Reasons for A+B computed in REW not matching the actual measurement?

[andrejsali]

I just started to use REW, and enjoy learning about my stereo system with LR speakers and two subs, and room a lot. I have a question. I used the REW A+B function to compute the sum of two spectra previously measured (20-300 Hz) individually for each LR speaker and sub. The function will often produce exactly what is measured by REW for this pair. But not always. What are possible causes of this discrepancy when it occurs? Can it be used to diagnose problems that need to be treated to improve the sound of the system? Room modes, SBIR, phase effects between the speakers, ...? Does the REW function simply add the powers of the two sources, without anything else fancy, such as consideration of the phases measured for the sources individually (sorry if I missed this information in the manual)?

Thank you very much!

 

[John Mulcahy]

Trace arithmetic uses magnitude and phase (when both are available). For meaningful results measurements should have the same timing reference. It is generally better not to restrict the frequency range of the measurements used in trace arithmetic.

 

[andrejsali]

Thanks a lot. This makes sense!

How would I go about using the same time reference when one measurement is a full range speaker (I used Acoustic Time Reference for it so far) and the other one is a sub (I used Set t=0 at IR start so far)? I am running REW on a Mac laptop with a MiniDSP mic and 2x4 HD interface. The interface feeds a stereo preamp, which drives both 2 main speaker amps and, indirectly via an analog sub controller, 2 sub amps.

I presume that when the same time reference is used for a sweep measurement, the computed A+B should always be the same as the actually measured one for both A and B active during the measurement (Since both magnitudes and phases are used)?

A supplemental question: I presume also need to have the same timing reference for sweep measurements of individual speakers and subs in a system in order to compare their phases playing music. Is this correct? I am hoping to use these measurements to better understand the main speakers and subs integration.

Thanks so much again!

 

[John Mulcahy]

The output channel selected for the timing reference should always go to a speaker capable of reproducing a high frequency signal. It should stay the same for all measurements whose timing you wish to compare. The measurement output channel can be the same as the timing ref output or different, so the measurement output could be connected to the sub while the timing ref continues to go to a speaker.

 

[andrejsali]

The output channel selected for the timing reference should always go to a speaker capable of reproducing a high frequency signal. It should stay the same for all measurements whose timing you wish to compare. The measurement output channel can be the same as the timing ref output or different, so the measurement output could be connected to the sub while the timing ref continues to go to a speaker.

Yes, this also makes sense. So I guess I would need to be able to get an interface with at least 3 input channels, not just 2, to be able to connect a small dedicated timing speaker to the third channel and conveniently use the L and R channels for the different LR and sub combinations in my system (I am trying to avoid switching cables; I envision doing lots of measurements over a long period of time as my system evolves). For example, with the current setup, I cannot easily measure both subs (one on L and the other one on R) with a timing reference. What would be the best way to implement the same timing reference for all combinations of main speakers and subs? Could I use the mac analog headphones output for a separate timing speaker (used only for timing)? If not, any solutions involving Loopback input? If not, what interface hardware would you recommend (I wish miniDSP had a 4x4HD equivalent of 2x4HD)? I suppose using the third output on the 2x4HD to feed the timing speaker would pollute the L+R measurement (unless I manually disconnected it each time during the measurement phase, which is inconvenient), so that is not a good solution.

 

[andrejsali]

Yes, this also makes sense. So I guess I would need to be able to get an interface with at least 3 input channels, not just 2, to be able to connect a small dedicated timing speaker to the third channel and conveniently use the L and R channels for the different LR and sub combinations in my system (I am trying to avoid switching cables; I envision doing lots of measurements over a long period of time as my system evolves). For example, with the current setup, I cannot easily measure both subs (one on L and the other one on R) with a timing reference. What would be the best way to implement the same timing reference for all combinations of main speakers and subs? Could I use the mac analog headphones output for a separate timing speaker (used only for timing)? If not, any solutions involving Loopback input? If not, what interface hardware would you recommend (I wish miniDSP had a 4x4HD equivalent of 2x4HD)? I suppose using the third output on the 2x4HD to feed the timing speaker would pollute the L+R measurement (unless I manually disconnected it each time during the measurement phase, which is inconvenient), so that is not a good solution.

