Making sense of REW MLP measurements

[klettermann]

My adventures in DSP continue my understanding declines. For some reason I decided to compare all my historical REW curves for the MSP. I kept 6 curves each for left and right speakers. What I discovered was strange. I found that the respective L and R curves both fell into 2 distinct curve groupings, i.e., 3 right curves very closely resembled each other (group 1) while the other 3 (Group 2) very closely resembled each other, but there were significant differences between the groups. Put differently, it seems that I'm doing repeatable measurements, yet with some kind of systematic and repeatable error that puts measurements into one bucket or the other. I've tried to ensure that MLP has been the same in all cases, though there might be small differences in mike placement. Still, they wouldn't more than an inch or 2 away from each other.

This effect is shown below. Both are for the right channel and both can be greatly improved by REW EQing. Finally, the mains XO to subs at 80Hz. Anyway, green is right "Group 1" while the orange is right "group 2." Curves aren't level matched, so ignore that. The point is that there are some substantial differences from ~150-250Hz, the gap at ~325Hz and then the green peak ~1.3kHz. Can this be the result of anything but a tiny difference in mike position causing some significant differences? And if so, then presumably using the MMM method would smear out these differences? At the same time, how to determine what's right for EQing? Or just take an average, EQ that and call it a day? To be clear, I'm really trying to tweak things for the MLP. 99% of music listening is just one person at the MLP. Comments greatly appreciated as always. Thanks and cheers,


from~150H-

carefully others each showed.

 

[olieb]

I've tried to ensure that MLP has been the same in all cases, though there might be small differences in mike placement. Still, they wouldn't more than an inch or 2 away from each other.

So, speakers, room, mic and position are the same and yet you get significant differences in a range where wavelength is 1m-2m. Maybe there are objects that change place (window, door, table, curtains, ...)?

And if so, then presumably using the MMM method would smear out these differences? At the same time, how to determine what's right for EQing?

MMM will smear "something" out, as it is a spatial averaging, but this might be something different. The most annoying thing to me would be the 1.3kHz peak. It is very common that the FR above Schroeder frequency (150-600 Hz) is uneven, that is just how room acoustics is, but this peak ...

 

[klettermann]

So, speakers, room, mic and position are the same and yet you get significant differences in a range where wavelength is 1m-2m. Maybe there are objects that change place (window, door, table, curtains, ...)?

I wish! But no, room is sparsely furnished, just couch and ottoman. And they haven't moved at all. Neither have speakers - I don't think. Some of the curves do go back several months and I may have moved things a tiny bit in that time. Come to think of it, I did series of measurements varying toe-in. But that would only be an inche or 2 at most. Still, I guess that could be enough to make a difference.

MMM will smear "something" out, as it is a spatial averaging, but this might be something different. The most annoying thing to me would be the 1.3kHz peak. It is very common that the FR above Schroeder frequency (150-600 Hz) is uneven, that is just how room acoustics is, but this peak ...

Yes, troubling. However, given the apparent measurement problem it might not even be real. The left channel has a similar peak, but I know that it can be easily EQd away in both channels. OTOH, would EQing a peak away actually be creating a dip? I don't know.

Maybe the lesson here is to do a fresh set of measurements whenever doing EQ work AND do a bunch of repetitions. And then also do set of measurements bracketing the MLP by a few inches in all directions and using their average for EQing? Thanks for the comments and cheers!

 

[Keith_W]

My adventures in DSP continue my understanding declines. For some reason I decided to compare all my historical REW curves for the MSP. I kept 6 curves each for left and right speakers. What I discovered was strange. I found that the respective L and R curves both fell into 2 distinct curve groupings, i.e., 3 right curves very closely resembled each other (group 1) while the other 3 (Group 2) very closely resembled each other, but there were significant differences between the groups.

Your post is very confusing. To me, you are saying that all the measurements of your right speaker look the same, and all the measurements of your left speaker look the same, but comparison between left and right speakers are different.

 

[klettermann]

Your post is very confusing. To me, you are saying that all the measurements of your right speaker look the same, and all the measurements of your left speaker look the same, but comparison between left and right speakers are different.

Sorry about that, I'll try to clarify. Actually I wish your interprettion were the case, cause then I wouldn't be bothered or surprised. Anyway, for each speaker there were 6 different measurements. Of those 6, 3 were the virtually identical (group 1) and the other 3 were also the same (group 2). The curves shown are the averages of these respective groups for the right speaker only. There are clear differences between the respective groups. I didn't bother to show the left curves, but they do the same thing. However, the left and right aren't identical and I wouldn't expect that since the room isn't perfectly symetrical. I just showed right to minimize the number of curves.

In any case, I think I may have figured it out, and it should be a lesson, maybe, for other DSP beginners like myself. My suspicion is that there were indeed differences between the respective measurements. AND, my labeling/notation on the different measurement series was lousy to nonexistant. I discovered an old series of measurements looking at toe-in and listening position. The effect if toe-in was actually quitr substantial, more than the differences discussed here. So, I'm thiunking that I did indeed slightly change positioning at some point without noting what it was. Lesson: always do a fresh set of measurements even if yhou can't remember changing anything. Thanks and cheers,