Room correction, speaker correction anything above Schroeder a mistake?

[jhaider]

Besides, what’s meant by «crude frequency responses»? It’s recommended that speakers have a deviation of less than 0.5 dB from a flat curve in the mid to higher frequencies to be used in professional applications (broadcasting, film, I think it is).

"Crude" in the passage refers to the curves that trained listeners were asked to draw. I'm going to guess that they used a sheet of paper or a screen with discrete bands, i.e. 1/3-octave, 1/2-octave, etc.

What Toole extracted from this study is that below the transition frequency humans perceive the room response, and in the statistical field we perceive the speaker's actual frequency response rather than the speaker+room response. The implication is that systems which modify the in-room statistical field response to fit some preset curve based on listening position measurements are probably mucking with the speakers' actual frequency response. Thus, they are more likely to do audible harm than good.

The Olive study has not, to my knowledge, been replicated.

I don't really agree with your hypothesis.

For one thing, I don't think your group 1 exists, outside of people who swap a cable and hear more liquidity or whatever.

For another what's a "flat speaker?" Is it a speaker with flat axial response? Flat power response? Flat DI, and if so how high?

Also, the task of "fool[ing] the senses into believing a story where audio is one of the ingredients" (emph. added) is as a much easier task audio-wise than fooling the senses into believing a sonic event is happening when audio is the sole ingredient. Similar to what Amir pointed out comparing mono vs. multi-channel, but several orders of magnitude more.

 

[j_j]

This is a bit more complicated than I want to go into at 23:51 on a weekend night.

It's also true that both measurement AND correction need to look both to human perception AND to the room response AND to the speaker respose. Amir, if you recall the stuff we did for Vista that was headed down the right road. (as much as one could with the obligatory "cheap microphone")

More I can't talk about at the minute.

 

[Cosmik]

For another what's a "flat speaker?" Is it a speaker with flat axial response? Flat power response? Flat DI, and if so how high?

If one believed the idea that the human listener deconvolves the room from what they are hearing, a flat speaker would be a point source that simply duplicated the signal in terms of SPL, whether on or off axis - what the room acoustics then did to this and how it measured in-room in terms of simplistic magnitude-only frequency response would be immaterial.

Some people e.g. John Watkinson mentioned above, or this speaker company, believe that the neutral speaker should be omnidirectional, but I think it is possible that a directional speaker would also be 'valid' as long as it had uniform dispersion at all frequencies - you could still 'deconvolve' back to it, but with correspondingly quieter 'ambience' 'stream'.

If the speaker isn't neutral in this way, you are hearing varying quantities of various frequencies being pumped into the room so your hearing would register this as the wrong tonal balance for the source and a discrepancy between the direct sound and the ambient - making the result more ambiguous and less 'real'.

 

[Cosmik]

I see two 'camps' who come at this problem from opposite ends:
(a) Most people start from where we are now: existing speakers; laptops and microphones; frequency response measurements; an assumption that human hearing is primarily a frequency response analyser; consequently that changing the EQ of the source can perfectly pre-compensate for the effects of the room; a conviction that phase and timing are much less important than tonal balance; an assumption that the room is a problem; 'room correction' algorithms. This view tries to justify and tie all these things together - with difficulty it seems to me.
(b) People like me who start from the idea that we can't second guess exactly how human hearing works, and that it is possible that a system may only sound 'real' when the subtleties of phase, timing and frequency response all gel appropriately. We are profoundly grateful that stereo seems to work. The room is an asset, not a problem. We observe that we don't seem to notice some acoustic phenomena that simple measurements would suggest should be colouring the sound strongly; we deduce that human hearing is not just a frequency response analyser. We take on board other people's clever ideas that might lead to more neutral transducers than 'traditional' ones. We don't continually modify the system in response to listening to it, or even 'science', but take neutrality of the source and transducers (as best we can) to the bitter end. Thankfully, we then observe that it sounds different, better than 'traditional' systems...

