Sunday, November 1, 2009

The Subjective and Objective Evaluation of Room Correction Products

In a recent article, I discussed audio’s circle of confusion that exists within the audio industry due to the lack of performance standards in the loudspeakers and rooms through which recordings are monitored. As a result, the quality and consistency of recordings remain highly variable. A significant source of variation in the playback chain occurs from acoustical interactions between the loudspeaker and room, which can produce >18 dB variations in the in-room response below 300-500 Hz.


In recent years, audio manufacturers have begun to offer so-called “room correction” products that measure the in-room response of the loudspeakers at different seating locations, and then automatically equalize them to a target curve defined by the manufacturer. The sonic benefits of these room correction products are generally not well known since, to my knowledge, no one has yet published the results of a well-controlled, double-blind listening test on room correction products. To what degree do room correction products improve or possibly degrade the sound quality of the loudspeaker/room compared to the uncorrected version of the loudspeaker/room? Can the sound quality ratings of the different room correction products be explained by acoustical measurements performed at the listening location?


A Listening Experiment on Commercial Room Correction Products


To answer these questions, we conducted some double-blind listening tests on several commercial room correction products [1]. I recently presented the results of those tests at the 127th Audio Engineering Society Convention in New York. A copy of my AES Keynote presentation can be found here.


A total of three different commercial products were compared to two versions of a Harman prototype room correction that will find its way into future Harman consumer and professional audio products. The products included the Anthem Statement D1, the Audyssey Room Equalizer, the Lyngdorf DPA1, and two versions of the Harman prototype product (see slide 7). Included in the test was a hidden anchor: the same loudspeaker and subwoofer without room correction. In this way, we could directly compare how much each room correction improved or degraded the quality of sound reproduction.


Each room correction device was tested in the Harman International Reference Room using a high quality loudspeaker (B&W 802N) and subwoofer (JBL HB5000) (slides 8 and 9). A calibration was performed for each room correction over the six listening seats according to the manufacturer’s instructions. Two different calibrations were performed with the Harman prototype: one based on a multipoint six-seat average, while the second calibration used a six-microphone spatial average focused on the primary listening seat. The different room corrections were level matched for equal loudness at the listening seat.


The Listener's Task


A total of eight trained listeners with normal hearing participated in the tests. Using a multiple comparison method, the listener could switch at will between the six different room corrections, and rate them according to overall preference, spectral balance, as well as give comments (see slide 14). The administration of the test, including the design, switching, collection and storage of listener responses, was computer automated via Harman’s proprietary Listening Test Software. A total of nine trials were completed using three different programs repeated three times. The presentation order of the program and room corrections was randomized.



Results: Significant Preferences For Different Room Corrections


The mean preference ratings and 95% confidence intervals are shown above in Figure 1 (or slide 17). The room correction products are coded from R1 through R6 in descending order of preference. The identities of the products associated with the results are not relevant for the purpose of this article. Three of the five room corrections (RC1-RC3) were strongly preferred over no room correction (RC4). However, one of the room corrections (RC5) was equally rated to the no correction treatment (RC4), and one of the room corrections (RC6) was rated much worse. Overall, the sound quality of R6 was rated "very poor" based on the semantic definitions of the preference scale.


Perceived Spectral Balance of Room Corrections


Listeners rated the perceived spectral balance of each room correction across seven equal logarithmically spaced frequency bands. The mean spectral balance ratings averaged across all listeners and programs are shown in slide 18. The more preferred room corrections were perceived to have a flatter, smoother spectral balance with extended bass. The less preferred room correction products (R5 and R6) were perceived to have too little bass, which made them sound thin and bright.



Listener Comments on Room Corrections


Listeners also gave comments related to the spectral balance of the different room correction products. Slide 19 shows the number of times a particular comment was used to describe each room correction. The bottom row indicates the correlation between preference rating and the frequency of the comment. The most preferred room corrections were described as "neutral" and "full," which corresponded to flatter, smoother and more bass extended spectral balance ratings. The least preferred room corrections (R4-R6) were described as colored, harsh, thin, and muffled, which corresponded to less flat, less smooth, and less bass extended spectral balance ratings. Slide 20 graphically illustrates the same information in slide 19.


Correlation Between Subjective and Objective Measurements


In-room acoustical measurements were made at the six listening seats using a proprietary 12-channel audio measurement system developed by the Harman R&D Group. Slides 23 and 24 show the amplitude response of the different room corrections spatially averaged for the six seats (slide 23), and at the primary listening seat (slide 24). The measurements are plotted from top to bottom in descending order of preference, each vertically offset to more clearly delineate the differences. A few observations can be made:


  1. The six-seat spatially averaged curves (slide 23) of the room corrections do not explain listeners' room correction preferences as well as the spatially averaged curves taken at the primary seat (slide 24). This makes perfect sense since all of the listening was done in the primary listening seat.
  2. Looking at slide 24, the most preferred room corrections produced the smoothest, most extended amplitude responses measured at the primary listening seat. The largest measured differences among the different room corrections occur below 100 Hz and around 2 kHz where the loudspeaker had a significant hole in its sound power response. The room corrections that were able to fill in this sound power dip received higher preference and spectral balance ratings.
  3. A flat in-room target response is clearly not the optimal target curve for room equalization. The preferred room corrections have a target response that has a smooth downward slope with increasing frequency. This tells us that listeners prefer a certain amount of natural room gain. Removing the rom gain, makes the reproduced music sound unnatural, and too thin, according to these listeners. This also makes perfect sense since the recording was likely mixed in room where the room gain was also not removed; therefore, to remove it from the consumers' listening room would destroy spectral balance of the music as intended by the artist.


