Big news coming from Sound United in 2023!

[chych7]

That applies only to a minimum phase system. A loudspeaker is usually not that so you need also the phase response. What dirac does is that it corrects for the phase shift due to the crossovers of the speakers. That's why it's imulse looks cleaner.

I still wonder about the whole phase correction part that Dirac does. From what I understand, we can't hear a speaker's phase. But, we can hear phase differences or misalignment of phase across multiple emitters, like between a speaker and sub, or left and right speaker (inverting the polarity of one speaker easily confirms this). From my study, I looked at the phase mismatch between the left and right speakers, and found that Audyssey did a better job at reducing error between the left and right (and also was better at matching SPL between left and right speakers). Subjectively Audyssey sounded better. Well, this was the case in my system and speakers, YMMV.

 

[GalZohar]

Well, there’s a thread on that. To summarize, aside from the unacceptable licensing terms and high cost…it doesn’t even run on macOS, has a UI that goes back to the 1990s with faux parametric EQ abstractions instead of direct target curve drawing, and doesn’t offer the bass management or multisub integration capabilities one should expect from an expensive paid room correction upgrade. It also seems dependent on REW - aren’t the optics of a paid software with restrictive and onerous licensing terms presumptively free-riding* off of donationware off-putting?

*I suppose it is possible there is a non-public fee sharing agreement or other compensatory mechanism. There should be!

I meant why it doesn't count while the mobile app does? Mobile app is limited to Android/IoS, PC app is limited to PC, so both have similar limitations there. Licensing is shit, sure, but it doesn't make it a bad product if it fits the budget. The UI isn't good in any way, but it's way superior to the mobile app. Bass management has slight advantage over mobile app, which also doesn't have the features you desire. So all in all I agree with you about the shortcomings of MultEQ-X, but if you like the mobile app I don't see why you'd hate MultEQ-X aside from the price. The only thing is you don't draw the target curve directly, but I find that better because you know exactly what kind of function you are creating and can modify it precisely without breaking everything with a single bad touch.

I still wonder about the whole phase correction part that Dirac does. From what I understand, we can't hear a speaker's phase. But, we can hear phase differences or misalignment of phase across multiple emitters, like between a speaker and sub, or left and right speaker (inverting the polarity of one speaker easily confirms this). From my study, I looked at the phase mismatch between the left and right speakers, and found that Audyssey did a better job at reducing error between the left and right (and also was better at matching SPL between left and right speakers). Subjectively Audyssey sounded better. Well, this was the case in my system and speakers, YMMV.

I wonder the same, because I haven't seen any proof that this is actually an advantage of Dirac, even though they market it as such.

 

[hmt]

they correct for phase shifts caused by IIR crossovers. The IR response is actually a proof for that. Due to the phase shift the impulses of the different drivers are not aligned. With dirac they re better aligned. You seem to confuse overall polarity of a speaker with its phase response.

 

[dlaloum]

Whether by measurement or ear, how do we determine "a winner" with software that's designed to modify the
incoming signal?
I'm no EE but I guess there's a way to measure any THD or noise introduced? SINAD?
When Amir reviewed Audyssey + Editor, he did so only in a basic UI subjective way.
That's not a criticism, simply an observation.

Exactly, what do we compare?
Which makes a flatter FR at the listening position, auto-magically?
Which is getting the distance for each speaker more accurately determined?
Who's is easier to set a desired room curve with?
I could go on but the software's job is to modify a signal, not be transparent.
So who's does a better job at not being transparent? LOL

Sure - but it is only relevant if you are looking at it in a perfect environment.... ie: an anechoic chamber, where there will be no reflections, no room effects etc...

The point of Room Correction software, is to take an imperfect environment, and adjust it for improved psycho-acoustic performance.... the key being "psycho acoustic".

The first thing one has to accept, is that the room is imperfect (non transparent, if you prefer!) - once you are past that, the question is, what can be done to ameliorate that inherently flawed environment.

You can perfect the environment, and then use the signal "RAW" - but that environment (anechoic) is a truly horrid place to listen to stuff in!

In reality stereo and surround recordings, are designed to be listened to in a living space or movie theater, environments that are imperfect, and have inherent flaws.

So, transparent to what!? - Transparent how!?