 

[oivavoi]

My hunch is that 2 channel systems would benefit even more from wideband correction than multichannel systems since they are more dependent on the distribution of reflections for the simulation of the stereo image. No idea whether it's indeed the case, just a hunch.

Thanks! Makes good sense

 

[oivavoi]

The research shows the opposite. In that the more the channels, the less picky we become in how the individual speakers sound.

Here is research that Dr. Olive et. al. performed in effectiveness of room equalization as the number of channels is increased: https://secure.aes.org/forum/pubs/conventions/?elib=14622

Comparison of Loudspeaker-Room Equalization Preferences for Multichannel, Stereo, and Mono Reproductions: Are Listeners More Discriminating in Mono?

View attachment 9859


X axis is different Equalization methods. I have highlighted the surround ratings in yellow. We see there that there is very little difference there between No Eq and the other methods. Whereas in mono (blue), No Eq rated far lower than other methods.

I think the surround presentation is so captivating and effective that we forego other problems with fidelity.

Thanks. very interesting.

 

[oivavoi]

If one believed the idea that the human listener deconvolves the room from what they are hearing, a flat speaker would be a point source that simply duplicated the signal in terms of SPL, whether on or off axis - what the room acoustics then did to this and how it measured in-room in terms of simplistic magnitude-only frequency response would be immaterial.

Some people e.g. John Watkinson mentioned above, or this speaker company, believe that the neutral speaker should be omnidirectional, but I think it is possible that a directional speaker would also be 'valid' as long as it had uniform dispersion at all frequencies - you could still 'deconvolve' back to it, but with correspondingly quieter 'ambience' 'stream'.

If the speaker isn't neutral in this way, you are hearing varying quantities of various frequencies being pumped into the room so your hearing would register this as the wrong tonal balance for the source and a discrepancy between the direct sound and the ambient - making the result more ambiguous and less 'real'.

It should be mentioned that Don Morrison actually recommends room correction, even though he doesnt say so on his website, performed with the pro driverack box that also functions as an active crossover. As I understood it, this is a very "lightweight" room correction, that only works in the frequency domain through IIR filters, and gently evens out any very large response deviations. I would assume most of the work it does is below the schroeder frequency. I briefly asked him what he thought of other products such as Dirac Live and Acourate, and he seemed skeptical. I would love to discuss this more with Don though, as his manufacturing partner Mark Kravchenko (who makes his drivers and plays ball with him on speaker design) is very skeptical of the idea of room correction above the schroeder frequency.

 

[svart-hvitt]

"Crude" in the passage refers to the curves that trained listeners were asked to draw. I'm going to guess that they used a sheet of paper or a screen with discrete bands, i.e. 1/3-octave, 1/2-octave, etc.

What Toole extracted from this study is that below the transition frequency humans perceive the room response, and in the statistical field we perceive the speaker's actual frequency response rather than the speaker+room response. The implication is that systems which modify the in-room statistical field response to fit some preset curve based on listening position measurements are probably mucking with the speakers' actual frequency response. Thus, they are more likely to do audible harm than good.

The Olive study has not, to my knowledge, been replicated.

I don't really agree with your hypothesis.

For one thing, I don't think your group 1 exists, outside of people who swap a cable and hear more liquidity or whatever.

For another what's a "flat speaker?" Is it a speaker with flat axial response? Flat power response? Flat DI, and if so how high?

Also, the task of "fool[ing] the senses into believing a story where audio is one of the ingredients" (emph. added) is as a much easier task audio-wise than fooling the senses into believing a sonic event is happening when audio is the sole ingredient. Similar to what Amir pointed out comparing mono vs. multi-channel, but several orders of magnitude more.

Jay, please allow me a response of a somewhat polemic nature.

I’d say many if not most hifi people are still in group 1 (i.e. no correction). This is due to lack of exposure to newer stuff, habit (old people!) and belief. A friend of mine, a very knowledgable guy, has an analog purist way of doing things. This is an informed choice that he’s taken. So he’s undoubtedly in group 1.

