"Speaker Profiling" - Non-Linear and Fully Dynamic - %99.6 Perfect => IR's are no good ?

BenIfin

BenIfin

Roadie
Messages
941
This is new to me ... Quantum Non-Linear Speaker Profiles ... apologies is this is old news

Some pretty big claims here. ... in short, they claim:-

=> " IR's " = static, linear and non-dynamic and therefore pretty-crap
=> " Speaker Profile's " = fully non-static, fully non-linear and fully dynamic and therefore [%99.6] perfect

In their words " ... we have achieved null test results showing our profiles to be up to a 99.6% match to a recording of the mic and speaker they are profiling, this is only possible due to them capturing the full non-linear dynamics and behavior of the speaker and mic ... "

Real or snake-oil ... anything to this (?)
 
BenIfin said:
This is new to me ... Quantum Non-Linear Speaker Profiles ... apologies is this is old news

Some pretty big claims here. ... in short, they claim:-

=> " IR's " = static, linear and non-dynamic and therefore pretty-crap
=> " Speaker Profile's " = fully non-static, fully non-linear and fully dynamic and therefore [%99.6] perfect

In their words " ... we have achieved null test results showing our profiles to be up to a 99.6% match to a recording of the mic and speaker they are profiling, this is only possible due to them capturing the full non-linear dynamics and behavior of the speaker and mic ... "

Real or snake-oil ... anything to this (?)
Click to expand...
Haven't tried them to be honest.
I think anybody that's been doing their own NAM profiles could do something similar. A very linear solid state poweramp's needed though - not sure how many folks have those.
The demos I heard of their stuff didn't really impress me that much. One thing I did like in their vids was their custom sub-bass transducer thingy which they hook up to the cab to get those low lows captured.
 
IMO good IRs that can be combined without phase issues and good room reverb modeling should be more than good enough.
 
These are full rig models of a solid state poweramp+speaker, AFAIK the non linearities and response of the poweramp is not accounted for in the model, so you are essentially running an amp DI back through a SS poweramp and cabinet+mic.

There’s also a bit of a question mark for me on how well you can really seperate the poweramp+cabinet response of one amp model, and slap on a totally different one after the fact. The model can’t factor in anything about the state that the amp was modelled under or change its response based on the amp being used.

I think there’s definitely merit in exploring different ways of capturing the sound of cabinets and machine learning tech is a great candidate in a lot of ways. I’m not sure the trade offs are necessarily worth it, even if they can sound good.

I’d be curious if the deeper parameters of NAM were explored to allow for interpolating captures and for capturing more detail in the time domain. There’s still going to be issues of aliasing, CPU, and support outside of a DAW (where currently you have to stack up several instances). There’s also the whole debate about input and output levels as really you’d want the amp model to “excite” the speaker in the same way that would happen in the real world.

As I said, happy to see it explored and I wish the guys behind it the best of luck. I just don’t think the trade offs are worth it right now.
 
Yeah that’s a good point. If the amp modeler is already modeling power amp and speaker interaction you don’t want to double down on the IR side. It might be better for preamps only, or maybe some modelers that don’t get the amp and speaker side.
 
This was the video that came up on for me a couple of days ago



I get the concept but from what I hear I didn't really like it - the top end seemed really weird (maybe aliasing?) and in the video above some quite extreme filters were being applied to tam the high end.

It has been a while since I've looked at NAM but I was under the impression it wasn't that suited to speaker cabs?
 
Also I’m not sure how the real dynamics of the speaker can be isolated from either the amp they are captured with, or accounting for the negative feedback of a valve amp driving the cab. It’s quite messy to think about.
 
MirrorProfiles said:
Also I’m not sure how the real dynamics of the speaker can be isolated from either the amp they are captured with, or accounting for the negative feedback of a valve amp driving the cab. It’s quite messy to think about.
Click to expand...
I think you can work around those with a solid state amp. IIRC those don't interact the same way as a tube amp would with a guitar cab.
 
2dor said:
I think you can work around those with a solid state amp. IIRC those don't interact the same way as a tube amp would with a guitar cab.
Click to expand...
Solid state amps have a high damping factor so the negative feedback interaction is less relevant. But we’re used to hearing speakers being driven by an amplifier that DOES get influenced by that, so IMO the captured speaker model should be captured under real world conditions, so it overloads in the same way. There’s also the assumption that because the solid state amp driving the cab is solid state, that it’s also “invisible”. While many solid state amps are designed to be very clean and high performing, there will always be limitations with what can be done
 
I was wondering how this worked. I'll check it out when the concept matures a bit.

It's an interesting concept, and the training must be unbearable if these are being done at high volumes. The time to put into an IR sweep plus processing is so much shorter than the NAM training.
 
Saw the video yesterday waiting to be amazed by the results and all I had was a guy making a crappy tone and say it’s was amazing.

I’m sure there’s improvements possible for IRs but was really disappointed
 
Sounds snake-oily to me, IIRC those who knows their speakers and IR's like Jay says an actual speaker is not all that "dynamic" in itself, and a good IR does the job great.
 
I've done quite extensive measurements and speakers aren't terribly nonlinear. They are somewhat nonlinear, especially at very high volumes but the nonlinearities they DO exhibit aren't particularly musical anyways.

What they do exhibit is thermal compression and a "profile" won't capture that anyway for a couple reasons:
1. The time constant is far too long for the neural net.
2. The thermal compression causes the real part of the speaker impedance to increase. This interacts with the output impedance of the amplifier. This is modeled in our products. Our power amp modeling models the increase in impedance due to thermal compression. You can adjust this using, unsurprisingly, the Speaker Compression parameter. This wouldn't be captured accurately with a profile for reasons I'm not at liberty to divulge.
 
FractalAudio said:
Our power amp modeling models the increase in impedance due to thermal compression. You can adjust this using, unsurprisingly, the Speaker Compression parameter. This wouldn't be captured accurately with a profile for reasons I'm not at liberty to divulge.
Click to expand...
I’ve had so much fun tweaking settings like speaker compression to change the sound of guitars in a mix, it’s great. Of all the ways to really tweak a sound, that page in the amp model in the FM9 is one of the most fun and interesting. It can turn a palm mute into a huge clanking sound or vice versa in interesting ways.
 
You must log in or register to reply here.
Back
Top