Audyssey's Next Generation of Room Correction (MultEQ-X)

[GalZohar]

What you are missing is that separate filters are only possible when setting directional bass which is not the mono bass you and mso are expecting. Audyssey (nor dirac for that matter) simply isn't equipped to apply (most of) what mso is trying to do. When you set to mono bass - only one set of filters is possible which applies to both subs. Besides, they aren't actually filters but rathe target curve adjustments, which may not result in the filters you were expecting it to apply.

 

[tengiz]

What you are missing is that separate filters are only possible when setting directional bass which is not the mono bass you and mso are expecting. Audyssey (nor dirac for that matter) simply isn't equipped to apply (most of) what mso is trying to do. When you set to mono bass - only one set of filters is possible which applies to both subs. Besides, they aren't actually filters but rathe target curve adjustments, which may not result in the filters you were expecting it to apply.

I get that, and I already said so before. It won’t work if the recording has stereo bass - which, incidentally, is also the exact case that breaks MSO optimization when your measurements are based on L+R summation only.

So if you're okay with MSO optimizing strictly for mono bass, then the MultEQ-X restriction imposed by the directional subwoofer option shouldn't be a problem.

But the real dealbreaker, of course, is only implementing the magnitude response. I can understand the technical reasons behind such an implementation, but I wish they were more explicit about it.

 

[andyc56]

I don't know, perhaps not, as the predicted response from MSO's global EQ doesn't align with the practical result in my setup.

What works for me is to run MSO's "As flat as possible at all listening positions" optimization without the shared input filters. After entering in the individual filters (which do align well with the predicted result) I EQ globally through MQX.

That's not normal. What kind of DSP device are you using? If the device doesn't have an input channel ("input PEQs" like, say, the miniDSP 2x4 HD) there are special considerations when setting up and also exporting biquads in MSO. See the post here for further explanation in the context of the miniDSP DDRC-88BM.

In short, if your device has no input PEQs, for shared MSO filters to work properly you have to set up MSO to tell it the count of input biquads is zero, so that it can create their equivalents replicated in the output channels. This in turn requires that you export biquads of all channels of a configuration, rather than the per-channel export.

 

[KenMasters]

That's not normal. What kind of DSP device are you using? If the device doesn't have an input channel ("input PEQs" like, say, the miniDSP 2x4 HD) there are special considerations when setting up and also exporting biquads in MSO. See the post here for further explanation in the context of the miniDSP DDRC-88BM.

In short, if your device has no input PEQs, for shared MSO filters to work properly you have to set up MSO to tell it the count of input biquads is zero, so that it can create their equivalents replicated in the output channels. This in turn requires that you export biquads of all channels of a configuration, rather than the per-channel export.

I'm using the inbuilt DSP of my subs in combination with MQX. I use MSO to first optimise the subs, entering the individual PEQ values into my subs manually, then applying light manual global EQ through MQX.

If I try using MSO with the shared filters and enter those into MQX, the resulting response deviates from the target curve with a slight S shape - concave from about 30-80Hz and then convex (I've looked and unfortunately I don't still have those measurements).

Whether it's down to MQX, my measurements, my room or a combination of, I get the best results by running MSO after removing the shared filters and doing minor global adjust manually after the fact (the below graph is a comparison of the raw response vs the applied individual sub filters alone):

 

[andyc56]

I'm using the inbuilt DSP of my subs in combination with MQX. I use MSO to first optimise the subs, entering the individual PEQ values into my subs manually, then applying light manual global EQ through MQX.

If I try using MSO with the shared filters and enter those into MQX, the resulting response deviates from the target curve with a slight S shape - concave from about 30-80Hz and then convex (I've looked and unfortunately I don't still have those measurements).

That result indicates that the mathematical form of the MultEQ-X PEQ filters in the shared sub path does not match what MSO is using for these filters. MSO uses IIR filter transfer functions in its computations. But on page 26 of the MultEQ-X manual, it says this:

A Biquad is an engineering term for a 2nd order IIR (feedback) filter. MultEQ NEVER uses IIR feedback filters for ANY channel. All target curve components are used as frequency magnitude only. While they do still affect phase, MultEQ filters are always finite filters that attempt a best match (given the available computational power) to the target frequency response.

Many biquad configurations are available, but the most common is the “Parametric” EQ. This is a method of adding a boost or dip over a specified frequency range (the “bandwidth” of the filter). Audyssey calls this a “Parametric Peaking Filter” and it is the default type of biquad. The filter interpretation used here exhibits positive/negative gain symmetry. If you are importing parameters from another Parametric EQ implementation, make sure to check that the frequency response of your source parametric EQ has the same as that shown in MultEQ-X for the same parameters.

So their filters are FIR filters that emulate the corresponding IIR filters. Strangely, they call them "biquads", but there is no such thing as a FIR biquad.

This difference in implementation explains the discrepancy between MSO's predicted results and what you were getting. On the positive side, getting good results for the per-sub PEQs indicates that the ones in the subs do have an implementation that matches the mathematical form used by MSO.