I wrote «flat speaker» with «» signs to indicate that this is not a formal term. I wanted to describe a speaker that measures flat with only small errors frequency response wise. Amir reminded me that a smooth target curve, which is probably not a 180 degrees flat one, is what I meant to communicate.

 

[Cosmik]

Yes. I extend what Cosmik declares,

to saying that the stereo effect is completely invariant, as you move from the ideal position - there is no "progressively diminishes", that registers as a subjective experience. This can only occur with a setup in extremely capable tune, which in turn is extremely rare - and perhaps this perception can only occur for a certain percentage of the populace.

Do you get the same disappearing speaker phenomenon with a single speaker, or does it have to be two? Does it work with multi-channel in just the same way?

Would it work with a mono signal pumped to two or more speakers scattered around the room? How about a multi-channel recording with the various speakers placed 'out of order' randomly in three-dimensional space?

The implication of the speakers always disappearing* and there being no diminishing of the stereo effect as you move away from the central position between the speakers is, to me, that the stereo or multi-channel aspect is being lost and that the system is just creating an indistinct wash of sound..?

*I do believe you that stereo can give a plausible 'scene' that is very resilient as you walk around the room and exists separate from the physical speakers, but that it doesn't behave precisely as a hologram of a 3D scene would as you move - possibly dependent on whether the recording is a 'purist' one or multi-mic'ed, etc. Is it an intimate recording, or from a hundred feet away in a cathedral? What exactly should happen as you move two feet to your left?!

 

[j_j]

Your ears do not deconvolve a room, rather first arrivals are emphasized via basic cochlear dynamics. It's a very nonlinear effect, nothing like deconvolution.

Different people like different direct/reverberant ratios. The pattern of your loudspeaker along with the room determine this in any one given room. Hence, when you pick speakers in your own room, you can meet your preference, or not.

Some people prefer no direct sound at all. That sounds muddy to me, but it does make for a wider sweet spot.

 

[Cosmik]

Your ears do not deconvolve a room, rather first arrivals are emphasized via basic cochlear dynamics. It's a very nonlinear effect, nothing like deconvolution.

Well, the human brain doesn't literally compute the trajectory of a cricket ball as it runs to catch it, but the end result in that application is equivalent. Are you saying that the end result in the human's brain is nothing like the separation of the direct sound from the ambience, or simply that the biological mechanism that achieves it is nothing like the mathematical process of deconvolution?

 

[svart-hvitt]

Well, the human brain doesn't literally compute the trajectory of a cricket ball as it runs to catch it, but the end result in that application is equivalent. Are you saying that the end result in the human's brain is nothing like the separation of the direct sound from the ambience, or simply that the biological mechanism that achieves it is nothing like the mathematical process of deconvolution?

FWIW,

Genelec has a very pragmatic attityde to this: They use colours plus a clear limit to indicate when direct sound is overwhelmed by reverberant sound for every speaker model. So they give users the choice to find a seating based on their direct-reverberation preferences. Their recommendation, however, is to sit closer to the direct sound.

 

[Fitzcaraldo215]

Well, the human brain doesn't literally compute the trajectory of a cricket ball as it runs to catch it, but the end result in that application is equivalent. Are you saying that the end result in the human's brain is nothing like the separation of the direct sound from the ambience, or simply that the biological mechanism that achieves it is nothing like the mathematical process of deconvolution?

I think you are still lumping all room reflections together as though they were all the same. It has been shown scientifically that early reflections are not directly distinguishable from direct sound unlike later arriving ones. The perception of early reflections is masked by direct sound, hence there is nothing for the ear-brain to convolve. The ear-brain does not have sufficient perceived information necessary to do so. However, those hidden early reflections still can modify the perceived direct+early reflected sound in tonality, timing, direction, etc. In other words, I agree with jj.

It is a different matter with later reflections, perceptably sensed as reflected energy and separable from direct sound.

