Originally Posted by markus767
it isn't helpful to make it look like if the frequency response is equalized flat then the time domain is automatically corrected too. This is only correct if a) the system is minimum phase and b) it is only correct at a single point in space. Room responses are NOT minimum phase and humans have TWO ears spaced a couple of inches apart hence equalization has to work at least at these two different points in space. This will become an even larger number of points when we want good sound for multiple listeners.
So we are back to EQs don’t work argument? Fortunately listening tests, measurements and published science says they do. Let’s review the listening test data once more since that message seems to have been lost already even though we drilled down into it for so many pages just now.
Harman Listening tests used an “anchor” (reference) of no EQ in controlled, double blind tests. It also tested two modes for their EQ: optimized for 6 seats simultaneously in one instance (RC1) and biased toward the sweet spot in another (RC2). These are some of the subjective results.
On whether the sound was “colored,” No EQ garnered 27 votes. RC1 optimized for one seat got only 2. RC2 optimized for all six seats received just 5. As a contract, RC6 with its time domain tricks and fuzzy logic, received 35 negative votes here.
At the other end of the spectrum for positive attribute such as Neutral, No EQ received just 4 votes. RC1 got a whopping 35 votes. RC2 likewise got 25. So dramatic improvement in creating improvements there. Once more, RC6 did poorly receiving precisely zero (0) votes for producing natural sound.
Frequency response measurements backed all of these findings. Optimizing for one seat (RC1) generated somewhat smoother response than optimizing for six (RC2). So there is a small compromise there. But either way, we were way ahead of no EQ. And this is a box that you earlier in this discussion called “a box of EQs.” Well, clearly a box of EQs in controlled tests generates superlative improvement over doing nothing. It also did substantially better than another solution which as your argument, hangs its hat on mathematical terms, rather than listening tests as its final proof point.
There is another published listening test from Harman, this time just comparing different approaches for their own algorithm. This one used four subs in optimized placement so the evaluation focused on results above 400 Hz:
Once more we have conclusive results that even when we change the approach to optimizing frequency response in the room, we still manage to statistically do better than no EQ. As I noted, we clearly hear frequency response anomalies and anything that reduces it, will convince us to give it thumbs up.
It would be hard to imagine if all of this is based on mathematics that says it won’t work per your argument. Fortunately such is not the case. You say that the room transfer function is not minimum-phase. I am assuming you are inverting the situation and arriving at that. That is, you are taking the existence of non-minimum phase response as meaning the transfer function is never that. Well, that is not the right way to look at it. It is like saying if I have a basket full of apples and I put a few pears in it also, you can say there are no apples in there! Yes, you have to sort through the basket if you want to find only apples. Hence the reason I mentioned there has to be expertise applied to this part of the system. It is not the filter knobs that are doing the work but the logic behind them.
Let’s review the literature on this:Dr. Toole:"For now, it is sufficient to say that low-frequency resonances in rooms behave as minimum-phase phenomena, meaning that if there is a prominent “bump” in the frequency response, it is probable that this will be heard as excessive loudness at that frequency and that for transient sounds, there will be bass “boom” at that frequency. Using equalization to reduce the bump also attenuates the ringing so both problems are solved simultaneously.
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. This suggests that 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. In minimum-phase systems, the magnitude versus frequency response (henceforth simply “frequency response”) contains enough information to enable the phase response to be computed, and from those two data sets, the transient response can be computed."
Here is the Craven and Gerzon
And Rubak and Lars Johansen
reference:”As pointed out by Craven & Gerzon , room equalizers based on Digital Signal Processors are able to reduce the reverberation time considerably, even if we only use minimum-phase equalizers. Our preliminary test results are in agreement with this important potential for DSP based equalizers. Therefore we have put focus on objective test methods concerning the improvement of the room acoustics using equalizers.”
I explained all of this in a previous post: http://www.avsforum.com/t/1425262/are-audio-companies-all-involved-in-a-huge-conspiracy/630#post_22478314
. All of the above published reports are in the context of room optimization and reflect the knowledge and experimental data of the researchers.
In my post above, I showed measurements that demonstrated how correcting frequency response errors generates corresponding time domain improvements.
Putting it all together, the trick here, and there is one, is that you need to do spatial averaging. When you do that, the minimum-phase aspect of the room falls out of the equation and what I showed in my previous post comes true. RC6 does away with that and uses fuzzy based pattern matching to accomplish the same. In their research paper Audyssey says that is preferable to spatial averaging in smoothing the frequency response. Clearly that did not work well here but everyone is attempting to do the same thing: using multi-point measurements to find the most optimal frequency response for multiple seats.
To borrow your own phrasing it is “unhelpful” to keep making assumptions about us not being familiar with such work when I am relying so much on published research. I have read that report and read it when it when it first came out. I like it as a fragment of the knowledge it parlays in filter choices. So much so that I put it in our technical library on WBF Forum back in 2010: In case there is doubt, click on the first link of Google search: https://www.google.com/search?q=site%3Awhatsbestforum.com+WBF+Library%3A+Audio+signal+Processing+Papers%2FPresentations&rlz=1C1SNNT_enUS374US375&oq=site%3Awhatsbestforum.com+WBF+Library%3A+Audio+signal+Processing+Papers%2FPresentations&sugexp=chrome,mod=0&sourceid=chrome&ie=UTF-8
So I have read it. I also know the context for which it was written. Dirac’s reason for existence originally was for car audio, not listening room acoustics. Today they have such an offering and I have had extensive discussions with their designers there. For now I like to draw your attention to this paragraph from it:”It is clear that even in a good listening room with good speakers a substantial improvement is possible using a careful mixed-phase design. The minimum-phase filter is clearly doing a good job as well; not nearly as good as the mixed-phase design it nevertheless improves the time-domain behavior. In a large well-designed listening room the impulse response would preferably be nearly minimum-phase and that is also the case in this room. Therefore we can cause improvements by just using plain minimum-phase filters. This unfortunately does not carry over to trickier environments such as car cabins”
So the very reference you provide backs the fact that room responses in low frequencies are comprised mostly of minimum-phase phenomena agreeing with others I provided. Please look at the last two words there: the emphasis on car audio. The volume of air in a car is far lower than our listening spaces. That pushes up the transition frequency to nearly 1000 Hz where psychoacoustics starts to enter the equation and your point regarding distance of two ears and having two of them comes into play. It is a challenging problem but fortunately outside of the scope of what we worry about here, namely, home listening spaces (although as they say, EQ is a powerful technique there just the same).
To be sure, we are dealing with an area of acoustics where some uncertainty exists. But there is an *integrated* path through it. This starts with well-designed speakers that perform well above transition frequency, room reflections and all. If you have a bad speaker, EQ can help some but the real solution is a good speaker. It is below transition frequencies where we need to focus our energy, pun intended, in using an EQ. As I show in my article on bass optimization (http://www.madronadigital.com/Library/BassOptimization.html
), it is critical to understand this fundamental point:
We see that seat to seat variation simply does not exist above transition frequencies (if you show them using correct psychoacoustics resolution). Where it shows up is in lower frequencies and that is what you want to correct with your EQ. If you apply enough smarts there, you can get superlative results. Here is the JBL SDEC-4500, the same box of EQs, doing that. Before (with four subs):
That "non-minimum phase" null just disappeared courtesy of a simple notch filter and power of computers applying large number of iterations to arrive at optimal performance not at just one seat, but multiple. And this is without EQ! But for sure includes electronic correction (of each sub).
I will be blunt: there is no way you can build an accurate sound reproduction system in your room without EQ. It is a critical link in the chain. At least the well implemented versions are.