The No.1 Website for Pro Audio
Why do Cab IR's seem lacking?
Old 3 weeks ago
  #31
Lives for gear
Quote:
Originally Posted by chris661 View Post
It can't be. An IR is derived from a frequency response sweep, which was taken at a particular drive level.

Therefore, the IR is only ever correct at that particular drive level. The frequency and phase response generally stay pretty much the same until you hit mechanical limits (and then the bottom end compresses quickly), but harmonic distortion will rise with power input.
A 1W sweep won't tell the full story of what happens when we hit the speaker with a 100W "chug" and really give it some pain.
But dynamic convolution is already a thing for some years now.
Focusrite did that 15 years ago. Acustica is doing it in most of their plugins.

And you can use other signals than a sine sweep.
Some to specifically look for intermodulation distortion for example.

https://www.google.com/url?sa=t&sour...UtC2u7f507MBEd
Old 3 weeks ago
  #32
Lives for gear
Here's my write-up: https://www.grimshawaudio.com/guitarspeakers
I actually managed to hit over 100% distortion on the last sweep, but that was being abusive.

Chris
Old 3 weeks ago
  #33
Lives for gear
It'd be easy enough to put something together.

How about:
- Record a riff DI'd,
- Re-amp at varying levels
- Normalise afterwards, so if I drive the speaker 6dB harder, I take 6dB off the resulting recording, etc.

I might have chance to do this at some point, but it would have to be a short riff. I'm recording these things at home, and dumping 256W into a >[email protected] speaker was unpleasant for everyone.

If anyone else fancies having a go, I'd be interested to hear.


IMO, an IR-based approach gets us pretty close. However, this is a forum where we argue over which version of a U87 sounds better. We're constantly chasing the last percent or two of performance, and that appears to apply to guitar cab emulation, too.

Chris
Old 3 weeks ago
  #34
Here are a few free ones for you.

Ranging from the most natural to a softer one. Note that these don't have a lot of high-end (it cuts off at around 6kHz). If you need more high end above that, I would recommend some saturation in the DAW.
Attached Files

Aaltotarha_GrandOldMan.wav (996.8 KB, 460 views)

Aaltotarha_GrandOldMan2.wav (118.8 KB, 342 views)

Aaltotarha_GrandOldMan3.wav (118.8 KB, 347 views)

Old 3 weeks ago
  #35
Gear Addict
 

Quote:
Originally Posted by chris661 View Post
Here's my write-up: https://www.grimshawaudio.com/guitarspeakers
I actually managed to hit over 100% distortion on the last sweep, but that was being abusive.

Chris
Thanks for doing this, there's a lot of food for thought. Would it be reasonable to estimate that the softest test signal was playing at around 100dB? I also wonder if one could consider microphone distortion negligible (e.g. if the results would look very different for a cleaner speaker).

It seems that there's fairly little compression going on, but in these figures it seems to kick in at the higher frequencies first. The distortion curve seems to roughly follow the shape of the frequency response in the mids, lagging by one octave due to the dominating 2nd harmonic. This makes me wonder if some of the distortion could be emulated by distorting the signal before it goes to the speaker (which is, of course, what a preamp does).

The low frequency ripple at high levels is a bit mysterious to me.
Old 3 weeks ago
  #36
Lives for gear
I suspect the LF ripples are down to the very fast sweep - it's possible that the speaker's mechanical clipping of the peaks of the waveform meant some frequencies were reported as quieter than others.

A slower sweep would probably show us more detail there, but as I mentioned in my previous post, SPLs were high and it wasn't particularly pleasant for the speaker or anyone here.


The first test signal would've been around [email protected], yes. 4V RMS into the test speaker (link below) would be 1W input, so about 99dB for most of the range, going above 100dB over the kHz range.

https://www.eminence.com/speakers/sp...l=Man_O_War_16

At the mic (a few inches from the cone), SPL would've been considerably higher - of the order of 20dB more than at 1m.
The final sweep would be at about [email protected], or ~144dB at the mic.

Chris
Old 2 weeks ago
  #37
Lives for gear
Yeah, could do.

For now, here's a link: https://drive.google.com/drive/folde...tz?usp=sharing

Where you can download the IR files for the different drive levels.

I hope that works - I just exported the impulse response as a 32-bit WAV file straight from REW.

