Redefine your budget room EQ 'flat' target curve to Harman's pro curve

[amirm]

I have a lot of respect for JJ Johnston and feel his presentation on the Acoustic and Psychoacoustic Issues in Room Correction that can be downloaded from here: http://www.aes-media.org/sections/pnw/pnwrecaps/2008/jj_jan08/ represents the state of the art understanding of room correction.

Since you like JJ's work, you may be interested in the history of it. I think it was 1999 when I saw a review of Tact Processor/EQ in an audio magazine (stereophile?) where it received a rating of 10 out of 10 in sound quality. So I proceeded to order one at a cost of $10,000. I sat through the onerous interface of TacT and activated it once all the filters were programmed. I was shocked at how seemingly the walls of my small theater had collapsed -- just like the review said. For the first time, I had correct blending of my sub and the rest of the speakers (powered/active Paradigms).

Then I did another experiment. I hooked it up on the output of my PC on the way to cheap powered stereo speakers for my computer. You know, the plastic little cups they call speakers and crappy box they call subwoofer. I was shocked and amazed and the difference it made there. This beyond cheap system had superb imagine and again, smooth bass response.

I set up meeting for my signal processing group and repeated the demo for them with PC speakers. Everyone was shocked at the difference it made. So I said let's put this in the audio pipeline of Windows. Resources being what they were, this sat there until we got lucky and AT&T blew up a music delivery system they had. And with it JJ became available. A number of companies went after him but we were lucky to persuade him to join us as our audio architect. We put revamping of the Windows XP pipeline and addition of room EQ on JJ's plate.

During the process, JJ was worried about the CPU requirements of an EQ. He explained that if we just matched the gain and delay between loudspeakers we would get most of the benefits. As much as I was sympathetic to that, I insisted that we get a proper EQ in there. Courtesy of our top of our optimization team members like Serge in that article, the final system barely used any CPU.

Alas, my team was responsible for the audio pipeline. The rest of the Windows including the explorer/shell was in another group. So the only way we could expose the interface was through crude Properties control panel. In other words, very difficult for anyone to discover. And easy for it to get lost as it did in the follow on releases when I, JJ and others left Microsoft.

Sadly Vista took so long to release that by the time it came out, consumer level Room Eq was available and hence we lost the opportunity to really showcase the power of PC to perform such optimizations.

 

[Fitzcaraldo215]

I have not really obsessed over target curves, and frankly, the "right" target curve, especially in the bass, might be a function of your typical listening levels. I am convinced that some smooth gradual measured roll off should be there in the upper octaves in most any room, and most target curves agree with this. So, I am a diehard believer in full range EQ, particularly since I think a lot of even hi hi end speakers are too bright in most rooms. They need to have nice flat anechoic measured response for marketing reasons but perhaps at the expense of a better sense of musical realism on most recordings.

With Dirac, which uses a curve similar to B&K, I have been quite happy since day 1 with the stock curve. I use that for both music and movies. All I can say is that EQ with that curve positively sonically beats the pants off of no EQ. I would never be without it, even in an extensively treated room, which I do not have and honestly do not feel the need for. Many audiophiles are extreme purists and simply refuse to believe in or will not settle for the obvious improvement DSP EQ can make. Most have not tried it and are convinced that room correction needs to be done the hard way, the more perfect way or not at all. Forgive me, but I am a pragmatist. So, when I can get obviously much better sound with a relatively small investment in the DSP EQ technology to do it, count me in.

I have the view that the listening experiments most audiophiles could do themselves with different curves are problematical in many ways. Listening to test tones ourselves would be ridiculous. And, trying to use music as a reference for tweaking the curve is also fraught with problems. It is subjective, of course, but since we do not know exactly what the recordings are supposed to sound like, we can easily spin our wheels trying to make one recording sound "ideal" at the expense of other recordings. It is not a tone control, for heavens' sake, easily adjusted to make this recording sound "right", and the next one also with a few tweaks to the EQ. (I do not use tone controls myself.). I have a good friend who is doing that with Dirac, and he is hopelessly lost, tweaking and retweaking. So, I am just enjoying the fruits of my 20 minute calibration, laid back and listening happily, while he is still tweaking and fiddling with the EQ.