Conclusions


There are significant differences in the subjective and objective performance of current commercial room correction products as illustrated in these listening test results. When done properly, room correction can lead to significant improvements in the overall quality of sound reproduction. However, not all room correction products are equal, and two of the tested products produced results that were no better, or much worse, than the unequalized loudspeaker. Room correction preferences are strongly correlated to their perceived spectral balance and related attributes (coloration, full/thin, bright/dull). The most preferred room corrections produced the smoothest, most extended in-room responses measured around the primary listening seat.


More tests are underway to better understand and, if necessary, optimize the performance of Harman's room correction algorithms for different acoustical aspects of the room and loudspeaker.


References


[1] Sean E. Olive, John Jackson, Allan Devantier, David Hunt, and Sean Hess, “The Subjective and Objective Evaluation of Room Correction Products,” presented at the 127th AES Convention, New York, preprint 7960 (October 2009).

53 comments:

  1. Dear Mr.Olive,

    "A flat in-room target response is clearly not the optimal target curve for room equalization. The preferred room corrections have a target response that has a downward slope with increasing frequency."

    How does this finding correspond these studies?:
    http://i78.photobucket.com/albums/j92/udauda/er4/K-2-1.gif

    BTW, I appreciate you sharing these great up-to-date scholarly contents with the public. It is really hard for the general audiophiles to get hands on JAES-materials.

    ReplyDelete
  2. Hi Udauda,
    Thanks for your comments. I'm not sure where those graphs came from so if you can point me to their source that would be helpful.

    Your graphs don't show the in-room target below 100 Hz where the largest differences where found among the room corrections we tested. Your curves do show a roll-off at higher frequencies, which confirms our results - although the magnitude of the slope and the frequency at which it starts are not the same.

    In truth, the optimal in-room target curve may depend on the loudspeaker directivity and reflectivity of the listening room. If the room is acoustically dead with few reflections and/or the directivity of the loudspeaker is quite high, the in-room response will represent a higher proportion of the direct sound, which should be flat. Using a target curve with large downward tilt will make the loudspeaker sound too dull.

    Therefore, the better room correction products will be more intelligent how they respond to variations in loudspeaker directivity and room reflectivity.

    ReplyDelete
  3. Measurement slide 24 doesn't tell the RC product numbers. How can we correlate which preferences coincide with which objective measurements? RC1 is the favorite and it subjectively measures flat according to your sample. How did it measure objectively?

    ReplyDelete
  4. Hi J Iannone
    Sorry for not making this more clear. In the blog article I point out that "the measurements are plotted from top to bottom in descending order of preference". So for slides 23 and 24 the curves from top to bottom are: RC1, RC2, RC3, RC4, RC5, and RC6

    ReplyDelete
  5. Sean

    what is the relationship of this in room eq target with sound power (i find it too similar to lsr measurement techniques curves)

    is the job of the correction system just to balance out this sound power depending on the acoustic caracteristicas of the room.

    have you done some tests with speakers that do not have a nice flat and smooth radiation pattern
    did room correction help them (a suppose not very much) in this case the 802n have a radiation pattern that imitates the axis frequency repose but that is not always the case.

    ReplyDelete
  6. Which one is which? I want to buy one but not waste my money.

    ReplyDelete
  7. Hi Valentin,
    Yes it seems that most of these room correction devices try to fill in the peaks and holes of the spatially averaged in-room response curve. Of course, the extent to which the in-room response curve matches the sound power response depends on room acoustics. Usually it's something in between the sound power and the direct sound.

    I would consider this loudspeaker (802N) as an example of " not having a nice flat smooth radiation pattern." In this case, the room corrections provided by RC1 through RC3 significantly improved the sound quality.

    ReplyDelete
  8. thanks

    so what happens when you do the same to a very well behaved loudspeaker like the LSR6332

    will the correction systems make them very hard to differentiate them once calibrated or will there be still a good difference between the well behaved and the irregular speaker

    so the b&W are not smooth es that i can see but i meant more like the Zu Essence or druid which are acclaimed by the press and have afoul measurements Essence (http://www.soundstagemagazine.com/measurements/zucable_druid/) or (http://stereophile.com/floorloudspeakers/zu_essence_loudspeaker/index5.html)

    ReplyDelete
  9. The JBL LRS6332 is the standard loudspeaker playback system in all of our Harman Reference Rooms located in Northridge, Farmington Hills, UK and Germany. All of these rooms are equipped with the Harman room correction system described in this paper. Since the JBL LSR6332 is already a good loudspeaker, we are mostly doing room correction at below 300 Hz. We also use Sound Field Management applied to four subwoofers in the corners, which largely takes care of the odd order modes. Because the LRS6332 has well-behaved direcitivity we have the option of equalizing its direct sound without adversely affecting its sound power response, and vice versa.

