Understanding the State of the Art of Digital Room Correction

[Rufus T. Firefly]

I want to delve further into better DSP for my system, however I'm not sure how to implement it. I feel like this all assumes someone is using a PC. If anyone has an idea how I can put this to use in an AVR-based 5.1.4 surround system I'm all ears. I play an equal amount of stereo music and movies and have 2 subs.

I'm not knowledgable in home theater but I was under the impression that a lot of those electronics had some sort of DSP capabilities. I would suggest shagging that info first. If your gear doesn't I'm sure there is hardware/app solution.

 

[GDK]

I would suggest shagging that info first.

There’s a figure of speech that may not cross the Atlantic particularly well…

 

[Chromatischism]

I'm not knowledgable in home theater but I was under the impression that a lot of those electronics had some sort of DSP capabilities. I would suggest shagging that info first. If your gear doesn't I'm sure there is hardware/app solution.

I am using Audyssey XT32 very successfully for my subs and speakers up to 400 Hz (this cutoff changes from time to time from 300-500).

I tried Dirac briefly and came back to Audyssey.

It uses FIR filtering with roughly 16,000 taps, so it's very good. And you get the full 16,000 taps on the subwoofer channel, despite its small frequency range. We need that because of how sensitive we are to frequency changes in the bass range. Basically, our auditory system's "resolution" is much greater there than at higher frequencies.

But, there is still better, especially in terms of customizability of the curve and other things. However, if it's too difficult to implement in my system I'll just settle for what I have and wait for Audyssey's next piece of software they are cooking up for PC. I have extremely neutral speakers with very good directivity so I'm not in dire need, but I'm always looking to take things to the next level, especially with sub + speaker integration and the minute changes that you can make surrounding that which make a big difference.

 

[tecnogadget]

I want to delve further into better DSP for my system, however I'm not sure how to implement it. I feel like this all assumes someone is using a PC. If anyone has an idea how I can put this to use in an AVR-based 5.1.4 surround system I'm all ears. I play an equal amount of stereo music and movies and have 2 subs.

I’m sorry to inform you this is only possible by the means of a PC. I was in a similar place like you waiting for AVR solution or a MiniDSP black box of a sort, and after waiting for years had to understand the hard way this level of customization and processing power is way above what an AVR can achieve…so I don’t see that happening in a near future in the AVR scene.

I'm not knowledgable in home theater but I was under the impression that a lot of those electronics had some sort of DSP capabilities. I would suggest shagging that info first. If your gear doesn't I'm sure there is hardware/app solution.

Of course nowadays most AVR’s (and even from 15 years ago) implement some sort of DSP, even stand-alone subwoofer do. But they are miles below the level discussed here by Mitchco, this is State of The Art DSP, AVR’s built in processors can’t handle this kind of filters. AVR’s are many times limited in taps per channel and usually don’t correct for timing like step response or excess phase.

I tried Dirac briefly and came back to Audyssey.

It uses FIR filtering with roughly 16,000 taps, so it's very good. And you get the full 16,000 taps on the subwoofer channel, despite its small frequency range.

I’m interested as where/how did you get that info/confirmation about 16.000 taps on Audyssey. Is it 16K total or per channel ? Is there any interesting whitepaper I could read ?

 

[markus]

The exact impulse response is nonlinear, but the filters clearly exist, and knowing the bandwidth sets an absolute limit for the narrowest that an impulse response could be, linear or not.

Bear in mind the nonlinearity is compressive. Stimulii that come just after the initial stimulii are effectively ignored. I'm not trying to be obtuse here, it's a complex system, and remember that there is a filter, effectively, for every inner hair cell. that means a lot of heavily overlapping filters, and the structure of the filter causes a rapid phase shift at the center point.

Please do watch the tutorial I cited above. Really. There is a lot to handle there, yes, I know.

1. I'm not disputing that the inner and outer part of our ears massively filters the physical input before the brain gets to decipher "the mess". Quite the opposite actually. What I'm questioning is a) that we know enough and b) the direct applicability of all those findings to room correction. They are not as clear (to me) as one would hope. In your talks you seem to suggest the very same.
One example would be modelling the basilar membrane as a set of gamma-tone filters. This suggests application of FDW. But the values that somewhat worked for me are nowhere near (source: Schnupp, Auditory Neuroscience):

Admittedly it's been a while since I went through the literature and there are certainly newer findings (yes, I've watched this and that)?

