Audyssey Room EQ Review

[Jim in Florida]

I'm just not sure how you'll get room EQ working without converting to digital.
.

I have Roon as a source. There is a room correction APP --- "REW" which aids in Room correction with a calibrated mike. You can actually bypass the receiver and go directly from the DAC to the amps if you want.

 

[Weeb Labs]

My concern was not noise but loss of source in the conversions. My DAC cost almost as much as my AVR receiver.

I wasn’t specifically referring to noise but to all potential non-signal products introduced by the conversions or volume control.

 

[Sal1950]

My concern was not noise but loss of source in the conversions. My DAC cost almost as much as my AVR receiver.

I hate to break it to you, but if your AVR has a halfway decent DAC in it and you could separate it's output from the rest of the AVR's DRC , etc. Then ran a bias controlled, level matched DBT, you'd be had pressed to identify any differences.
Been there, done that.

 

[markus]

Nice review @amirm The number one issue I have with (all) H/W solutions like this is the limited number of FIR filter taps to actually provide real low frequency room correction. Whether it is Audyssey, miniDSP, Trinnov, DEQX, or other H/W processors they all suffer from the same issue - not enough FIR filter taps below 100 Hz to really be called room correction.

Sorry for digging this up but I just listened to a podcast where you state the same thing and I think it's "inacourate" information so to speak Audyssey has shown years ago (when they introduced XT32) that low frequency FIR filter resolution can be substantially increased by clever signal processing, even at a low tap count (keyword "warped FIR filter").

Unfortunately all the graphs I've posted about 10 years ago on avsforum.com are lost but a quick search here brought up this: https://www.audiosciencereview.com/...vs-multeq-xt-vs-multeq-xt32.14786/post-460948

 

[Rebelhifi]

I want to thank everyone for the information. I applied Amir's estimated correction curve (Audyssey XT32) on my Marantz AV7705. The sound has been transformed with my Klipsch Ref Premier speakers. Also, I was able to boost the dual SVS Subwoofer curve (about 5 db). It has been very helpful and sounds great.

 

[Chromatischism]

Nice review @amirm The number one issue I have with (all) H/W solutions like this is the limited number of FIR filter taps to actually provide real low frequency room correction. Whether it is Audyssey, miniDSP, Trinnov, DEQX, or other H/W processors they all suffer from the same issue - not enough FIR filter taps below 100 Hz to really be called room correction.

As a maths example, the miniDSP 2x4 HD product datasheet talks about 4096 taps. But this is the total number. For 4 channels you have 1024 taps each channel available. The frequency resolution of a 1024 taps filter @ 48 kHz samplerate is 48000/1024 = 46.875 Hz. So e.g. below 100 Hz there are just 2 frequency bins at 46.875 and 93.75 Hz. This clearly means that you have no control over the lower frequency range.

As a graphic example, here is a Trinnov Altitude 16 versus Audiolense. Audiolense is a software based DSP/DRC product that is not limited by the number or FIR filter taps for low frequency control. In this case, Audiolense has generated a FIR filter with 65,536 taps, hosted in a software based convolution engine on a PC. We can clearly see the difference with the measured Trinnov correction on top and the measured Audiolense correction on the bottom using the same speakers same room, mic, etc:

View attachment 59916

Here is another example using Audiolense with a FIR filter length of 131,072‬ taps for the ultimate in low frequency control with 600ms of excessphase correction at 10 Hz:

View attachment 59917

Note that this measurement was taken at 9ft at the listening position using REW's default 500ms window and no smoothing. Meaning low frequency room reflections (i.e. standing waves, resonances) are getting into the measurement. But as one can see in the phase response, it follows the speakers minimum phase response, with no excessphase (i.e. no low frequency room reflections) disturbing the bass response. The result is crystal clear, even sounding bass response. To learn more, I wrote an article on the subject or one can hear my talk on it.

The point I am making is that not all DRC products are the same. Some have serious limitations on how much low frequency room correction can realistically be accomplished and folks should be aware of that.

I missed this post, but it is definitely not a concern for Audyssey as stated by @markus. Here are some measurements I've taken over time that show there is quite good resolution in the bass region:

Lots of squiggles there showing fine manipulation of the response. You can also see that Audyssey detects and avoids putting energy into true cancellations.

Multi-sub merge:

I would like to improve the 22 Hz dip but otherwise it is very smooth. And this is with a magnified 10-200 Hz view.

Another good one that shows that Audyssey has plenty of capability in the bass region. 2 subs. This included Audyssey's setting of delays + EQ to blend them as in the prior screen shot, plus the crossover into the FR speaker:

The proof of pudding is in the eating.

 

[mitchco]

@Chromatischism Not bad How about turning off all smoothing, go down to 5 Hz, and show the phase response as well:

 

[Chromatischism]

How about turning off all smoothing

I don't use any smoothing for bass except sometimes Variable (seen in shot 2), which doesn't really do anything under 100 Hz.

