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

[Dj7675]

Well ... I´ll be damned ... wow ... they really delivered . ... REW support with the latest version (1.1.420.0), which just came in this morning. Unfortunately, I am on my way out, so I cant try it.

View attachment 206600

That is a nice feature. The most promising thing is the continued development in features and fixes. I think that bodes well.

 

[peng]

That is a nice feature. The most promising thing is the continued development in features and fixes. I think that bodes well.

In addition to features and fixes, I hope they will also work with D+M's engineering to improve on impulse response (relative to Dirac Live) for the 2022/23 models, assuming they will still use Audyssey.

 

[Senior NEET Engineer]

It looks to me like a combination of some nulls at <=40Hz and large peaks above. The nearfield looks fine down to 20Hz so it's probably not the fault of your subwoofer itself but rather the room.

You should probably analyze the room modes with test tones at the relevant frequencies while walking around the room. What are the room dimensions? Looks like your first major room mode is at 50Hz, but there are less significant ones below 40Hz. Nulls rarely would account for lack of bass at the entire 20-40Hz range, so it's probably mostly just the peaks above 40Hz combined with a few nulls below 40Hz.

If you can, try find a listening position where you will get a better response at the 20-40Hz range. You can listen to test tones to see where the peaks and dips are for each frequency, and try to avoid placing your seat where there are severe peaks/dips, or at least where there are too many of them. Of course you should also take into account the placement of the speakers (and screen), so it ends up being a bit difficult to find a position that works for everything, and you'll have to find the best compromise.

The room is 21 x 13.5 x 10.5'.

I moved the subwoofer so that the subwoofer driver was inches away from the speaker bass drivers. Bass response was nearly identical from 40-110hz. Comparing to the mid and high frequency response of the speaker, I see that < 40hz looks fine and 40-100hz is elevated. I think this confirms that there isn't any bass suck out issue. Just some huge peaks starting at 50hz like you said.

 

[HarmonicTHD]

The room is 21 x 13.5 x 10.5'.

I moved the subwoofer so that the subwoofer driver was inches away from the speaker bass drivers. Bass response was nearly identical from 40-110hz. Comparing to the mid and high frequency response of the speaker, I see that < 40hz looks fine and 40-100hz is elevated. I think this confirms that there isn't any bass suck out issue. Just some huge peaks starting at 50hz like you said.

Did you use a room mode calculator? Or alternatively the room Simulator as part of REW? Gives you at least an initial idea what might be going on.

Calculate Room Modes

www.mh-audio.nl

 

[MarkI]

Hello,

I just became a member. Was reading this tread last night. I’ve been a member over at the AV science form for quite some time.
Audio Science Review seems like a better fit for me.
A bit of background, I got multEQ-X several months back. I’ve spent well over 100 hours playing around with it. Done dozens of different profiles, played with every PEQ filter there is, gotten some great results and obviously some horrible ones.

A few weeks back on some of the forms / threads, appears that Audessy / Marantz, Denon had a time alignment issue.
So I immediately took some readings using REW with my Umik-2.
The results I got were odd. So like always, because this is a hobby for me and I enjoy it. I proceeded to try to get the actual measurements not what multEQ-X is or what My Marantz receiver has. I know there’s some secret sauce around all of this. I’m not trying to start some controversy Or create some scandal.

If you look at the screenshot of my notes. I ended up spending quite a bit of time to force MultEQ-X / my Marantz 7705 to get the actual real distances from the main listening position. After it produced results that I’ve never had before. I could hear my rear speakers. Meaning I’m in a full bubble of sound now. Before when listening to stuff in Neural:X - Dolby Atmos, was like being in Half to 2/3 of a bubble.
When using the profile that has the corrected distances the sides and rears have a seamless pan affect. And again the rears are significantly more present.
And yes of course I’ve turned up the rears and played around with different volumes settings. it’s not the same.

If you’d like please take a look at my dedicated theater https://www.avsforum.com/threads/sunken-treasure-theater.3243916/
it’s in a huge sunken living room. The volume of space around me is gigantic. When listening or watching anything it already sounds massive me. So I can only surmise that Audessy think or wants it to be big/small space, who knows.

I’ve been told before that Audessy’s eight Mike positions takes in to consideration all of this stuff. ???

So have any of you played around with this. Maybe tried moving the distances around. Have you gotten better or worse results. Have you also re-run the first mic position. Meaning you can use the new app and right click each of the speaker, then retake that measurement. move the mic around until you get the exact distance needed. This way the delay is close to what it should be. See picture below
This is the profile I saved where It's corrected. Because my understanding is the first mic position is what does the distance and delay.??? I also have another profile that just has the distances changed but the delays are off, which sounds awesome?

 

[KMO]

If I'm reading your numbers correctly, there's almost no change there (ignoring the sub). The addition of 30cm to every speaker doesn't have a net effect. Mainly Rs, Lrs and Rrs are now somewhat closer, so more delayed, while everything else has stayed basically the same.

Which is odd, as you initially seemed to be showing L, C and R as being the odd 3 out - was that from physical measurements?