You didn't address some of my other points which I think are on topic here when talking about "state of the art" room correction:

2. Active absorption
This somewhat is a misnomer as sound waves don't really cancel in terms of "energy is removed". You made several statements in the past that it's generally a bad idea to add energy to the room when there's already too much in it. Obviously this is not generally true otherwise a DBA or the reduction in point-to-point variance by adding low frequency sources together with matching filtering (which requires multiple in-room measurements) wouldn't work. But it does. It doesn't matter if points outside the listening area get worse (which isn't necessarily happening anyway).

3. Single mic location
In your point of view is a single measurement with an omnidirectional mic enough to do "state of the art" room correction? You seem to suggest that more aspects of the sound field need to be known even when the measurement is taken only at a single location. This suggests the use of a mic array?
I'd like to read through your AES convention paper but I don't have access to it. Would need to reactivate my AES membership but I think (old) convention papers aren't accessible anyway with a regular membership? At least that's how it worked in the past.

 

[markus]

I’m interested as where/how did you get that info/confirmation about 16.000 taps on Audyssey. Is it 16K total or per channel ? Is there any interesting whitepaper I could read ?

The number of taps regarding filter frequency resolution is somewhat misleading here as MultEQ XT32 uses warped FIR filters to get good low frequency resolution.
Google brings up this: https://www.ee.columbia.edu/~dpwe/papers/KarjHLH97-warp.pdf

 

[j_j]

1. I'm not disputing that the inner and outer part of our ears massively filters the physical input before the brain gets to decipher "the mess". Quite the opposite actually. What I'm questioning is a) that we know enough and b) the direct applicability of all those findings to room correction. They are not as clear (to me) as one would hope. In your talks you seem to suggest the very same.
One example would be modelling the basilar membrane as a set of gamma-tone filters. This suggests application of FDW. But the values that somewhat worked for me are nowhere near (source: Schnupp, Auditory Neuroscience):

View attachment 162162

Admittedly it's been a while since I went through the literature and there are certainly newer findings (yes, I've watched this and that)?

You didn't address some of my other points which I think are on topic here when talking about "state of the art" room correction:

2. Active absorption
This somewhat is a misnomer as sound waves don't really cancel in terms of "energy is removed". You made several statements in the past that it's generally a bad idea to add energy to the room when there's already too much in it. Obviously this is not generally true otherwise a DBA or the reduction of point-to-point variance by adding low frequency sources together with matching filtering (which requires multiple in-room measurements) wouldn't work. But it does. It doesn't matter if points outside the listening area get worse (which isn't necessarily happening anyway).

3. Single mic location
In your point of view is a single measurement with an omnidirectional mic enough to do "state of the art" room correction? You seem to suggest that more aspects of the sound field need to be known even when the measurement is taken only at a single location. This suggests the use of a mic array?
I'd like to read through your AES convention paper but I don't have access to it. Would need to reactivate my AES membership but I think (old) convention papers aren't accessible anyway with a regular membership? At least that's how it worked in the past.

Well, although I'm not very fond of gammatone filters (they have what I am forced to conclude are the wrong envelope shape), look at the length of the various filters.

Now, consider, if the energy is there around any one inner hair cell (recalling there are many per ERB), that inner hair cell will fire. Then the sensitivity to that particular spot on the basilar membrane is very low for 1 millisecond minimum (nerve must reload), and at that point, the outer hair cells will be starting to depolarize, also reducing sensitivity for substantially longer time. So the first approach is one major part of the timbre detected, more than one would expect via energy arguments by quite a bit. Now, after the neurons have settled, you're into the "not early" situation. That measures the timbre of the room reflections (assuming some dynamics are present).

When that does not "make sense" in regard to the direct, yes, strange things happen Elevation can go wrong, directional perception can get fuzzy or poor (angle is likely to close to right if the source is properly delayed for arrival at the two ears one way or another, but not necessarily stability, etc), and lots of bad stuff can happen.

This is just a few of the issues. There is a whole lot more involved, of course.

