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

[GalZohar]

As far as I'm aware the AVR crossover filters and Audyssey filters are not related (at least not directly). What one expects from the other isn't 100% clear, but supposedly it would make sense that the AVR expects the speaker and/or Audyssey to roll off at 12db/oct past the crossover, which in reality usually doesn't happen exactly like that, so maybe there is something to gain with manual modifications, which is why I was wondering if anyone had success.

I don't have the MultEQ-X, but seems like after tweaking the subwoofer distance I don't really have any serious cancellations below the crossover, just a small dip at ~70Hz for 1 speaker which I'm not sure if it can be helped. I see it's a cancellation because that frequency has higher SPL if I disconnect either the subwoofer or the speaker.

 

[Reverend Slim]

You can measure with REW to see what your response looks like. What does it show?

Well, as previously mentioned, I can't... because I don't have a laptop or UMIK-1. I run MultEQ X off a desktop PC in a whole other room and just walk back and forth to move the mic around on my tripod.

I guess I could get a general idea with noise tones and Spectroid on my phone, at least for sake of comparison (since it obviously won't be accurate).

 

[Chromatischism]

As far as I know, Audyssey doesn't make filters below the detected F3 of your speaker. That doesn't mean it's rolling off the speaker, but there could be room interactions that come into play to give one that impression. The crossover filter is implemented entirely in the AVR and will combine with the natural roll off of the speaker to get 24 or 36 dB / octave, give or take, since some speakers have a tapered bass rolloff (KEF, Klipsch). From Audyssey's point of view, all speakers would ideally be mostly straight lines after EQ and before handing them off to bass management. This is how Dirac works as well. Audyssey makes decisions on what to EQ based on the detected roll-off of the speaker, and that can in some cases combine with room effects to cause the system to make the wrong decisions, but it mostly does a good job.

 

[Reverend Slim]

As far as I know, Audyssey just doesn't make filters below the detected F3 of your speaker. The crossover filter is implemented entirely in the AVR and will combine with the natural roll off of the speaker to get 24 or 36 dB / octave, give or take, since some speakers have a tapered bass rolloff (KEF, Klipsch). From Audyssey's point of view, all speakers would ideally be mostly straight lines after EQ and before handing them off to bass management. This is how Dirac works as well. Audyssey makes decisions on what to EQ based on the detected roll-off of the speaker, and that can in some cases combine with room effects to cause the system to make the wrong decisions, but it mostly does a good job.

As I said before, that doesn't seem to be the case here. See pp. 24-25 of the MultEQ-X User Guide. Also, see p. 19 for how it shows the detected cutoff of the speaker. Hence the confusion.

 

[GalZohar]

Yes, the documentation is confusing, and there seem to be 2 separate options, 1 for the frequency below which no corrections are applied, and one for the rolloff. It's not clear how they interact. Seems like unless some appropriate explanation is found, a measurement would be needed to settle this confusion.

 

[Reverend Slim]

Yes, the documentation is confusing, and there seem to be 2 separate options, 1 for the frequency below which no corrections are applied, and one for the rolloff. It's not clear how they interact. Seems like unless some appropriate explanation is found, a measurement would be needed to settle this confusion.

Like I said earlier in the thread... If you set a cutoff on the target curve page, you have to leave that channel on Auto on the following page because it's using the filters to shape that cutoff. If you change it from Auto to a frequency to stop EQ'ing, that attempt at creating that rolloff goes away.

Jeff Clark's gonna be live with Audioholics tomorrow night regarding either the ACM1-X or the line level to mic level adapter that's showing up on Amazon for people to use their old Audyssey Pro mics. Maybe someone can hit him with the question during that.

 

[anotherhobby]

Yes, the documentation is confusing, and there seem to be 2 separate options, 1 for the frequency below which no corrections are applied, and one for the rolloff. It's not clear how they interact. Seems like unless some appropriate explanation is found, a measurement would be needed to settle this confusion.

If I were to guess, I'd put my money on that it's just the cutoff frequency below which no corrections are applied. I was calling it a rolloff because that's what it looks like in the MultEQ-X, they call it a Cutoff, and what I measured was my sub rolling off earlier than I wanted. However, if you look closely below with Cutoff set to Auto vs Disabled, some of my corrections do indeed overlap with the 20 Hz cutoff it was applying. This is visible on the right after I disable the cutoff.

And so nobody has to go back and look, here was the resulting measurement difference at the primary listening spot by disabling the cutoff:

I think disabling the EQ along that cutoff curve accounts for the difference I saw in measurements. The point of my original point was to mainly say that I am happy that MultEQ-X lets me disable this, because I was able to get the result I wanted. All of the EQ in the miniDSP for the subs by MSO are EQ cuts that are bigger than the small bumps I added back in to smooth it out more. I'm just effectively cutting them slightly less and with a little more precision than 5 MSO filters can do.

