Something Cool I've Been Working On

zenaxe said:
This doesn't make any sense given that IRs are, in fact, used to reproduce reverb and their entire purpose is to describe an LTI system as a function of time (via convolution).
Click to expand...
You are correct. I should have said that IRs, as captured and used for cab emulation, are all about EQ. Room reflections affect that EQ.
 
zenaxe said:
LOL. This doesn't make any sense given that IRs are, in fact, used to reproduce reverb and their entire purpose is to describe an LTI system as a function of time (via convolution). Maybe you guys should hold off on commenting further on technical details or correcting people wrt this topic. ;)
Click to expand...

Thanks for explaining this. I've always thought if cab IRs were just EQ curves, they couldn't possibly ever be able to completely emulate an actual cab because along with simple EQ curves, speakers are physical objects, so wouldn't you also have to take into account variables like speaker momentum, inertia, minimum time to full speaker excursion relative to minimum time to full positive or full negative amp output energy, that sort of thing?

After all, a speaker is limited to how quickly the cone can be moved, whereas an electrical signal is basically only limited to slightly less than the speed of light, which is quite a bit faster. The difference means all other variables eliminated, cabs and speakers likely affect tone over time instead of simply being a complex EQ curve... right?
 
Last edited:
GreatGreen said:
I've always thought if cab IRs were just EQ curves, they couldn't possibly ever be able to completely emulate an actual cab because along with simple EQ curves, speakers are physical objects, so wouldn't you also have to take into account variables like speaker momentum, inertia, minimum time to full speaker excursion relative to minimum time to full positive or full negative amp output energy, that sort of thing?
Click to expand...
All of those things affect the speaker's frequency response, and they're accounted for in the IR.


GreatGreen said:
After all, a speaker is limited to how quickly the cone can be moved, whereas an electrical signal is basically only limited to slightly less than the speed of light, which is quite a bit faster. The difference means all other variables eliminated, cabs and speakers likely affect tone over time instead of simply being a complex EQ curve... right?
Click to expand...
Electrical signals are also limited in how fast they can respond. Capacitors and inductors limit how fast an electrical signal can change. Those things, and the speaker variables you mentioned, determine the frequency response. They limit how fast a speaker can move, but they don't vary over time.


What an IR can't model is the frequency-dependent impedence that the cab presents to the amp. That's why the cab impedence is modeled in the Amp block instead of the Cab block.
 
Rex said:
^^^This.

IRs are all about EQ, and only about EQ. They're just way more detailed than yout typical EQ.

@Thomas Larsson: If IRs had decay, they'd sound like reverb.
Click to expand...

IRs do sound like reverb. All the IRs in the axe fx would have originally contained audible reverb tails. However, these have been cropped so that only the first 42ms are actually heard - which is why these IRs cannot fully reproduce the sound of the room they were captured in (lo rez IRs are even shorter, at 21ms), and will not be perceived as reverb. What the IRs in the axe do capture is the sound of the speaker, and whatever early room reflections are present in the first 42ms. As you lengthen an IR, the sound of the room it was captured in will become more prominent as it will contain more reflections.
 
Rex said:
All of those things affect the speaker's frequency response, and they're accounted for in the IR.



Electrical signals are also limited in how fast they can respond. Capacitors and inductors limit how fast an electrical signal can change. Those things, and the speaker variables you mentioned, determine the frequency response. They limit how fast a speaker can move, but they don't vary over time.


What an IR can't model is the frequency-dependent impedence that the cab presents to the amp. That's why the cab impedence is modeled in the Amp block instead of the Cab block.
Click to expand...

More on this (i.e., cab impedence is modeled in the Amp) and how one tweaks the cab impedance to align with different IRs.
 
jasonwb007 said:
Have you considered using Daubechies wavelets to model the response?
Click to expand...

