Something Cool I've Been Working On

[AlbertA]

Out of curiosity about the Axe I really would like to know what happends in the Cab block.
I totally understand if that is secret though!

It's mainly just convolution Convolution - Wikipedia, the free encyclopedia

 

[FractalAudio]

The cab block is direct convolution. You can't do it in the frequency domain via FFT as that would add excessive latency. All modelers use direct convolution for this reason.

 

[Andrew Male]

Forgive my total noob question, but it's something I've always wondered. Does a loudspeaker's frequency response or compression characteristics change depending on amplitude? e.g. if a speaker is driven hard does the response/compression change?

If so do IRs capture this? Or are IRs only representative of the particular volume setting that the speaker at the time was asked to produce when capturing the IR?

Just saw this paper, but this seems like a very complex simulation of a loudspeaker, its materials and components.

http://www.klippel.de/uploads/media/Prediction_of_speaker_performance_at_high_amplitudes_01.pdf

 

[barhrecords]

Forgive my total noob question, but it's something I've always wondered. Does a loudspeaker's frequency response or compression characteristics change depending on amplitude? e.g. if a speaker is driven hard does the response/compression change?

If so do IRs capture this? Or are IRs only representative of the particular volume setting that the speaker at the time was asked to produce when capturing the IR?

Just saw this paper, but this seems like a very complex simulation of a loudspeaker, its materials and components.

http://www.klippel.de/uploads/media/Prediction_of_speaker_performance_at_high_amplitudes_01.pdf

An IR does not contain non-linear information like compression or distortion.

I believe some of the AxeFx CAB block parameters (in addition to the IR) are designed to add some non-linear speaker / cabinet behavior though.

 

[Andrew Male]

OK I understand, I assume that's what the Motor drive setting is there for.

 

[AlbertA]

If that's true it means that the axe ir
Works different than convlotion reverbs.

Nope, it's all the same. Convolution is the keyword here. That's what is used in the cab block. An impulse response is specified with time domain coefficients (in this context the domain of interest is time)

That would mean that it could be worth using
Other impulse-applications in my DAW to get
The time domain information as well.

Even if the "tail " is 20ms it still should
Sound more interesting than just an eq.

You guys are getting to hanged up on the word EQ. In an audio context, an EQ means you are changing the frequency response somehow (often just focusing on the magnitude). In any other context it means you are flattening the spectrum and cancelling any group and phase delays.

A linear equalizer can be characterized with an impulse response. I think the confusion is that maybe you think of an EQ as an 8-band Graphic EQ, or a 5-6 band parametric EQ, i.e. a broad brush kind of application.
Some people like to think of a Finite Impulse Response in the context of speaker cabinet simulation as sort of N-band EQ (if thinking in a frequency domain context) - in the case of the Axe-FX II a 2048 band EQ.

BTW, impulse response of equalizers will "tail" in the time domain

But i'm still not 100 % convinced that the cab block
Is just an eq.

Hopefully what I wrote above helps.

How can we then explain that
A cab block sounds several times louder when
activated compared to bypassed.
I smells like a bunch of very short dubbs
contributing to louder appearance.

I didn't get what you were trying to say here. The cab block is mainly a "convolution" block that accepts at finite impulse response with a maximum length of 2048 (actually 2040 I think since last 8 samples are for the IR name). That impulse response could be the description of a speaker cabinet system or a 3-band graphic EQ or whatever linear time invariant system of your choice.

However, see above why you could interpret a speaker cabinet impulse response as a sorts of N-band EQ.

In fact , I´m not shure if Cliff was talking about the cab-block or just explaining different theories.

He was describing some mathematical transforms.

 

[AlbertA]

Forgive my total noob question, but it's something I've always wondered. Does a loudspeaker's frequency response or compression characteristics change depending on amplitude? e.g. if a speaker is driven hard does the response/compression change?

If so do IRs capture this? Or are IRs only representative of the particular volume setting that the speaker at the time was asked to produce when capturing the IR?