I don't know with any certainty at all which way it works, and I don't have time to test this theory with measurements any time soon. However, I would guess that if I disabled all of the PEQs for the subs in MultEQ-X, and then compared cutoff modes, it would probably measure the same with either cutoff mode set. It'd be interesting to try though and see what happens.

 

[Chromatischism]

As I said before, that doesn't seem to be the case here. See pp. 24-25 of the MultEQ-X User Guide. Also, see p. 19 for how it shows the detected cutoff of the speaker. Hence the confusion.

I see what you mean; that section should be re-written to give a lot more detail and context, with graph examples.

But if you think about it, it's not inconsistent with what I said. You can apparently override the auto-detection of the low-frequency roll-off. That is, what Audyssey assumes about the roll-off in order to determine where to stop applying EQ. It looks to be using a trendline, not unlike a moving average that crosses the unsmoothed line. The goal is to not over-boost a speaker, especially ported speakers below their bass reflex tuning frequency.

@anotherhobby above shows that when he tells Audyssey to not even assume he has a rolloff, it continues to EQ his subs. He's essentially boosting the low bass up to flat, which is putting a lot of power into those sealed subs because they naturally roll off by -10 dB at 17 Hz. They do give a warning in the User Guide when using it in this way just so they aren't liable if anyone blows something up.

So TL;DR, he's not disabling a software roll-off, he's boosting his subs.

 

[GalZohar]

That's why I think it would make more sense to measure higher cutoff frequencies than the default, as that might have some use in improving speaker-subwoofer integration. The guide does mention the 2nd order for best integration with AVR's 2nd order crossover.

 

[anotherhobby]

@anotherhobby above shows that when he tells Audyssey to not even assume he has a rolloff, it continues to EQ his subs. He's essentially boosting the low bass up to flat, which is putting a lot of power into those sealed subs because they naturally roll off by -10 dB at 17 Hz. They do give a warning in the User Guide when using it in this way just so they aren't liable if anyone blows something up.

So TL;DR, he's not disabling a software roll-off, he's boosting his subs.

There is only an EQ boost in the Audyssey curve, but that's on top of a larger cut already applied by miniDSP. There is no effective boost applied anywhere to the subs. Clearly you don't get from the uncorrected curve to the corrected one by EQ boosting.

 

[Reverend Slim]

I see what you mean; that section should be re-written to give a lot more detail and context, with graph examples.

But if you think about it, it's not inconsistent with what I said. You can apparently override the auto-detection of the low-frequency roll-off. That is, what Audyssey assumes about the roll-off in order to determine where to stop applying EQ. It looks to be using a trendline, not unlike a moving average that crosses the unsmoothed line. The goal is to not over-boost a speaker, especially ported speakers below their bass reflex tuning frequency.

@anotherhobby above shows that when he tells Audyssey to not even assume he has a rolloff, it continues to EQ his subs. He's essentially boosting the low bass up to flat, which is putting a lot of power into those sealed subs because they naturally roll off by -10 dB at 17 Hz. They do give a warning in the User Guide when using it in this way just so they aren't liable if anyone blows something up.

So TL;DR, he's not disabling a software roll-off, he's boosting his subs.

Bear in mind also that he's talking about on the subs (which default to no cutoff in MultEQ-X from what I can tell) whereas I'm talking about how it behaves on every other channel (where the AVR's crossover would interact with any filtering). For me (and probably many others), it makes sense not to have any cutoff on the subs done by Audyssey, since my subs already have their own built-in protection. The question is: Do you need Audyssey's "protection" by way of the rolloff it tries to apply to the other channels if they're already being filtered by the AVR's crossover?

I mean, the guide mentions 2nd order for best integration... so are Audyssey's filters being applied to the channels AFTER the AVR's filtering or before? I'm just wondering how aware Audyssey is of that interaction.

 

[GalZohar]

I'm not sure there is any importance to the order of filters, as it's just addition in the end, isn't it? Supposedly the "standard" crossover is 24db/oct, while the AVR provides only 12db/oct, so forcing the speaker to roll off at 12db/oct from the crossover point down (assuming it's capable of at least that much) could be beneficial. I was wondering if that actually works as expected and if anyone managed to actually improve final subwoofer integration by using this feature.

 

[Chromatischism]

Bear in mind also that he's talking about on the subs (which default to no cutoff in MultEQ-X from what I can tell) whereas I'm talking about how it behaves on every other channel (where the AVR's crossover would interact with any filtering). For me (and probably many others), it makes sense not to have any cutoff on the subs done by Audyssey, since my subs already have their own built-in protection. The question is: Do you need Audyssey's "protection" by way of the rolloff it tries to apply to the other channels if they're already being filtered by the AVR's crossover?