Cheers,
Chris
Old 2 weeks ago
  #38
Lives for gear
 
Yuri Kogan's Avatar
 

Quote:
Originally Posted by tINY View Post

Glad you found some one in a position to do this easily.

I do hear some differences. It is subtile.

And, we still aren't capturing the non-linearities.....



-tINY

It is far from subtle when you are playing, and especially dynamically playing. And yes compression is only one artefact. Cone distortion is another. None of this can be done by IR on the fly. Its the same Kemper situation, but less predictable.
PLAYERS know about it and use it. But if Chris wants to run tests for every non-linearity - all good. Have to add that we recorded a quad with greenbacks a couple of days ago - every speaker had a different sound and response. In the ballpark, but not a subtle difference. To a point where all 7 people present after a discussion have chosen the top/left one. Musicians notice those things.
Old 2 weeks ago
  #39
Quote:
Originally Posted by Yuri Kogan View Post
But if Chris wants to run tests for every non-linearity - all good. Have to add that we recorded a quad with greenbacks a couple of days ago - every speaker had a different sound and response. In the ballpark, but not a subtle difference. To a point where all 7 people present after a discussion have chosen the top/left one. Musicians notice those things.

Yes, musicians notice. I expected the Man of War to be less different than that, but, it's a heavy cone and huge motor structure. Still, without the non-linearities in the high powered samples, that could explain how similar they sound.

Remember, too, that this was driven with a solid state amp with a high damping factor... So, even the frequency response effects of the tube output stage are lost.



-tINY

Old 2 weeks ago
  #40
Lives for gear
Okay, I did a really quick test of something interesting.
I took an exported IR file, and imported it back into REW.

Guess what?

No distortion shown. Only frequency & phase response.

Since the response shape stayed pretty much the same at high power, of course you'll never tell the difference.

So, this comes back to the idea that I put forward at the start: an IR file cannot fully describe the behaviour of a speaker. With no harmonic distortion transcribed, we're missing out on a key aspect.

At high power, the distortion was over 100% THD in places - that would be a huge and obvious change in the sound of the speaker in the room, and a cabinet simulator is missing that entirely.

Chris
Old 2 weeks ago
  #41
Lives for gear
Quote:
Originally Posted by chris661 View Post
Okay, I did a really quick test of something interesting.
I took an exported IR file, and imported it back into REW.

Guess what?

No distortion shown. Only frequency & phase response.
This could well be only a option in visual representation as with a logarithmic sweep, its possible to separate the linear part from the distortion.
Old 2 weeks ago
  #42
Lives for gear
 
Yuri Kogan's Avatar
 

Quote:
Originally Posted by tINY View Post

Yes, musicians notice. I expected the Man of War to be less different than that, but, it's a heavy cone and huge motor structure. Still, without the non-linearities in the high powered samples, that could explain how similar they sound.

Remember, too, that this was driven with a solid state amp with a high damping factor... So, even the frequency response effects of the tube output stage are lost.



-tINY

Yes damping factor of the tube amp is low, and its affect is most felt in low end. Its especially noticeable with class A tube amps cause the output impedance is so unstable and frequency dependant. So you lose that "bloom" when you go to AB and get a tighter bottom end as well. Less back EMF. with SS amps you get a lot more control due to a very high damping factor - ie more predictable performance. That's why these are used in studio monitoring and not so much as guitar amps - they sound boring.
Old 2 weeks ago
  #43
Lives for gear
 
Yuri Kogan's Avatar
 

Quote:
Originally Posted by chris661 View Post
Okay, I did a really quick test of something interesting.
I took an exported IR file, and imported it back into REW.

Guess what?

No distortion shown. Only frequency & phase response.

Since the response shape stayed pretty much the same at high power, of course you'll never tell the difference.

So, this comes back to the idea that I put forward at the start: an IR file cannot fully describe the behaviour of a speaker. With no harmonic distortion transcribed, we're missing out on a key aspect.

At high power, the distortion was over 100% THD in places - that would be a huge and obvious change in the sound of the speaker in the room, and a cabinet simulator is missing that entirely.

Chris
No disagreement here. But is there more to your life? Or is it just Corona-fuelled?
Old 2 weeks ago
  #44
Lives for gear
Quote:
Originally Posted by Yuri Kogan View Post
No disagreement here. But is there more to your life? Or is it just Corona-fuelled?
Oh, it's entirely Corona-fuelled. Not the alcoholic drink, either.

