The Subjective and Objective Evaluation of Room Correction Products - Page 9

I don't think Dennis meant that it was predetermined, since he didn't describe it as such. From what I remember of Harman's speaker tests, both groups of listeners (trained and untrained) ended up arriving at the same preference eventually. Just took fewer trials for the trained listeners, making them more convenient to use in tests. Nothing more complicated than that.

Hi Kal,

I agree, and would add that Audyssey does such a poor job of displaying their results that an independent measurement system is a necessity regardless of which of your objectives someone chooses to pursue.

With regard to confirming/critiquing Audyssey's overall effectiveness, it should also be emphasized that the intent of Audyssey's fuzzy logic approach to weighing results is to improve the sound quality at multiple listening locations without unduly impacting the sound quality at the primary listening location. So to test the effectiveness of this departure from approaches that use simple spatial averaging would require subjective listening tests in locations other than the primary location, as well as at the primary location.

Sean's objective was to rank how competing room correction products sound to trained listeners and to draw conclusions based on how they specifically measure with regard to frequency response. To do "apples to apples" comparisons all of Audyssey's frequency response processing, Dynamic EQ should be included in future subjective listening tests and objective measurements.

I didn't notice any other measurement parameters mentioned in the testing other than frequency response, but other measurements, such as time domain measurements, might be useful in ascertaining differences. It would also be instructive if a wider selection of content, including multichannel content with a full complement of speakers were included in future testing to supplement the monaural testing.

Larry

Thanks much for the detail.

Do you put any stock in Chris K's contention that even though one might create an identical frequency response correction filter using either PEQ, IIR or FIR, only FIR need apply due to optimal phase properties?

If these various implementations do in fact result in different phase yet similar freq response, might not there be some resultant audible difference?

I've also seen multiband octave equalizer that allows independent adjustment of level and phase. As if it matters.

I realize that none of this proves anything. It just raises the question.

My understanding is that if you are willing to tolerate latency (delay from sound going in to sound appearing at the output), FIR can give you arbitrary phase relationships with identical amplitude response. In some cases, this can be useful, and other cases, this can be hurtful. IIR is just like analog: minimum phase relationship where the amplitude is exactly determined by phase (modulo all-phase filters).

People who do bass correction use IIRs because they can (in theory) exactly correct bass peaks due to room modes.

--Andre

It doesn't strike me as a marketing campaign although I suppose some of that is inevitable when the person posting is in the industry. I would never use any automatic EQ system unless it allowed full manual override. My main interest in this thread is which in-room target curve the listeners found most pleasing and how to duplicate the measurements. It's just another data point for setting up my own system manually so Sean won't be selling me anything. I appreciate how forthcoming he is about what's going on. It's a refreshing change from some of the other manufacturers' reps who go out of their way to obfuscate what their product does.

Back to the Audyssey thinly veiled marketing presentation thread?

Just want to chime in and thank Tom for all the info he disclose. The dynamics behind his efforts are irrelevant to me as long as I look at the data/results objectively. I do enjoy reading his articles in his blog along with the free linked AES papers! I place Harman articles on bass reproduction along with Tool book on the same rank with Everest books in terms of educational merit.

Now back to the topic

Considering the ear/brain ability to compensate for tonal imbalances when given enough, time, I am interested in knowing if listeners we capable of switching among different RCs at will or some kind of waiting period was mandatory. The latter is prefered since the first would give a preference advantage to FR with mean performance among the FR RC candidates.

I think I'm going to have to agree with you Dennis, my ARC experience really left a bad taste. Now if we could just get digital crossover/EQ systems with arbitrary control over the transfer function instead of being limited to analog filter functions. It would also be nice to have the ability to control below 20 Hz but I suppose if I ever really want to get rid of my 18 Hz room mode bump I can always put an analog notch in front of the woofer amplifiers. My system is tuned up pretty well using ARTA with an Earthworks M30 measurement mic and the Behringer crossover. Let me know if you ever come across a crossover/EQ with arbitrary control. It would be nice to be able to just give it a transfer function right out of mathcad or something and have it do that.

mk

The Dolby/Lake processor is probably what you'd like. Too bad they stopped making them.

Quick question, judging from the discussion of on-axis and off-axis frequency response, you have the left or right front speaker, which are the ones I presumed you tested, toed in to directly face the listener? Is this the preferred method of setup for this speaker?

Unless I'm missing something it looks like its still based on analog filter functions. You have options for Graphic or Parametric EQ with selectable parameters but that's not what I want. I would like to abandon analog filter functions entirely and be able to have it do what ever I want as opposed to for example creating a band pass filter with a level and Q. So far I have not found a crossover/EQ that can do that.

mk

I'd be curious what makes up the Predicted In-room Response. Specifically what angles are used for Direct and Early Reflected Response, and what gating (if any) is used for the Sound Power metric. Wondering how Toole's findings on integration time are utilized when doing the sound power/in room response, or maybe it is not and it's just a steady state RTA measurement type?

Also what do the top 10% of loudspeakers have for AAD and NBD?

I've read Sean's (excellent) paper on this and had some correspondance with him over these questions. Eventually he (politely) brushed me off as I think he thought I was a competitor in China or something So I'll take this opportunity to ask the those questions publically in hopes that he feels more compelled to answer them. Sly I know, but I really am just a hobbyist speaker designer that is VERY interested in the applying the findings of Toole and Harman

There's an interesting thing in play with room correction solutions. That being two schools of oppostion to them, one is the "audio purist" who believes anything in the chain that alters the original signal is blasphemy and the other being the "tweaker" who dismisses auto EQ because of lack of configurability and their belief they can do better.

The proof is in the pudding and for every one of those I'd wager there's at least 20 that absolutely prefer to use these solutions, otherwise known as the "target demographic" for the product design. How many of the "trained listeners" fall into that category?

and that doesn't include loudspeakers or rooms? Both of them can massively alter the original signal....typically far more than any other part of the playback chain.


Most autoEQ that I've seen can be tweaked by the user...to some degree.


Why the scare quotes around 'trained listeners'?

I think one of the interesting things in play is the effort being expended by some quarters to portray the Harman work as something that doesn't apply to *them*...or as a nothing more than a highly biased marketing ploy.

I can't speak for others, but I could care less if it is or isn't a marketing ploy. ( I don't think it is FWIW.) I'm questioning the validity of testing something out of the element of it's intended design parameters, not the testers themselves. My understanding is that tests like these should be subject to peer review. Does that mean only consenting opinions?

So, what's your point?

In the demo I saw, the user could draw a filter on a tablet PC screen and it would make a filter to fit.

Hi Monte,

The nice thing about the IIR filters used in the used in the popular digital boxes is they use much less CPU power and have much lower latency than FIR filters. Have you tried importing a transfer function into REW and having it calculate the filters? I've only messed with it a little but it seemed to come pretty close. TI has some software that does the same thing but I haven't tried it.

Edit: I'd had the TI software, ALE (Automatic Loudspeaker Equalizer), on my drive forever but I had never gotten around to trying it. It's pretty slick; it generated these filters in a few seconds using example measurement and target files. The only thing is you'd need to build a little spreadsheet to convert bandwidth in Hz to Q.

http://www.speakerpower.net/PDF/DSP-...lete-Rev-B.zip

The TI software can do that too. Maybe Lake was using a TI chip? You just click on the screen to draw the target curve and there's a user setting for your +/- dB tolerance.

Sounds familiar.

The Pioneer and Yamaha EQs can be configured manually. Audyssey lets you choose a target curve but it doesn't give you any control over (or knowledge of) the actual filter created. And none of them make the measurements available to the user so you don't know if the measurements even make sense. Getting good measurements isn't that easy.

LspCAD can generate it exactly. I just don't want to have to use a PC and sound card as a crossover. The Behringer crossover has a feature similar to what Roger described where you can manipulate the response by drawing a line but it's still limited to the same filter functions. This is probably good enough for most room EQ problems but I would like something that can do the linear phase digital crossover functions in Horbach and Keele's paper on symmetric pairs.

http://www.xlrtechs.com/dbkeele.com/...20Part%202.pdf

My new CBT center speaker will not use this design approach like my existing one does but I would still like a crossover with the capability to create the curves exactly. I want to do more playing around with this approach to crossover design, it seems like the ideal solution.

mk

If you want linear phase, I kinda like Bruno's approach with the new UcD plate amps. He uses IIR filters for the crossovers and driver EQ and a short FIR filter that's an inverse allpass to unwrap the phase. You can download a demo version of Jan's Phase Arbitrator and decide for yourself if unwrapping the phase is worth the hassle. I decided for myself that it isn't (even though it makes nice square waves). If you really need FIR filters, the DEQX is probably the best bet but it's spendy.

Edit: I just read the paper you linked and I don't think any of those solutions could do that. Oh well.....

http://www.diyaudio.com/forums/class...as2-100-a.html
http://www.thuneau.com/arbitrator.htm
http://www.deqx.com/

No, David McGrath of Lake did all the filter design. Nothing else like it.

The symmetric pair approach requires the crossover to create a slope to a point and then rapidly change to a brick wall. No more than two pairs are ever playing at the same time. There is no "pass band" in the crossover, just a single point where it is at full output. (you could say the pass band is 1 Hz wide) On each side of that point, there is a linear phase slope to the single points where adjacent pairs are at full output. Once the full output point (critical frequency) of the next pair is reached, the slope rapidly transitions to a brick wall. The spacing of the drivers determine the polar response. Each pair is spaced at the same fraction of a wavelength at its critical frequency as all the other pairs. The linear phase transition between pairs creates a phantom pair that is the summation of the two pairs and always maintains the same fraction of a wavelength spacing. The result is perfect control of the polar response of the speaker within the boundaries of the pairs. Eventually the wavelength is too short to have spaced pairs so the polar response becomes a function of a single driver in the center. On the low end you can maintain polar control as low as you want as long as you don't mind the size it takes to keep the same fraction of a wavelength. I can approximate the crossover with the Behringer but so far I can't find anything that can do it exactly other than LspCAD.

We should probably move this discussion back toward room correction systems now but hopefully eventually someone will have a crossover/EQ that can create the necessary curves for this. It seems to be the best approach to polar response control yet. Of course individual driver polar response has some effect but if you model it using ideal point sources then the polar plot at all frequencies within the pair boundaries are identical. Polar response becomes completely independent of frequency. Again, this is because all the driver pairs both real and imaginary always maintain the exact same fraction of a wavelength spacing.

mk

Hey Monte,

I read the paper and edited my post while you were typing. I agree that none of the commercial products will do that kind of crossover. A PC running Linux might be a reliable way to go but it's still a pain.

Back to our regularly scheduled room correction discussion with the Audyssey users feeling abused and thinking we 'just don't get it'....

OK, that's it. We are not going to share the secret handshake with you.

AudysseyPro lets you see the measured response and lets you control the target curve (within limits).

They know. They don't care.

Can one of you Audyssey fans post some before/after curves, done with an external measurement system, since you seem to be implying Sean rigged the results?