With my vented subs I don't have a lot of output at 5 Hz. With the sealed subs though it was tremendous:

There is no smoothing applied; the response is so smooth because I always place my subs against walls.

 

[mitchco]

@Chromatischism can Audyssey correct for excess phase at low frequencies? Can you show the phase response like I have shown in my chart measured at the listening position?

See section on "A common cause of non-minimum phase behaviours in rooms" in John's paper on minimum phase. For myself and others using excess phase correction at low frequencies the subjective listening difference is not only smooth bass, but clear bass.

 

[Chromatischism]

can Audyssey correct for excess phase at low frequencies? Can you show the phase response like I have shown in my chart measured at the listening position?

See section on "A common cause of non-minimum phase behaviours in rooms" in John's paper on minimum phase. For myself and others using excess phase correction at low frequencies the subjective listening difference is not only smooth bass, but clear bass.

Phase:

We aren't using the same scales – this measurement was not taken to 200 Hz so I don't have that data. Starting from 100 Hz, the first full phase rotation is at 25 Hz.

I would be curious what audible difference this makes, if any. The bass in this room is subjectively excellent until a few issues arise at some resonant frequencies in the 80-100 Hz range that are room-related. Floyd Toole and others make it clear that we don't hear phase, only the effects of phase on frequency response. Thoughts?

 

[mitchco]

Based on your graph, it would appear that Audyssey does not provide excess phase room correction capabilities.

I am taking about the audible effects of low frequency room reflections like in John's paper I linked to above under the section: "Axial modes in a rectangular room. To provide a simple example of how the summation of the signals in a room can make it non-minimum phase, even at low frequencies, we can look at the behaviour of axial modes in a perfectly rectangular room."

Myself and others are saying there is a clarity improvement in the bass response in the room when using excess phase room correction versus without. Like 800ms of excess phase correction starting at 10 Hz using a frequency dependent window. The effect is not as dramatic as smoothing out the magnitude response, but clearly audible when switching between a correction filter with time domain correction and one without. Here is a measurement example to demonstrate:

Some 28 milliseconds later in the room one is hearing a maximum phase reflection (additive peak) of the same magnitude or larger than the direct sound. From an audible perspective the bass sounds somewhat unclear or blurred or doubled.

This is a 3 way digital XO system, the DSP software also time aligns the individual drivers and uses a linear phase digital XO to sum properly in both the frequency and time domain plus flips the polarity of the subs, in addition to the magnitude and separate excess phase correction. Yields:

Basically ideal minimum phase response arriving at one's ears. Accurate sound.

As mentioned in my post you quoted not many room correction programs have the capability to correct for the rooms full transfer function.

Anyway, all in good fun

 

[Sancus]

Like 800ms of excess phase correction starting at 10 Hz using a frequency dependent window.

Doesn't that mean 800ms of delay though? That would cause a lot more problems than it solves for most normal uses of an Audyssey-containing AVR, I'd guess. You'd need to be able to turn it on and off very easily and quickly.

 

[mitchco]

Of course there are tradeoffs and it depends on one's use case. As mentioned in an earlier post, I use JRiver which accounts for the filter delay when watching movies. My original post and the later one's here are discussing what makes for the state of the art in room correction. Not all room correction systems are the same.

 

[ernestcarl]

To me, even correcting some minor but "fixable" abrupt phase rotations and extended group delay spots in the time domain -- of low bass frequencies -- can turn out to be a night and day difference. It's relatively simple enough to A/B test this for onself.

If you are just flattening the phase in-post after minimum phase crossovers have already been applied, there may be no visible difference in the magnitude frequency response at all. However, applying frequency dependent windowing (FDW) with a low number of cycles might reveal a different story.

While it's true that miniDSP doesn't quite have enough taps available to fix some serious low bass issues, one might be able to nudge the phase just enough to achieve wider coherency across the desired crossover range. Don't get me wrong, I agree that the low number of taps seriously limits the device's capabilities, but I've found myself able to squeeze the heck out of it to gain maximum utility.


miniDSP partial xo phase correction in the left. Additional phase correction via JRiver's convolution engine in right.

See that dip around 80 Hz? It only appears in the graph with frequency dependent windowing applied.


Yeah, the bass looks kind of low below 30 Hz since (in this particular scenario) I was equalizing for the overall room response -- which also maximizes the limited dynamic headroom of my sub & speakers -- i.e. so I can play the drivers as loud as possible without compression or significantly increasing subwoofer bass distortion.

So the overall room -- call it "power" -- response is nevertheless balanced.

We can see that flattening group delay also increases measurable "clarity" traces in the bass at our critical xo junction.