 

[dallasjustice]

That's a very Interesting story. I bet Mitch would also be interested in it because he works for Microsfoft now. Maybe he can convince them to put Media Center back into Windows 10.

Maybe Mitch can volunteer his services help Microsoft design a time coherent active speaker system driven by low wattage Class A pass labs amps.

Since you like JJ's work, you may be interested in the history of it. I think it was 1999 when I saw a review of Tact Processor/EQ in an audio magazine (stereophile?) where it received a rating of 10 out of 10 in sound quality. So I proceeded to order one at a cost of $10,000. I sat through the onerous interface of TacT and activated it once all the filters were programmed. I was shocked at how seemingly the walls of my small theater had collapsed -- just like the review said. For the first time, I had correct blending of my sub and the rest of the speakers (powered/active Paradigms).

Then I did another experiment. I hooked it up on the output of my PC on the way to cheap powered stereo speakers for my computer. You know, the plastic little cups they call speakers and crappy box they call subwoofer. I was shocked and amazed and the difference it made there. This beyond cheap system had superb imagine and again, smooth bass response.

I set up meeting for my signal processing group and repeated the demo for them with PC speakers. Everyone was shocked at the difference it made. So I said let's put this in the audio pipeline of Windows. Resources being what they were, this sat there until we got lucky and AT&T blew up a music delivery system they had. And with it JJ became available. A number of companies went after him but we were lucky to persuade him to join us as our audio architect. We put revamping of the Windows XP pipeline and addition of room EQ on JJ's plate.

During the process, JJ was worried about the CPU requirements of an EQ. He explained that if we just matched the gain and delay between loudspeakers we would get most of the benefits. As much as I was sympathetic to that, I insisted that we get a proper EQ in there. Courtesy of our top of our optimization team members like Serge in that article, the final system barely used any CPU.

Alas, my team was responsible for the audio pipeline. The rest of the Windows including the explorer/shell was in another group. So the only way we could expose the interface was through crude Properties control panel. In other words, very difficult for anyone to discover. And easy for it to get lost as it did in the follow on releases when I, JJ and others left Microsoft.

Sadly Vista took so long to release that by the time it came out, consumer level Room Eq was available and hence we lost the opportunity to really showcase the power of PC to perform such optimizations.

 

[amirm]

I didn't know Mitch had joined Microsoft. Alas, it is a different company than the one I worked on. Then, from Bill Gates on down there was strong support and huge investment in audio/video technologies. My signal processing team alone was nearly 50 people. All of that disappeared when Sinofsky moved from Office to run Windows. Media Center unfortunately is dead and buried, something that Sinofsky did. I don't think it will be put back in business.

Mitch which group are you in? Feel free to PM me if you don't want to share in public.

 

[Mitchco]

Thanks for the kind words Audiophile Bill. We are already seeing the future of DSP in products like: http://kiiaudio.com/en/tech-acou.html and http://www.tonmeister.ca/wordpress/2015/10/06/beolab-90-behind-the-scenes/ where DSP is being used to control the directivity of the speaker. Probably see more of this along with more speakers bundled with DSP.

Fitzcaraldo215, I don't obsess over target curves either as I have been using the same one for 2 years and sounds good with any music. What I wrote was a quick procedure on how folks can tweak, if so inclined, to find what sounds best to their ears and providing industry references to preferred target responses, which are mostly the same.

Amir, thanks for the story. I have been with MSFT for 5 years, not in the audio division, but know quite a few of the audio guys. Unfortunately, audio in the big scheme of things is not a priority. Windows 10 reduced audio latency, which is great, and introduced a new C# API called AudioGraph to take advantage of that, but it is not targeted towards audiophiles: https://msdn.microsoft.com/en-us/library/windows/hardware/mt298187.aspx In order to bypass the Windows audio subsystem (i.e. mixer, resampler, effects, etc.,) one still needs to code in C++.

