FAQ: Room Correction Systems (Audyssey, etc.)

[Boomzilla]

OK - Allow my ignorance to show (again), please... All "room correction" systems that I currently see require a calibrated microphone, placed in the listening position. The usual "roar" plays over the speakers, and the microphone and electronics (normally in a receiver or AV preamp) correct the frequency response and speaker distance (phase) for the listening position. Why do it this way? The disadvantages include:

1. The correction applied is specific to the microphone position only
2. If the speaker position changes, the calibration is no longer valid
3. The system does nothing to defeat room echo except as it affects frequency response

Instead of using a microphone at the listening position, why couldn't one put the room correction electronics and the microphone inside a self-powered loudspeaker? The speakers would need to be toed in toward the listening position for calibration. Then the speaker, rather than roaring, could emit a single pulse and listen for echoes. Fast Fourier transform could calculate the frequency response (in room) and distance from the listening position (by first arrival & strongest echo). The secondary echoes, being room boundaries (walls, floors, ceilings) could be notched out or inverse-cancelled. This would have several advantages:

1. The correction could be easily repeated at the push of a button without having to reconnect an external microphone (which the user has probably damaged or lost).
2. Electronics are getting cheap and the speaker cost would not be significantly increased.
3. The speakers could be moved to other venues and instantly calibrated without use of external electronics.

This is probably NOT feasible for home use. However, for touring bands, DJs, or other users who must constantly adapt to new (and sonically hostile) environments, a self-calibrating speaker would be GREAT. There are downsides, of course:
- Rather than a single computer for all speakers, each speaker must now have its own computer & calibrated microphone built in
- The calibration must be VERY versatile and able to calibrate anything from small rooms to stadiums
- The speakers would HAVE to be self-powered

Ignoring, for the moment, the actual practicality of such a system, is this idea theoretically possible? Is there any reason why the speakers MUST be calibrated from a "listening position microphone" or could they self-calibrate?

Obviously, the time for the "listening position echo" to return to the speakers will be twice the time that it takes the sound to travel to the listening position (since the sound must leave the speaker AND return from the listening position), but that can be easily calculated.

One major obstacle would be if the strongest echo were NOT from the listening position (for example in a large room where a listening chair was centrally located). In that case, the speaker would have no way to determine WHERE the listening position was unless the user overrode the default and entered a distance.

[jdskycaster]

I think eventually we will see multi-channel powered speaker "systems" with built in DSP and RCS but currently cost and complexity would be very high. Using current RCS software how would the speaker know where you are sitting in the room and for what positions are you trying to achieve the best response? Audyssey will do its best to correct for as large an area as needed, and more mic positions means better response over a larger area, but it is not possible to get perfect response across 8 seats. The area alone might cover anywhere from 30-50% or more of the room. I have tried it multiple ways in my HT and the best response is always taking more measurements across a smaller number of seats (or total area). Basically you cannot have your cake and eat it too. Which could be a reason why many people have less than positive experience with it. Measurements still have to be taken with care. You cannot just place the mic on the back of each seat, take eight or more measurements and call it good.

Also, with the correction built into the speaker (as a system) moving any one of them would still invalidate the measurements and you would have to recalibrate. I understand that not having to position a mic manually each time would be easier but I am not sure that it would or even could generate better results.

[KeithL]

The answer is actually very simple - and sort of obvious once you think about it: There is NO SUCH THING as "electronic room correction".

The room is what it is; and the only way to fix it is to determine (then correct) any problems it has (move walls, add padding, add diffusers); ditto for the speakers themselves.

What systems like Tact, and Dirac, and EmoQ do is to try and "pre-modify" the signal in such a way that their modifications "cancel out" most of the major problems.

The reason that the different systems produce different results, and one or another may produce the best results with a certain set of speakers in a certain room, is that doing this effectively is a very complex process - with lots of compromises and choices to be made. (And, in case you were wondering, doing it anywhere near perfectly is absolutely impossible - even theoretically.) The way they work is to send out a known signal, measure what it sounds like WHEN IT ARRIVES AT YOUR EARS, and then calculate a signal that is "corrected" to produce a "perfect" result at that location. One of the major limitations of such systems is that their ability to produce an accurate correction is inversely related to the amount of area they have to cover (you can have a really accurate correction that works if you sit within one inch of the "X" on the floor, or you can have a much less accurate correction, but one that works for two or three seats). If you were to put the microphone in the speaker, then you would end up with a correction setting that made everything sound right if you were sitting with your head resting on the speaker.

What EmoQ, or Audyssey, or TacT does is to play a known signal with a known expected frequency response; it then measures what that signal really sounds like (at the listening position); calculates a "correction factor"; applies that correction factor; and measures the result. Sophisticated systems like EmoQ may then apply more detailed corrections and try again to get even better accuracy. When the system is done, it "knows" what correction factors were required to make the known test signal sound right at the listening position, and applies those correction factors to everything you play. (Actually, that describes the frequency response settings; another similar process is used to detect and set the speaker distances.)