One possibility that would allow you to use all of MSO's available optimization types is to put shared PEQs in the MSO configuration but just ignore them when done. You'd end up with a very non-flat response for the combined subs, but you could use the MultEQ-X automatic room correction to flatten that, rather than manual entry of emulated PEQs. Changing the filtering in the shared path does not affect either the seat-to-seat variation or the SPL penalty at all, so you could use this approach with the MSO multi-stage optimization if desired.

 

[tengiz]

That result indicates that the mathematical form of the MultEQ-X PEQ filters in the shared sub path does not match what MSO is using for these filters. MSO uses IIR filter transfer functions in its computations. But on page 26 of the MultEQ-X manual, it says this:

...

So their filters are FIR filters that emulate the corresponding IIR filters. Strangely, they call them "biquads", but there is no such thing as a FIR biquad.

This settles it for good. I can understand the reasons behind the implementation, but not the naming, of course.

Andy - never had a chance to thank you for your awesome work, although I used your creation for years. So - thank you!

 

[Newman]

That result indicates that the mathematical form of the MultEQ-X PEQ filters in the shared sub path does not match what MSO is using for these filters. MSO uses IIR filter transfer functions in its computations. But on page 26 of the MultEQ-X manual, it says this:



So their filters are FIR filters that emulate the corresponding IIR filters. Strangely, they call them "biquads", but there is no such thing as a FIR biquad.

This difference in implementation explains the discrepancy between MSO's predicted results and what you were getting. On the positive side, getting good results for the per-sub PEQs indicates that the ones in the subs do have an implementation that matches the mathematical form used by MSO.

One possibility that would allow you to use all of MSO's available optimization types is to put shared PEQs in the MSO configuration but just ignore them when done. You'd end up with a very non-flat response for the combined subs, but you could use the MultEQ-X automatic room correction to flatten that, rather than manual entry of emulated PEQs. Changing the filtering in the shared path does not affect either the seat-to-seat variation or the SPL penalty at all, so you could use this approach with the MSO multi-stage optimization if desired.

I am trying to visualise how to use MSO and MULTEQ-X with an AV10 (4 sub outputs), with each sub being driven by a Hypex plate amp (which has some DSP). Can you offer some thoughts on how I could best use those resources, please?

 

[andyc56]

I am trying to visualise how to use MSO and MULTEQ-X with an AV10 (4 sub outputs), with each sub being driven by a Hypex plate amp (which has some DSP). Can you offer some thoughts on how I could best use those resources, please?

MSO does not directly support AVRs and pre-pros with multiple sub outs. The assumed hardware setup is described in a brief way in the Hardware Requirements section on the home page, and in a lot more detail on the Details of Configurations page.

I reckon you'd want to connect each balanced sub out straight into the Hypex balanced in. If you could arrange for all the sub out signals to be either exactly the same, or differ only by a delay, over which you have full manual control, that would work. If the sub outputs were all exactly the same, you'd use delay on the Hypex to control the relative sub delays. That might be better.

MSO is compatible with Hypex PEQs, since the original beta tester in 2015 used a Hypex PSC2.400 plate amp. When specifying PEQs in analog format (Q, center frequency and boost/cut), it's important to recognize that there isn't a single standard definition of Q. I wrote an article about that.

As mentioned in post 1445, you could use the Audyssey automatic sub EQ as a surrogate for the shared filters of MSO, provided you could keep it from messing with the individual sub delays and levels, and force all sub outputs to be identical.

I'm not familiar with Audyssey, other than what I've read in the MultEQ-X manual. I'm a two-channel guy, having been reluctantly sucked in to supporting the HT world. I had a brief look at the AV 10 manual, but when I saw it was 342 pages, I bailed.

 

[Newman]

MSO does not directly support AVRs and pre-pros with multiple sub outs. The assumed hardware setup is described in a brief way in the Hardware Requirements section on the home page, and in a lot more detail on the Details of Configurations page.

I reckon you'd want to connect each balanced sub out straight into the Hypex balanced in. If you could arrange for all the sub out signals to be either exactly the same, or differ only by a delay, over which you have full manual control, that would work. If the sub outputs were all exactly the same, you'd use delay on the Hypex to control the relative sub delays. That might be better.

MSO is compatible with Hypex PEQs, since the original beta tester in 2015 used a Hypex PSC2.400 plate amp. When specifying PEQs in analog format (Q, center frequency and boost/cut), it's important to recognize that there isn't a single standard definition of Q. I wrote an article about that.

As mentioned in post 1445, you could use the Audyssey automatic sub EQ as a surrogate for the shared filters of MSO, provided you could keep it from messing with the individual sub delays and levels, and force all sub outputs to be identical.

I'm not familiar with Audyssey, other than what I've read in the MultEQ-X manual. I'm a two-channel guy, having been reluctantly sucked in to supporting the HT world. I had a brief look at the AV 10 manual, but when I saw it was 342 pages, I bailed.