 

[Rodney Gold]

I like a hump in the bass ..here's one of my target curves .. no significant downward tilt in the treble as my room is not lively and absorbs a fair amount of hf
All subjective.. I often tweak it , currently the whole bass hump is about 2db down , as the curve Im showing was a tad bass heavy.. good for electronica .. but not that good for natural sounding bass recordings.
I have been using DSP right from the tact days..used many many units..latest stuff is dirt cheap compared to the TACT
Currently using acourate or Dirac.

 

[emcdade]

This was the most enlightening discussion regarding room acoustics and EQ I've ever heard from Paul Hales on HTG.

https://twit.tv/shows/home-theater-geeks/episodes/164

He believes EQ should only be applied responsibly below 200 hz.

The crux of his argument is that most decent speakers are designed to measure relatively flat in an anechoic environment, because speaker designers know that if the speaker measures flat in the anechoic room it will sound good in a typical living room. The intent is not such that speakers will measure flat in living room.

When you apply EQ to the speaker, you alter the direct sound of the speaker that was painstakingly designed to measure flat in an anechoic room, which we already know is desirable.

When we listen and cue into the direct sound of the speaker, we are now hearing a colored version of that wonderful flat measuring speaker thanks to EQ.

It is a great listen.

 

[Wayne A. Pflughaupt]

This was the most enlightening discussion regarding room acoustics and EQ I've ever heard from Paul Hales on HTG.

He believes EQ should only be applied responsibly below 200 hz.

The crux of his argument is that most decent speakers are designed to measure relatively flat in an anechoic environment, because speaker designers know that if the speaker measures flat in the anechoic room it will sound good in a typical living room. The intent is not such that speakers will measure flat in living room.

When you apply EQ to the speaker, you alter the direct sound of the speaker that was painstakingly designed to measure flat in an anechoic room, which we already know is desirable.

When we listen and cue into the direct sound of the speaker, we are now hearing a colored version of that wonderful flat measuring speaker thanks to EQ.

And yet, from what I see in reviews, speakers rarely ever deliver flat response above 200 Hz, even relatively expensive ones. Are we to believe a little electronic correction wouldn’t yield an improvement in sound quality? (Graphs courtesy of Sound and Vision Magazine.)

Bryston Mini-A ($4700)

Paradigm Prestige 15B ($6100)

MartinLogan Motion 60XT ($6700)

PSB Imagine T3 ($11,800​

Regards,
Wayne A. Pflughaupt

 

[emcdade]

And yet, from what I see in reviews, speakers rarely ever deliver flat response above 200 Hz, even relatively expensive ones. Are we to believe a little electronic correction wouldn’t yield an improvement in sound quality?

That depends, how did you decide which speaker to purchase? Did you audition them and pick which one you liked? Or did you just pick one and imagine what it would sound like with EQ applied?

And again how do you know you're even "correcting" the speaker to your target curve when a mic=\= human hearing?

I would argue that if you need to adjust the direct sound of the speaker to enjoy it that you simply picked the wrong one.


I can also tell you that when Audyssey or Dirac corrects the dip in the midrange response as seen on the Motion 60xt graph that the character of the speakers sounds considerably worse. The speaker was designed to have a leaner midrange that gives vocals a more upfront sound. When you "correct" that out the speaker does not handle that well. Suddenly vocals feel strained and compressed. The "correction" in FR essentially ruins the character of the speaker that I picked out, because I liked that sound.

 

[Fitzcaraldo215]

Here is a very good discussion relating to the subject:

http://seanolive.blogspot.com/2009/11/subjective-and-objective-evaluation-of.html

Having calibrated or participated in calibrating a fair number of rooms on very high quality systems with different EQ systems, speakers, etc., I can only say that I agree. Full range EQ with a downward sloping target with increasing frequency was preferable in all cases, not just to me, but to others, most importantly to the system owner. All of those involved are frequent classical concert goers who listen primarily to hi Rez recordings.

Yes, omni calibration mikes do not hear like people do. The downward sloping target is one compensation for this to achieve flat perceptual response rather than flat measured response, though it is an approximation and it might need fine tuning adjustments. Some tools offer a range of target curves for different room volumes, also an approximation.