    Room correction will improve poor loudspeakers like the ones you have indicated, but why put an expensive bandaid on a scab if you can easily avoid getting the scab in the first place?

    ReplyDelete
  10. We did a study of the best 6 hamburgers and we can tell you that some hamburgers were much better than others.

    Just don't ask us which ones....

    LAME

    ReplyDelete
  11. "Room correction will improve poor loudspeakers like the ones you have indicated, but why put an expensive bandaid on a scab if you can easily avoid getting the scab in the first place?"

    Indeed :)

    ReplyDelete
  12. Hi Anonymous:

    Being anonymous yourself, I would think you'd appreciate the fact that the identities of the room corrections are also anonymous :)

    The top two hamburgers we tested were made by Harman. The other 4 hamburgers have been identified as having tainted meat. So buy your hamburgers at Harman if you want to avoid getting sound poisoning.

    THe Big Cheese

    ReplyDelete
  13. Sean,

    LOL. You are too funny. You know though that first you have to have the hamburgers safe for human consumption and available for sale.

    Poste as Anonymous, but not
    Randy Bessinger

    ReplyDelete
  14. Sean

    As I recall when the JBL tech came over to calibrate my system, the equipment included two laptops, five mics, a router, and a DACS box. I realize a Synthesis® calibration may be more ambitious than any of the products tested here, especially given that all the parts of the system are "known" by the calibration software and that the SDEC gives amazing control that's unlikely to be found in a receiver-based system.

    Having written all that (and I did), will the Harman products be an extension of the JBL/Lexicon/BSS ecosystem? Or will they be more like Audyssey and others that by nature must be more generic, "one-size-fits-all" applications because of uncontrollable variables like speaker selection?

    Perhaps as an in between point, if we have Harman brand pickles and mustard, will the new Harman hamburgers taste better than with Bose onions and Monster cheese?

    Doug

    ReplyDelete
  15. Hello Sean,

    thanks for sharing al those infos.

    A question about the pdf :
    - slide 24 and 23 show that the best performer is the RC1 (red curve, Harman 2 ?) that has been optimised for one seat over RC2 (green curve, Harman 1 ?) that has been optimised for 6 seats. Can you confirm ?

    Comes another question : how do you average measurements ? Do you average only 6 or many measurements to avoid artefacts from reflexions ? Do you use long gating windows ? But maybe those infos are in the AES preprint.
    Thanks a lot.

    ReplyDelete
  16. So, Sean, tell me, what do you equalize? Time? Midrange Amplitude, first arrival?

    ReplyDelete
  17. Hi JJ Ohl,

    Yes, I can confirm RC1 is optimized for primary seat, RC2 is optimized for the 6 seats. In this case, we did spatial-averages over 6 microphone positions to avoid equalizing spectral artifacts that are very localized to the position of the microphone. These artifacts can be from room resonances or acoustical interference effects from the direct/reflected sound. We used steady-state measurements in this study.

    ReplyDelete
  18. Hi Sean,

    Why didn't you choose more diverse music? If I remember correctly, music program can have large influence on listener preferences (as seen in "Sound reproduction" figure 8.14). Also, I think it was wrong decision to choose "mono" for a RC tests; in real world room (which is not perfect and symetrical as harman's)each stereo channel contributes differently to the percived response and it would be interesting to see how RC interacts with that - of course one would need less perfect room for that.
    I had some expirience with Copland DRC, but, in my room, it was better without it (potential problem is that Copland is A/D-DSP-D/A machine).

    Thanks for those most interesting blogs.
    Best regards,
    Vuki

    ReplyDelete
  19. I have a suggestion for an article topic: have there been any good blind tests of systems that convert 2-channel stereo into surround sound? I'm talking about things like Dolby Pro Logic II, SRS circle surround etc. It would have to cover different kinds of music and recordings.

    Also, another less important topic: Harman uses blind testing to ensure speakers are judged only on sound quality during the engineering process, but people don't usually buy speakers blind. Do they use similar "focus group" type testing to judge different aesthetic designs, ("deaf testing?") or to do sighted listening and judge people's overall impressions of the speakers? I'd think those would be important if you're actually trying to sell the things, and to make them acceptable in homes.

    ReplyDelete
  20. Dear Mr. Olive,

    thank you very much for sharing this interesting report.

    The problem of flat target response sounding usually too bright is probably as old as equalization is. So far the better explanation I found for this is the fact that our ear works by detecting "spectral envelopes" and not a smoothed average spectra like the one you get by traditional fractional octave smoothing.

    Am example of what you get when you evaluate a room response by a spectral envelope may be found here:

    http://drc-fir.sourceforge.net/doc/drc.html#htoc28

    I'd be glad to here your opinion on this issue. So far I found no reference in the literature about the use of the spectral envelope for equalization, but its use is taken as a given in the speech synthesis and analysis field, which is where I found it for the first time.