The point, however, is that the entire process is ONLY valid for the position where the listening takes place - which is why they all have a microphone that must be placed at the listening position. Systems that allow multiple measurements can sometimes produce better results, but they still cannot necessarily produce a correction that works for all the positions where measurements are taken. Instead, what they can do is use the extra information to produce a better compromise (hopefully) which allows a reasonably good result for a wider target area.

You could indeed get SOME useful information from a microphone located at or near the speaker, but not nearly as much as from one located at the listening position.

Jul 18, 2013 7:52:08 GMT -5 Boomzilla said:

OK - Allow my ignorance to show (again), please... All "room correction" systems that I currently see require a calibrated microphone, placed in the listening position. The usual "roar" plays over the speakers, and the microphone and electronics (normally in a receiver or AV preamp) correct the frequency response and speaker distance (phase) for the listening position. Why do it this way? The disadvantages include:

1. The correction applied is specific to the microphone position only
2. If the speaker position changes, the calibration is no longer valid
3. The system does nothing to defeat room echo except as it affects frequency response

Instead of using a microphone at the listening position, why couldn't one put the room correction electronics and the microphone inside a self-powered loudspeaker? The speakers would need to be toed in toward the listening position for calibration. Then the speaker, rather than roaring, could emit a single pulse and listen for echoes. Fast Fourier transform could calculate the frequency response (in room) and distance from the listening position (by first arrival & strongest echo). The secondary echoes, being room boundaries (walls, floors, ceilings) could be notched out or inverse-cancelled. This would have several advantages:

1. The correction could be easily repeated at the push of a button without having to reconnect an external microphone (which the user has probably damaged or lost).
2. Electronics are getting cheap and the speaker cost would not be significantly increased.
3. The speakers could be moved to other venues and instantly calibrated without use of external electronics.

This is probably NOT feasible for home use. However, for touring bands, DJs, or other users who must constantly adapt to new (and sonically hostile) environments, a self-calibrating speaker would be GREAT. There are downsides, of course:
- Rather than a single computer for all speakers, each speaker must now have its own computer & calibrated microphone built in
- The calibration must be VERY versatile and able to calibrate anything from small rooms to stadiums
- The speakers would HAVE to be self-powered

Ignoring, for the moment, the actual practicality of such a system, is this idea theoretically possible? Is there any reason why the speakers MUST be calibrated from a "listening position microphone" or could they self-calibrate?

Obviously, the time for the "listening position echo" to return to the speakers will be twice the time that it takes the sound to travel to the listening position (since the sound must leave the speaker AND return from the listening position), but that can be easily calculated.

One major obstacle would be if the strongest echo were NOT from the listening position (for example in a large room where a listening chair was centrally located). In that case, the speaker would have no way to determine WHERE the listening position was unless the user overrode the default and entered a distance.

[Boomzilla]

Hi Keith -

You're completely correct in what you've said. My question is whether the speakers could either be "told" where the listening position was prior to calibration or whether the listening position might be calculated from the ranging (parallax) of the two signals? In other words, is it absolutely necessary to place a single microphone at the listening position? If one speaker could listen to the other & vice versa, the room could be modeled from the differences. A "calculated listening area" could be determined, and the speakers optimized for that area. It wouldn't be as precise as having a microphone at X-marks-the-spot, but it might allow for a bigger "sweet spot" although possibly with slightly less focus.

Note that I'm not interested in whether this is the BEST way to accomplish room correction, merely asking if it is theoretically feasible. As computing horsepower increases per Moore's law, the hardware to do something like this becomes possible (and probably possible even with today's technology).

[paintedklown]

Speaking of room correction systems.

Are there any on the market that allow you to select that bands that are EQ'd for each channel?

For example: Say my left speaker is too close to the wall and I need to cut the bass a bit right at 90hz, so I need a P-EQ that allows me to select 90hz instead of having to choose from a pre-set list of frequencies that may (or may not) have 90hz available to tweak.

I suppose I am looking for user selectable EQ bands vs. fixed preset bands. I hope this question makes sense.

[Topend]

^ Sounds like you want a PEQ to me?

Dave.

[paintedklown]

Jul 24, 2013 12:11:03 GMT -5 Topend said:

^ Sounds like you want a PEQ to me?

Dave.

Yes Dave, that is exactly what I am getting at. However, with my Yamaha, the EQ bands to select from are fixed bands. I was wondering if there was a room correction system that allowed you to punch in ANY frequency for adjustment.

