Why I caution against the use of REW (or similar) to calculate correction filters above transition

Matthew J Poes

AV Addict
Thread Starter
Joined
Oct 18, 2017
Messages
1,903
I figured this would be better as a post rather than an article, as it allows for even-keeled discussion.

Early in my time learning to use room measurement software and take in-room measurements, I was attracted to the idea of using EQ to compensate for a non-flat response. In time I read the work of folks like Floyd Toole and Earl Geddes and took on what I now think is a common misconception, that my goal should be a flat in-room response. The misconception being that this can be achieved through EQ after the fact, rather than focusing on the response of the speaker itself in free-space conditions. Subjectively I found that improvements to the bass were readily obvious, but the overall sound often worsened. In fact, before learning about room curves I often assumed a flat listening position response was ideal, but always thought it sounded bright and bass shy.

The more I studied this topic and reread the articles from Toole (and the researchers he cited) I began to realize I had missed really important points. Like the measurements he indicated as being critical are anechoic measurements. That he specifically cited problems with in-room measurements. For example, Toole says "Steady-state in-room measurements may be indicative of certain problems that are audible, but they are of little use in assigning corrective measures" in Loudspeakers and Rooms-Scientific review. In an earlier article, Toole cited Brociner and Von Recklinghausen (I'm paraphrasing) who note that a non-smooth power response in a reverberant room will show a similarly irregular response in an anechoic response, that the same can not be said of the reverse. A smooth response in the reverberant room (a normal listening room) is not a guarantee of good sound or a smooth pressure response in the anechoic chamber. In other words, a speaker that measures bad in your room likely sounds bad and measures bad in an anechoic chamber. However, a speaker that appears to measure well in-room doesn't necessarily measure well in an anechoic chamber and may not sound good.

Of course, we can delve farther in the science of this, but the point is, in-room measurements, while able to potentially highlight problems, can't really tell you if things are good. Further, that using them as a means to provide corrective measures is problematic. Those measurements contain comb filtering (which shouldn't be corrected), diffraction from various objects or barriers near the speaker (which should not be corrected), resonances (which maybe can be corrected), and speaker problems (which maybe can be corrected). It also contains problems which can be corrected but not by EQ. Problems with the speaker's polar response for example.

Further, REW generates minimum phase filters. These are perfect for applying correction in the bass range, and if done carefully, even an acceptable solution in the transition zone. However, most rooms and speakers are not minimum phase above the transition zone and as such the correction is not an accurate inversion of errors. These inaccurate filters mixed with the inaccurate measurements can lead to a system that is worse off than if no correction was applied. Yet the response will measure, in room, totally flat. For this reason, I've concluded that manually applying PEQ (minimum phase filters) across a full bandwidth of a speaker based on in-room measurements is not a good idea.

Now, that isn't to say that I dislike full-bandwidth room correction, I just think it needs to be a special kind. It needs to use multiple measurements taken across a range of spaces so that it can figure out important things. Distinguish between diffraction and actual response errors. Correct things that don't move with the position. It needs to generate its filters using a method that matches the mixed phase of the room response in the first place. This is the primary reason why I like DIRAC so much. Of all the systems I've ever used, its the only one that seems to get all of this right.

A few References:
http://www.wghwoodworking.com/audio/loudspeakers_and_rooms_for_sound_reproduction.pdf
http://mariobon.com/Articoli_storici_AES/Toole/AES_1986_Toole_01.pdf
https://static1.squarespace.com/static/55c8a274e4b09cb562cd3ea0/t/57d159a69de4bbd9d306ac8f/1473337780806/Dirac+Room+Correction+(Audio+Engineering+Society,+AES+Sweden+lecture).pdf
 
Top Bottom