And the wavelet spectrogram looks more linear after fixing that very abrupt (likely room induced) phase rotation / group delay peak:

My EQ strategy here is actually very simple -- a subtractive one -- primarily reducing excess energy in the room:

High pass and low pass filters only for the subwoofer and none applied (or full-range) for the front LR mains.

Crude phase correction curves I came up using rePhase:

Some settings & additional info after generating filters in rePhase

SUBWOOFER Channel
taps: 2036 samples
windowing: hann
sample rate: 96000 Hz
impulse delay: 1018 samples, 10.604 ms
max response: 0 dB, max impulse -0.14 dB

Lt & Rt FRONT MAIN Channels
taps: 1024 samples
windowing: hann
sample rate: 96000 Hz
impulse delay: 511.857 samples, 5.322 ms
max response: 0 dB, max impulse -5.4 dB

JRiver stereo LR (optional) secondary FIR filter
taps: 12228
windowing: hann
sample rate: 48000 Hz
impulse delay: 6113.999 samples, 127.375 ms
max response: 0 dB, max impulse: -0.09 dB



Here are the raw phase responses using only miniDSP's (2x4 HD) severly limited filter tap capabilities

Well I say, that's not too bad at all!


Our Sub+Mains Summation Result

No dips created thanks to the largely coherent phase curves.


Lastly (and finally!), the step response:

 

[markus]

Doesn't that mean 800ms of delay though? That would cause a lot more problems than it solves for most normal uses of an Audyssey-containing AVR, I'd guess.

That's exactly why long FIR filters can only be used for music reproduction. A sample first has to go through the whole FIR filter until it can be heard. This introduces a significant delay but low latency is crucial these days for virtually all relevant applications like gaming, movie watching, streaming.

 

[Steve Dallas]

I recently replaced the plate amps in both my SVS S2000 subs and needed to run XT32 again to re-calibrate. I happened to have the house to myself for a couple days this weekend and finally had the right circumstances to do just that.

After much measuring with REW, moving subs, changing phase, manually setting crossover points and speaker / sub levels, correction limits, correction curve, etc, I finally managed these measurements, which are the best I have seen from XT32 in this terrible room, IIRC: (Dirac does a little better job with similar intervention, but without sub integration in my setup.)

This is with 2 subs set to 180 phase, crossed at 80Hz, correction curve at +3dB @20Hz, and correction limited to below 1100Hz. There are a few obvious places I wish I could have coaxed XT32 to smooth more, but it is not bad compared to many other measurements I have seen. I may pull those low frequency rises down a bit, to average things, depending upon how my listening tests go. A 1dB reduction in the curve editor may do the job, but things are unpredictable with XT32, so adjusting a measuring and adjusting and measuring is the name of the game. I simply ran out of quiet time with this round.

For comparison, here is what Audyssey did on the left side in letting it correct the entire FR with all default settings except for turning off midrange compensation:

A little thin in the bottom and managed to make above 1K look a little worse. Still miles ahead of my uncorrected situation.

All-in-all, XT32 is an invaluable tool, even for the typical consumer. It can yield measurably better results with additional knowledge and equipment.

 

[Chromatischism]

@Steve Dallas show us a graph with limits of 10-300 Hz and a vertical scale that gets you to 5 dB increments. On my screens that is either 45-105 dB (the 1080p laptop) or 50-100 dB (the 1440p PC).

Also I'm surprised you are using subs with 180 phase. Normally the best result and the general recommendation is to put them at 0 before Audyssey.

 

[Chromatischism]

That's exactly why long FIR filters can only be used for music reproduction. A sample first has to go through the whole FIR filter until it can be heard. This introduces a significant delay but low latency is crucial these days for virtually all relevant applications like gaming, movie watching, streaming.

What is the practical implication here? I haven't heard of gamers, etc, noticing issues with Audyssey (FIR) EQ.

 

[markus]

What is the practical implication here? I haven't heard of gamers, etc, noticing issues with Audyssey (FIR) EQ.

Short delays usually aren't a problem (see image below) but the filters user mitchco is promoting are way longer and therefore work only if there's no accompanying video content. In my mind a universally applicable room correction system needs to build around IIR and very short FIR. Dirac Live does that but I'm not sure their filters are as good as they could be. If I find the time I might write down my thoughts on it in detail.

Source: https://www.prosoundtraining.com/site/wp-content/uploads/2014/02/Lip-Sync-Errors.pdf

 

[Steve Dallas]

@Steve Dallas show us a graph with limits of 10-300 Hz and a vertical scale that gets you to 5 dB increments. On my screens that is either 45-105 dB (the 1080p laptop) or 50-100 dB (the 1440p PC).

Also I'm surprised you are using subs with 180 phase. Normally the best result and the general recommendation is to put them at 0 before Audyssey.

I tried it with phase set to 0, 90, and 180. 180 measured best.

Here are the zoomed measurements: