The Subjective and Objective Evaluation of Room Correction Products - Page 6

2 channel recording and playback mucks up sound too (except perhaps for ambisonic). Perhaps that's why sometimes (often actually) I prefer 'upmixed' to straight 2 ch (or could also be due to decreasing the psychoacoustic influence of room reflections in my rather bright space). I like to be able to have both 'pure' 2 ch (2.1 actually) and surround synthesis as my options.

Since nearly all of us measure our systems in room rather than anechoically, and since room gain varies a lot from room to room, what would be a ballpark FR target when the system is measured in room?

Flat? It seems the answer is no.

What would be the ballpark FR target of speaker in room?


Here is a suggested X-curve target that I got elsewhere. The FR in the 100 to 400 Hz area is subject to inaccuracies in room measurement systems.

In fact you would be right, since the clustering is "fuzzy", meaning the responses do dot belong exclusively to one or another cluster. Still, you cannot get a better result in terms of simple spectral error than by using a spatial average and equalizing it flat.

Yes, but it mucks it up in the way that the original sound mixer designed it to be mucked up.

Kind of interesting side bar, but I have been playing around with my system to improve the sound of my mono DVDs (DD 1.0). You can even improve mono with a little work!

Hi Palmer,

If I understand correctly the ideal acoustic environment for 2 channels is rather different, i.e., livelier, than the ideal acoustic environment for multichannel content. Therefore, if your room acoustics is relatively lively, and then you use surround processing to upconvert it to multichannel then it is reasonable to expect a less than optimum result.

I'd be the first to admit that I'm not an audiophile, but I always upconvert two channel content to multichannel, and it always sounds better than left as two-channel in my relatively dead home theater. Better yet, I simply listen to multichannel music.

Larry

It used to be aggravating 15 years ago; now it's reflex. It helps to have a processor that allows for quick and easy adjustments. Somewhere in the first song I'll usually adjust tilt and surround processing to my tastes, and forget about the hardware for the rest of the album.

Unfortunately, I don't envision a push button that will make those adjustments automatically. I use my pre-pro mostly for listening to 2-channel music in surround (movies are a distant second priority). Since there is no standard with music, I could end up listening to one CD at -20 and then have to bump it up to -9 for the next album, just to keep the same volume level.

As Roger alluded to earlier, how is even a simple feature like Loudness Compensation supposed to work properly when each album averages a different level from reference? And even if you could get the music industry standardized overnight, what am I supposed to do about the 1500 CDs sitting on my shelf?

So I'm with you for almost wanting a second volume control for the sub, just to deal with variances from recording to recording. Heck, I'm doing something similar already. Minor pain, but I prefer that to the alternative (using the same setting for all music recordings).

Hi Randy,

I believe the original poster conceded that he appreciates that certain types of critical listening may be easier to achieve with monaural content rather than multichannel.

However, I believe a key point he is making, that seems to be repeatedly overlooked, is that some important functions of competing room correction products only have relevance in a multichannel setting. For example, some or all of the products may have features specifically designed to timbre match speakers in order to facilitate more realistic panning of sounds from one channel to another. One can not reasonably infer the effectiveness of a multichannel feature by listening to monaural content.

Obviously, the focus of this particular test was not to consider these type of multichannel specific equalization features, nor to determine listener preferences applied to movies.

Larry

Right, for instance I believe that Audyssey's target curve, (which is what they exclusively came to market with upon inception) was developed with a lot of subjective tests in varying environments to emulate the intended use of the end product.
I'm not sure they can do anything about timber matching - that's over my head - but I believe, using Audyssey as an example, that if there's a finite number of filters available, they can be moved and used where it best suits the multi-channel outcome - and that result may not be the best outcome for a single speaker.

I couldn't find it except that it says something like "measurements were done according to product instructions", but if the measurements were done using all speakers and then they used the mono set up for evaluation wouldn't that guarantee a skewed result? If it was done with just the sub and left speaker then it would add credibility, but it's still not an app the product was designed for.

Hi Sean,

Thanks for the remarks.

I think I understand. Our ears are constructed in such a manner to filter identical sounds coming from the rear to create a different "timbre" than when coming from the front. These tonal differences usually helps us tell whether sounds are coming from the front or rear, particularly for on-axis sounds. Our brain then processes these differences into spacial cues and interprets the similar sounds as being the same timbre.


Obviously, Audyssey isn't going to reveal the details of their processing, and I'm sure if they did I would never fully understand it.

