all right. It was a... Jump to Conclusions processor. You see, there would be this guitar processor... that you would put on the floor. And it would have different... conclusions written on its scribble strips.... that you could jump on!

 

Phil: That is the worst idea I have ever heard in my life, DI.

[brue58ski]

Digital Igloo: You know, I had an idea like that once. A long time ago.

 

Phil: (nonplussed) Oh really? What was it, DI?

 

Digital Igloo: Well all right. It was a... Jump to Conclusions processor. You see, there would be this guitar processor... that you would put on the floor. And it would have different... conclusions written on its scribble strips.... that you could jump on!

 

Phil: That is the worst idea I have ever heard in my life, DI.

I think my wife had one of your processor's last night. (Uh oh. She's here.)

Uuuuuuhhhhhhh. Just kidding honey!

[glideman]

Is that your "O" face?

[ozbadman]

For example the team behind Revalver IV (  a highly respected amp-sim VST/AU plugin ) say in their promo blurb that in the LATEST version - IV - they moved to using an open-source commonly used electronic circuit emulation-design program called SPICE  ( which has been around for a long time - and used to model circuitry - originally not in real-time ) because NOW CPU power allows them to use SPICE to model everything in the circuit.... which means prior to this--- not everything WAS being emulated at component level. Presumably certain circuit groups ( tone-stack for example ) were implemented through some other way.. 

 

 

Just curious !

Well, if you are implying this gives a more accurate model, that assumes that (1) the spice model is correct and (2) the spice models, and all their interactions are identical to the interactions that are happening in the particular amp you are modeling, including the exact impedances (since resistors in any 1 specific amp have an exact value, not a value +- tolerance) and component interactions.

[ozbadman]

I wouldn't expect much of a response. The powers that be are not big on discussing anything even vaguely technical. They still refuse to divulge specs as innocuous as the recommended height for the mag pickups on the Variax guitars.

Well, recommended height is personal. They could have a factory setting height.

[tommasi]

The question is very interesting beyond the "mere" amp modeling scenario, and more into an epistemological one.

 

I believe in older days "amp sim" had a more holistic approach. I don't care WHY something sounds the way it sounds, I just want to recreate its sound. Hence, you'd record the amp as a whole, and made sure the signal transformation from input to recorded output was well approximated (fitted) by your model.

 

I like the holistic approach, maybe because I come from artificial intelligence, and we gave up trying to make something "intelligent" a long time ago, trading for something that "behaves intelligently".

 

The holistic approach has a few disadvantages -- namely, that parameters are hard to implement: it's easy to create a model for an amp on a specific setting, recorded by a specific mic in a specific position, it's difficult to create a generic model that varies along those parameters accurately. On the other hand, this "holistic" approach is based on minimal (if not zero) knowledge of what we're modeling: we need to know absolutely nothing of the amp. This is still very much in use -- think of IR -- that's exactly how that works. I cannot model a "space" or a cabinet, for the physical variables are impossible or very hard to master and take into account, so I just see how it reacts to an impulse. Or, think of the profiling of an amp that the Kemper does, that's the main idea behind it.

 

At any rate, it seems that amp modeling moved more towards the "reverse engineering" and modeling components -- alas, from my perspective, losing the snobbish ivory-tower approach of "I don't care how that works. that's for electricians". I believe the first time I heard Line6 mentioning something along those lines was for the HD line. It amazes me that this approach actually pays off -- given how an amp is a very complex system, trying to reproduce it as a whole, even approximately, would seem to have a better chance at capturing its essence than decomposing it into bits whose sum might diverge significantly from the actual thing.

 

So, not only am I wildly off topic, I am also wrong.

[willc]

Reading a lot in the L6 forums since quite some years, I did not write here so far. In this case I'd like to explore a little bid on improvements and developments.

I'm considering myself as a "Power User" in the way that I was very much going into details of sound creation, learning what the models and what the sw-controllers do. Creating lots of sounds using a big range of amp, cabinet, effect etc models. (appr 1000 single and dual sounds - mainly by song)

I don't doubt the quality of the development steps e.g. from POD X3 to HD and now to Helix, but I still look at it from what comes out in the studio and in the live PA. Comparing X3 to later units I always found ways to make the X3 sound close to the originals and really good enough in my mind. At the end of the day I'm still on X3 so far. Why?

