Enhanced simulation of non linear CAB behavior

DvD

New Member
Hi all,

Defenitely the AXE-FX exells in the simulation of pre/power amps. It captures the essential subtleties one needs to get a good pre/power amp sound. I am extremely happy with my ULTRA.

Although it is a matter of opinion, I somehow am not sonically convinced of the CAB simulations. Surely, the impuls response method to capture the subtle frequency response of a CAB, is really great and can't be much better. Nevertheless, it only captures the linear frequency and phase behavior of a CAB.

Yes, there is a parameter that allows you to simulate speaker break-up characterics, however, I am not convinced. It seems like it just another drive type distortion stage. However, I could be wrong. The simulation of non linear characteristics could therefore be better, in my opinion.

Especially, when you deploy the AXE-FX in wide band flat response monitor situations, cabinet simulation is of utmost importance, and I would argue that it is as important in determining your overall sound as is the pre/power amp itself. Especially when you really krank it up and put a virtual 150 watts in your virtual cabinet, it will (and should) sound totally different then when you put a misely 10 watts in it. These non-linear characteristics are very subtle and may prove to be very complex to simulate, I haven't got a clue about that. I am sure Cliff has some ideas about this.

Recently, I came across a unit called Torpedo from a French company Two Note. In essence, it is a high-end digital stereo cabinet simulator for studio use, designed to be fed from a real power amp or from a pre-amp. It also utilises an impuls respons technique, but also claims to capture the "true" non-linear characteristics of a cabinet. I do not exactly know how it is done, my best guess is that several impuls responses at various levels are captured, and the unit "interpolates" between these IRs.

In essence, the Torpedo is "just" a DSP with simulation software, as is the AXE-FX. So anything the Torpedo can, the AXE can do also, provided with the proper software. So, here is my suggestion for upcoming AXE software releases. I would like to have the non linear aspects of the CAB simulations to be (significantly) enhanced.

Perhaps, Cliff would like to throw in his 0,02 cnts about this topic.

Greetz
Dave
 
DvD said:
Especially, when you deploy the AXE-FX in wide band flat response monitor situations, cabinet simulation is of utmost importance, and I would argue that it is as important in determining your overall sound as is the pre/power amp itself.
I agree with your general statement. I do not agree that nonlinear speaker behavior is a significant contributor to the sound of the vast majority of amp/cab combinations. Most speakers, even the cheap cone transducers typically used used in guitar cabs, produce lower levels of distortion than the amps being used with them.

I would further point out that whenever loudspeaker distortion is a significant contributor to the sound you hear, the loudspeaker is very close to its thermal limits. If it is operated continuously in this regime, it will fail in a relatively short time. The upshot of this is that very few guitar sounds rely on nonlinear speaker behavior for their character, popular myths to the contrary notwithstanding.
 
I have no intention of turning this into a big discussion, but I was referring to the non-linear behavior of a loudspeaker, not necessarily to the distortion as we know it from amps, and certainly not to the distortion when driven to its thermal limits. I have never found that sound to be of any use.

Non-linear subtleties are for example the frequency response of a heavily driven speaker significantly differing from a mildly driven one. I was also referring to the gradual changes in K2 and K3 harmonics spectrum when the load on the speaker is increased. Or the gradual (non-linear) compression that occurs when loading the speaker. This is not something happening when driven to its thermal maximum, it is already occuring much earlier.

I believe you are right that in comparision to a heavily (metal) overdriven pre/power amp it is sufficiently masked to be of no relevance. But what about a very mildly overdriven pre amp, but cranked up to a loud but fairly clean level, there speaker non-linearity is of importance. The best example I have is to just take your amp and cabinet, set it to a mild level, listen to it, then crank it up but keep the sound clean, and observe how the tonality changes. You can not attribute all of this to the pre and power amp, the speaker adds to it.

Rgrds
-Dave
 
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.
 
I was just going to saying to myself "Fletcher-Munson response curves" as I was reading this post! :cool:
Cliff is on top of it as usual.
IMO it is also worth considering the tonal response of the room that this speaker cabinet is in. The room plays an increasing role in the sound of the cabinet as the cabinet is turned up. It would be really difficult for the human brain to separate the non-linearity of the room response from the speaker/cab response to the increasing power.
 