    Regards,

    Denis Sbragion

    ReplyDelete
  21. Another thing I don't understand is: how can it be that flat is not the prefered curve? Than either the whole recording chain is wrong (tilted) or the listeners are "trained" wrong? Shouldn't (in theory at least) the flat curve be ideal?

    ReplyDelete
  22. Hi Sean,

    Thank you for posting your slides from the AES convention. Between reading your blog and Dr. Toole's book, I was inspired to learn how to use FFT software and bought an equalzier. I used your slides to try to create the preferred curve in the single seating postion. I did spacial averaging and have been playing with various filters to reduce peaks, but mostly seem to need to fill nulls. I am experiencing the sound power issues you mention, particularly at 2000hz. By the way, I have a two-channel system.

    I am curious about the use of filters above 200hz. There seems to be a general consensus that it is not so useful, or it degrades sound quality in the higher frequencies. I'm making the assumption, given your conclusions in the blog, that increasing the sound power around 2000hz is desirable with filters?

    I have been trying to create a sloping curve like the red one in slide #24. I'm thinking that using fewer filters is better, but I certainly am finding it imposible to create such a flat curve. Would you mind sharing any general guidelines you recommend when using PEQ?

    I hope you don't mind a question about speakers. I like two channel listening, but I am curious about multi-channel as well. I have read a fair amount about several of the Revel speakers and I'm curious how one might go about deciding how to choose speakers (without sound power issues) given specific room dimensions.

    Thanks for a great Blog,
    Nick

    ReplyDelete
  23. Hi Nick,
    Thanks for your comments. Equalizing above 200 Hz did indeed result in an improvement in sound quality for this loudspeaker by filling in the 2 kHz hole in its sound power response. The danger in equalizing above 500-600 Hz is when the loudspeaker has a non-contant directivity: you might improve the sound power response at the expense of the direct sound.

    The flat curve we created (RC1 in slide 24) was achieved with 12 filters. We use some psychoacoustic rules in prioritizing which peaks and dips we go after. The spatial averaging is important in avoiding equalizing peaks or dips that are very localized to one microphone location. You don't want to equalize something that changes when you move your head a few inches.

    I would contact Kevin Voecks at Revel with any specific questions you have about Revel loudspeakers. The main issue in selecting a specific loudspeaker for a room is the volume of the room - not its dimensions. The larger the room, generally the more SPL output capability you need from the loudspeaker to fill the room with sound.

    Cheers
    Sean

    ReplyDelete
  24. The anechoic-based PIR response is it not flat, but rather has a downward slope with increasing frequency. This is because the loudspeaker becomes increasingly directional at high frequencies. When you put the loudspeaker in a semi-reflective room, the measured response will naturally have an in-room response that looks like the PIR. If we equalized the in-room curve to be flat, it would boost the HF of the direct sound, and make the loudspeaker sound too bright and thin.

    The only case where a flat in-room response would be desirable was if the room were the acoustic equivalent of an if anechoic chamber and the listener is hearing just the direct sound.

    ReplyDelete
  25. Dear Mr. Olive,

    regarding the in room response being close to the PIR response, and so causing a bright direct sound when equalized to flat, well, this is what I thought too since a couple of years ago.

    Then I built my own listening room, which is by design really damped. This is the (equalized) response I get in that room, showed with many different windows applied to the impulse response:

    http://itchef.infotecna.it/ListeningRoom/Misure/index018.gif

    As you see the response is flat both for the direct sound and the stationary field. So in theory this should sound neutral. Unfortunately it wasn't so, the subjective balance was still bright, quite a lot.

    Again, when I checked the spectral envelope of the room response I got a non flat, somewhat raising, response, and equalizing for a flat spectral envelope provided the expected subjective neutral balance.

    Regards,

    ReplyDelete
  26. Thank you, Sean. I hope my question below is a simple enough one to answer. If there is another discussion or any research that addresses the topic that you know about I would very much appreciate being directed to it.

    You said:

    "The danger in equalizing above 500-600Hz is when the loudspeaker has a non-constant directivity: you might improve the sound power response at the expense of the direct sound."

    I'd like to understand this comment better. I'm thinking 'non-constant directivity' means that there is a dip in the frequency response somewhere, i.e. the dip at 2000Hz?

    I am unclear about why improving the sound power response -- boosting the spot where there is non-constant directivity -- would cause the direct sound to be negatively altered?

    I'm stuck here because it seems that boosting the area around 2000Hz would increase the direct sound from the loudspeaker?

    ReplyDelete
  27. Hi Nick,

    Suppose the direct (on-axis) sound is flat but the sound power response has a big hole at 2 kHz. When you measure the loudspeaker in a reflective room you will measure a combination of the direct sound and the sound power response.

    What you will mostly see in the in-room measurement is the sound power response of the loudspeaker, since the reflected sounds make up a greater proportion of the total energy at the microphone. In this scenario, filling in the 2 kHz hole in the sound power will at the same time boost the on-axis sound at 2 Khz as well. So the danger here is that if the quality of the direct sound mainly determines your impression of the loudspeaker sounds, the 2 kHz boost of the direct sound will make the loudspeaker sound too bright.

