Question for Cliff about the IR Technology and the future of Cab Simulation

SynFactory

SynFactory

Inspired
Cliff,
I've read somewhere that you guys wanted to have a new Cab Simulation algorithm but it was too heavy on the CPU at that time (maybe it was back when there are only the Standard and The Ultra).
Don't get me wrong, I really like the IR technology, the work of the guys at Ownhammer and Redwirez is really great, but at least IMHO the IR thing is the only non "proprietary" and outdated algorithm on the Axe FX.
There are some little problems inherent the IR that let me feel that the amp and the cab are not "connected" together.
Frequencies that get stucked and whistle, a static feel that do not let the cab breathe in conjunction with the master volume (breakup) and output frequencies that are too far from the real world (too low or too high).
I know that it will be a really hard process and maybe it will be really heavy on the CPU, because there will be a lot of factors to add, the wood used, the size of the cab, the speakers, the mic, the preamp and most important of all the "Air".

I wrote you this because I think that the cab+mic+pre is at least IMHO 70-75% of the sound and a solution to this will be the next step to perfection of the entire system.

Hope that you understand what I say, because speaking of technical things in another language can be hard :)

Cheers
 
SynFactory said:
Cliff,
I've read somewhere that you guys wanted to have a new Cab Simulation algorithm but it was too heavy on the CPU at that time (maybe it was back when there are only the Standard and The Ultra).
Don't get me wrong, I really like the IR technology, the work of the guys at Ownhammer and Redwirez is really great, but at least IMHO the IR thing is the only non "proprietary" and outdated algorithm on the Axe FX.
There are some little problems inherent the IR that let me feel that the amp and the cab are not "connected" together.
Frequencies that get stucked and whistle, a static feel that do not let the cab breathe in conjunction with the master volume (breakup) and output frequencies that are too far from the real world (too low or too high).
I know that it will be a really hard process and maybe it will be really heavy on the CPU, because there will be a lot of factors to add, the wood used, the size of the cab, the speakers, the mic, the preamp and most important of all the "Air".

I wrote you this because I think that the cab+mic+pre is at least IMHO 70-75% of the sound and a solution to this will be the next step to perfection of the entire system.

Hope that you understand what I say, because speaking of technical things in another language can be hard :)

Cheers
Click to expand...

I think the cab sim. technology outputs excellent results.

Too much low/high output is not caused by the IR technology. It's the result of dealing with close-mic'd cabs, as opposed to listening to real cabs (no mic).
Breakup is handled using the Drive parameter in the Cab block.
Air is handled using the room simulation parameters.
Stuck frequencies?
 
"There are some little problems inherent the IR that let me feel that the amp and the cab are not "connected" together."

This is accounted for in the amp block (look at the low and high freq resonance parameters).

from the manual:

"LF RESONANCE, LF RES FREQ, LF RES Q – Guitar loudspeakers have a low-frequency resonance,
typically about 100 Hz. This typically shifts up slightly when the speaker is mounted in an enclosure. This
resonance causes an increase in the power amplifier response due to the finite output impedance of the
power amp. These parameters set the magnitude of the resonance and its center frequency and Q, or width.
HF RESONANCE, HF RES FREQ, HF RES Q – A loudspeaker voice-coil presents an inductive load to the
power amp at high frequencies. This inductive load, in conjunction with the output transformer capacitance,
creates a high-frequency resonance. This resonance causes an increase in the power amplifier response due to
the finite output impedance of the power amp. These parameters set the magnitude of the resonance, its
center frequency and Q, or width."

Also the cab size was added in the last update.
 
yek said:
I think the cab sim. technology outputs excellent results.

Too much low/high output is not caused by the IR technology. It's the result of dealing with close-mic'd cabs, as opposed to listening to real cabs (no mic).
Breakup is handled using the Drive parameter in the Cab block.
Air is handled using the room simulation parameters.
Stuck frequencies?
Click to expand...