And of course all of this through those least transparent of all components... speakers - with most people's speakers having THD of over 1% (especially in the bass)

What some of the Room EQ systems are now starting to demonstrate (eg: Dirac Live ART), is that active correction, which is effectively another step away from purist "transparency" - gets substantially closer to achieving perceived transparency.

I have no idea what a system of measurement to identify perceived transparency would look like. But our current measurements are designed to identify signal transparency, in one dimension only... and now we are talking about perceiving the end result of that signal in a three dimensional domain.

To some degree, this will force us back to subjective evaluations, while the tools for objective measurement are developed.

 

[jhaider]

I didn't intentionally match at 4.5 kHz. This was the default result after Audyssey calibration, without any additional adjustment, at the same volume level on the receiver for both calibration runs.

A few points, but first a preliminary question: are you confident the individualized calibration was not applied for your other mic during the measurements? That's one of those easy things to miss.

• I don't think anyone is claiming that Audyssey mics are are exactingly sensitivity matched, only that their frequency response is generally good enough for room correction purposes. Your two seem to have a broad-band sensitivity difference of around 1dB for the most part. That's within reasonable error tolerance for acoustic measurements. In fact, even great microphones can vary a bit in sensitivity. For example, my four iSemcon EMX-7150 measurement microphones vary in sensitivity by about 0.75mV for 94dB output at 1kHz (using a 48V mic preamp) according to their calibration sheets. Doing the math explained here amounts to just over 1dB difference in measured broadband level. (All are similarly flat in response.) When I calibrate levels to use them simultaneously in REW Pro, the microphone preamp setting ends up slightly different for each one, as expected.
  
• Given the broadband level was slightly higher with your calibrated microphone, your expression of preference for the calibrated-mic run is as expected. Also no surprise that you ascribed the differences you heard to something other than level differences. We all seem to do that! One must wonder if the individual calibration files intentionally result in slightly louder results so as to make people perceive "the calibration" as better. Or it could just be sample variation.
  
• I think you're assuming a level of precision in acoustic measurements that is very hard, if not impossible, to achieve in an anechoic chamber let alone a normal human home. There's some good discussion on this in the third Neumann KH80 Klippel NFS measurement thread, if memory serves.

So again, if you want to compare the microphones' relative FR, which is what really matters, you should align their SPL at a reasonable reference point, because the absolute SPL is not material unless you’re multiplexing. And allow for some HF error tolerance because they won't be placed exactly the same. That's not to say your two microphones don't read slightly differently and thus will lead to slightly different results. By eyeball - which is difficult because of the lack of useful SPL alignment - they seem to exhibit some variance on the top and bottom ends. Is it enough variance to make a difference even with levels matched? I don't assume that's impossible. Your listening position response looks excruciatingly hot either way for my tastes. Here's what I consider great in-room performance.

N.B. The point of ART is to extend the region of low mean spatial variation (MSV or “spread” in Dirac-ese -the shaded area in the graph above) to the octaves above the subs. Here, you see very good MSV in the bass - generally under 5dB across a 6’ x 4’ area - thanks in large part DLBC. But the MSV gets much wider above the subwoofer crossover. With ART, one hopes to see similarly low MSV in in the octave above the subs, and into the octave above that.

I meant why it doesn't count while the mobile app does? Mobile app is limited to Android/IoS, PC app is limited to PC, so both have similar limitations there.

Because it can't be run natively within macOS (i.e. a macOS software, or run through a web browser). That is IMO a threshold requirement for relevance for any commercial consumer software. Note that all of Audyssey's peer and aspirational competitors can be used natively on Macs - Dirac, ARC Genesis, and Trinnov have native macOS software, as do Genelec and Neumann; RoomPerfect runs on the AVP through a web browser, as do the pre-correction processing sections in Storm and Monoprice.

but if you like the mobile app I don't see why you'd hate MultEQ-X aside from the price. The only thing is you don't draw the target curve directly,

Well, faking PEQ filters instead of just seeing and drawing the damn target curve is pointless and really dumb in my view. PEQ filters are perhaps on occasion useful before room correction measurements, e.g. to apply equalization based on anechoic data. That is how they can be used on, e.g. Storm or Monoprice AVPs in conjunction with Dirac. (Storm also lets you use PEQ as part of active crossover filters.) But in lieu of a target curve? No thanks.