    ReplyDelete
  28. Hi Denis,

    Based on the measurements in your link, the in-room response is very flat. In our room correction paper, we found that flat in-room target curves sound too thin and bright. The preferred in-room target has a negative slope.

    I'm not sure what you mean by room's spectral envelope. How did you measure this?

    ReplyDelete
  29. Hi Vuki,

    The programs were chosen on the basis of their ability to reveal audible differences among the room corrections: the important program characteristics include spectrally dense, wide bandwidth, low bass and are relatively constant and homogeneous over time.

    The choice of mono playback was based on a previous room correction study we did that found listeners were more discriminating of room correction benefits in mono versus stereo and surround. We used a so-called ITU-R symmetrical room and loudspeaker setup in that study, and I would agree that the benefits of room correction might be even greater for non-symmetrical rooms and setups. That doesn't mean the results from our study are invalid.

    ReplyDelete
  30. Hi Jimmy Olson:

    Thanks for your suggestions. We are doing some research and testing of up-mixers, and hopefully I can talk about this in a future article.

    Regarding your question on focus groups and testing consumer preference for different product industrial design: Our market research people are doing that type of research, in addition to the sound quality tests that I do.

    This research is very important, since the look and feel of the product (as well as brand and marketing) often drives then product purchase decision before the consumer even gets an opportunity to hear the product.

    This is why companies like Apple are so successful. People will pay a premium for a product if it looks expensive, beautiful and is easy to operate.

    ReplyDelete
  31. Dear Mr. Olive,

    I completely agree with you that a flat room reponse (I mean flat by fractional octave, usualy 1/3, smoothing) sound thin and bright. This is an established fact since the old era of 1/3 of octave analog graphic qualizers. I'm only questioning that the main cause of this is the fact that most of the times the direct sound becomes bright when the stationary field response get equalized to flat. It may have a part, but I think it is just part of the story.

    First of all, some clarification. I'm an hobbyist, who started working on room equalization something like 10 years ago, just for fun. I found this issue of the spectral envelope almost by accident, exactly because I was getting the same problems you already found (much quicker than me) in your experiments. Of course, as an hobbyist, I don't even remotely have the knowledge and experience of people like you and Mr. Toole. That's why I'm asking for your opinion. Furthermore, my English isn't that good, so forgive me if some comments might appear a bit harsh. It's just that I'm a bit in troubles finding the right words.

    Regarding the spectral envelope, again, it's something I found almost by accident. Before finding it, I was using a tilted response too, following the usuale Moeller curve. On most "normal" rooms, it worked pretty well, as expected. Then I found that in my room neither the usual Moeller curve, or other kinds of tilted curves, nor a completely flat response were going to work.

    So I started searching some reference about how a given unsmoothed magnitude response will be perceived by our ear. By pure luck I found some papers in the field of speech synthesis and analysis explaining how our ear works when evaluating timbre and balance. If you look for the "warped minimum variance distortionless response" or "discrete cepstrum" keywords you'll find a lot of papers on algorithms for spectral envelope computation.

    These algorithms are optimized for speed, noise resilience and the voice range, so they are not really suited for the HiFi field. So I developed my own algorithm, checking of course that it was providing results close to the algorithms above, at least in the applicable voice range. I also checked that, with normal rooms, it was providing something close to the usual expected tilted response, as it should. Then I applied it to my quite uncommon room, and, to my surprise, it worked unbelievably well. Since then I tested it also on many other rooms and I ensure you that I never have been able to achieve a balance as uniform and similar sounding as the one I got after introducing the spectral envelope computation.

    So it seems to work pretty well, but to my surprise I found almost no reference about the spectral envelope in the standard audio literature. Don't know why. If you read the papers in the speech synthesis and analysis field the fact that our ear works on the sectral envelope, and not the traditional smoothed response, is taken for granted. It's not even questioned. The only issues are about the best and fastest way to compute it. I even found a medical paper analyzing by MRI the parts of the brain involved in the spectral envelope "computation" and analysis. So I really don't know why there's no trace of it in, say, the JAES or other audio related journals.

    Regarding my own algorithm, it's pretty simple. Having no speed and real time performance issues it has been quite easy to develop a variation of the usual fractional octave smoothing but based on a different power averaging and so providing a sort of "envelope detection" instead of simple smoothing. If you want I will provide you some Matlab/Octave code which may be used to test and evaluate it. Just write me by e-mail (d.sbragion@infotecna.it) and I'll send it to you with some explanation on how to use it.

    Thanks again for your kind attention.

    Regards,

    Denis Sbragion

    ReplyDelete
  32. Sean

    do you now if the JBL Performance BassQ processor uses the sound field management algorithm or is it only available threw synthesis sdec/bss systems

    ReplyDelete
  33. I greatly enjoyed reading and learing from this article.

    What would you say to my 'self help tip' that I took from this article, that it would be a good starting point for me to equalise my home system to -1dB per octave from 20Hz to 20kHz? (using my digital 1/3 octave equaliser, pink noise, averaged response, with microphone correction).

    Grant S

    ReplyDelete
  34. Hi Dr. Olive,

    This is another great entry! Thanks very much!