I do not think it is possible to hear precisely "what the speaker was supposed to sound like" or the sound "that you bought" in your room, which is likely quite different from the anechoic chamber, rooms the designer used or even the room in which you auditioned it.

 

[amirm]

The thesis of that video is correct in that if you have a well designed speaker, you don't want to start the process with full EQ correcting for dips and such. I would stop the correction at 500 Hz.

That however is a different topic than target curve. When you pull down the peaks in bass, invariably the subjective response is "lack of bass." The sound will be "correct" but from preference point of view, we like to have more impact in low frequencies. So we apply an overall curve that provides that tilt down from low to high frequencies. This has been shown in double blind studies to be beneficial and I would not hesitate to use it. You can sit there and keep trying to convince yourself that the sound is correct otherwise but I rather have my ears say that .

And on topic of ears, for sure you want to use that and not be married to any kind of rule. I explain this topic in my article from last year: http://www.madronadigital.com/Library/Room Equalization/Room Equalization.html

Here is a graph from it of my actual theater:

We can see that the massive peaks and dips continue to almost 1 Khz. So the notion that we should stop at 200 Hz doesn't seem right to me. That would be correct for a very large room/open space.

 

[Wayne A. Pflughaupt]

I would argue that if you need to adjust the direct sound of the speaker to enjoy it that you simply picked the wrong one.

You certainly have a point. Personally, I expect that I could hear the anomalies in response in many of the graphs above, and perhaps would pass over those speakers. However...

And again how do you know you're even "correcting" the speaker to your target curve when a mic=\= human hearing?

Well yes, in some ways the mic can indeed “hear” the same way you do. If the measurement mike shows a wide honkin’ peak in the midrange as bad as what’s seen on the green and red MartinLogan graphs, then someone with trained ears should be able to hear it. This isn’t really complicated. The goal is to get an improvement in sound quality, not achieve a ruler-flat curve. That typically means addressing the response issues that are wide and large. If you don’t think things sound better after appropriate equalization, then don’t use it, or try dialing it back a bit.

I can also tell you that when Audyssey or Dirac corrects the dip in the midrange response as seen on the Motion 60xt graph that the character of the speakers sounds considerably worse. The speaker was designed to have a leaner midrange that gives vocals a more upfront sound. When you "correct" that out the speaker does not handle that well. Suddenly vocals feel strained and compressed. The "correction" in FR essentially ruins the character of the speaker that I picked out, because I liked that sound.

I haven’t used either myself, but I expect that dissatisfaction with Audyssey or Dirac comes from using their recommendation of multiple measurement locations all over the place. Wayne Myer (AudiocRaver) over at Home Theater Shack has done extensive experimentation with these auto EQ systems and suggests instead that single location, or multiple tightly-grouped locations, is the way to go.

Another issue that can arise with auto EQ systems is their vertical mic orientation, which can give problematic results in some rooms.

Regards,
Wayne A. Pflughaupt

 

[emcdade]

I do not think it is possible to hear precisely "what the speaker was supposed to sound like" or the sound "that you bought" in your room, which is likely quite different from the anechoic chamber, rooms the designer used or even the room in which you auditioned it.

Well Hales believes we cue into the direct sound of the speaker, and I agree. It's the reason why your voice still sounds like you when you walk from the living room into the dining room. The same applies to a loudspeaker. This of course assumes you have a decent room that isn't wall to wall hard surfaces.

Your results with EQ are quite different than mine. So far MCACC, Audyssey xt32, and Dirac live have all made my speakers sound worse doing a broadband EQ. I have finally settled on running Audyssey in LR Bypass which which only EQ's my sub, and it does an excellent job at that.

As far as target curves go, I agree that a downward slope would be preferable to a flat response. However I agree with Cheryl's thoughts earlier that a more novel approach would be to flatten bass response from 200hz down, and then "smooth" the rest of the speakers frequency response measured in situ.

I may play around with a custom target curve in Dirac that follows the measured response and see if there's any audible benefit to be had over Audyssey LR bypass.