- Maybe newcomers, who start with HD/Helix or those who did not intensively use previous model ranges will say "never mind" BUT I think,

- as a power user of x3 I would expect a high end successer unit to deliver downwards compatibility as Priority 1, while
  1a is, NOT to deliver less/worse, 1b is to deliver more/better than the predessessor. 

  The HD/Helix model range has dramatically decreased and even if some sounded awkward, there are many discontinued models which were great.

  I'm not ready to through away 10 years of my work on sounds just because L6 introduced a new technology which soon will be old.
  .. and if I have to, I'd be open to switch to competitors - although I'm now used to L6 concepts.

 

I wanted to switch to HD but didn't, now I would like to switch to the Helix Generation, but I can't redo all the work from scratch, finding new ways to make my sounds. If the L6 guys would find a way to import X3 models (and sounds) into the Helix at the same (not improved) quality, I'd be fine and go from there.

But for now, I got stuck!

[ozbadman]

So, not only am I wildly off topic, I am also wrong.

 

:P Best post ever.

[ozbadman]

- as a power user of x3 I would expect a high end successer unit to deliver downwards compatibility as Priority 1, while

  1a is, NOT to deliver less/worse, 1b is to deliver more/better than the predessessor.

 

While I am sure it's disappointing, newer products are often not backwards compatible with previous products. That includes product features, not just patches etc. This includes Windows no longer running some old DOS and XP stuff, my PC no longer having a DVD player, my new Roland mid-range V-drums not having all the features of the TD-12, my new girlfriend not... wait. I digress.

 

In the case of the X3, when the HD was released, the X3 had probably about 10 years of development behind it (it was XT based). Useful I'm sure, but old algorithms. L6 wanted to start from scratch, with a low-cost system. L6 specifically said a number of times when they released the HD series that it was not a direct replacement for the X3. Many features and amps, and some effects that were available on the X3 are still not available on the HD. But, the amps and effects that are on the HD sound much, much better.

 

If you're happy with your X3, keep it. The HD is not a direct replacement for a lot of X3 features. Just buy an HD or HELIX as well if you want it to sound good. :)

[PeterHamm]

- as a power user of x3 I would expect a high end successer unit to deliver downwards compatibility as Priority 1...

 

With respect... I am reasonably sure that "downward compatibility" would be insanely limiting on the ability to create a whole new far more realistic sound engine, so not only do I guess it wasn't priority 1, but am hopeful it will never even be priority 10.

 

The only way to successfully deliver downward compatibility in a unit like this is to cripple it sonically compared to what would be possible. If Line 6 did that, I guarantee you they would have had a flop on their hands.

 

there are a handful of situations where I think my old HD still is a great solution, and so for the short term I am keeping it, but when I want to make the best sound possible, the HD 500 is not the solution I'll reach for now.

 

By comparison, for the very brief time I had the X3 and HD side by side, I realized quickly that I would never ever want to use the X3 again, the HD was THAT much better. At least the HD still sounds good to my ears next to Helix.

 

But backward compatibility? No. Please. Don't ever give us that...

[HonestOpinion]

With respect... I am reasonably sure that "downward compatibility" would be insanely limiting on the ability to create a whole new far more realistic sound engine, so not only do I guess it wasn't priority 1, but am hopeful it will never even be priority 10.

 

The only way to successfully deliver downward compatibility in a unit like this is to cripple it sonically compared to what would be possible. If Line 6 did that, I guarantee you they would have had a flop on their hands.

 

there are a handful of situations where I think my old HD still is a great solution, and so for the short term I am keeping it, but when I want to make the best sound possible, the HD 500 is not the solution I'll reach for now.

 

By comparison, for the very brief time I had the X3 and HD side by side, I realized quickly that I would never ever want to use the X3 again, the HD was THAT much better. At least the HD still sounds good to my ears next to Helix.

 

But backward compatibility? No. Please. Don't ever give us that...

 

If the configuration and resource files from an older unit was available I suppose someone could write a translation "engine" that would translate old settings into new Helix ones. That would be a huge and time consuming project though and would be best left to an Open Source community that had the will and the manpower. i definitely would not want Line6 diverting their development resources to this. I think at least theoretically it could be done however, especially if the products were designed with this in mind from the ground up.

[d0stenning]

Well, if you are implying this gives a more accurate model, that assumes that (1) the spice model is correct and (2) the spice models, and all their interactions are identical to the interactions that are happening in the particular amp you are modeling, including the exact impedances (since resistors in any 1 specific amp have an exact value, not a value +- tolerance) and component interactions.