    I have one question though. It is clear to me that the B&W has issues with its directivity and that it can benefit from slight EQ to boost the dip in its power-response. However, what should you do if you have loudspeakers with very well behaved frequency response and directivity - like the Revel Studio - but there is a peak in the steady-state measurements in the midrange? It can't be the speaker, so it must be the room that causes it. Should you reduce the peak by means of EQ and if yes, to what extent?

    If you can find the time, an answer would be much appreciated!

    Martijn

    ReplyDelete
  35. Hi Dr. Olive,

    Since my last message here I have done some experiments myself. My loudspeaker system is a digitally crossed and equalized dipole with a small baffle size. This ensures a very smooth and constant DI-curve up to high frequencies. The frequency response is very flat and smooth too. However, in my (living-) room the steady-state spatial average response at the listening position is not as smooth as I would expect. In the midrange there is a peak in the response (centred a little above 1 khz). The placement and acoustics in my room are sub-optimal, but don't seem particularly bad. EQ'ing the peak completely flat gives the best subjective sound quality.

    I've set up my system in a friends' dedicated listening room. The placement and acoustics were a lot better. In his room the response at the listening place is a smoother, without the peak. There was no need for midrange EQ.

    My preliminary conclusion is that, contrary to common believe, room-EQ in the midrange can be beneficial.

    I'd still very much like to hear your opinion.

    Martijn

    ReplyDelete
  36. Hi Sean,

    Very interesting paper or DSP room EQ.

    I have considerable experience with the Snell RCS1000 (predecessor to the Tact and Lyngdorf) and some experience with the Audysey.

    Aside from the curve of system 6, the difference between the other curves seems to be the choice of target curve. That is #1,2,3 and 5 were nearly equally good at smoothing out the curves. In fact in slide 24 there is very little difference between the result for system 2 and system 5. Would your care to speculate on why such a difference in listener preference for those two?

    Judging from the correlation between your room averaged curves and the EQ results these look like steady state EQ approaches. As you know, the problem with that is that you have to guess at what the best in-room target curve should be. A you point out EQing to a flat response would make the (more important) direct field too bright. The trouble is that the ideal room curve to achieve a good balance with the early sound is fairly indeterminate. This was a frustration with the Snell unit and I’ve observed the same with the Audysey guys. There was always an element of listening to trial EQs and then fudging the target curve to get the final balance you wanted.

    JBL seems to have chosen a 1dB per octave downslope as a target, but what if the speaker they were EQing had a different d.i.curve (or the room a different absorption curve)?

    Any reason why they didn't attempt to do room correction based on a time windowed measurement? Lots of research shows that variable time window (long at LF and short at HF) gives a measurement closer to the perceived balance.

    Clearly if we want automatic EQ to work consistently then an appropriate measurement needs to be the starting point.

    (By the way, Lyngdorf doesn’t believe in EQ for the top end of the speaker, so I’m guessing they are number 6.)

    D. Smith
    Toronto

    ReplyDelete
  37. Hi Dave,

    Thanks for your post. I agree with much of what you say.

    It is true that much of the difference between these different room correction products is related to the choice of target function.

    Slide 24 in the presentation shows the in-room responses of the 6 room corrections tested (including no EQ). measured at the primary listening seat. Here you can see significant differences below 100 Hz and around the sound power dip in the loudspeaker. There are also significant differences in the overall slope of the curves measured over several octaves.

    You asked why Room corrections 2 and 5 scored so differently based on preference. If you carefully compare the two in-room measured curves there are significant differences in low frequency extension below 40 Hz; but more importantly, the slopes of the two curves over several octaves are sufficiently large that the difference in perceived spectral balances between two loudspeakers is quite different. This is evident in the perceived spectral balance ratings and listener comments given to RC 2 and RC5.

    The Lyngdorf product we tested most definitely applied a full band room correction (at least up to 4 kHz) since it fixed the sound power hole in the loudspeaker around 2 kHz. Room Correction 6 didn't fix the sound power problem, and in fact made it worse.

    You suggest that a variable time window applied to the HF and LF portions of the curve give the best indicators to the perceived balance, yet I am not aware of published studies that definitely support that claim. I only know of only 1 or 2 previous studies on room correction that have published results based on controlled listening tests. If you could provide me those references, I would appreciate it.

    Cheers
    Sean

    ReplyDelete
  38. Dr. Olive,

    Please excuse me, but is there something wrong with my two posts that you choose not te reply? I thought you might not have read them yet, but you did reply to David's post above.

    David,

    I would also very much like to see those papers. I find it very plausible that you should treat different parts of the time window differently, as reflected sounds are processed in our brain differently than the direct sound. All in all I think this kind of research would be difficult to do if you are aiming for good external validity.


    Kind regards,

    Martijn

    ReplyDelete
  39. Hi Sean,

    Thanks for the response. Yes, once you mention it I can see that there are small but significant differences between #2 and #5.