More questions (very grateful for your time!): Could I use the speaker / sub / mic distances to manually assign the Timing offset in the Measurement window to all measurements, especially for the subs? I suppose that would suffer from the unknown delay in my analog sub processor (which includes a phase setting), and I am interested in precise measurements? Any solutions involving the "Use loopback as timing reference", given my current (or some new) hardware?

 

[andrejsali]

More questions (very grateful for your time!): Could I use the speaker / sub / mic distances to manually assign the Timing offset in the Measurement window to all measurements, especially for the subs? I suppose that would suffer from the unknown delay in my analog sub processor (which includes a phase setting), and I am interested in precise measurements? Any solutions involving the "Use loopback as timing reference", given my current (or some new) hardware?

I tried to user the MiniDSP 2x4HD interface to mute the third timing speaker between the timing signal and the measurement sweep, but there was not enough time to do that. I suppose increasing the time may mess up the accuracy of the measurement? Anyway, in case I need a new USB audio interface to add a convenient Timing reference, would Focusrite Scarlett 18i8 work?

 

[andrejsali]

I tried to user the MiniDSP 2x4HD interface to mute the third timing speaker between the timing signal and the measurement sweep, but there was not enough time to do that. I suppose increasing the time may mess up the accuracy of the measurement? Anyway, in case I need a new USB audio interface to add a convenient Timing reference, would Focusrite Scarlett 18i8 work?

For a second, I thought using HDMI with my surround processor would do the trick (the LR speakers and the subs are part of my home theater system, too), allowing me to use say the center speaker as the acoustical timing reference as I measure L, or R, or L+R speakers. But this seemingly does not work as both L and R cannot be selected in REW when soundflower is used (only individual channels can). Maybe there is some software that allows mixing channels to achieve this goal, used together with soundflower?

So now I am not sure anymore a multichannel USB interface would allow using a 3rd output channel for timing reference, to measure L, R, or L+R channels?

Unless there is some weird timing randomness when using two audio interfaces, one for the speakers to be measured (say, a miniDSP 2x4HD) and the other one for timing reference (say, the built in Mac sound card via an external speaker connected to the headphone output or even the Mac speakers directly (though they may not be loud enough?)), that still seems the easiest solution, if REW supported it?

 

[andrejsali]

For a second, I thought using HDMI with my surround processor would do the trick (the LR speakers and the subs are part of my home theater system, too), allowing me to use say the center speaker as the acoustical timing reference as I measure L, or R, or L+R speakers. But this seemingly does not work as both L and R cannot be selected in REW when soundflower is used (only individual channels can). Maybe there is some software that allows mixing channels to achieve this goal, used together with soundflower?

So now I am not sure anymore a multichannel USB interface would allow using a 3rd output channel for timing reference, to measure L, R, or L+R channels?

Unless there is some weird timing randomness when using two audio interfaces, one for the speakers to be measured (say, a miniDSP 2x4HD) and the other one for timing reference (say, the built in Mac sound card via an external speaker connected to the headphone output or even the Mac speakers directly (though they may not be loud enough?)), that still seems the easiest solution, if REW supported it?

Sorry for all the spam - I am silly and slow, indicating my novice status with these things! But I did finally figure out the best solution, which is really simple: in REW, use L for the measuring signal and R for the timing reference and never change these settings; then use the miniDSP mixer to send the L and R to the right outputs (L to whatever whatever speaker/sub combination I wish to measure; and R to the separate timing reference speaker that is not part of the stereo system I wish to measure (which happens to be the center channel speaker in my home theater)).

With this setup, my original query is also resolved: all spectra summing sanity checks are passed beautifully now that the phases are measured accurately. So I guess there is not even a need to measure anything but individual speakers, as I can accurately compute any combined response. Beautiful! On to addressing room modes and SBIR effects a little more rationally, I hope!

 

[John Mulcahy]

Sorry for all the spam - I am silly and slow, indicating my novice status with these things! But I did finally figure out the best solution, which is really simple: in REW, use L for the measuring signal and R for the timing reference and never change these settings; then use the miniDSP mixer to send the L and R to the right outputs (L to whatever whatever speaker/sub combination I wish to measure; and R to the separate timing reference speaker that is not part of the stereo system I wish to measure (which happens to be the center channel speaker in my home theater)).

Glad you worked it out. The timing signal doesn't overlap with or affect the measurement sweep and there shouldn't be any need to mute the speaker used for the timing reference.