[flamingeye]

Jul 18, 2013 16:06:26 GMT -5 Boomzilla said:

Hi Keith -

You're completely correct in what you've said. My question is whether the speakers could either be "told" where the listening position was prior to calibration or whether the listening position might be calculated from the ranging (parallax) of the two signals? In other words, is it absolutely necessary to place a single microphone at the listening position? If one speaker could listen to the other & vice versa, the room could be modeled from the differences. A "calculated listening area" could be determined, and the speakers optimized for that area. It wouldn't be as precise as having a microphone at X-marks-the-spot, but it might allow for a bigger "sweet spot" although possibly with slightly less focus.

Note that I'm not interested in whether this is the BEST way to accomplish room correction, merely asking if it is theoretically feasible. As computing horsepower increases per Moore's law, the hardware to do something like this becomes possible (and probably possible even with today's technology).

I 'm probably wrong on this but it sounds like what you want is Dirac unison

[KeithL]

Yes, actually, you've got it sort of backwards.

Dirac Live treats each speaker as a separate entity - so measures each, and applies corrections to each, independently.
The IDEA of Dirac Unison is that, when configured to do so, speakers can be used to "augment or correct" each other.
So, for example, if your left front speaker doesn't make enough bass, you could ask the left surround speaker to "help it out".
This could be in the form of simply adding more output, corrected for phase and distance, or of a correction signal, intended to cancel out excess output.
(If there was bump in the frequency response of the left front speaker, as measured at the listening position, perhaps an out of phase signal, coming from the left surround, could be used to partially cancel it.)

What you're talking about is not practically possible at this point in time.
The problem is that you're dealing with a whole bunch of "simultaneous assumptions" - which make the math absurdly complicated.
For example, picture the geometry of a more or less sane and normal listening room in your head........
If you play a sound from any speaker, with a microphone at the listening position, the sound that arrives at the microphone FIRST will be the sound that "went straight from the speaker to the microphone".
This way you can make calculations based on the first sound being "direct" and later sound being "reflected".
However, this assumption is really based on having speakers arranged around the edges of a room, and the microphone in the middle.
(Because that geometric arrangement ensures that, for whichever speaker is playing, the shortest path to the microphone is the straight line - which includes no surface reflections.)

You most certainly COULD figure out the locations of each speaker using a microphone in each speaker.....
But you would STILL need readings from the listening position to do a reasonably good job of room correction.
So, why bother to add microphones, and extra electronics, in each speaker, IN ADDITION to the microphone you need for room correction?
Remember that the whole point of room correction is to correct what you hear at the listening position..... so taking a bunch of measurements IN OTHER PLACES really won't do the job by itself.
At most, after figuring out the acoustics of the listening position, those extra readings will give you somewhat useful additional information.
(At least one company did offer a set of speakers where each speaker had some sort of built-in "ultrasonic range finder" that figured out all the locations by triangulation......
I don't recall the manufacturer... and the fact that you've never heard of it either sort of tells us how successful the product was. )

Notice that Dirac Live extends this to taking several measurements spread over "the normal listening AREA".
While taking additional measurements at the speaker locations would provide even more data, which might be somewhat useful, it would not eliminate the need for the current set of measurements.

Bear in mind that you cannot "correct every place in the room" - even with an infinite amount of information and an infinitely powerful computer...
(Because you run into an infinite regress - where each correction creates new errors which must themselves be corrected.)
You are also currently in a situation where you are balancing approximations of several unknowns and very few knowns.
- you DON'T know the radiating patter of each of your speakers
- you DON'T know the reflectivity, absorption, or diffusion, of the surfaces in your room

Imagine trying to determine what the light in a room would look like when you shine a flashlight into it....
- without knowing what the beam of the flashlight looks like
- and without knowing how shiny the walls are or what texture or color they are
(Your best hope is going to be to replace your eyeball with a light meter - located at the particular spot whose lighting you want to figure out.)

Feb 12, 2018 13:43:00 GMT -5 flamingeye said:

Jul 18, 2013 16:06:26 GMT -5 Boomzilla said:

Hi Keith -

You're completely correct in what you've said. My question is whether the speakers could either be "told" where the listening position was prior to calibration or whether the listening position might be calculated from the ranging (parallax) of the two signals? In other words, is it absolutely necessary to place a single microphone at the listening position? If one speaker could listen to the other & vice versa, the room could be modeled from the differences. A "calculated listening area" could be determined, and the speakers optimized for that area. It wouldn't be as precise as having a microphone at X-marks-the-spot, but it might allow for a bigger "sweet spot" although possibly with slightly less focus.

Note that I'm not interested in whether this is the BEST way to accomplish room correction, merely asking if it is theoretically feasible. As computing horsepower increases per Moore's law, the hardware to do something like this becomes possible (and probably possible even with today's technology).

I 'm probably wrong on this but it sounds like what you want is Dirac unison