Originally Posted by AustinJerry
Measuring one speaker, then the other, and averaging the results does not produce an accurate depiction of the actual response when both speakers are playing together. There is interaction between the two speakers that is not captured by the way you took the measurements. The individual speaker measurements should be useful, though, just not the average.
There's several interesting and subtle issues that come into play when trying to predict the combined response of several speakers and/or subs from their individual responses. For more information, there's a good thread on HTS titled "Can REW predict 2sub performance from single sub measurements?
". John Mulcahy answered that question in post #3 of the thread
. It is possible, but a timing reference is required for the technique to work, and it's not possible to have a timing reference with a USB mic. When a timing reference is available, the calculation of the sum will take into account the phase relationships between the multiple speakers and/or subs because it uses a summation of complex numbers.
The reason the summation doesn't give the correct result without a timing reference is software-induced time shift. In the absence of a hardware timing reference, REW and other measurement programs use various techniques to set a timing reference in software. The usual method is to shift the impulse response in time such that its peak occurs at t=0. This is done separately for each measurement when this mode is selected. When the frequency response is computed from the FFT of this time-shifted impulse response, the software-induced time shift will affect the phase of the frequency response. The trouble is that in general, the software-induced time shift using this method will be different for each measurement, obscuring the correct phase relationships between the various speakers and/or subs. This makes the complex summation of the individual measurements different from the measurement of all taken together. These should be the same because superposition applies to linear systems.
The measurement program HolmImpulse has a feature called "time locking" which is capable of providing a consistent timing reference for all measurements, even with a USB mic. It works as follows. The timing reference for the first measurement is established by measuring in the mode that shifts the impulse response peak to t=0. Then "time locking" is enabled, which tells the software to use the same software time shift for all subsequent measurements. There's a thread about this feature
in the Parts Express forums.
After I found out about this feature of Holm, I made a request for the feature in REW
, as it would be extremely useful for a multiple-subwoofer optimization program I'm working on. A couple of people seemed interested, but I don't know if John has any interest in doing it. (BTW, for some reason, the forum software over there put in an erroneous "USB adapter" link in my post. I didn't put that in there).
I'm suspicious that this effect may be the cause of the multiple-sweep oddities experienced by another poster earlier in the thread as well.Edited by andyc56 - 8/22/13 at 7:20pm