 

[chych7]

A few points, but first a preliminary question: are you confident the individualized calibration was not applied for your other mic during the measurements? That's one of those easy things to miss.

• I don't think anyone is claiming that Audyssey mics are are exactingly sensitivity matched, only that their frequency response is generally good enough for room correction purposes. Your two seem to have a broad-band sensitivity difference of around 1dB for the most part. That's within reasonable error tolerance for acoustic measurements. In fact, even great microphones can vary a bit in sensitivity. For example, my four iSemcon EMX-7150 measurement microphones vary in sensitivity by about 0.75mV for 94dB output at 1kHz (using a 48V mic preamp) according to their calibration sheets. Doing the math explained here amounts to just over 1dB difference in measured broadband level. (All are similarly flat in response.) When I calibrate levels to use them simultaneously in REW Pro, the microphone preamp setting ends up slightly different for each one, as expected.
• Given the broadband level was slightly higher with your calibrated microphone, your expression of preference for the calibrated-mic run is as expected. Also no surprise that you ascribed the differences you heard to something other than level differences. We all seem to do that! One must wonder if the individual calibration files intentionally result in slightly louder results so as to make people perceive "the calibration" as better. Or it could just be sample variation.
• I think you're assuming a level of precision in acoustic measurements that is very hard, if not impossible, to achieve in an anechoic chamber let alone a normal human home. There's some good discussion on this in the third Neumann KH80 Klippel NFS measurement thread, if memory serves.

So again, if you want to compare the microphones' relative FR, which is what really matters, you should align their SPL at a reasonable reference point, because the absolute SPL is not material unless you’re multiplexing. And allow for some HF error tolerance because they won't be placed exactly the same. That's not to say your two microphones don't read slightly differently and thus will lead to slightly different results. By eyeball - which is difficult because of the lack of useful SPL alignment - they seem to exhibit some variance on the top and bottom ends. Is it enough variance to make a difference even with levels matched? I don't assume that's impossible. Your listening position response looks excruciatingly hot either way for my tastes. Here's what I consider great in-room performance.

Yes I'm 100% sure I didn't accidentally apply the calibration on the stock mic. It's a pretty obvious setting in MultEQ-X.

To clarify, what I did was a very specific test to compare the two mics; this does not reflect my normal calibration (I normally put in a -1 dB/dec target, similar to your Dirac plot, which follows the natural speaker response). For the test, I had removed all target curve options in MultEQ-X, which means Audyssey targets flat. I ran the Audyssey calibration at a single measurement position with L/R speakers only, and the two mics were exactly at the same position when the calibrations were run (I used a fixed tripod at the MLP). There was no subjective assessment done in this test. No level matching done, because I wanted to see the default output of the calibration.

Now in my past calibrations with the stock mic, I had always noticed that the highs sounded bright. I also checked the response with REW/UMIK-1, and it kept showing boosted highs, even though the Audyssey default curve is supposed to put a high frequency roll-off. To compensate, I forced a stronger high freq roll-off (by changing the target curve) to get a reasonable high frequency response/downward slope. The mic calibration issue was validated when I got the ACM-1X and that boosted high went away by default, and I didn't need to put in that extra high frequency roll-off.

 

[peng]

they correct for phase shifts caused by IIR crossovers. The IR response is actually a proof for that. Due to the phase shift the impulses of the different drivers are not aligned. With dirac they re better aligned. You seem to confuse overall polarity of a speaker with its phase response.

I am not too sure how to interpret the phase thing vs audible effects, so I plotted the following to see what's happening in the crossover point, that is 80 Hz as shown:

It seems that with Audyssey on, the phase vs frequency curves look similar but more linear, at 80 Hz with Audyssey on, it was 23 degrees, with Audyssey off, it was 32 degrees. Is that a problem, significant enough to use MSO/minidsp to fix that? I also read, such phase change does not matter much audibly speaking but I really don't know in theory is it something audibly detectable by people with normal hearing when listening to music.

 

[peng]

Well, faking PEQ filters instead of just seeing and drawing the damn target curve is pointless and really dumb in my view. PEQ filters are perhaps on occasion useful before room correction measurements, e.g. to apply equalization based on anechoic data. That is how they can be used on, e.g. Storm or Monoprice AVPs in conjunction with Dirac. (Storm also lets you use PEQ as part of active crossover filters.) But in lieu of a target curve? No thanks.