Ok let me further explain. And let me say that you know better than me about the subject, I'm not a sound or software engineer and I don't have a II at the moment.
About the IR. Using test tones that goes from 0 to 20000 HZ seems a little bit strange for a really limited device like a guitar cab that usually goes from 60-80hz to 5k-6k. It's this approach that let me think that the technology is a little bit strange, guitars don't produce sine or pink noise sweeps.
I know it's a way to evaluate the cab frequencies spectrum but this is the reason why IMO it sounds a bit strange.
You say that the problem is inherent to the fact that using close mics cause that.
Yes, in some ways, but I think that the cab in the beginning didn't produce those extreme frequencies, the mic only translate what the cab output is, it didn't create frequencies that aren't there.

As far as Drive and Room parameters, I don't have the II so I cannot comment on the late one, but I think that the problem is at the beginning.
An impulse is a "photo" of a cab in an exact moment of time, breakup is how a cab/speaker reacts to different solicitations and "stress".

It's a bit odd because the Impulse is static and these algorithms are trying to recreate the dynamic "movement" of the speakers.
For me it's like convert a 44100 HZ file to 96000 HZ. Doesn't make much sense.

For the "Stuck Frequencies" I mean strange, resonant peaks that occur sometimes in a recorded tone, especially relevant with high gain sounds, that creates harsh, hollow and clashing frequencies especially when playing more than one note at the time.

Maybe with an example it will be easier to understand what I say.

Mesa Boogie Dual Rectifier - Axe FX comparison - YouTube

I linked this video of the Axe FX because I think that the "base" sound is almost identical but when it plays those power chords and let them ring there is a difference. It starts around 1:00 with the real rectifier and cab and then around 1:26 it plays with the Axe FX. I know that there are other variables like different cab, mic etc. but as I said I noticed this in my previous Axe Standard patches too and I think that was caused by the IR.

Hope that this make sense, and let me say that I've asked this because I've read that phrase from Cliff in the past, but I'm not able to find it. I will search it again and see what I found.
 
javajunkie said:
"There are some little problems inherent the IR that let me feel that the amp and the cab are not "connected" together."

This is accounted for in the amp block (look at the low and high freq resonance parameters).

from the manual:

"LF RESONANCE, LF RES FREQ, LF RES Q – Guitar loudspeakers have a low-frequency resonance,
typically about 100 Hz. This typically shifts up slightly when the speaker is mounted in an enclosure. This
resonance causes an increase in the power amplifier response due to the finite output impedance of the
power amp. These parameters set the magnitude of the resonance and its center frequency and Q, or width.
HF RESONANCE, HF RES FREQ, HF RES Q – A loudspeaker voice-coil presents an inductive load to the
power amp at high frequencies. This inductive load, in conjunction with the output transformer capacitance,
creates a high-frequency resonance. This resonance causes an increase in the power amplifier response due to
the finite output impedance of the power amp. These parameters set the magnitude of the resonance, its
center frequency and Q, or width."

Also the cab size was added in the last update.
Click to expand...

Sean, thanks for the reply!
But I don't understand one thing that maybe is misleading me.
It's a "cab" characteristic but it's in the "amp" block.
So is it something inherent to the Amp used or the Cab?
 
i think in regards to the amp block being involved, it just showcases that the Axe Fx makes us rethink our old school way of making tones. It isn't just static amp into static cab, maybe with some static pedals in front. Now, it is an interaction between many adjustable things that allow someone to mold their tone anyway they want.

The last thing I have done, and I have ZERO desire to do, is compare axe to a real amp. I don't care if my axe sounds 100% like a triple rectifier, or shiva, or fender twin....i just want it to sound freaking awesome and give me the sound i hear in my head. And by adjusting settings in amp block, to dial in a better sound, is the same thing to me as dialing in a cab block to get the sound. I just want the sound :).

I think sometimes we get caught up in trying to use the axe 100% like a real amp/cab/pedals set-up, and not fully utilizing the power of creation. I fell into that trap with my ultra, which is probably why i didn't connect that well with it. With the axe-II i decided to force myself to change my ideas about tone creation, and I have been loving tones left and right.

just my $.02, but once i came to these terms my love of the Axe went up exponentially
 
The interaction between a tube amp and the speaker occurs naturally, there is just no way to change these parameters that are available in the advanced amp block when you are playing through a traditional tube amp / speaker. The Axe-FX enables you to adjust these parameters that are not immediately available in the real world.