No that wasn't the point at all. I merely used the Revolver ( which is a respected sim - regardless of which is better ) blurb as an illustration of how sim marketers all seem to end up implying that their current version "actually" simulates down to component level - as if earlier versions didn't.   And if one looks back to the early versions of these sims - in the blurb they all say/suggest the same thing... "we emulate/simulate the circuitry" 

[d0stenning]

The question is very interesting beyond the "mere" amp modeling scenario, and more into an epistemological one.

 

I believe in older days "amp sim" had a more holistic approach. I don't care WHY something sounds the way it sounds, I just want to recreate its sound. Hence, you'd record the amp as a whole, and made sure the signal transformation from input to recorded output was well approximated (fitted) by your model.

 

I like the holistic approach, maybe because I come from artificial intelligence, and we gave up trying to make something "intelligent" a long time ago, trading for something that "behaves intelligently".

 

The holistic approach has a few disadvantages -- namely, that parameters are hard to implement: it's easy to create a model for an amp on a specific setting, recorded by a specific mic in a specific position, it's difficult to create a generic model that varies along those parameters accurately. On the other hand, this "holistic" approach is based on minimal (if not zero) knowledge of what we're modeling: we need to know absolutely nothing of the amp. This is still very much in use -- think of IR -- that's exactly how that works. I cannot model a "space" or a cabinet, for the physical variables are impossible or very hard to master and take into account, so I just see how it reacts to an impulse. Or, think of the profiling of an amp that the Kemper does, that's the main idea behind it.

 

At any rate, it seems that amp modeling moved more towards the "reverse engineering" and modeling components -- alas, from my perspective, losing the snobbish ivory-tower approach of "I don't care how that works. that's for electricians". I believe the first time I heard Line6 mentioning something along those lines was for the HD line. It amazes me that this approach actually pays off -- given how an amp is a very complex system, trying to reproduce it as a whole, even approximately, would seem to have a better chance at capturing its essence than decomposing it into bits whose sum might diverge significantly from the actual thing.

 

So, not only am I wildly off topic, I am also wrong.

You're not wildly off topic.... 

 

and what are you also wrong about ?   :)

 

Some friends call me "curious Dan"..... i'm just curious from an engineering standpoint.  I think simulation - whether in games, physics, VR and especially the realms of sound, be it acoustic instruments or amps.....  its all fascinating and amazing stuff. 

 

We live in a decade in which Einsteins theory has just been vindicated by a twin-black hole collision about a billion light years away ... AND ..... we get to play with virtual amps which most of us don't and could never own... 

 

What a great time to be alive.  HELIX and LIGO :)

[mileskb]

Indeed, and if Helix had four SHARC DSPs, or next-generation SHARCs, I'm sure we'd have no problem finding more granular elements of amps and pedals to model. I'll betcha $100 Ben and co. already have a list of things they'd love to do that Helix won't currently let them. No snake oil here.

 

It's not just about measuring stuff, tho'. You need a DSP and sound design team that understands why an amp or pedal component sounds and feels the way it does. Behavior modeling is much, much more complicated than measuring points in a circuit. Spreadsheets are easy; blending math and art is hard.

 

My posts are vague because I'm a product designer, not an engineer.

 

Slightly helpful insight might be had on Helix's blog page.

 

Without saying much DI really said a mouthful here that some folks have tic'd on, but it's worth repeating...

 

It's one thing to recreate components in digital form, it's something completely different to get those digital components to act as they do in the real world when they interact with one another.   I have been spending time creating some very simple patches... AMP + CAB and then an effect.  Maybe Reverb or Distortion, or Chorus etc..  Even WAH..  but only one or two effects on at a time.   And I've been playing with the effects  placement... before, parallel or after the amp and/or cab.  The way the blocks interact in some cases is just scarey accurate.  We've all played with turning the guitar up on certain amps and you can hear the clean-to-breakup/edge as you increase the volume... but other subtle stuff too like the way a the top of the WAH can cause a little breakup while the rest of the sweep stays clean.  Think the theme from Shaft style.  And even further, you can get just that breakup to affect a 60's Spring Reverb, just like it does in real life.   It borders on fuzzy logic or AI.   I know of no other modeler that acts like this to this detail.

[glideman]

Hey, Beavis...  he said "Theme From Shaft"...    aCk!!!   :D

 

Don't even think about getting into effects of temperature, etc. on components.   LoL!