 

[audioguy]

[
I haven’t used either myself, but I expect that dissatisfaction with Audyssey or Dirac comes from using their recommendation of multiple measurement locations all over the place. Wayne Myer (AudiocRaver) over at Home Theater Shack has done extensive experimentation with these auto EQ systems and suggests instead that single location, or multiple tightly-grouped locations, is the way to go.

Dirac gives three options for correction: chair, row and room. I ALWAYS use chair and get much better results [sonically and measurably]. I did the same with Audyssey. I never moved the mic outside the physical dimensions of the MLP chair. Same thing: audibly and measurable superior. And I agree that using a larger area does give worse results.

Another issue that can arise with auto EQ systems is their vertical mic orientation, which can give problematic results in some rooms.

Regards,
Wayne A. Pflughaupt

Depends on the mic. The mic I use for Dirac measurements (Earthworks M30) does have only a very slight difference when measured vertically vs horizontally. They will calibrate the mic in either orientation. The recommend Dirac mic (Earthworks M23] is purposely calibrated in the horizontal direction, so that is the way I had my M30 re-recalibrated.

 

[Rodney Gold]

All you are doing with a target curve is bringing what the room would do back in a controlled manner
The room will have bass gain and treble droop
its all about taste anyway.. whatever hits your ears and makes you happy is good.

 

[Fitzcaraldo215]

Another issue that can arise with auto EQ systems is their vertical mic orientation, which can give problematic results in some rooms.

Of course, I cannot swear that all omni mikes are the same. But, I have looked at the published response plots for some pro omnis in all dimensions. They are somewhat less sensitive gradually in their bottom hemispheres, as you would expect from the blockage by their bases. But, they tend to be fairly close in response between 0 (head on at top dead center) and 90 degrees (from the sides).

For stereo measurements, I do not think it will make much difference if you position them with 0 degrees facing forward toward front dead center of your room, or if you point them up at the ceiling with their sides facing the speaker. For Mch, the pointed up with 90 degree side incidence to all direct sound from speakers is clearly the way to go.

Personally, I would always prefer the pointed up orientation in order to best capture the direct sound of the speakers and most of the relevant reflected sound from the ceiling and walls, including the wall behind you. They all influence the sound you hear. I happen to think that especially ceiling reflections can be important in determining how things will sound in the higher frequencies. Most floors are carpeted, or should be. So, while there still might be some floor bounce of mid/upper bass frequencies (the mike will still pick those up even if at 90 degrees), mid/HF energy should usually not be a problem there.

Thus speaketh the acoustic measurement amateur.

 

[Wayne A. Pflughaupt]

Depends on the mic. The mic I use for Dirac measurements (Earthworks M30) does have only a very slight difference when measured vertically vs horizontally. They will calibrate the mic in either orientation. The recommend Dirac mic (Earthworks M23] is purposely calibrated in the horizontal direction, so that is the way I had my M30 re-recalibrated.

The room can make a difference as well.

Since measurement mics become increasingly directional the higher up the frequency range you go, a 90° calibration file (quite naturally) boosts the high end to compensate for the droop in response you’d get with the mic pointing up. It’s common for off-axis response to start sagging at 2 kHz.

Common Off-Axis High Frequency Response of a Measurement Mic​

But if your room happens to be one with a hard ceiling, such as wood or (God forbid) glass, that doesn’t absorb and diffuse the high frequencies as well as a more common textured drywall ceiling, the highs probably won’t attenuate sufficiently.

For example, let’s say the ceiling in the room in question, for whatever reason, has poor high frequency absorption, and the sound reflecting off the ceiling is attenuated only a couple dB at 8-12 kHz when it arrives - essentially on-axis - at the vertically-oriented measurement mic. The 90º calibration file has boosted on-axis response by 5 dB or more in the 8-12 kHz range. This means the measurement platform (REW, Audyssey, Smaart, etc.) will see the reflected signal as 3 dB hotter than the direct signal. The mic can’t tell a direct signal from reflected; it’s only going to “see” the one with the higher SPL, and thanks to the 90° calibration file, the resulting measurement will show a boost in the upper frequencies that really isn’t there.

I realize the above is a bit simplistic and based on a certain amount of assumption, but hopefully you can see how 90º mic orientation can foul up an auto EQ or even a manual EQ process in some rooms. Fortunately such situations seem to be few and far between, but I have seen it happen.