Are you saying you adjusted by looking at REW measurements?

Audyssey indeed uses the first mic position to determine distances. And you're free to modify manually, you don't have to go via the app.

 

[MarkI]

If I'm reading your numbers correctly, there's almost no change there (ignoring the sub). The addition of 30cm to every speaker doesn't have a net effect. Mainly Rs, Lrs and Rrs are now somewhat closer, so more delayed, while everything else has stayed basically the same.

Which is odd, as you initially seemed to be showing L, C and R as being the odd 3 out - was that from physical measurements?

Are you saying you adjusted by looking at REW measurements?

Audyssey indeed uses the first mic position to determine distances. And you're free to modify manually, you don't have to go via the app.

Thank s for the response, I messed up please look again. I had imported the original numbers wrong, I carefully hand measured all speakers from #1 Mic position. That's what got me going down this track. If you look again multEQ-X is giving different measurements, not that much different but different than the actual physical measurement. So one would think that I could go into MultEQ-X, just type in the actual numbers. However it doesn’t work that way and that’s the part I don’t understand but that may be the secret sauce thing that I’ve heard other people talk about. And yes I understand I can go into the AV directly and simply move things a foot back or a foot forward. Don’t fully understand how that affects the delay/timing. I suppose it can’t hurt. The unaffected consequence of what I did was again I got a better very noticeable sound bubble around me.

 

[HarmonicTHD]

Thank s for the response, I messed up please look again. I had imported the original numbers wrong, I carefully hand measured all speakers from #1 Mic position. That's what got me going down this track. If you look again multEQ-X is giving different measurements, not that much different but different than the actual physical measurement. So one would think that I could go into MultEQ-X, just type in the actual numbers. However it doesn’t work that way and that’s the part I don’t understand but that may be the secret sauce thing that I’ve heard other people talk about. And yes I understand I can go into the AV directly and simply move things a foot back or a foot forward. Don’t fully understand how that affects the delay/timing. I suppose it can’t hurt. The unaffected consequence of what I did was again I got a better very noticeable sound bubble around me.

Ah no don’t type in the actual numbers. The head of development for Audyssey stated eg in an interview with JoeNTell that there are differences in the distances and one should not adjust manually.

However, and I think KMO was essential to that discovery, it was found some two weeks ago, that there is an inaccuracy in the speed of sound value, which the amp uses for distance calculation. I forgot, but somewhere here it was described in detail what is wrong and how to adapt.

Please KMO, I am quoting from memory here, correct me if I am wrong.

 

[KMO]

Thank s for the response, I messed up please look again. I had imported the original numbers wrong, I carefully hand measured all speakers from #1 Mic position. That's what got me going down this track. If you look again multEQ-X is giving different measurements, not that much different but different than the actual physical measurement. So one would think that I could go into MultEQ-X, just type in the actual numbers. However it doesn’t work that way and that’s the part I don’t understand but that may be the secret sauce thing that I’ve heard other people talk about.

Okay, MultEQ-X is adding an average 13cm over "actual" in the first section, possibly due to some sort of overall latency. But a constant offset like that doesn't matter, as it's only the relative distances that are significant.

Subtracting off the constant 13cm, can help visualise the changes:

Speaker	MultEQ-X	Actual	MultEQ-X - 0.13cm	Difference
L	3.85	3.68	3.72	+0.04
R	3.85	3.68	3.72	+0.04
C	3.57	3.47	3.44	-0.03
Ls	3.24	3.10	3.11	+0.01
Rs	2.40	2.21	2.27	+0.06
Lrs	2.63	2.46	2.50	+0.04
Rrs	2.61	2.46	2.48	+0.03
Ltf	3.45	3.34	3.32	-0.02
Rtf	3.44	3.34	3.32	-0.02
Ltr	3.30	3.22	3.17	-0.05
Rtr	3.32	3.22	3.19	-0.03

Those are pretty small adjustments, so find it hard to find it is that significant. Have you looked at the REW results to confirm whether your measurements are producing more-aligned results? Note that the AVR's delay function only works in 3cm steps itself - we're quite close to its resolution.

When working at this level of precision, I begin to wonder how much different speakers might matter - can group delay due to different crossover types start to show up? It might be worth a couple of centimetres equivalent. I wouldn't be totally confident about saying that a tape measurement is better than MultEQ-X's audio measurement without further testing, unless the speakers are actually identical.

The bigger "secret sauce" correction is that the numbers (actual or MultEQ-X displayed) get reduced by a factor 0.875 when entered into the AVR. If you have a real FL measurement of 3.85 or 3.68, then they need to be entered into the AVR as 3.37 or 3.22. And MultEQ-X is already doing that. That correction is bigger than the changes you made - for the most extreme speaker pair of L versus Rs, it's equivalent to an 18cm shift. Your biggest shift was 11cm between Rs and Ltr.

 

[KMO]

Ah no don’t type in the actual numbers. The head of development for Audyssey stated eg in an interview with JoeNTell that there are differences in the distances and one should not adjust manually.

