Left and Right Measurements vs Combined

[Bama214]

An REW noob here, but read the help files and watched enough tutorial videos to be dangerous.

I’ve successfully measured (individually) my left and right speakers, computed calibration filters, and loaded those filters into my RME-ADI-2 DAC. Then remeasured each speaker individually. Great result!

My question relates to now measuring the combined response of both speakers. When I do this the new measurement (right+left) bears little resemblance to what I was expecting. I assumed it should be reasonably similar to the average of the right and left individual measurements, with an overall level boost. Instead, the combined measurement introduced multiple new peaks which weren’t in either of the individual measurements. I would also say that those individual measurements were quite similar, with no significant difference in frequency peaks or dips. (My room is quite symmetrical).

Hoping for some clarification here. For now I’m ignoring the combined measurement and the result sounds quite good. But, you know, audiophile nervosa.

Appreciate any comments!

 

[thothsong]

Sounds like it's due to phase differences. Measure L, R, and L+R sweeps all made with the same timing reference. Then use REW trace arithmetic to calculate A+B for L+R, and compare that with the measured L+R sweep. Assuming they match, then take a look at phase for both L and R sweeps, to see how they compare. Or, when you've made sufficient posts to allow it, post an mdat file containing those three sweeps. Edit: you should also compare the L and R impulse responses, to see if they're time aligned.

 

[Bama214]

Thanks for the response. Can you please define “trace arithmetic to calculate A+B“. I have the sweeps.

 

[thothsong]

It's important that the sweeps were done with a timing reference. If you're using the latest REW 5.20.14 early access, go to the All SPL graph, click on Actions, select Trace arithmetic Select L and R traces for A and B, A+B for the operation, and click Generate. If you're using 5.20.13, Trace arithmetic is under Controls instead.

 

[Bama214]

Ok - found trace arithmetic. Computed L+R as you suggested. Result does match my combined measurement.

So should my procedure have been to first measure the combined L+R frequency and use that to generate the EQ filters?

Thanks again!

 

[thothsong]

No, you need to figure out why they don't match, because they should match. Just to be clear, you used a timing reference? It would be useful to see the mdat file, but I forget how many posts you have to make before you're allowed to upload one. Try uploading?

 

[Bama214]

Yes acoustic reference.

Yes the A+B sum of the right and left channel matches the measurement of L+R

No the individual L and individual R measurements (which are quite similar) don’t match the L+R measurement

Thanks for all the help. I’ll try to append the mdat soon

 

[Bama214]

Success - mdat attached

 

[thothsong]

OK, guess I misunderstood you. Go to the Impulse graph in Overlays, select L and R, and see if they're aligned.

 

[thothsong]

Apparently not success, I don't see it attached.

 

[Bama214]

Too bad - looked like the mdat was accepted.

Using the combined L+R measurement shows a 100% peak at zero and an 85% peak at 250 usec, and 25% peaks at 635usec and 880 usec (looks like the same 250 usec offset). The overlay screen shows the same

 

[thothsong]

I mean compare the two individual L and R sweeps in Overlays, to see if one is offset from the other. I suspect you need 10 posts to attach files, but you should be able to insert images. Perhaps in the meantime you could provide a screen capture (use the camera icon in REW) of Overlays Impulse with the individual L and R sweeps selected, one of Overlays Phase with the individual L and R sweeps selected, and one of the All SPL graph with L, R, L+R, and the computed A+B all present.

 

[Bama214]

Photos attached

 

[thothsong]

The Impulse overlay is showing that L and R aren't quite time aligned. Is that offset consistent, if you take several pairs of measurements? It could be because your mic isn't exactly equidistant from both speakers, or it could be that you need to add a delay to one speaker to bring them into alignment. The third pic is missing the measured L+R. The Phase overlay is busier than I had expected, hard for me to digest, but it is showing differences between the two speakers, which affects the combined L+R.

 

[Bama214]

The measured L+R is essentially identical to the A+B

The mic position can certainly be adjusted. The 250usec difference is only 3 or 4 inches.

A photo of the

room is attached. Microphone was centered (as best I could by eye).

Not sure how to proceed. I’ll try to be more precise with the mic position (it’s on a stand).

I’m using the Umik-1 oriented horizontally and pointed directly at the centerpoint between the speakers at ear level. Does it have a preferred rotational alignment?

 

[thothsong]

If you point the mic at the ceiling and use the 90-degree cal file, how do the measurements compare?

 

[Bama214]

I can try later - away from the house now. I have used the vertical orientation when I adjusted the theater system, but understood horizontal was preferred for 2 channel.

I might also try the RTA with white (or pink?) noise to see what that might show.

Thanks again for taking so much time with this!

 

[thothsong]

To see if the impulse offset is causing any meaningful difference to the combined response, go to All SPL, select just the individual L and R measurements, and use the Align IR Start control/action. Look at Overlays Impulse to verify that they're now aligned. Then compute a new A+B, and compare that with the measured L+R. And with your next post you might be able to upload an mdat file.

 

[Bama214]

Excellent!
Did the time alignment of the individual L and R

Verified impulse now aligns

Did the A+B with the aligned L and R

The resultant A+B matches closely the contour of the aligned L and R (with an approximate 5dB level change).

Not sure where I go from here. Given there is a time delay between L and R which seems to produce a L+R anomaly compared to the individual L and R measurements, which plots do I EQ with? Can I reasonably use the new time aligned A+B as the basis for the EQ filters? (note: all the plots below used the original L and R plots Rms averaged to generate the EQ filters. The new time aligned A+B would be used to update those filters slightly)

This begs the question of the reality of the time delay and what effect that would really have

At any rate this has been a real learning experience and I’m very grateful for the support

 

[thothsong]

Can you attach an mdat file now?

 

[Bama214]

Attempting to send the file

 

[Bama214]

looks like a fail

 

[Bama214]

Will pause to rerun using modified mic position to attempt to eliminate the time offset

 

[thothsong]

How large is the file?

 

[thothsong]

As you've said, the distance involved is small. Consider how much your head can move just while sitting. Is the mic where your head would be at the MLP? If not, that's where I'd place it. If there's an offset when measured from there, then a small shift of one of the speakers could fix it, or if your system supports setting per-speaker delays, then setting a delay could fix it. Or, measure L+R in several places, an inch or two left/right/forward/back of your center head position, to mimic head movement, to understand the range of impact.