I run a live sound business, so work has ground to a complete halt with the virus. I have lots of nice equipment around (and some not-so-nice), plus a degree in Physics.

I'm a fairly curious person, so if I see a discussion where I think "hey, I've got the equipment here to actually test the theory", then I'll give it a go and post the results.

Sometimes it's interesting, like this one. Another one was where people were wondering if speakers "lose detail" (presumably stiction at play - where if the coil is receiving very very small signals, it won't bother to move) at low SPLs. After putting an MC930 up close to a small HiFi speaker, I found that the mic could reliably pick up sweeps that I couldn't hear in the room. The curves were still very linear down there, so it looks like speakers work fine at very low SPLs, but our ears do not. Not surprising, but Fletcher & Munson would be pleased to hear it.

Back to the guitar speakers, it's now obvious that IR modelling misses out on the harmonic distortion that speakers contribute.

Looking at the 16V RMS (16W into this speaker) curve, we've got:

- 40% THD at the bottom end, mostly 2nd harmonic with a bit of 3rd.
- A narrow peak of 3.5% THD (mostly 2H) around 400Hz
- A rise up to 6% THD (mostly 2H) at 2.5kHz

IR files won't capture that, but a real speaker in a real room will act that way.

It appears to me, then, that we can reach a conclusion here. IR-based simulators don't capture the whole picture, so real cabinets will always sound different. There's no way anyone can claim they wouldn't notice 40% THD.

Once the software developers figure out a way to include distortion (harmonic, as we've discussed here, as well as inter-modulation, as another poster has mentioned), then cabinet simulation ought to get a lot closer to sounding like a real cabinet.

Chris
Old 2 weeks ago
  #45
Lives for gear
Quote:
Originally Posted by Yuri Kogan View Post
Told you so. It is not hard to understand though
And no you are trying to hi-jack the thread into god knows what. We are discussing why speaker non-linearities are not reproduced in IRs.
The claim was that the speaker is very nonlinear with regards to levels, while the development director of Celestion states that guitar speakers are quite linear in that regard and most of the nonlinearities are coming from the power amp.

So if you want to test the nonlinearities of the speaker you have to use an amp that is much more linear than a guitar tube amp.
Which was done here.
Old 2 weeks ago
  #46
Lives for gear
Quote:
Originally Posted by Yuri Kogan View Post
Told you so. It is not hard to understand though
And no you are trying to hi-jack the thread into god knows what. We are discussing why speaker non-linearities are not reproduced in IRs.
And I already posted several studies that showed that they are reproduced.
Old 2 weeks ago
  #47
Lives for gear
 
Yuri Kogan's Avatar
 

Quote:
Originally Posted by chris661 View Post
Oh, it's entirely Corona-fuelled. Not the alcoholic drink, either.

I run a live sound business, so work has ground to a complete halt with the virus. I have lots of nice equipment around (and some not-so-nice), plus a degree in Physics.

I'm a fairly curious person, so if I see a discussion where I think "hey, I've got the equipment here to actually test the theory", then I'll give it a go and post the results.

Sometimes it's interesting, like this one. Another one was where people were wondering if speakers "lose detail" (presumably stiction at play - where if the coil is receiving very very small signals, it won't bother to move) at low SPLs. After putting an MC930 up close to a small HiFi speaker, I found that the mic could reliably pick up sweeps that I couldn't hear in the room. The curves were still very linear down there, so it looks like speakers work fine at very low SPLs, but our ears do not. Not surprising, but Fletcher & Munson would be pleased to hear it.

Back to the guitar speakers, it's now obvious that IR modelling misses out on the harmonic distortion that speakers contribute.

Looking at the 16V RMS (16W into this speaker) curve, we've got:

- 40% THD at the bottom end, mostly 2nd harmonic with a bit of 3rd.
- A narrow peak of 3.5% THD (mostly 2H) around 400Hz
- A rise up to 6% THD (mostly 2H) at 2.5kHz

IR files won't capture that, but a real speaker in a real room will act that way.

It appears to me, then, that we can reach a conclusion here. IR-based simulators don't capture the whole picture, so real cabinets will always sound different. There's no way anyone can claim they wouldn't notice 40% THD.

Once the software developers figure out a way to include distortion (harmonic, as we've discussed here, as well as inter-modulation, as another poster has mentioned), then cabinet simulation ought to get a lot closer to sounding like a real cabinet.

