Inconsistent top end in measurements

[WillDoLikeMildew]

Sometimes when making measurements, the high frequencies get measured as 2.5dB louder than other recordings, despite both recordings taking place in the exact same conditions: mic in the same position, receiver with the same settings, all speakers in the same positions. Additionally, I live in a quiet place and take great care to reduce any unwanted noise during sweeps. These "boosted treble" measurements are not outliers, but happen for many measurements in a row. In fact, lately they have been happening more frequently than my "lower treble" measurements, which has me wondering which measurement is correct, or if neither of them are. Here is an example of two measurements that were taken back to back in the exact same conditions.

I use a UMIK-1 mic with the calibration file provided by miniDSP. I do the sweeps with a Macbook Pro plugged into my receiver via AUX. If I recall correctly the pink noise level test before sweeping measures 79 dB according to REW. These are both measurements of L,R, and Sub playing simultaneously. Octave smoothing is 1/3. I would really love to know what I'm doing wrong.

 

[thothsong]

Do you get the same variation when measuring an individual speaker?

 

[WillDoLikeMildew]

Do you get the same variation when measuring an individual speaker?

I noticed that when I took my individual speaker measurements and virtually summed them with the alignment tool, the result had about 2.5 dB more top end than the measurement I got by actually playing both speakers at the same time. As I recall, I repeated this whole process and it happened the second time too. But to answer your question, I didn't personally notice individual speaker measurements with the same conditions showing different frequency responses. But to be fair, I didn't do that many measurements of individual speakers and I also wasn't looking for this problem at the time.

I could be way off here, but I'm suspicious that the lower the recording level, the more likely the treble is to show up too high. This could explain why my measurements of individual speakers showed up as brighter than measuring both speakers playing at once (which would be a louder measurement).

 

[thothsong]

I'd suggest taking more single speaker measurements, to see if they are stable. Did you use a timing reference for the individual speaker measurements that you summed?

 

[WillDoLikeMildew]

I'd suggest taking more single speaker measurements, to see if they are stable. Did you use a timing reference for the individual speaker measurements that you summed?

Yes, I use the left speaker as a timing reference for everything.

edit: Also for the record, the individual speaker measurements were both brighter than the measurement of both playing simultaneously, so any excess of treble was there prior to being virtually summed.

 

[thothsong]

I'd see if single speaker measurements are stable. Perhaps there's some timing variability between channels that's resulting in sufficient phase shift at upper frequencies to cause their sums to vary?

 

[John Mulcahy]

Here is an example of two measurements that were taken back to back in the exact same conditions.

Can you attach the mdat file?

 

[WillDoLikeMildew]

Can you attach the mdat file?

 

[John Mulcahy]

Pretty dramatic differences in the impulse responses for those two measurements, but with both speakers playing even a couple of mm of mic shift could have a large effect. The second measurement is much cleaner.

 

[WillDoLikeMildew]

Pretty dramatic differences in the impulse responses for those two measurements, but with both speakers playing even a couple of mm of mic shift could have a large effect. The second measurement is much cleaner.

Thank you for looking at that, John. The mdat I sent is simplified to two measurements because I thought the full mdat I was working on would be confusing to someone who wasn't me but the full story is that before doing either of those measurements, I made a measurement that lines up pretty much exactly with the last measurement there. I'm very careful not to touch my mic or my mic stand between measurements. I could potentially believe that the arm of my mic stand might sag over time but that wouldn't explain how the measurement reverted again to perfectly match how it was before. Maybe the stand somehow swiveled one direction and then the other? Beats me.

If such a small difference in mic position could cause such large effect as you say, then would it be ill-advised to look at an L+R+Sub measurement for the purpose of finding the right EQ to match a target curve? And if it's not advisable to look at that measurement, which measurement should I be looking at? I would be tempted to answer with "the left and right channels individually with the sub turned on" but part of me also feels like that's the wrong answer because sub would be 6 dB louder when both channels are playing because there's just one sub and it's perfectly in phase with itself, while the speakers would be less than 6 dB louder when both channels are playing because there's two of them and they're not perfectly in phase.

I'd see if single speaker measurements are stable. Perhaps there's some timing variability between channels that's resulting in sufficient phase shift at upper frequencies to cause their sums to vary?

Well, yesterday I did a bunch of measurements of individual speakers and I was unable to get any discrepancy, but I also did a bunch of measurements of both speakers playing and I got a very tiny amount of discrepancy that I am willing to ignore. Nothing like the discrepancy I first showed here.

5 Left speaker measurements:

5 Right speaker measurements:

7 L+R measurements:

While doing this, I happened to notice the right measurements had a little more treble than the left measurements and I don't think this was an error in the measurement because it was so consistent so I went ahead and adjusted the EQ in my receiver to correct this.

Finally, I tried L+R+Sub simultaneously and weirdly enough, I started getting larger discrepancies in the high end between these measurements. It's still a really small discrepancy and nothing like the discrepancy I first showed in this thread, but it is notably larger than the discrepancy of the L+R measurements I just did, which is weird, because you wouldn't think a sub would do anything to the highs, but perhaps this just isn't enough data to make that call.

Afterwards I did a couple of left speaker measurements which were consistent with each other and then I did couple of right speaker measurements and these were consistent with each other, but I cannot compare these to the individual speaker measurements I did prior to changing the receiver because obviously those will be different because of those changes.

Then I did 3 more measurements of L+R+Sub and they were inconsistent with each other.

All 9 L+R+Sub measurements I did:

So I was unable to replicate such a large discrepancy as the one I originally posted so you would think that one was an outlier but I swear at one time I was consistently getting discrepancies that large, but I would have to go through old mdat files and my madman's scrawling notes for definitive proof of that to post here and right now, it's more convenient to believe these recent measurements are accurate, especially the ones of individual speakers. But idk