Bottom line, for those inclined to use 90º orientation as default, I recommend initially taking both 0° and 90° measurements in any new room you measure in, with the appropriate calibration files in place for each. If they both look the same, then all is well. However, if they look significantly different, I’d put more confidence in the 0° measurement.

For the record, Herb Singleton of Cross Spectrum labs only recommends 90° measurements as appropriate for acoustics or calibration measurements, not frequency response. REW author John Mulcahy makes the same recommendation for frequency response measurements.


Regards,
Wayne A. Pflughaupt

 

[Fitzcaraldo215]

Well Hales believes we cue into the direct sound of the speaker, and I agree. It's the reason why your voice still sounds like you when you walk from the living room into the dining room. The same applies to a loudspeaker. This of course assumes you have a decent room that isn't wall to wall hard surfaces.

Your results with EQ are quite different than mine. So far MCACC, Audyssey xt32, and Dirac live have all made my speakers sound worse doing a broadband EQ. I have finally settled on running Audyssey in LR Bypass which which only EQ's my sub, and it does an excellent job at that.

As far as target curves go, I agree that a downward slope would be preferable to a flat response. However I agree with Cheryl's thoughts earlier that a more novel approach would be to flatten bass response from 200hz down, and then "smooth" the rest of the speakers frequency response measured in situ.

I may play around with a custom target curve in Dirac that follows the measured response and see if there's any audible benefit to be had over Audyssey LR bypass.

I am not sure I trust Hales on this. As a speaker guy, he might wish to weigh the contribution of his or other's speakers over the contribution of the room. Both are important. Yes, the Haas/precedence effect says direct sound predominates our perception as far as localization is concerned, but the reflected sound, whose presence is masked by that, also substantially contributes to the sound in tonality (frequency response), and other things.

Yes, the ideal speaker, per Toole, has similar frequency response on or off axis, so ideally reflections above the transition frequency have a similar frequency balance to direct sound. But, even then, do your room and its surfaces always nicely cooperate with that ideal, unless carefully treated? I think my room or most others I know with their mixed surface materials do not cooperate ideally as to reflected energy.

Your voice might indeed sometimes be quite recognizable in the next room, but it will not sound the same as actually being there. A major part of that is due to the fact that much of the reflected sound field in the original room is cut off from entering the second room, in addition to attenuation caused by distance with increasing frequency and other differences. So, it is hard to find the analogy illuminating or useful.

Personally, I just do not get the idea of EQing response carefully above the bass to follow the measured in-room response. It seems like a lot of work aimed at preserving the peak/valley or plateau response of the speakers in the room, assuming that is something worth preserving. In Dirac, you can very easily just wall off any part of the target curve you wish, preserving original un-EQed frequency response there. Of course, you also lose Dirac's time domain corrections in whatever range you eliminate from the calibration.

The Sean Olive paper I cited earlier indicates listener preference for smooth response vs. less smooth. Most preferred was a more or less continuous downward slope toward the highs. Audyssey's curve with flat bass and the BBC dip/midrange compensation, was least preferred. I used Audyssey for a long time, and it sounded much better to me when I was able to eliminate the BBC dip via Audyssey Pro. That is history now, as I find Dirac Live much, much better across the frequency spectrum.

 

[emcdade]

You misunderstood the analogy. I'm saying that we cue into the direct sound of someone's voice regardless of what room we're in. You sound like you, not a room. Same applies to a loudspeaker.

And the entire point of running a Dirac curve that matched measured response would be to gain the time domain corrections. I haven't had much time to listen yet but I doubt there will be an audible difference.

If you're ok modifying the direct sound of your speaker then more power to you. It's what sounds best to you in the end that matters. I will just say Dirac or any other room EQ is no sure bet to improve the sound in your room. In my case, it made it considerably worse.

 

[Fitzcaraldo215]

Well, I have looked at the measured in-room response of a number of speakers in the $10k and up ballpark. I have not yet seen response that is worth preserving or that sounds better uncorrected in the mids and highs, as well as the bass, where the room predominates.