Chris
Looks like you guys are in the same situation we are here. Worst, they closed the Arts ministry and made it a part of industrial development department Meaning the guys are missing out on all the grants and support. Where is Spartacus when you need him?
Yes, in agreement on all you said. I use IRs but not as a substitute for speakers,, ingeneral. Occasionally maybe - if the tone is right. But it misses out on the dynamic bits so guess it depends on what is being recorded. For general live stuff, it maybe sufficient.
Old 2 weeks ago
  #48
Lives for gear
Quote:
Originally Posted by tenderboy View Post
And I already posted several studies that showed that they are reproduced.
Hold up. Which non-linearities are we talking about here?

An impulse response can only contain frequency and phase response data. That's how the maths works.
IR files cannot contain harmonic or inter-modulation distortion, which I'd suggest is an important aspect.


Sascha, I'll see what I can put together. You'll have to forgive my guitar playing, though.

Yuri, I've ended up copying this guitar cab's frequency response using EQ before now. Works okay when the guitarist's amp stops working - just take a DI at the pedalboard and set the stage monitors to "stun"

Chris
Old 2 weeks ago
  #49
Lives for gear
Quote:
Originally Posted by chris661 View Post
Hold up. Which non-linearities are we talking about here?

An impulse response can only contain frequency and phase response data. That's how the maths works.
IR files cannot contain harmonic or inter-modulation distortion, which I'd suggest is an important aspect.

Chris
Here is a method described to capture intermodulation distortion in IRs :
http://www.diva-portal.org/smash/get...FULLTEXT01.pdf

And here is a paper that seems to adress IRs and nonlinear systems (although I cannot check as my AES account ran out)
http://www.aes.org/e-lib/browse.cfm?elib=16158

And than there is dynamic convoution that takes varoius IRS at different levels and maps them regarding to the input signal...as done by Focurite in the Liquid Channel or Acoustica Plugins.
Old 2 weeks ago
  #50
Gear Addict
 

Quote:
Once the software developers figure out a way to include distortion (harmonic, as we've discussed here, as well as inter-modulation, as another poster has mentioned), then cabinet simulation ought to get a lot closer to sounding like a real cabinet.
I think it's worth noting that we don't know if current software includes some kind of cabinet/speaker-related distortion modeling. Of course, just convolving with an IR doesn't do this. If the distortions don't depend heavily on the type of speaker and cabinet chosen in the software, one might be able to emulate them after the fact. However, I haven't seen any plug-in manufacturers advertise this type of modeling.

Not necessarily a guitar plug-in, but IIRC IK Sunset Sound uses some kind of a dynamic convolution variant, which technically could capture speaker non-linearities in the chambers.
Old 2 weeks ago
  #51
Lives for gear
Quote:
Originally Posted by tenderboy View Post
Here is a method described to capture intermodulation distortion in IRs :
http://www.diva-portal.org/smash/get...FULLTEXT01.pdf

And here is a paper that seems to adress IRs and nonlinear systems (although I cannot check as my AES account ran out)
http://www.aes.org/e-lib/browse.cfm?elib=16158

And than there is dynamic convoution that takes varoius IRS at different levels and maps them regarding to the input signal...as done by Focurite in the Liquid Channel or Acoustica Plugins.
I went through the first paper twice, and can't find anywhere where they say that harmonic distortion data is captured within an impulse response.

Quote:
The results in chapter 7 indicate that it is possible to estimate both room impulse response and loudspeaker distortion from a single measurement, using the exponential sine sweep method
I used a swept sine sweep for my measurements, and found that I could also get an impulse response as well as distortion data.
The point that you seem to be missing is that making an IR out of a sine sweep involves discarding the harmonic distortion data. An IR contains only frequency and phase information.

The dynamic convolution work sounds promising, and I'll be sure to read into that.


Sascha, feel free to post up some audio clips for me to run through the speaker. Would prefer that you keep them <10s - 120dB+ isn't pleasant for anyone here, so I'd like to keep things as short as possible.

Chris
Old 2 weeks ago
  #52
Quote:
Originally Posted by thenoodle View Post
Why are cab ir's lacking? The technology is young and new. The sos article serves as an illustration to the complexity challenges.