Thanks, much appreciated. (Sorry that I got the impression you were much into Audyssey and MULTEQ-X.) I will study and learn.

cheers

 

[GalZohar]

If your device supports DLBC just use that and forget MSO. Seems like MSO is a poor man's dlbc.
If you use Audyssey then the best way is to run MSO for minimizing seat to seat variation, tricking Audyssey to choose the MSO delays and levels from MSO, apply MSO filters to each sub internal DSP, remove the internal gain and delay you set in the subs DSP to trick Audyssey, then let Audyssey complete its measurement and EQ. Hopefully that works.
But seriously, just use DLBC in your case.

 

[OCA]

If your device supports DLBC just use that and forget MSO. Seems like MSO is a poor man's dlbc.
If you use Audyssey then the best way is to run MSO for minimizing seat to seat variation, tricking Audyssey to choose the MSO delays and levels from MSO, apply MSO filters to each sub internal DSP, remove the internal gain and delay you set in the subs DSP to trick Audyssey, then let Audyssey complete its measurement and EQ. Hopefully that works.
But seriously, just use DLBC in your case.

FYI, DLBC’s sub alignment algorithm is a joke compared to MSO. It’s primitive, requires you to fit your measurements to the algorithm rather than the algorithm adapting to your setup. It also gets easily confused by corner sub placements. Sure, it's better than Audyssey’s built-in sub alignment, but that doesn’t make it good.

Just my 2 cents...

 

[GalZohar]

FYI, DLBC’s sub alignment algorithm is a joke compared to MSO. It’s primitive, requires you to fit your measurements to the algorithm rather than the algorithm adapting to your setup. It also gets easily confused by corner sub placements. Sure, it's better than Audyssey’s built-in sub alignment, but that doesn’t make it good.

Just my 2 cents...

Can you be more specific?
I'm having a hard time finding actual results of people running DLBC, even less so comparing it to anything else.

 

[OCA]

Can you be more specific?
I'm having a hard time finding actual results of people running DLBC, even less so comparing it to anything else.

Neumann MA1 vs. Dirac Live Bass Control (DLBC) - Dual KH750 subs

You are avoiding my question. I asked those question to creator of Audiolense, Acourate and Deqx. I dont own Neumann Systems but I am Pręty confidential it is no diffrent to mentioned above. You may prove me wrong.

audiosciencereview.com

 

[dlaloum]

Can you be more specific?
I'm having a hard time finding actual results of people running DLBC, even less so comparing it to anything else.

Look at the Dirac forums on here and on AVS... the threads are now enormous, and will require some searching, but lots of responses and measurements from DLBC users.

 

[Acerun]

I thought I'd write a little experience that I've had. I'm still a novice with REW and MultiEQ-X but I'm keeping at it, always trying to improve the experience. I recently have engaged ChatGPT and it's been like having a private tutor. Chat will analyze your REW and offer suggestions, then comment on the revised versions, offering small experiments that one might try and in other cases pointing out some slight variances and many times calling out that they aren't worth chasing unless clearly audible. It helped me improve the crossover and tame a little bit of midbass bloat that I was experiencing. Here is my most recent REW in 1/12 graph for fun. ChatGPT thinks it's terrific!

 

[mmares056]

I thought I'd write a little experience that I've had. I'm still a novice with REW and MultiEQ-X but I'm keeping at it, always trying to improve the experience. I recently have engaged ChatGPT and it's been like having a private tutor. Chat will analyze your REW and offer suggestions, then comment on the revised versions, offering small experiments that one might try and in other cases pointing out some slight variances and many times calling out that they aren't worth chasing unless clearly audible. It helped me improve the crossover and tame a little bit of midbass bloat that I was experiencing. Here is my most recent REW in 1/12 graph for fun. ChatGPT thinks it's terrific!

You forgot to measure L+R Playing at the same time. Yes measuring LR + subs individually helps get a general idea but seeing the response of both speakers plus the subs playing at the same time is more important for phase alignment in the XO region. If you don't mind measuring the L + R playing at the same time with the subs on and post the results. I'm curious to see how the response is in the XO region and left and right of the XO region.

 

[Acerun]

Really good catch. Measurement attached. I shared with ChatGPT too and it thinks there is a phase misalignment between left and right mains or between mains and subs in that overlap zone that didn't show up individually. Would love any ideas on how to address! I'm putting ChatGPT to work on ideas as well!

 

[Acerun]

FWIW, I'm crossing at 80, using MultiEQ-X with a -0.45 tilt and this bass shelf

 

[mmares056]

FWIW, I'm crossing at 80, using MultiEQ-X with a -0.45 tilt and this bass shelf View attachment 470902

MQX is not very good at phase aligning subs to mains and you need to do some tweaking post EQ If REW is the only audio measurement tool you use, you can do the Sub Distance Tweak to phase/time align your sub to mains and find the best XO for your room. You may find that 80 Hz XO is not the right XO for your speakers in your room.

 

[Acerun]

MQX is not very good at phase aligning subs to mains and you need to do some tweaking post EQ If REW is the only audio measurement tool you use, you can do the Sub Distance Tweak to phase/time align your sub to mains and find the best XO for your room. You may find that 80 Hz XO is not the right XO for your speakers in your room.

Thanks. Ironically, the system sounds better than it ever has before...right now...I've never heard to measure Left and Right with subs. I was always told to measure separately. I have historically used 60 as the crossover, but now I think 80 sounds best.