However, I gather from a high-level response to one of my questions that, for better or worse, it does more than equalize to an identical target curve.


Perhaps the feature is merely intended to make non-identical speakers sound similar and not render a theoretically identical timbre. That is, they are marketing a feature to deal with a less than ideal situation, but none the less a practical concern in many home listening environments.

Larry

I see, thanks, Sean.

I sent you a PM.

I have a stupid question: why wouldn't a product like the BeoLab 5 or another well-designed speaker with extremely wide (180) or even horizontally omnipolar dispersion pattern be the most preferred by listeners? I couldn't ever quite figure that out while reading "Sound Reproduction." Thanks so much!

Do RC systems do anything to address the reverberation time of the room? If for instance RT60 is reference, does room correction make it sound like you're listening in an RT60 room? If not, then isn't that a key variable in sound quality of "the room" that RC fails to equalize?

Also, couldn't another large portion of the variances observed as preference be due to individual hearing? It would be interesting if each person was given a hearing test to determine the FR of their ears and factor that into the correction curve of each test. Instead of ending up with an average curve preferred by most people, maybe there's a more specific better sounding target curve if variances in human hearing were compensated for and eliminated as a variable.

Waves Audio sell a plug-in for this called "PS22 Stereo Maker" which you may find interesting. It was invented by Michael Gerzon. Assuming you can't afford this (or Pro Tools to run it under), you can download the users' manual from their website as a PDF.

I am not allowed to post the URLs as links, so just search the Waves website for "PS22" and click on "Manuals". Michael Gerzon is in Wikipedia.

Regards,
Martin

Hi,

Audyssey did a lot of subjective testing and developed their own version of equal loudness curves. This data was used to develop their Dynamic EQ add-on to MultEQ. Dynamic EQ considers the actual real-time listening level and adjusts the flat bass region of its target curves, tilting the bass upward. It would have been interesting to see how MultEQ XT Pro plus Dynamic EQ would have placed in the preference test, however, that feature is not available in the Sound Equalizer and therefore was not tested.

Audyssey also continued Tom Holman's work at THX on re-equalization and developed their own versions of high frequency roll-offs and incorporated them in the Audyssey Reference Target curves. This feature would have been available in Sean's testing and so Audyssey was not at a disadvantage in this regard.

Although Sean is dubious as to the value of doing more than equalizing each satellite speaker to the same target curve, nevertheless Audyssey claims to do a more comprehensive type of timbre matching than just frequency matching. However, that feature only has utility if more than one satellite speaker is playing at the same time.

Apart from timbre matching I don't believe the basic equalization of the subwoofer and a single main speaker would be compromised by running with only one satellite speaker. Although it measures all speakers when it does its measurements, Audyssey performs the equalization on each speaker individually. That is, it runs a test signal to each individual speaker, one at a time, and then calculate the filters for each channel. So the measurements do not require a multichannel setting where all channels are active simultaneously.

In summary, the most fundamental disadvantage to Audyssey in Sean's testing wasn't that the testing was done with a single satellite speaker. It was that Dynamic EQ was not considered as Audyssey's means of intelligently altering its Reference Target curve which is flat below 4 kHz. Sean and others of course are correct, a flat target curve may not always be the optimal curve.

Larry

Admittedly, my head is swimming and if my question seems misplaced, I apologize.

At least one of the RC's has a target curve intended for cinema content, but yet the three pieces were all music. Well recorded, perhaps, as was acknowledged, but not to cinema standards. Would an RC with a cinema correction curve not have a leg down right out of the gate? If the answer is that they ALL had cinema curves, then ... nevermind.

As an aside, what does the purveyor of a cinema-tuned RC suggest their users do when playing music? If it's inappropriate to use the cinema curve, there must be some other means to accommodate music within the RC envelope, I'd think.

Rhetorical, right?

Someone earlier in the thread posted "a work in progress" wrt RC schemes and that got an "amen" from me.

What RCs have a "cinema" curve? I know of the BassQ having that, but I can't remember any others off the top of my head.

Audyssey's "default" target curve emulates in the home the curve mixing stages are EQ'd to. Some gear has a flat alternative, and some others have a workaround .. of sorts.

Thank you. Somehow I have missed that. Given that most of that target is flat, I have always been under the impression that the gentle roll off done by the Audyssey curve was a target based on research that resulted in a roll off common to all "good" rooms using "good" speakers (as opposed to purely flat out to 20KHz). I thought that this didn't attempt to translate the cinema response to the home as does THX post-processing.

Sean,

Can you comment on whether or not the Harmon RC system (RC1 and 2) correct in the time domain? If yes can you explain how the system would correct in the time domain?

I have always been curious as to how a system would correct in the time domain. Any explanation here would be appreciated.

I was initially irritated that this study did not identify which signal alteration systems (I heartily agree that calling them "room correction" is a misnomer) produced which results. After perusing this thread and thinking about it, I realize that this was the right thing to do. Moreover, the degree of identification you have done is excessive. The reason for this is that people tend to reify studies like this, when in actuality, this study has virtually zero relevance for the experience of listeners enjoying music or movies in their home.

In delineating the reasons for my conclusion, I will acknowledge that some of this repeats points that others have made or alluded to.

1. The n was woefully small and the description of the subjects quite limited. This makes it impossible for any reader to determine whether they were sufficiently similar to him or her to make comparisons relevant. These people positively correlated "boomy" with their overall satisfaction with their listening experience. Is this true for most people? Certainly not for audiophiles. I'm also skeptical that you had sufficient power for the ANOVA to be valid.

2. You conducted the study in a room tweaked to within an inch of its life. Do you have a dollar figure on the cost of the acoustic treatment employed? Far beyond the means of the average listener, and unlikely to be encountered in the real world.

3. A limited and homogeneous sample of music. It would be hard to find 3 more similar artists. There is no jazz or classical, let alone death metal, ambient, rap, etc.

4. A lack of specificity to the subjective dependent variables. Do you have any data on whether people can reliably determine whether music is "thin" or "muffled"? Without reliability, of course, the question of validity is moot. I had to smile when I saw you employ the terms "forward" and "colored". When audiophiles use such terms, AES types greet them with sneering and derision.

5. The music employed was downmixed to mono. Nobody listens to mono any more except audiophiles and people who listen to music on clock radios. Audiophiles are, of course, self deluded and those who listen on clock radios have no need for the signal processing tested.

6. One group of components are used. There are myriad types of speakers that real people use. Would you have achieved similar results with Magnepans? Lowthers? Klipschhorns? How about different amplifiers? Oh, wait. All amplifiers sound the same, right? So where does Harman get off charging so much for Levinsons?

I understand that some of this is endemic to the process. You have to have sufficient control of variables to rule out confounds in your results. However, the greater this control, the less the relevance to real-world experience. I also understand the need to limit the factors assessed. Other approaches are impractical. Who wants to try to interpret a 10-way interaction?

So the responsible stand is to not overinterpret these results. It would be more than "bad manners" to reveal the results more fully, it would have been misleading. I did notice that a concern for manners didn't prevent you from tooting the horn of your own system or bum-rapping a competitors speaker (and a discontinued model at that) in this thread.

In other areas of Psychology research (which is what you are doing whether you recognize it or not) both laboratory research and naturalistic observation are considered important. The strengths and weaknesses of each approach are considered, with the hope that, somewhere in the middle, consistent and generalizable conclusions can be reached. It is therefore puzzling that places like the AES are so dismissive of the experiences of audiophiles. This isn't just engineering that you are dealing with, it's human perception. This is a far more ephemeral, quixotic, and emotionally tinged endeavor.

You seem unaware that this study wasn't the first time anyone had ever tested human audio perception at Harman. It rests on decades of studies by Olive and others -- studies concerning many of the issues you seem to think were only arbitrarily addressed, if at all.

As opposed to the equipment under test, which would not be "far beyond the means of the average listener." As a point of reference, the room correction products cost $2500-$7000, and the B&W 802Ns retailed for about $8000. For $10000+, one could buy a substantial number of ASC Tube Traps (the larger half-round ones cost a little more than $1000 in the 8' tall version), RPG Skylines (<$200 each), and absorption panels (variable in cost, but should not consume the remainder of the budget).


These measurements suggest that the current model also has significant directivity problems: http://stereophile.com/floorloudspea...bw/index4.html

(edited to remove potentially inflammatory content)

could it be time to unsubscribe in such an amazingly short period of time?

I think it depends on the compromises the fuzzy logic has to make when considering the whole, including mulptiple seats. Would the end result of just EQ'ing the speaker and one sub at one listening position be different than if those two were EQ'd as part of the 7.1 set up at multiple seats?