Why would one use wavelets? A wavelet is a transform from the time domain to the "time-frequency" domain. They are typically used for analysis, de-noising and compression. They are related to the STFT but allow you to adjust the time vs frequency tradeoff.

I don't understand what use a Daubechies wavelet (or any wavelet, for that matter) would be in modeling loudspeaker response. But I'm willing to listen. Perhaps you are on to some new technique that the industry is not aware.
 
A Fourier Transform is a decomposition of a signal from the time domain to the frequency domain. Once you perform the transform all time localization is lost. For example, you might have a signal that has a duration of 10 seconds. In that signal there is a short pulse at, say, 1 kHz. If you take the FT of this signal you'll see a spike at 1 kHz. However there is no information as to when that pulse occurred. You can break the signal up into smaller pieces and perform the FT on the pieces. This will then tell you roughly both the frequency and when the pulse occurred. The smaller you make the pieces the more you can localize the time at which the pulse occurred but the frequency resolution decreases (since the frequency resolution is the sample rate divided by the number of samples). If you make the pieces longer you increase the frequency resolution but decrease the time resolution.

Wavelets are based on this property. You define a basis set that gives you the desired tradeoff in time vs frequency resolution. It's a bit more complicated than that but this is the basic theory. Therefore wavelets are a "time-frequency" representation of a signal whereas a Fourier Transform is just a frequency representation. If you break the signal up into pieces and do the FT on the pieces this is the Short-Time Fourier Transform (STFT). Wavelets are superior to the STFT because they give good time resolution for high-frequency information and good frequency resolution for low-frequency information which is what we are usually interested in for real-world signals.

Regardless I don't see any application of wavelets to speaker modeling.
 
Well, it's taken us some time to understand all this, but now that we've got it, I'd say we're ready for ultrares whenever you are. J/K, I can't say I really understand this, but it is definitely interesting stuff, and I appreciate you cluing us in on it Cliff. I'm excited for ultrares. I hope it will be coming "soon".
 
smcrosby said:
More on this (i.e., cab impedence is modeled in the Amp) and how one tweaks the cab impedance to align with different IRs.
Click to expand...
Cab impedance is modeled on the SPKR page of the Amp block. The graph you see there is a plot of the speaker impedance that the amp model sees. You set the high-frequency and low-frequency resonant peaks on this page.
 
antcarrier said:
IRs do sound like reverb. All the IRs in the axe fx would have originally contained audible reverb tails. However, these have been cropped so that only the first 42ms are actually heard - which is why these IRs cannot fully reproduce the sound of the room they were captured in (lo rez IRs are even shorter, at 21ms), and will not be perceived as reverb. What the IRs in the axe do capture is the sound of the speaker, and whatever early room reflections are present in the first 42ms. As you lengthen an IR, the sound of the room it was captured in will become more prominent as it will contain more reflections.
Click to expand...
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.
 
zenaxe said:
LOL. This doesn't make any sense given that IRs are, in fact, used to reproduce reverb and their entire purpose is to describe an LTI system as a function of time (via convolution). Maybe you guys should hold off on commenting further on technical details or correcting people wrt this topic. ;)
Click to expand...
Maybe you should hold off too, see what Cliff explained... ;)
 
Rex said:
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.
Click to expand...

The cabs currently in the axe fx have not been transformed into the frequency domain. They are IRs. IRs exist in the time domain.
 
If that's true it means that the axe ir
Works different than convlotion reverbs.
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.

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

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


Rex said:
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.
Click to expand...
 
Last edited:
Rex said:
Maybe I've got it wrong. Doesn't the FFT happen in the cab block?
Click to expand...

My understanding is that CONVOLUTION happens in the cab block, which uses the captured IR, which is (effectively) completely in the time domain.

Cliff's description above is just to explain the difference between time-domain and frequency-domain processing. I don't believe he said that there's anything frequency-domain in the cab block (apart from the hi/low-cut filters.)
 
You must log in or register to reply here.
Back
Top Bottom