Although it is a matter of opinion, I somehow am not sonically convinced of the CAB simulations. Surely, the impuls response method to capture the subtle frequency response of a CAB, is really great and can't be much better.
It can't be better actually but you can have a different result, more in the room and more clear.


I do not exactly know how it is done, my best guess is that several impuls responses at various levels are captured, and the unit "interpolates" between these IRs.
They provide an utility that allows users to "capture" their own IRs. This utility only uses one sweep at one level and nothing more.


BTW french owners of the Axe-Fx whose have tried the Torpedo have not been impressed. I remember one in particular saying that it didn't make any miracle.
 
Six String Symphony said:
IMO it is also worth considering the tonal response of the room that this speaker cabinet is in. The room plays an increasing role in the sound of the cabinet as the cabinet is turned up.
Worth considering? That might be the next big thing. L6 tries to do it but IMO just diffuses the sound. Even so, getting a FRFR at modest volume to sound like a guitar cab exciting a room is likely where things need to go. That's really what I was hoping for when I first ran across the "Air" param on the cab sims.
 
fremen said:

Thxs, I will defenitely check this out further. Nebula deploys a Volterra series type of convolution, which gives you the possibility of simulating non-linearity.

Anyhow, I was aware of the method, but not that someone already got so far with implementing it, in a freely usable package. I only was aware of this method to be used in some advanced room accoustics correction systems.

I am puzzled how they handle parametrization. Imagine only two parameters (that is, two knobs to dial in). One would have to measure reponses for each possible position of the knobs. It would result in a tremendous amount of data to store. Let alone the complexity of the algorithm.

Anyway, this method is perfect for enhanced (static) CAB simulation. It would be possible to not only simulate the CAB itself, but also the characteristics of the room it is sitting in at various excitation levels.

Rgds
-Dave
 
From what I've read about Nebula on various forums, it promises a lot but fails to deliver anything better than with static IR methods regarding cab simulation. That was months ago, it may have evolved since
 
aleclee said:
Six String Symphony said:
IMO it is also worth considering the tonal response of the room that this speaker cabinet is in. The room plays an increasing role in the sound of the cabinet as the cabinet is turned up.
Worth considering? That might be the next big thing.
Not really. In order to simulate a desired room, you first must come up with a way to cancel the acoustic environment in which you are playing. There are a number of companies that have claimed success at that over the years. There's a tiny fly in the ointment: it is physically impossible to do that. :shock:

When you play an Axe-Fx through an FRFR speaker, you are almost always in a room. There is therefore no need to simulate one, nor would that be possible with a rig that any guitar player could afford.

Even so, getting a FRFR at modest volume to sound like a guitar cab exciting a room is likely where things need to go.
That's already possible, and not at all difficult. I do it every day.
 
fremen said:
From what I've read about Nebula on various forums, it promises a lot but fails to deliver anything better than with static IR methods regarding cab simulation. That was months ago, it may have evolved since

I went through those posts and it seems they have come a long way. I will check it out further and if any good I will let you know. Just for fun, I will set something up in order to be able to compare, if the Voltarra kernel IRs produce convincingly better (soundwise) results than plain IRs. If not, the technique might be superior but in the end it would be just killing the bug with an elephant, as a matter of speech.

Rgds
-Dave
 
Jay Mitchell said:
When you play an Axe-Fx through an FRFR speaker, you are almost always in a room. There is therefore no need to simulate one, nor would that be possible with a rig that any guitar player could afford.
Yes and no. Thing is that many folks go to an Axe/FRFR combo to help manage volume. If you're not cranking the FRFR up to regular guitar amp volumes, then it's not interacting with the room the same way as a cranked guitar amp. The effect might be more subtle than the F-M curve but it's real and it's missing in many FRFR applications. What I'm talking about is taking a step away from the "miked cab on a recording" sound and back towards the "amp in a room" sound. Never mind that, when recording, adding some "room vibe" would help the device take another step towards sounding even more like a traditional amp in a room rather than depending on FX like reverb, delay, and modulation.
Jay Mitchell said:
Even so, getting a FRFR at modest volume to sound like a guitar cab exciting a room is likely where things need to go.
That's already possible, and not at all difficult. I do it every day.
You have me confused. On the one hand you say it's possible and you do it every day but above you claim that even it's neither desirable nor possible in an affordable rig. Can you elaborate a bit? As for how you accomplish the effect of a guitar cab exciting a room, are you doing it by driving your FRFR at guitar levels (not really modest volume, IMO) or via FX?
 