KG
 
SynFactory said:
Sean, thanks for the reply!But I don't understand one thing that maybe is misleading me.It's a "cab" characteristic but it's in the "amp" block.So is it something inherent to the Amp used or the Cab?
Click to expand...
it is an interaction between the cab and amp. it is an effect the cab has on an amp. Also, speaker drive has been moved to the amp block as well because it causes aliasing in the cab block.as was said, another reason for putting those in the amp block is if you use a real cab, it is the only way for the virtual amp to react like it had a particular cab assignec to it.it would be nice if we had a listing of resonances for particular cabs.i think what you are asking for is a physical model of a cab, mic, and room. i dont see that happening with this generation of axefx. that is alot of processing. you dont need to do that to get accurate results. you are assuming whatever shortcoming you are hearing are a result of using impulse responses for cab simulation. i dont think the case has been made that, that is the cause for the shortcomings you are experiencing.
 
Not trying to be a jerk, but you say you don't have an AFII yet. So how can you be so sure that these problems you encountered in the past, still exist with the AFII, the new algorithms, new modelling and most importantly all the new parameters that you can tweak?
I mean, maybe you should just wait to receive your AFII, learn how to use it and after that, if you still have a problem with the cab emulations, begin a thread about it...
I'm just saying cause from what I've read in most users' comments and from what I've heard in various clips, I don't think that there really is a problem here... It's just a matter of tweaking and taste me thinks...
 
Impulse responses model the linear characteristics of a speaker/cab. Distortion and breakup are non-linear and cannot be captured by an IR.

An impulse, at least in theory, contains all frequencies. It is easier to generate a sweep and use deconvolution to get equivalent results; for audio devices where we don't care about frequencies outside the range of human hearing, it is common to limit the sweep to 20-20 kHz, but that does not imply anything about the frequency response of the item being analyzed.

Plus it is common to filter the results further. With the Axe there is often some additional processing applied that transforms the IR into a minimum phase signal. This may account for some of what you are hearing.
 
Last edited:
javajunkie said:
it is an interaction between the cab and amp. it is an effect the cab has on an amp. Also, speaker drive has been moved to the amp block as well because it causes aliasing in the cab block.as was said, another reason for putting those in the amp block is if you use a real cab, it is the only way for the virtual amp to react like it had a particular cab assignec to it.it would be nice if we had a listing of resonances for particular cabs.i think what you are asking for is a physical model of a cab, mic, and room. i dont see that happening with this generation of axefx. that is alot of processing. you dont need to do that to get accurate results. you are assuming whatever shortcoming you are hearing are a result of using impulse responses for cab simulation. i dont think the case has been made that, that is the cause for the shortcomings you are experiencing.
Click to expand...

Thanks for the explanation Sean, now it makes sense. Yeah, that's what I've asked if it was possible, a physical model of cab+mic+room.
Do you think that a some sort of multilayered Impulse Response can be made to at least trying to replicate the breakup? (using impulses at different volumes).
I think that it is what Nebula programs are doing if I'm not wrong.

And I'm assuming that the IR is the problem because I've also encountered this with a tube amp head+loadbox+IR.
So it's not a problem of the Axe FX or its ability to emulate amps.

johna156 said:
Not trying to be a jerk, but you say you don't have an AFII yet. So how can you be so sure that these problems you encountered in the past, still exist with the AFII, the new algorithms, new modelling and most importantly all the new parameters that you can tweak?
I mean, maybe you should just wait to receive your AFII, learn how to use it and after that, if you still have a problem with the cab emulations, begin a thread about it...
I'm just saying cause from what I've read in most users' comments and from what I've heard in various clips, I don't think that there really is a problem here... It's just a matter of tweaking and taste me thinks...
Click to expand...