I think most listening rooms have more early reflections than later ones due to their size. Concert halls have more late reflections than early ones. This may partly explain the riddle of why a Steinway tends to sound like a Steinway even in different halls.

 

[Fitzcaraldo215]

Is it possible that multichannel systems benefit even more from wideband correction than 2 channel systems? No idea whether it's indeed the case, just a hunch. Feels like all those channels can easily create a mess in a room, and that a central DSP/EQ command center could do wonders for keeping it tidy? That said, I have never listened to a dedicated multichannel system for music, so I don't know.

Empirical data indicates that is not true, as Amir's post shows.

But, I think it may be true in some Mch systems that calibrating all main, center and surround channels to an identical target curve can be beneficial in terms of overall sonic integration. If speaker frequency response differs somewhat because all speakers are not identical, or even if identical speakers all around exhibit different response due to room placement and early reflections, the result might well be better if all are EQed to the same target curve.

 

[amirm]

But, I think it may be true in some Mch systems that calibrating all main, center and surround channels to an identical target curve can be beneficial in terms of overall sonic integration.

The Harman ARCOS system is the only one I know that will try to do that with multiple subs. See https://audiosciencereview.com/forum/index.php?threads/computer-optimization-of-room-acoustics.12/

 

[fas42]

Do you get the same disappearing speaker phenomenon with a single speaker, or does it have to be two? Does it work with multi-channel in just the same way?

Have never specifically tried single, or multi. But my very strong feeling would be that it should - I've effectively done the single speaker trial by listening very closely to a single speaker, on one side, so the output of that would almost completely exclude the sound of the other. If the system is not working sufficiently well, then the drivers are very obvious; if good enough, then the sound comes from "behind the speaker", the cabinet just happens to be something in the way, in the direct line of sight to the musical event.

Would it work with a mono signal pumped to two or more speakers scattered around the room? How about a multi-channel recording with the various speakers placed 'out of order' randomly in three-dimensional space?

Can't say. Would all be good experiments to try out, once a system is working well enough to achieve this quality.

The implication of the speakers always disappearing* and there being no diminishing of the stereo effect as you move away from the central position between the speakers is, to me, that the stereo or multi-channel aspect is being lost and that the system is just creating an indistinct wash of sound..?

*I do believe you that stereo can give a plausible 'scene' that is very resilient as you walk around the room and exists separate from the physical speakers, but that it doesn't behave precisely as a hologram of a 3D scene would as you move - possibly dependent on whether the recording is a 'purist' one or multi-mic'ed, etc. Is it an intimate recording, or from a hundred feet away in a cathedral? What exactly should happen as you move two feet to your left?!

Most definitely not an "indistinct wash of sound"! Let's say there are three distinct instruments in the picture: a drum kit on the left, vocalist in the centre, and lead guitar on the right - these all exist extremely solidly as tangible, pointable to, sources of sound; you can "see" the sound maker as a definite object, the same as in real life, say listening in the normal position. Then, moving around as much as you like does not disrupt this illusion one iota; the ear/brain appears to have locked on to the positioning cues; it's as if the musicians exist in the real space, behind the speakers, with the appropriate lateral and depth positioning that matches what one hears in the ideal listening position - the 'spooky' variation of this is with true mono material, over stereo speakers; the musicians "follow you", they always remain in the same lateral (or is it depth??) plane as yourself.

The space you hear is completely dependent upon the recording: it can be cavernous, or intimate; or a combination of the two, if different elements are recorded with distinct spatial characteristics. It can be "a 3 ring, or N ring circus" in terms of how it comes across.

 

[fas42]

What exactly should happen as you move two feet to your left?!

What "should" I can't say; what does happen matches the concept that people use of the system being a 'window' to the event; everything exists as sound makers past the plane of the speakers, in a particular place on a stage as big as it needs to be, to "hold everything" - and you then moving around and noting how the sound changes mirrors what would happen if that stage actually existed.