Oh, and a single point capture of a soundfield involves four (not one) variables, 3 volume velocities and one pressure.

 

[Chromatischism]

I’m sorry to inform you this is only possible by the means of a PC. I was in a similar place like you waiting for AVR solution or a MiniDSP black box of a sort, and after waiting for years had to understand the hard way this level of customization and processing power is way above what an AVR can achieve…so I don’t see that happening in a near future in the AVR scene.

I understand we don't have AVRs that can run this software, but that isn't what I was asking for. Just ideas in general about how to implement it, like possibly running everything through a PC but not using it as the source, for example.

I’m interested as where/how did you get that info/confirmation about 16.000 taps on Audyssey. Is it 16K total or per channel ? Is there any interesting whitepaper I could read ?

Jeff (I can't remember his last name at the moment) at Audyssey explained so in one of the more recent Sound United videos. The one where he revealed that they are working on a new PC software. That count is per-channel.

 

[Rufus T. Firefly]

Curious to know what the range of industry pricing is for residential SOTA DSP. Not looking for exact numbers just ballpark.

 

[tw99]

So Dirac may not be "state of the art" per Mitchco's opinion...

I can see how these other products could make more sense if you have a PC based replay system.

However, Dirac via a Minidsp box improves the sound of my system significantly, and has the advantage of not requiring a PC, so personally I'll be sticking with it.

 

[markus]

Oh, and a single point capture of a soundfield involves four (not one) variables, 3 volume velocities and one pressure.

What would be the next best thing? I guess the answer is multiple measurements? Where does that leave the single mic optimization process proposed in post 1 of this thread?

 

[j_j]

What would be the next best thing? I guess the answer is multiple measurements? Where does that leave the single mic optimization process proposed in post 1 of this thread?

You are aware that one can use, say, 4 omni's in different, known locations, and the understanding of the wave equation to get the same result. Yes?

 

[jhaider]

What troubles me is that this site is allegedly a bunch of objective people.

Perhaps. That may be why some of the author’s previous missteps, such as conflating in-his-room-at-one-point measurements with useful anecohic data, engender skepticism.

If you're willing to put the effort into this, the result is stellar to a point I can't describe it to you...it's really something you need to hear and measure for yourself (not in that order).

How about some measurements? You have the ability to take gated (quasi-anecohic) measurements, right? I’m curious what his process does to the direct field above the transition region compared to what Mr. Sprinkle et al. intended. Those two measurements would be much more interesting than a 2 hour advertorial video.

 

[LumbermanSVO]

I have a few Acoustic Power Labs processors and love them, despite the 2-channel one testing poorly because they do this:

That was a very early test, I've since figured out the low end hump.

I also use them in my home theater setup and do not have any lip sync issues, and that is with two of the units having 4096 taps on both the inputs and outputs.

 

[Alice of Old Vincennes]

No matter which EQ is the "best" or most sophisticated, if lip sync issues arise, I`ll be in trouble at the home front.

 

[markus]

You are aware that one can use, say, 4 omni's in different, known locations, and the understanding of the wave equation to get the same result. Yes?

Yes Thanks for confirming that a single measurement with an omnidirectional microphone isn't enough.

 

[Alice of Old Vincennes]

Look at Amir`s responses to his room correction review.

 

[markus]

Look at Amir`s responses to his room correction review.

Link?

 

[Phorize]

A deep dive presentation on the fundamentals of "proper" Digital Room Correction (DRC). Includes hands-on DSP FIR Filter Designer demos using Acourate and Audiolense.

Having participated in many audio forum discussions, having watched online videos on Digital Room Correction (or DRC), and having reviewed over a dozen DRC products over the past 11 years, I have come to two conclusions. One is that there is considerable misunderstanding about DRC, how it works and even what problems DRC is trying to solve. And, just as important, understanding what is possible using the SOTA of DRC. I hope you find the content educational and practical.

Great work, and much appreciated!

 

[thorvat]

Link?

Let me propose another link. In this thread folks quickly agreed that when comparing different mics you have to do it in the same position as even few cm of difference in position makes a difference in measurement.

What better argument than that would you need that even in the single seat scenario multipoint measurement is needed in order to get reliable spatially averaged input for room correction?