What part of "I noticed this in my previous Axe Standard patches too and I think that was caused by the IR" you don't understand?
Since they use the same technology to cab emulation that's why I'm asking.
Glad you're trying not to be a jerk, maybe try to be a better reader next time.

LMO said:
Impulse responses model the linear characteristics of a speaker/cab. Distortion and breakup are non-linear and cannot be captured by an IR.An impulse, at least in theory, contains all frequencies. It is easier to generate a sweep and use deconvolution to get equivalent results; for audio devices where we don't care about frequencies outside the range of human hearing, it is common to limit the sweep to 20-20 kHz, but that does not imply anything about the frequency response of the item being analyzed. Plus it is common to filter the results to a further. With the Axe there is often some additional processing applied that transforms the IR into a minimum phase signal. This may account for some of what you are hearing.
Click to expand...

Yes, that's why I've asked what are these new ideas Cliff thinked of.
 
LMO said:
With the Axe there is often some additional processing applied that transforms the IR into a minimum phase signal. This may account for some of what you are hearing.
Click to expand...

Minimum-phase processing, by definition, does not alter the frequency response, only the phase. Since speakers are inherently minimum-phase (or very near MP) to begin with all the minimum-phase processing does is "trim" the IR automatically. It is, therefore, inaudible. The benefit of MP processing is that multiple IRs can them be combined without worry of phase problems.
 
Here was what Cliif stated earlier about the next generation of cab modeling. I don't think this will make into the 2 though:

"The technique for simulating that is dynamic convolution and I've been working on that for a few years. It is beyond the scope of the Axe-Fx though as it requires a LOT of memory.

The technique involves sampling the DUT at multiple excitation levels and recording the impulse response for each level. The actual convolution then interpolates between responses based on the input signal level. Typically you use, say, 256 excitation levels which means you generate 256 impulse responses. That means one cabinet model requires 256 times the memory than regular convolution.

The drawback of dynamic convolution is that it is not parametric. Therefore you have no control over the model.

I'm not sure if the Torpedo uses dynamic convolution or not. I would hope so given its price but I don't see anything in the literature alluding to that.

I would be interested in hearing a comparison using dynamic convolution to regular convolution to see if the difference is noticeable. I've done tests using just a few excitation levels and, to my ears, haven't noticed any difference.

I believe that what you are hearing when turning the volume up is far more Fletcher-Munson related than any nonlinear behavior."
 
SynFactory said:
Thanks for the explanation Sean, now it makes sense. Yeah, that's what I've asked if it was possible, a physical model of cab+mic+room.
Do you think that a some sort of multilayered Impulse Response can be made to at least trying to replicate the breakup? (using impulses at different volumes).
I think that it is what Nebula programs are doing if I'm not wrong.

And I'm assuming that the IR is the problem because I've also encountered this with a tube amp head+loadbox+IR.
So it's not a problem of the Axe FX or its ability to emulate amps.



What part of "I noticed this in my previous Axe Standard patches too and I think that was caused by the IR" you don't understand?
Since they use the same technology to cab emulation that's why I'm asking.
Glad you're trying not to be a jerk, maybe try to be a better reader next time.



Yes, that's why I've asked what are these new ideas Cliff thinked of.
Click to expand...


Were you using a reactive load box?
Does Nebula have quite a bit of latency? I only used it briefly when it first came out so I don't remember. I know it uses significant cpu.
 
javajunkie said:
Here was what Cliif stated earlier about the next generation of cab modeling. I don't think this will make into the 2 though:

"The technique for simulating that is dynamic convolution and I've been working on that for a few years. It is beyond the scope of the Axe-Fx though as it requires a LOT of memory.

The technique involves sampling the DUT at multiple excitation levels and recording the impulse response for each level. The actual convolution then interpolates between responses based on the input signal level. Typically you use, say, 256 excitation levels which means you generate 256 impulse responses. That means one cabinet model requires 256 times the memory than regular convolution.

The drawback of dynamic convolution is that it is not parametric. Therefore you have no control over the model.

I'm not sure if the Torpedo uses dynamic convolution or not. I would hope so given its price but I don't see anything in the literature alluding to that.