That's what I was thinking too when watching that part of the video, and I anticipated that it would create confusion as there bound to be people who would hear what they wanted to hear. Not that it matters, but just seems silly. Jeff, the presenter did a good job to make it clear that it's not actually PEQ, but again, many who loved PEQs because they like to do their own tweak, ended up thinking that it was in fact a feature like Yamaha's, that allows you to run PEQ manually on top of the RC filters, or even just use it by itself. I don't know if you know about how Ratbuddyssey work, I wish Audyssey had taken the same approach though. That is, instead of doing the PEQ imitation thing, just prompt the user to enter some anchor frequency points (forget the fake "Q" entries), the intended boosts and cuts accordingly, and then let App do the math, re-calculate and implement the modified FIR filters behind the scene.

Prior to the launch of the X, I have communicated with Audyssey about the effectiveness of the Ratbuddyssey's simple approach on leveraging the D+M's Editor app, but I guess naturally, even if I managed to convince him (my contact at the time), they listened to the many online reviewers (including Audioholics) and the popular Youtubers (wish they have half of your in depth RC knowledge, but most if not all of them don't), that PEQ is great, tweakers like it, so please do it like Yamaha did, that they allow you to run YPAO, and then do your own PEQ blablabla, even importing them from REW.

You hit the nail on the head when you said "PEQ filters are perhaps on occasion useful before room correction measurements"!!

I thought it was so funny that there was one popular youtuber who told people on one video, that he preferred to use the X just to implement his own PEQ manually instead of using the A(auto) R. That is fine if he simply told us his reasons, but the way he said it, indicated that he knew little about how auto RC (in this case, Audyssey) works, let alone the pros and cons of PEQ vs Audyssey. If I remember correctly, he thought $200 was worth it just to be able to do his own PEQ manually without using Audyssey's auto/filters that he didn't like so much (understandably..)

 

[hmt]

I am not too sure how to interpret the phase thing vs audible effects, so I plotted the following to see what's happening in the crossover point, that is 80 Hz as shown:

It seems that with Audyssey on, the phase vs frequency curves look similar but more linear, at 80 Hz with Audyssey on, it was 23 degrees, with Audyssey off, it was 32 degrees. Is that a problem, significant enough to use MSO/minidsp to fix that? I also read, such phase change does not matter much audibly speaking but I really don't know in theory is it something audibly detectable by people with normal hearing when listening to music.


View attachment 259847View attachment 259848

Remember that the relationship is that from an imulse you can generate phase AND frequency response and vice versa. In a minimum phase system you also clear the phase response by cleaning up the FR. For the non minimum phase part this does not work. Regular passive loudspeakers have phase rotations at the crossovers. That can be cleared up by dirac with it's FIR. In fact dirac applies FIR only above a certain frequency response due to latency and processing power constraints. So in your case (<300Hz) there will be no difference. Is there an audible difference for speakers with minimum phase crossovers vs linear phase crossovers? The jury is still out but if there is any differency it's not supposed to be big.

 

[dlaloum]

Another interesting thing about phase and filters.... most "natural" systems (which would include speakers and standard crossovers I would assume) are minimum phase (as opposed to linear phase) - for these systems there is an relationship between frequency response and phase, which results in frequency response being "flat" or neutral when phase is corrected and vice versa - so if you get one of the two right, you can infer the other...

Linear phase filters seperate this relationship and allow the adjustment of frequency/amplitude seperately from phase.... which if the system you are correcting is a minimum phase one - will throw it "off kilter".

What I don't have clear in my head, is to what degree speakers and room interactions are minimum phase.... without the interference of digital processing, I would assume that the overall system is minimum phase - and therefore phase adjustments and frequency adjustments would go hand in hand.
(Note: this is the case for example with record player cartridges, where the loading and the cantilever resonances are all minimum phase.... and can be adjusted accordingly)

 

[hmt]

IIR crossovers (used in most speakers) result in the speaker not being a minimum phase system. They have phase rotations at the crossovers. Concerning the room interactions. First I thought at least in the low frequencies speakers in a room are a min phase system. Unfortunately this is not the case. There can be phase shifts that can't be explained by the frequency response (which makes the integration of subs and mains difficult). In REW you can separate the excess phase (or excess group delay) to see this.