Just saw this paper, but this seems like a very complex simulation of a loudspeaker, its materials and components.

http://www.klippel.de/uploads/media/Prediction_of_speaker_performance_at_high_amplitudes_01.pdf

Speakers are mostly linear.

Here's some previous posts on this:
http://forum.fractalaudio.com/axe-fx-ii-discussion/73859-question-v11-beta-testers-3.html#post904477
http://forum.fractalaudio.com/axe-fx-ii-discussion/73859-question-v11-beta-testers-3.html#post904487

 

[InsideOut]

It's mainly just convolution Convolution - Wikipedia, the free encyclopedia

Thanks a lot Bert!! I read that and now my brain is leaking something. :roll

 

[zenaxe]

Maybe you should hold off too, see what Cliff explained...

Agreed. I do not claim to be an expert. I am an EE, so I have some basic coursework on signals and systems and can follow some of the discussions but my work experience/specialization is way outside this area, so I only rank barely above amateur compared to some here.

It just struck me funny, since in the general sense, IR convolution is not decoupled from time and is frequently specifically used as a method to reproduce room/reverb effects. Of course, I will certainly defer to Cliff or Jay M or any of our other true resident experts for concrete technical details!

 

[Rex]

The cab block is direct convolution. You can't do it in the frequency domain via FFT as that would add excessive latency. All modelers use direct convolution for this reason.

Got it. Thanks for clearing that up.

 

[AlbertA]

Cliff addressed this a few posts above. Once the IR is transformed into the frequency domain, time information is lost, and the ability to model reverb (which is time-dependent) goes with it.

Not quite right. You can get back to the time domain from the frequency domain by the inverse Fourier transform. There's no information lost, simply the time information is not available in the frequency domain and frequency information is not available in the time domain.

Impulse responses do model reverberation effects quite well but they are quite a bit longer than an impulse response of a speaker system.

 

[Rex]

Not quite right. You can get back to the time domain from the frequency domain by the inverse Fourier transform. There's no information lost, simply the time information is not available in the frequency domain and frequency information is not available in the time domain.

Thanks for clarifying that. The last time I studied this stuff was decades ago. I've never used it in "real" life, so it's been rusting in a corner of my mind.

 

[sprint]

So, back to my question, will I hear any significant difference with this change?

 

[notalemming]

Since Cliff titled this thread "Something Cool I've Been Working On", I would think so!

 

[sprint]

Well...then put me down fur some utrarez!!!

 

[trancegodz]

Very informative posts from Cliff and Kevin in this thread. I learned a lot about IRs I didn't know before. Really looking forward to the Ultra Rez IRs.

 

[themaidenmaniac]

I'm guessing this is going to be heading for a "firmware 13" release. Haven't heard much on it over the last few months, so be nice to get the fire burning again

Also extremely interested in whether the new technology will be adaptable to the tone match block? Any ideas guys?

 

[deathsilver]

I'm guessing this is going to be heading for a "firmware 13" release. Haven't heard much on it over the last few months, so be nice to get the fire burning again

Also extremely interested in whether the new technology will be adaptable to the tone match block? Any ideas guys?

What I would love are side by side audio samples to see if we can spot the difference.

 

[themaidenmaniac]

Oh you will! At a res of 8000 everything is tighter and less "smeared" for lack of a better word. This is going from my experience of an EQ curve match at 8000 odd samples as opposed to 1-2000 samples. I'll post a link to some of my recorded examples.

 

[lauke-lux]

I'm guessing this is going to be heading for a "firmware 13" release. Haven't heard much on it over the last few months, so be nice to get the fire burning again

Also extremely interested in whether the new technology will be adaptable to the tone match block? Any ideas guys?

Absolutely share your point of view, IMHO it's a major change that must have quite some impact here & there on the FW, but I'm looking forward to it & think it will definitely send any need for regular use of real tube amps to the stone age.