 

[Bjorn]

@svart-hvitt:
It seems like you have a learned a thing or two since this thread was started
You argumentation at the norwegian forum says so.

Ivo, who's behind the ARTA software, says something related to this that is worthwhile paying attention to:

In my work I have found that over-equalized high order systems have lot of troubles to retain natural sound. I always prefer low order IIR equalization which introduces small number of resonances (in peaks and dips).
Why?
Let me note that ripples in loudspeaker response comes from two sources: resonances and sound wave reflections. It is not good thing to equalize ripple (notches) that are reflections based with filters which use pole-zero-resonance compensation. In that cases filters just becomes sources of new resonances.

 

[Snowpuppy]

You can only correct for the response at a single point. And unless all points in coverage are similar, then correcting this single point will not correct the total sound field.
The speaker's directivity alone, which varies as we know, is reason enough for room correction not to work. Not to mention that most of the response from the room isn't minimum phase but varies in both time and strength.

No acoustic problem can be fixed with EQ/DSP unless it's minimum phase. You can't deal with a three dimensional problem using a one dimensional solution.

With Audyssey (in its latest incarnations, such as XT32 at least) you can take measurements in more than just one location, providing a more overall level of room correction that should, at least in theory, be more pleasing to listen to when it’s more than just you in the room listening. If you live alone, or you’re the only person who uses the room to listen to whatever it is, be it music or movies, calibrating from the location where your ears will normally be will end up yielding better results though. I have a higher end Denon with this room correction but I haven’t bothered using it yet since I’m currently only using it in my bedroom occasionally (read very rarely), and since it sounds just fine as it is, I’ve left it alone for now. I don’t think there would be much to gain from room correction in there anyway TBH.

I’ve used Pioneer’s MCACC to great effect however, and I can attest to what a remarkable job it does in taming room response. It does correct for phase, time, distance and it obviously changes the EQ level for each speaker as well. I will say this though, it’s only as good as the room you put it in, and if your room sucks, it can only do so much. Case in point, I moved from one condo to another, both with identical layouts. The first room it was in was roughly 18x15 that had an adjoining dining room, a long hallway, and another small hallway/entry area, so it wasn’t a dedicated theater room by any stretch of the imagination, just a living room. The second room and layout were identical. The first had hardwood floors throughout the entire area, leather couches and chairs, and wooden blinds. The second is carpeted, with a large fabric upholstered sectional couch, and curtains (with blinds as well). In the first condo, room correction through MCACC was able to tame the “liveness” of the room for the most part, but it was less than perfect. In the second room, the results were stunning. It was able to accurately detect the distance of the speakers from the microphone location within a quarter of an inch, and it sounded like a whole new system (for the better), although not a single component had changed, and everything was placed exactly as it was in the other room. Pioneer claims to have developed their MCACC Advanced & Pro in collaboration with AIR Studios in London, and that it’s a 3D calibration. It can tell the size of the room, the ceiling height, the distance of all of the speakers, and adjust for room resonance (again, to a degree, the first room was horrible, but the 2nd was much better and easier for it to “fix”), room modes, delay, phase and more. In the first room, while my Velodyne HGS-18 was “fun”, it never sounded great. In the second room it’s like night and day, especially after running MCACC again. I’ve recently acquired a Velodyne SMS-1 which I plan on using to fine tune the sub even more, but after MCACC alone it’s pretty darn awesome.

I guess my point is twofold, the first being that it’s no longer an attempt to correct a 3 dimensional problem with a one dimensional solution, and that you can correct for response at more than one point. Perhaps when you wrote this that wasn’t possible, you were unaware of it being available, or maybe you’re talking about something different entirely. The second point is that the technology is out there, it’s greatly improved since it first became available, and it does an amazing job at fixing room issues. It can turn a bad room into a tolerable room, a tolerable room into a decent room, a decent room into a good room and a good room into a great room, but it can’t turn a bad room into a great room. It can make a very noticeable improvement, and it’s incredibly easy to use, but like everything it has its limitations. Some rooms need more help than a room correction program can offer, especially very “live” rooms. If your room is good (doesn’t sound like an echo chamber or have too many reflective surfaces), these RC programs can make them sound great!