I would be interested in hearing a comparison using dynamic convolution to regular convolution to see if the difference is noticeable. I've done tests using just a few excitation levels and, to my ears, haven't noticed any difference.

I believe that what you are hearing when turning the volume up is far more Fletcher-Munson related than any nonlinear behavior."
Click to expand...

Yes. that was it, finally I can read it again :)


javajunkie said:
Were you using a reactive load box?
Does Nebula have quite a bit of latency? I only used it briefly when it first came out so I don't remember. I know it uses significant cpu.
Click to expand...

Yes, I've used a 4 OHM, THD Hotplate Load Box with my previous amp.
There are only few free nebula programs, I know that Ownhammer also does some of these but you have to pay the full software to buy 3rd party programs and it's quite expensive, like 200 euro (on mac there is only version 3, version 2 for PC is cheaper).
Given the fact that on my Quad Core Mac I was able to instantiate on a blank project just 2-3 instances of the plugin, and then it starts to crackle and being not really responsive, sure it was not for me.

There are a couple of interesting tests done by a really good guy on Ultimate Metal forum to listen if you want:

The ultimate IR, NEBULA, REAL DEAL test - Ultimate Metal Forum
Part 2 of "The ultimate IR, NEBULA, REAL DEAL test" - Ultimate Metal Forum

javajunkie said:
Here is a discussion about physical modelling of cabs:
http://forum.fractalaudio.com/ultra-std-wish-list/24908-new-approach-cab-modeling.html#post390416
Click to expand...

Thank you, I will surely read it!
 
I'm only saying that you should try it FIRST before deciding that the AFII has the same "problem" with the gen I and therefor needs better IR technology or whatever.
If you're so touchy about it though, forget about it...
 
"Yes, I've used a 4 OHM, THD Hotplate Load Box with my previous amp."

no those are not reactive load boxes. They put a fixed load on your amp. Reactive loads keep the speaker/amp impedance and inductive/capacitive reactance relationship intact.
 
The Axe-Fx II models speaker distortion in the Amp block. There is a Speaker Drive parameter that allows you to set the desired amount of speaker distortion. IMO, the high-quality impulses available combined with the speaker distortion modeling obviate the need for anything more advanced. I am willing to be proven wrong if someone wants to record a sound clip using the Axe-Fx II built-in cab modeling vs., say, Nebula.

I recently did a study of speaker frequency response vs. drive level. Even at very high drive levels the IR is virtually identical. This proves that the fundamental frequency response of the speaker is not a function of drive level and, therefore, speaker distortion can be modeled separately.
 
javajunkie said:
"Yes, I've used a 4 OHM, THD Hotplate Load Box with my previous amp."

no those are not reactive load boxes. They put a fixed load on your amp. Reactive loads keep the speaker/amp impedance and inductive/capacitive reactance relationship intact.
Click to expand...

My fault, didn't know that

FractalAudio said:
The Axe-Fx II models speaker distortion in the Amp block. There is a Speaker Drive parameter that allows you to set the desired amount of speaker distortion. IMO, the high-quality impulses available combined with the speaker distortion modeling obviate the need for anything more advanced. I am willing to be proven wrong if someone wants to record a sound clip using the Axe-Fx II built-in cab modeling vs., say, Nebula.

I recently did a study of speaker frequency response vs. drive level. Even at very high drive levels the IR is virtually identical. This proves that the fundamental frequency response of the speaker is not a function of drive level and, therefore, speaker distortion can be modeled separately.
Click to expand...

Thanks for the answer. I have to say that with the free programs, there is one based on a Marshall cab, I didn't hear a difference at all.
I found this link from one of the posts that i put above on Ultimate Metal forum to download an IR and a Free Nebula program of the same cab:

Kalthallen-IMW-Impulse_2.zip - 4shared.com - online file sharing and storage - download

I don't have an Axe II at the moment so maybe someone else could give it a try, the Nebula plugin for the free programs it's, well, "free" and you can download it here:

Nebula3 Free Bundle
 
You must log in or register to reply here.
Back
Top Bottom