We now know how to adjust manually, so if you want to put tape distances into the AVR, just multiply by 0.875 (for metres) or 0.889 (for feet). But if you have MultEQ-X, and can manually enter numbers into that, then it will do that multiplication for you.

 

[KMO]

I'm still not clear what change MarkI actually made. Was it entering tape distances directly into the AVR? If so, that's mainly undoing the 0.875 adjustment, which would render the timing less precise. Which might be preferable for some reason, although I'm a bit doubtful about the human ability to clearly tell the difference.

Despite being involved in making these corrections, I don't claim I can tell the difference, apart from in an REW graph.

 

[Ruspamen]

suppose I use audyssey only to calculate distances, on the avr should I leave the modified ones or should I enter the ones you see in the multeq-x screen on the PC?

 

[KMO]

suppose I use audyssey only to calculate distances, on the avr should I leave the modified ones or should I enter the ones you see in the multeq-x screen on the PC?

If you used Audyssey MultEQ-X to get the distances, then the smaller numbers it put into the AVR are already correctly modified to account for the AVR's wonky distance->delay conversion.

You do not want to enter the real distances MultEQ-X shows you on the PC into the AVR.

All this talk of manual adjustment is for people who are not using MultEQ-X.

 

[Ruspamen]

If you used Audyssey MultEQ-X to get the distances, then the smaller numbers it put into the AVR are already correctly modified to account for the AVR's wonky distance->delay conversion.

You do not want to enter the real distances MultEQ-X shows you on the PC into the AVR.

All this talk of manual adjustment is for people who are not using MultEQ-X.

i understand thank you, so the measures in the multeq-x pc screen should never be used? are valid only those transferred and modified on the avr? thank you

 

[KMO]

i understand thank you, so the measures in the multeq-x pc screen should never be used? are valid only those transferred and modified on the avr? thank you

Well, they're real distances, and older versions of Audyssey would have entered them as-is. Entering real distances is good, but doesn't produce the best possible results. It's useful for MultEQ-X to show the real distance just to check whether it looks valid.

But it's a recent "discovery" (for Audyssey and us) that Denon AVRs have a weird distance->delay conversion, so you get better results by adjusting all the real distances by 0.875 to compensate.

 

[Ruspamen]

Well, they're real distances, and older versions of Audyssey would have entered them as-is. Entering real distances is good, but doesn't produce the best possible results. It's useful for MultEQ-X to show the real distance just to check whether it looks valid.

But it's a recent "discovery" (for Audyssey and us) that Denon AVRs have a weird distance->delay conversion, so you get better results by adjusting all the real distances by 0.875 to compensate.

So the measurements of the denon screen, match the measurement of multeq-x screen multiplied by 0.875? is this the right way?

 

[KMO]

So the measurements of the denon screen, match the measurement of multeq-x screen multiplied by 0.875? is this the right way?

They should match like that, yes, and we've established with our own REW measurements that this should produce the best results.

 

[peng]

Well, they're real distances, and older versions of Audyssey would have entered them as-is. Entering real distances is good, but doesn't produce the best possible results. It's useful for MultEQ-X to show the real distance just to check whether it looks valid.

But it's a recent "discovery" (for Audyssey and us) that Denon AVRs have a weird distance->delay conversion, so you get better results by adjusting all the real distances by 0.875 to compensate.

Do you know if this only applies to all D+M models of a year model or only since year X models? Thank you.

 

[KMO]

Do you know if this only applies to all D+M models of a year model or only since year X models? Thank you.

We suspect it's all D+M AVRs ever.

I was testing with my relatively-old AVR-4308 (2007 vintage), and others have been using current AVR-Xx700 models.

The behaviour arises fairly naturally from just making the maths really simple for a relatively dumb integer microcontroller, so it was probably a choice made back in the mists of time when the distance UI was first added. 3cm = 0.1ft = 0.1ms is easy.

So the chances of an even-older-than-2007 AVR having a better speed-of-sound constant are very low, and there's no sign of any change in new AVRs. And changing new AVRs would just add confusion, what with MultEQ-X and now us manually fudging the numbers.

 

[peng]

We suspect it's all D+M AVRs ever.

I was testing with my relatively-old AVR-4308 (2007 vintage), and others have been using current AVR-Xx700 models.

The behaviour arises fairly naturally from just making the maths really simple for a relatively dumb integer microcontroller, so it was probably a choice made back in the mists of time when the distance UI was first added. 3cm = 0.1ft = 0.1ms is easy.

So the chances of an even-older-than-2007 AVR having a better speed-of-sound constant are very low, and there's no sign of any change in new AVRs. And changing new AVRs would just add confusion, what with MultEQ-X and now us manually fudging the numbers.

Thanks again, I asked because I have had the AVR-4308CI, AV7005, AV8801 and AVR-X4400H (tried both the Editor App and AVR to run Audyssey) and never noticed any such differences as displayed on the screen vs physical tape measurements, may be 0.1 to 0.5 ft (<15 cm) or so that I would usually not bothered doing anything about. May be I would have noticed such differences if I checked it with REW but I had no reason to do so, until now I guess. Regardless, I doubt the difference would be audible anyway.., just being curious.