 

[Newman]

What I don't have clear in my head, is to what degree speakers and room interactions are minimum phase.... without the interference of digital processing, I would assume that the overall system is minimum phase - and therefore phase adjustments and frequency adjustments would go hand in hand.

Toole says that “At subwoofer frequencies the behavior of room resonances is essentially minimum phase (e.g., Craven and Gerzon, 1992; Genereux, 1992; Rubak and Johansen, 2000), especially for those with amplitude rising above the average spectrum level. (So) what we hear can substantially be predicted by steady-state frequency-response measurements if the measurements have adequate frequency resolution to reveal the true nature of the resonances.”

His general conclusion seems to be that for practical purposes, we can correct the frequency response and not be too concerned by phase, in the bass region ie up to 200-300 Hz.

However, above that, the wavelengths get too short in relation to room dimensions and we start hearing distinct reflections, and equalisation of the ungated frequency response degrades the direct sound frequency response, which is an overriding priority to keep flat and smooth.

Concerns about phase rotations at crossovers in the above-mentioned bass region are frequently overstated. We really cannot hear these phase errors except as a frequency response error, ergo, fix the frequency response error and job done.

cheers

 

[GalZohar]

Well, faking PEQ filters instead of just seeing and drawing the damn target curve is pointless and really dumb in my view. PEQ filters are perhaps on occasion useful before room correction measurements, e.g. to apply equalization based on anechoic data. That is how they can be used on, e.g. Storm or Monoprice AVPs in conjunction with Dirac. (Storm also lets you use PEQ as part of active crossover filters.) But in lieu of a target curve? No thanks.

I also don't see much point of the regulars PEQ "peak" filters for target curves, but they're probably more for those who want manual correction with REW generated filters.

You still have shelving and tilt/slope filters, which is probably what you would use for most target curve purposes, and isn't it nicer to do it parametrically rather than visual drawing? Plus with drawing you don't actually control the exact curve between the pointers, the application needs to decide that for you, while with parametric drawing you set exactly the function you want (from within a set of supported ones), rather than be forced with whatever the drawing chooses for you. With Dirac can you also modify the curve parametrically after you finish drawing?

Anthem also went with fully parametric target curve, and it's even more limited as it only has pre-set parameters that you can change.

 

[abdo123]

I would assume that the overall system is minimum phase - and therefore phase adjustments and frequency adjustments would go hand in hand.

This is actually the case below 100Hz or so, which is also the range where most people recommend to limit Equalization, but above that range Group Delay and excess phase go haywire.

 

[Chromatischism]

This is actually the case below 100Hz or so, which is also the range where most people recommend to limit Equalization

I think you meant to write to limit above that range (I would say more like 200 Hz, though I do anywhere from 300-600 on my speakers, depending which one it is – SBIR is a bitch).

 

[abdo123]

I think you meant to write to limit above that range (I would say more like 200 Hz, though I do anywhere from 300-600 on my speakers, depending which one it is – SBIR is a bitch).

I have a non-minimum phase blip in my room around 150 Hz but yeah if you have smooth sailing till 200 Hz then filter away.

 

[abdo123]

@Chromatischism @dlaloum

I explain filterability of systems more in detail in this post:

Room Treatments in the Real World

But having played in a wind ensemble and attended lots of concerts the concert halls aren't over dampened, usually they have some dampening combined with reflection panels. Those are diffusers, especially those with multiple various height surfaces. ...Which are usually a feature of concert...

www.audiosciencereview.com

 

[Newman]

I have a non-minimum phase blip in my room around 150 Hz but yeah if you have smooth sailing till 200 Hz then filter away.

You can still do it in your room, just get the FR flat and smooth.

 

[GalZohar]

I have a non-minimum phase blip in my room around 150 Hz but yeah if you have smooth sailing till 200 Hz then filter away.

If you have a non-EQ-able area at 150Hz, don't systems like Audyssey/Dirac know to avoid ineffective EQ in that area without disabling any EQ above that frequency?

 

[abdo123]

If you have a non-EQ-able area at 150Hz, don't systems like Audyssey/Dirac know to avoid ineffective EQ in that area without disabling any EQ above that frequency?

No but you can adjust the target so they’re left mostly alone.