    As to papers about time windowing:

    A New Psychoacoustically More Correct Way of Measuring Loudspeaker Frequency Responses
    Preprints 1871 (F-4) and 1963 (G-4)
    Jorma Salmi

    A Perceptual Criterion for Loudspeaker Evaluation
    James M. Kates
    JAES Vol. 32, No. 12, Dec 1984

    and: Samuel Bridges
    Effect of Direct Sound on Perceived Frequency Response of a Sound System
    AES preprint 1644 (H-6)

    These are pretty good on the subject (maybe not to your usual standard!). There are also some interesting papers that have good experimental evidence on the subject but "come to the wrong conclusion", such as Schulein.

    Send me your e-mail on LinkedIn and I'll send you a synopsis I wrote up for another project.

    Regards,
    David

    ReplyDelete
  40. Hi Martijn,

    Sorry for the delay in response. I don't always respond to comments in chronological order and I apparently missed yours.

    Regarding your above question about the room correction to a loudspeaker like the Revel Salon that is known to have a smooth frequency response and directivity, yet the in-room measurements indicate a peak in the midrange from the room?

    In cases like this, you have to ask yourself what caused the peak? Normally peaks associated with room resonances will not be visible above 300-400 Hz and at those frequencies the peak will vary from seat to seat.

    Constructive/destruction interferences between the direct and reflected sounds can cause a peak in the midrange but normally this disappears and changes with microphone location. If you do 3-6 spatial-averages, the peak should disappear, and the need to equalize goes away.

    If the peak survives spatial averaging then I suspect the problem is with the loudspeaker. If the loudspeaker has constant or smooth directivity you can equalize out the peak and improve the on and off-axis sound produced by the loudspeaker.

    If you don't know the on/off-axis behavior of the loudspeaker, you risk improving the reflected sounds at the expense of the direct sound.

    Finally, if you own Revel Salons, they don't need any equalization except below 300 Hz where the room dominates what you hear :) If you own good loudspeakers, you should focus on correcting the low frequency acoustical interactions between the loudspeaker and room.

    ReplyDelete
  41. Hi Martijn,

    Regarding your 2nd question about success in equalizing midrange peaks in your loudspeaker: I think your experiment at home confirms what we've found now in two room correction experiments using two different loudspeakers. If you have a loudspeaker with reasonably constant or smooth directivity you can equalize problems in the loudspeaker up to a fairly high frequency.


    Some dipoles (e.g electrostatic panels) do have relatively constant DI above a certain frequency-like Loudspeaker M in this link: http://seanolive.blogspot.com/2008/12/part-3-relationship-between-loudspeaker.html

    Although this speaker has a horrible frequency response, it is similarly horrible at different measurement positions. It has the potential of sounding much better with equalization since removing a peak on-axis will also fix the same peak off-axis.

    In your example you noted that the visibility of the mid peak changed depending on which room it was measured. One possibility is that the peak is not due to the room per se, but in the off-axis response of the loudspeaker. The peak would be more noticeable in the measurements taken in the more reflective of the two listening rooms. I don't know if this was the case or not.

    ReplyDelete
  42. Hi Sean,

    Thank you for your reply. I don't want to draw away too much from the topic, but just for reference, here is a plot of the response of de speaker. These are the curves from onaxis to 90 degrees in 15 degree steps. The green curve is the average of the response at these six angles.
    http://img709.imageshack.us/img709/2514/filex.jpg
    In my opinion this is an excellent measurement result, except perhaps for the range above 7 khz. The response at the rear looks a bit less good, but still the irregularities at the rear don’t seem to coincide with the midrange bump measured at the listening seat (spatial average of several measurements). The vertical response looks reasonably good too. Also, my living room does not sound more live than my friend's listening room.
    Therefore I don’t think the speaker to be the cause. The speakers are placed at about 1.4 meters away from the front wall and a little over a meter from the side-walls. However, I think the problems I measure in my room might be caused by early reflections against the furniture in close proximity of the speakers. I will have to investigate it further.

    Martijn

    Martijn

    ReplyDelete
  43. Hi Martijn,

    The measurements do indicate that the speaker is well-behaved in the horizontal axis. These measurements I presume have been time gated to only include the direct sound, so the bump above 1 kHz in the steady-state measurements must the caused by a reflection as you suggest. If you did enough spatial averages my moving the microphone to different seats the effects of the reflection off the furniture should disappear. You can also put some absorption on the reflecting surface to confirm/deny that it's the culprit.

    What you want to avoid is attempting to "equalize" the effects of a discrete reflection.

    ReplyDelete
  44. Hi,

    IMHO the discussion about flat or sloped target curves does not hit the right point. I wonder that people do not recognize how e.g. different smoothing algorithms lead to different slopes.

    At the end it is more important to study the resulting correction curve. A speaker with a good direct pulse response typically does not need too much correction, thus the correction curve should be quite flat.

    Now if a smoothing algorithm leads to a certain slope the correction will only be flat if the target follows the same slope.
    Because correction = target minus smoothed measurement
    (deconvolution in time domain = division in frequency domain = subtraction in logarithmic frequency domain)

    So watch out for the correction curve !

    ReplyDelete
  45. Having spent 10 years as the Sales and Marketing Director of the "original" digital room correction company (SigTech), I found this incredibly interesting.