I mean, the guide mentions 2nd order for best integration... so are Audyssey's filters being applied to the channels AFTER the AVR's filtering or before? I'm just wondering how aware Audyssey is of that interaction.

I am still not convinced Audyssey applies a roll-off to speakers. Measurements would easily clear this up though.

 

[Reverend Slim]

I am still not convinced Audyssey applies a roll-off to speakers. Measurements would easily clear this up though.

I'll be glad to take a screenshot of the filters it's applying on mine. As I said earlier, changing the point where equalization stops to anything other than Auto removes the rolloff resulting from the assigned cutoff on the target curve page.

Also, see p. 25 of the user guide:

Override – The user may override the cutoff Frequency and Order. Lowering the frequency can subject the speaker to too much boost, and too much power. Raising the frequency can blend better with manual bass management. Changing the order to a higher number increases the steepness of the cutoff. Changing to a very high order / steep cutoff may tax the limits of EQ to match desired target curve. It is recommended to only use orders above 4 in order to match the existing steep roll-off of the speaker. Note that by matching the roll-off of the speaker, the net outcome is there is no EQ needed.

If it isn't using the filters to implement a rolloff, why would it "tax the limits of EQ to match desired target curve"?

See also on p. 24:

Cutoff Mode is closely related to “Low Frequency EQ Limit” (Filter Settings page) as well as the Bass Management Speaker Size and Frequency (Calibration Settings page). Use these three settings together to achieve the desired low frequency filtering.

Just saying... We need a better explanation of the interaction between this and the filtering in the AVR itself.

 

[Andysu]

i don't use it as it would mess up additional fancy matrix decoding as the signal has to be in correct freq/level/phase .
has anyone taken the speakers that is going to be used indoors and taken them outdoors and ran , rew frequency sweeps to see what the actual frequency is ? then take them back indoors place them where they will be finally placed after moving them around a small bit with more repeated . rew frequency sweeps and looking at the overlays and then seeing how it looks in the room with some frequency been added and subtracted due to the room ? also running the auto-eq outdoors as well to see what the results are vs indoors see most have a gardens to do these tests .

 

[Reverend Slim]

I just fired off the following to Audyssey's support and will post their response when I receive one:

MultEQ-X detects some of my speakers (SVS Prime Satellites) at around what their known F3 is, but while the measured rolloff of the speaker is roughly in line with 4th order, the detected cutoff shown on the aggregate measurement is 2nd order (presumably due to a bump down near 50Hz if I had to guess). To make it line up better with the speaker's natural rolloff, I was using the cutoff override to manually assign those speakers a 4th order filter and tweaking the frequency until the filtering closely aligned with the detected rolloff when looking at the Filter Settings page.

My question is: How does the crossover in the AVR then interact with that rolloff? In other words, is MultEQ-X taking into account the AVR's known filter slopes so you don't have cascading filters? The user guide is unclear as to how these settings interact.

Related question: Does Audyssey's filtering take place before or after the crossover applied by the AVR in the signal chain?

 

[Reverend Slim]

Hats off to Audyssey's support. They answered my question fast!

Our recommendation is to move the crossover frequency up higher, and let MultEQ-X create the 2nd order roll-off. The 2nd order roll-off of the satellite should theoretically align better with the AVR bass manager, which also has a 2nd order high-pass filter. Between the two of them, it makes a Linkwitz-Riley (LR4) crossover that should be everywhere in-phase. That is, the cascaded filters are desirable and part of the design of the bass management system. The AVR uses cascaded 2nd order low-pass filters on the subwoofer to complete the LR4.

We would recommend using the 4th order roll-off only when the speaker is used as "large" - i.e., you want to match the speaker's native roll-off because bass management isn't used.

Audyssey's filtering is AFTER the crossover by the AVR. In MultEQ and MultEQ XT level systems, this is advantageous because the subwoofer signal is narrow bandwidth and can receive extra processing without needing extra DSP power.

 

[GalZohar]

Nice answer, so pore or less what I expected. Now the question remains if anyone managed to get some meaningful improvement by using the suggested settings rather than the default, as I would rather avoid spending 200$ if it doesn't help much.
In any case, seems like if one does as suggested, he should get similar subwoofer integration results as can be obtained with ARC or Dirac (DLBC excluded), but at a much cheaper price.

 

[sweetchaos]

New video...currently live streaming:

 

[2Bad]

Just finished watching the live stream... no commited Mac support so count me out.