aleclee said:
Six String Symphony said:
IMO it is also worth considering the tonal response of the room that this speaker cabinet is in. The room plays an increasing role in the sound of the cabinet as the cabinet is turned up.
Worth considering? That might be the next big thing. L6 tries to do it but IMO just diffuses the sound. Even so, getting a FRFR at modest volume to sound like a guitar cab exciting a room is likely where things need to go. That's really what I was hoping for when I first ran across the "Air" param on the cab sims.
If you're referring to L6's A.I.R., it's nothing else but adding early reflections to their cab sims. No mic movement or anything like that involved. To do that you'd need, let's say the parameter goes from 0-100, 100 different IRs for each step. Can be done, yes, but very tedious.

You can do the exact same thing in the Axe by putting a Reverb right after the cab.
 
aleclee said:
If you're not cranking the FRFR up to regular guitar amp volumes, then it's not interacting with the room the same way as a cranked guitar amp.
First, let's make clear what is really different at low volumes: the interaction between amp and room does not change with volume - room effects are very linear - but human hearing does change. At very low volumes, reflected energy may lie below your hearing threshold. This effect becomes more pronounced as you age (and as you damage your hearing by playing too loud).

The effect might be more subtle than the F-M curve but it's real and it's missing in many FRFR applications.
When it is missing, it has nothing to do with FRFR. It would still be missing if you were playing through a guitar cab at the same volume.

What I'm talking about is taking a step away from the "miked cab on a recording" sound and back towards the "amp in a room" sound.
My Axe-Fx/FRFR rig sounds exactly like an "amp in the room." I occasionally do A/B comparisons with one of my tube amps, and there is nothing missing from the Axe-Fx rig.

Never mind that, when recording, adding some "room vibe" would help the device take another step towards sounding even more like a traditional amp in a room rather than depending on FX like reverb, delay, and modulation.
Room reflections are the province of reverb, and I've got reverb settings that do an excellent job of creating the illusion of an acoustic space when I record direct.

As for how you accomplish the effect of a guitar cab exciting a room, are you doing it by driving your FRFR at guitar levels (not really modest volume, IMO) or via FX?
A good FRFR system will excite a room in similar manner to a closed guitar cab. If you really want the cab + room effect of an open cab (I never do), you can easily do that with an added rearward-firing cab, which need not be FRFR. Just operate it reverse polarity and set its level to taste. Instant acoustic dipole.

The quality of the IR and the monitor are the two most important ingredients in success in this endeavor. Close-mic'ed IRs are not the way to go.
 
bmi said:
Although it is a matter of opinion, I somehow am not sonically convinced of the CAB simulations. Surely, the impuls response method to capture the subtle frequency response of a CAB, is really great and can't be much better.
It can't be better actually but you can have a different result, more in the room and more clear.


[quote:3p19oucq]I do not exactly know how it is done, my best guess is that several impuls responses at various levels are captured, and the unit "interpolates" between these IRs.
They provide an utility that allows users to "capture" their own IRs. This utility only uses one sweep at one level and nothing more.


BTW french owners of the Axe-Fx whose have tried the Torpedo have not been impressed. I remember one in particular saying that it didn't make any miracle.[/quote:3p19oucq]
If it only uses one sine sweep, it may be Volterra convolution but not dynamic convolution. I wrote a post here a while ago about the specifics of Volterra convolution. Basically a Volterra kernel is an extension of the concept an impulse response to a class of nonlinear systems. It's very straightforward to take a logarithmic sine sweep and extract the diagonal Volterra kernels from it.

Dynamic convolution, like Cliff said, uses a series of impulse responses (so they're linear), and some nonlinearity comes from switching between them based on the input signal level.

Nebula I believe uses both -- it uses dynamic convolution, but it switches between diagonal Volterra kernels rather than just impulse responses. This is even more computationally intensive.
 
Back
Top Bottom