Let's see how things develop over the next ten years.
It's not just the newness of the technology - it's that because of the way IRs are created they do not and cannot reproduce all the nonlinearities of the speaker being driven with a (complex) guitar signal from a real amplifier. What guitar speakers do under real working conditions is quite complex and highly dependent on the content of the signal itself. It's like trying to hit a moving target with a gun that's fixed in place - every so often it might be spot on, but most of the time it won't be.

It's the difference between a fixed, static process and a dynamic, reactive one.
Old 2 weeks ago
  #53
Lives for gear
 
Yuri Kogan's Avatar
 

Quote:
Originally Posted by chris661 View Post
Hold up. Which non-linearities are we talking about here?

An impulse response can only contain frequency and phase response data. That's how the maths works.
IR files cannot contain harmonic or inter-modulation distortion, which I'd suggest is an important aspect.


Sascha, I'll see what I can put together. You'll have to forgive my guitar playing, though.

Yuri, I've ended up copying this guitar cab's frequency response using EQ before now. Works okay when the guitarist's amp stops working - just take a DI at the pedalboard and set the stage monitors to "stun"

Chris
Yes, I run my pre-amps into the 2 notes unit directly all the time. Mostly at home for mock ups. But try running a drum-kit through a pair of those - very interesting sound.
Old 2 weeks ago
  #54
Here for the gear
 

Machine learning

Quote:
Originally Posted by tINY View Post

While I posted this in another thread, I'm starting a new one that will hopefully be more technical and reasoned. I hate posting useful information in a thread that gets buried by Trollery.

For a lot of guitar sounds in recordings (or live on stage) the non-linearities of a guitar speaker are essential to the final sound. And, because impulse response modeling is a linear process, it can't capture the dynamics of an actual guitar speaker. In fact, everything about most electric guitar timbre is dependent on non-linear, low-fidelity phenomena.

Here is an actual quote by Ian White of Celestion:

"With hi-fi and pro PA speakers, we can develop finite element models and see on a screen what they're going to be doing,” Ian says. "But that's almost impossible with guitar speakers because they're so non-linear. It's difficult enough to predict how a guitar speaker is going to sound when you're feeding it a clean guitar signal, but it's the distorted stuff that really gets the cone modes going. It doesn't need to be that loud to do that — just the complexity of the signal really gets the tone humming."

https://www.soundonsound.com/techniq...uitar-speakers

I'm up for a meaningful discussion....



-tINY

The real realm in which to model this would be neural networks whereby via machine learning one could represent the nonlinearities of a speaker ... We have the science to do it but no one has attempted it yet.
Old 2 weeks ago
  #55
Lives for gear
 
Yuri Kogan's Avatar
 

Quote:
Originally Posted by Giuliano777 View Post
The real realm in which to model this would be neural networks whereby via machine learning one could represent the nonlinearities of a speaker ... We have the science to do it but no one has attempted it yet.
For that to happen you have to be able to define and quantify what you want to be observed and learned. With such a non-linear device which has not very repeatable performance, changing in time and with age - life is to short to model it and too expensive too.
Old 2 weeks ago
  #56
I can't hear any differences between speakers and IRs. Often the issue is our ears which are non-linear.

But you can test it. Use a transistor amp, send a sine wave through and see if you can hear or see distortion.
Old 2 weeks ago
  #57
Lives for gear
I did. Here's the write-up: https://www.grimshawaudio.com/guitarspeakers

IRs do not capture the harmonic distortion of a driven speaker, which ought to be very obvious in listening. At moderate levels, there was 40% THD in some areas.

Chris
Old 2 weeks ago
  #58
Lives for gear
 
Yuri Kogan's Avatar
 

Quote:
Originally Posted by Raaphorst View Post
I can't hear any differences between speakers and IRs. Often the issue is our ears which are non-linear.

But you can test it. Use a transistor amp, send a sine wave through and see if you can hear or see distortion.
Are you saying you used a guitar amp into a speaker and neverheard any distortion, compression...?
Old 2 weeks ago
  #59
Quote:
Originally Posted by Yuri Kogan View Post
Are you saying you used a guitar amp into a speaker and neverheard any distortion, compression...?
not from the speaker.
Old 2 weeks ago
  #60
Lives for gear
 
Yuri Kogan's Avatar
 

Quote:
Originally Posted by Raaphorst View Post
not from the speaker.
Interesting. I know of quite a few instances where it was used on purpose to get a certain tone/dynamic. Used it like that myself.
But fair enough
Topic:
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump