How much do modelers really "model"?

Ignored Member

Ignored Member

Neutered male
I was reading through some promo literature on modeling that talked about how they "modeled" the behavior of valves.

My BS detector went off and caused me to question how much of this stuff is really "modeling".

When someone says they are modeling the behavior of a valve amp, I think about transformer sag, impedance variances from the speaker, all those things that really get to what makes a tube amp respond the way it does.

I find it really hard to believe that a small processor could do that in any meaningful way without introducing noticeable latency. That seems like a lot of math.

Now carefully-tuned EQ? That I believe.

Does anybody have any inside info on this? To what extent are these devices modeling the behavior of components versus just adjusting their EQ behavior to be like those components?
 
When you consider that smart phones are more capable than the computers that ran the shuttle, is it that hard to beleive? :tongue:
 
Punchy said:
When you consider that smart phones are more capable than the computers that ran the shuttle, is it that hard to beleive? :tongue:
Click to expand...

Exactly.

How accurate they are is debatable, and varies by device and manufacturer.

But the processing speed and capability is WAY beyond what a human could interpret as latency.

The signal goes through those chips, and processed in less time than it takes for the original signal to even get to them from the guitar.
 
USian Pie said:
I find it really hard to believe that a small processor could do that in any meaningful way without introducing noticeable latency. That seems like a lot of math.
Click to expand...

Yeah, but it's analog math. :grin:



The first computers used valves, now that I think about it. :old:

I really am not versed in what's going on technically with modeling, but I imagine it could be an application of solid state devices, programmaticly controlled digitally, so that the impact on the signal happens in real time, without latency. As long as you don't put software in the path of the electrons, no A/D conversion malarkey, I think it's an achievable goal.
Does that sound reasonable?
I wish I had the Powerball winnings to afford myself the time to research this and give you a definitive answer, but not this week. Looking forward to more informed responses. Go.
 
I have my doubts too, but I have nothing to base that on beyond my own experience: I own a POD and have played around with modelling amps that others own.

Running a really hot pedal into a POD does not produce the same result as running it into an actual amp, and if it's not handling that properly, then how is it "modelling" an amp?
 
Tralfaz said:
I have my doubts too, but I have nothing to base that on beyond my own experience: I own a POD and have played around with modelling amps that others own.

Running a really hot pedal into a POD does not produce the same result as running it into an actual amp, and if it's not handling that properly, then how is it "modelling" an amp?
Click to expand...

Of course it isn't going to react to a pedal the same way an amp does, and that's kind of an absurd comparison since it's not designed to. Typically, an "Amp Modeller" is part of unit that models other devices as well, such as distortion/overdrive, modulation, delay etc... The "Amp" model is the last in line to be processed, and it is dependent upon the other "models" in the chain before it. Putting a pedal before it instead of providing it its own "model" of that pedal is like introducing a foreign antibody. It will not know how to "properly" respond.
 
Not sure how much modeling is enough but, I used to gig for years and played through a LOT of amps, back when they weren't vintage..... And if I had todays modeling technology.....THEN, I would had used it... Are modelers exact dups of vintage amps? NO...THANK GOD! I like the modelers better... they are at least consistent from day to day. Don't have to re-tube... and then chasing the tone as tubes age.

Modelers do a great job where it really matters.

AND you don't have to own $100k of amps to get so many cool amps under one foot.
Just imagine hauling all those amps to a gig....
 
Ya know, a lot of 'tone' in amps comes from inefficiency. That's hard to do in binary. Just sayin'...
 
altoidman said:
Ya know, a lot of 'tone' in amps comes from inefficiency. That's hard to do in binary. Just sayin'...
Click to expand...

There would logically follow, I'm assuming, from the concept of "artificial intelligence," its corollary, "artificial stupidity," which you could then program into all sorts of devices. :shrug: :lol:
 
OGG said:
Of course it isn't going to react to a pedal the same way an amp does, and that's kind of an absurd comparison since it's not designed to. Typically, an "Amp Modeller" is part of unit that models other devices as well, such as distortion/overdrive, modulation, delay etc... The "Amp" model is the last in line to be processed, and it is dependent upon the other "models" in the chain before it. Putting a pedal before it instead of providing it its own "model" of that pedal is like introducing a foreign antibody. It will not know how to "properly" respond.
Click to expand...

I don't think it's an absurd expectation at all. You're right that amp modelers can't handle that very well -- they're not designed to -- but they're not very useful to me without that ability since the built-in effects models are generally not to my liking.

And I would hope that eventually technology and ingenuity gets us closer to a modeling amp that can pass a guitarist's "Turing Test," so to speak, pedals and all. Maybe add in the smell of the tubes glowing and that echoey twang from the reverb spring when you pound on the top.
 
Well, they passed my test many, many years ago... right about the time I realized that I could have a good, consistent and reliable tone for both gigging and recording.

Many top name players have realized this as well. Which is why you see an ever increasing number of touring acts with no amps on the stage.

The fact of the matter, is that all of those little inefficient imperfections and subtle nuances that make tube amps so desire able to players, are almost always totally lost on the audience.

And even in the studio, by the time a record is compressed and mixed/mastered, most listeners wouldn't know the difference, or be able to pick out that little tidbit of tone.

Sure, there are exceptions, some music is more open and dynamic, allowing for subtlety to come through, and some listeners are more savvy than others. But, at a live rock show, at high volume... not so much.

If at all.

I for one believe that a great deal of amp love is either ego driven, or psycho-sematic.
 
A couple of things I should clarify:

1. I'm not saying modelers don't sound as good as tube amps. I've played tube amps that sounded like crap. All that great soft clipping and transformer sag isn't cool anymore when a tube goes microphonic.

2. Before going down the management route, my professional background was in programming. The computer science part of this is interesting to me. When someone says they are "modeling" something, I wonder what elements they are including in their algorithm. Some things can be made extremely detailed and complex but they require more processing time to complete. As fast as computers are today, there are realistic limits to how complex the algorithm can be before the lag is noticeable. That's what I'm curious about -- how complex are the models really?
 
USian Pie said:
I was reading through some promo literature on modeling that talked about how they "modeled" the behavior of valves.

My BS detector went off and caused me to question how much of this stuff is really "modeling".

When someone says they are modeling the behavior of a valve amp, I think about transformer sag, impedance variances from the speaker, all those things that really get to what makes a tube amp respond the way it does.

I find it really hard to believe that a small processor could do that in any meaningful way without introducing noticeable latency. That seems like a lot of math.

Now carefully-tuned EQ? That I believe.

Does anybody have any inside info on this? To what extent are these devices modeling the behavior of components versus just adjusting their EQ behavior to be like those components?
Click to expand...

The modellers I've looked at (Line6 POD HD and AxeFX II) have a mind-numbing array of adjustable parameters for each amp. Transformer sag, swapping speakers, tube bias, and Class A vs. A/B all immediately come to mind, but there are a lot more than that. The processors they have in them are pretty impressive. The latency is negligible IMO. I don't find it hard to believe at all, but I've actually had my hands on some of them.

Tralfaz said:
I have my doubts too, but I have nothing to base that on beyond my own experience: I own a POD and have played around with modelling amps that others own.

Running a really hot pedal into a POD does not produce the same result as running it into an actual amp, and if it's not handling that properly, then how is it "modelling" an amp?
Click to expand...

If you're clipping the input of a digital amp, then the modelling doesn't even have a chance to come into play before you're ruining the signal. You have to set up pedals in a way that works well with the modeller you are using. I use a bunch of dirt pedals in front of my POD HD500 with great results. It's even with the same settings I use on my amps. But maybe I use my pedals in a different way than you do.
 
If we are only asking if modern modellers have powerful and fast enough processors on board to do their business without noticeable lag, then the simple answer is...

Yes.

As complex as the algorithms may be, they are easily dispatched by the hardware.

Digital effects aren't exactly "new". They have been around for over 30 years now.

Some of the very first digital effect pedals and even rack mounts were capable of doing what they did without perceptible lag or latency, and they usually did so with a single main processor that was, itself, generally not at the state of the art.

Modern units employ multitudes of processors and sub-processors that are many multitudes more powerful and many multitudes faster.

And, they still mostly use hardware that would be considered beyond its shelf life in most other computing applications.
 
Tralfaz said:
I don't think it's an absurd expectation at all. You're right that amp modelers can't handle that very well -- they're not designed to -- but they're not very useful to me without that ability since the built-in effects models are generally not to my liking.

And I would hope that eventually technology and ingenuity gets us closer to a modeling amp that can pass a guitarist's "Turing Test," so to speak, pedals and all. Maybe add in the smell of the tubes glowing and that echoey twang from the reverb spring when you pound on the top.
Click to expand...

I see arguments like this as a variation of "I've adapted to conventional amplifiers & have developed a process to get the tones I want however I am unwilling to adapt this process to a different platform". I get that your last sentence is a bit tongue in cheek but it does reflect the mindset of many guitarists w/r/t modeling products; if they don't replicate every single aspect of an amplifier, tonally. physically & "spiritually", they just aren't good enough.
 
USian Pie said:
2. Before going down the management route, my professional background was in programming. The computer science part of this is interesting to me. When someone says they are "modeling" something, I wonder what elements they are including in their algorithm. Some things can be made extremely detailed and complex but they require more processing time to complete. As fast as computers are today, there are realistic limits to how complex the algorithm can be before the lag is noticeable. That's what I'm curious about -- how complex are the models really?
Click to expand...

I don't know if any of us here have enough details on the actual level of complexity of high end modelers like the AxeFx and Kemper to really answer your questions thoroughly. AxeFx commonly is cited as having 1ms of processing latency, which roughly equals having your ear about a foot away from a speaker. I'd bet that if you head over to the forums dedicated to these devices, you'll find some more technical discussion on their designs.
 
fused said:
Not sure how much modeling is enough but, I used to gig for years and played through a LOT of amps, back when they weren't vintage..... And if I had todays modeling technology.....THEN, I would had used it... Are modelers exact dups of vintage amps? NO...THANK GOD! I like the modelers better... they are at least consistent from day to day. Don't have to re-tube... and then chasing the tone as tubes age.

Modelers do a great job where it really matters.

AND you don't have to own $100k of amps to get so many cool amps under one foot.
Just imagine hauling all those amps to a gig....
Click to expand...

I am with you on this!
 
I think I may differ from the crowd on this whole modeller thing.
withh the Pod X3 Live, what I wanted was a large variety of amps/cabs/effects.
and that's just what I got.

having never actually played through 99% of the modelled amps, I haven't the faintest idea if they're close or not.
and I couldn't care less.

I'm able to get the sounds I want, and that's what matters to me.

and at a lower cost of what the effects alone would have run me.

add in the fact I can plug it into my pc for recording, and it was a win/win for me.

YMMV, of course.
 
You must log in or register to reply here.
Back
Top