    But I have one question: given that at least two of the products you tested (Audyssey and Tact/Lyngdorf) have the ability to have variable target curves, how did you determine which curve to select AND remain completely neutral given that Harmon was testing a Harmon prototype as part of the group. And by neutral, I am not suggesting that the preferred target response was not valid, but rather the acceptability of a particular product.

    (I am submitting this as 'anonymous" since when I tried some of the other choices, nothing worked (operator error!!)

    ReplyDelete
  46. After almost three months since the last comment, I´d like to (try and) bump this discussion.

    First of all, huge thanks for a great blog and especially this posting (and for the work you, Floyd Toole and all your colleagues have been and are doing).

    More specifically, I have two questions:

    Could you provide an evaluation of the account of "spectral envelope" as a guideline for room correction that Dennis brought up? From what I understand this might give useful insights into the psychoacoustic backgrounds of the correlation of room curve measurements and preference ratings.

    Second, as a recently converted DIY audio guy, I am working on a full range dipole design with the aims of constant directivity and smooth on- and off-axis frequency response. My question thus is this: Do I understand your research correctly if I suppose that room correction should only be needed below the Schröder frequency if dealing with well-behaved loudspeakers considering their on- and off-axis frequency response? From what I gathered, response artifacts caused by specific reflections shouldn´t be subject to EQ because their effect will vary with listener position. Am I right about this?

    I hope you will excuse the naiveté of my questions - I´m still learning...

    Thanks again for the insights you provide!

    Wolfgang

    ReplyDelete
  47. Dear Mr Olive,
    thanks for sharing your findings with us. How does the slope of the corrected frequency response depend on the directivity of the speaker?
    As I understand,a conventional speaker is focusing more and more whith rising frequency. Therefore the total energy it radiates into the room reduces with frequency. Since in 4 m distance the sound is dominated by the reverberant sound field, the frequency resonses measured at the listening position will have a downward slope with respect to frequency.
    What would be, if you test and correct a speaker with really onidirectional behaviour even at the highest frequencies. Would the slope be less inclined or flat?

    Markus

    ReplyDelete
  48. Hi Dr. Olive
    Thank you for sharing your work in this blog.
    I have read some of Dr.Toole and your work, and it really helps a lot.
    I know it's important for a speaker to have flat and smooth frequecy response for on-axis, early reflection, and sound power. However, how do I choose a good subwoofer? Since the frequecy range a subwoofer produces has low directiviy, does it mean that I only have to check the on-axix frequecy response? I check the spec of jbl pro subwoofer lsr4312sp for example, its low-end response does not extend as low as some other competitor's products that have similar size and price, ex. svs or velodyne. Do I miss some spec that is also important for a subwoofer?
    Thank you.

    ReplyDelete
  49. Hi Joey,
    You are correct. Subwoofers are generally omni-directional over their operating frequency range although LF ports can be highly directional. So a single frequency response curve can characterize their amplitude response, which should extend down 25 Hz or lower. That being said, the room will generally dominate what you hear below 200 Hz and therefore some control or equalization is necessary. This can smooth out and extend the frequency response of the subwoofer at a single seat as long as the subwoofer has the output capability to handle the EQ boost.

    And output capability is where most subwoofers differ in performance. The factors include the performance of the transducer(s), box volume, the size of the amplifier, and the type of electronic limiting used. Ideally, you need one or more subwoofers capable of reproducing bass down to 20 Hz or more at SPL's approaching 100-110 dB @ 1 meter with minimal audible distortion.

    There have been some attempts at characterizing the useful output capability of subwoofers such as the peak SPL in 1/3-octave bands using test signals like Don Keele's co-sine shaped tone bursts (a.k.a. boinks). These signals are very short in time and therefore don't heat up the voice coil. The main issue with all of these tests is that the results don't necessarily correlate with subjective listening results using music signals. More research is needed to characterize the performance of subwoofers that correlate measurements to subjective performance.

    ReplyDelete
  50. Hi Sean,

    Great presentation. Have you had a chance to eval some of the latest DRC software offerings? Some good discussions going on:

    http://yabb.jriver.com/interact/index.php?topic=69312.0

    http://yabb.jriver.com/interact/index.php?topic=68828.0

    http://www.computeraudiophile.com/blogs/What-tone-quality-your-audiophile-system

    Would be interested in your thoughts.

    Cheers, Mitch

    ReplyDelete
  51. Hi Mitch
    Thanks for the useful links. Very interesting discussions and I'm happy to hear there in consensus among us (you and B&K) in the choice of in-room target curves for equalization. I was not familiar with the B&K work until recently when someone brought it to my attention.

    The Bob Katz tonal balance chart was also interesting, and I generally agree with most of this terms and the ranges -- although the exact frequency ranges depend on the instrument,etc

    ReplyDelete
  52. Hi Sean,

    Thanks and the discussions continue. As an ex recording/mixing engineer, it was popular to have at least one set of monitors equalized to the B&K curve in most of the control rooms I frequented. Urei 813B time aligns were very popular, I am dating myself now ;-).

    The other interesting side effect using the B&K target (or same target as your findings) is that the soundstage was also the best. Meaning it was not too far back or not to far forward, but just right. I am wondering in your listening tests if you found the same correlation?

    Regards, Mitch

    ReplyDelete