As far as the original question of messing with anything above the Schroeder line, programs like MCACC and Audyssey are going to do that when they EQ the speakers. I suppose you could save the settings for all the other calibrations such as phase, delay, etc and return all of the EQ settings above 300hz back to flat and see if that sounds better to you, but I have a feeling it won’t. I’m not knowledgeable enough to know exactly how it all works, but if it corrects for resonance in higher registers, and you disable that correction, I can’t imagine it sounding better. Actual room treatments are honestly probably much better than trying to fix a sound issue through EQing, and a room with proper treatments will sound even better after using room correction, but for most people who aren’t highly critical listeners/audiophiles, room correction software is going to be all that they need to make them happy. It doesn’t require panels, bass traps, risers, or isolation, and it’s already built in to the receiver, so why not use it? From my experience it does a fantastic job, and it sounds much better after using it, so why not?

I really don’t know why I would want to make my speakers respond correctly for anechoic conditions. My room isn’t an anechoic chamber, and quite honestly I wouldn’t want it to be. A completely dead flat room is at the other extreme from a super “live” room, and both are not very pleasant in my opinion. In any event, whether or not I wished it were anechoic is irrelevant, it isn’t, and it never will be. I really don’t see how correcting a speaker for a completely different room could be of any advantage to me.

I’ll be honest, I don’t care about the Schroeder line at all. I care about what sounds good, and whatever gets me the best sound is what I’ll use. I understand that you guys like talking science around here, and I do find it pretty interesting, I love science! When I’m watching a movie or listening to music though, I don’t care much about the science, I care about what it sounds like to me. If something makes it sound better, that’s what I want. Room correction software like MCACC & Audyssey work very well, all things considered, and when you combine them with room treatments, they make a huge difference, and can really make a room sound great! I’m all for scientific discoveries though, so hopefully one or more of you guys can figure out a room correction system that’s even better.

 

[Tangband]

One has to remember that the microphone and ear/brain is functioning VERY different because of the HAAS and precedence-effect.

This happens in distances more than 5 ms soundwise from the mic/ear/brain to the loudspeakers.
Thats about 1,7 metres or more.

If you do measurements in a room with longer distances than about 1,7 metres from the loudspeaker, the result from the measurement do NOT correlate good with what the ear/brain experience.

After the distance of 5 ms, meaning 1,7 metre or more, the microphone takes up all sound, while the ear/brain starts to select the sound, because of the law of the first wavefront , and the precedence effect.

Room correction thats all beneficial has nothing to do with the loudspeaker placement, its a correction of the fundamental room-resonances , depending fully of the geometry of the room, and has nothing, as I said, to do with placement of the loudspeaker. Meaning resonances between floor/roof, wall/wall, wall/wall in a rectangular room.
All of those fundamental room-resonances are lower in frequency than 80 Hz.

There are only three fundamental room-resonances in a rectangular room. This can be corrected and the results are gonna be better everywhere in the room - and theres almost no drawbacks.

Correction of reflexes ( different than resonances ! ) higher than 80 Hz can sound better, but just in one listening spot.

If you want to correct the frequency response of a loudspeaker higher than 500 Hz, you can do this:

1. Put only one loudspeaker in the middle of the room. Turn of the other loudspeaker.

2. Measure with a good measurement microphone and a program like REW on axis about 50 cm up till 1 metre distance from the loudspeaker . Use a pulse window of 5 ms. Measure at 0 degrees, 15 degrees, 30 degrees. Use the ”average” function in the loudspeakerprogram.

3. Make the corrections needed in a dsp like minidsp.

4. Repeat with the other loudspeaker.

Stay away from programs like Audyssey and similar , or turn them of.