Phase issue causing dip in 13.5k range?

[datapata]

Hello! Im new to all of this, and I don't know all the terminology and thus have a hard time to search for an answer. But here goes!

I measured with a UMIK-1, 48k, 20-2000khz. No EQ have been applied. LR measurement is a "true" measurement, i.e not vector averaging L & R. But if I vector average I get essentially the same frequency response.

Ive got a pair on Genelec 8030s and I have this issue causing a big dip around the 13.5k. Just looking at the frequency responses of L & R it doesn't look like a dip in one channel would bring the LR measurement down.

1/3 Smoothing

Here is a 1/3 smoothed phase of L and R. As you can see the channels seems to be around 180 degrees, which if I understand correctly would essentially cancel out that frequency?

Non smoothed phase looks pretty all over the place?

And here is the LR phase

So.
* Is it common that L and R have such different phase response? Is it due to the room?
* What can I do to fix this? Do I have to use something like rephase or is one speaker defect?

Thanks!

 

[ernestcarl]

Just use an acoustic timing reference if you are using a USB mic.

Nowadays, I just use some pre-recorded tracks as I've found it gives far more stable results with my UMIK-1 and software media player (JRiver).

*full-range PN + Silence + Time Referenced Sweep

 

[datapata]

Just use an acoustic timing reference if you are using a USB mic.

Im using essentially the same settings, except the length.

Nowadays, I just use some pre-recorded tracks as I've found it gives far more stable results with my UMIK-1 and software media player (JRiver).

More stable how? How are you generating the pre-recorded tracks? When I generate a sweep in REW I don't get that stuff thats on both L/R in the beginning and cannot find a checkbox for it.

Thanks.

 

[datapata]

Playing a bit with this, it might just be a timing issue?

L has a system delay of 0.0011ms and R has 0.0942ms.

So using the alignment tool and applying a 0.01ms delay on L I get a response that is eerily similar to the LR measurement.

Perhaps I ought to do multiple measurements for each channel with FDWing and vector average them to not have the microphone position be so vitally important?

 

[staticV3]

Perhaps I ought to do multiple measurements for each channel with FDWing and vector average them to not have the microphone position be so vitally important?

At that point, just do MMM.
Here are the REW settings for that:

https://media.discordapp.net/attachments/860982405074649089/1041416797070557264/image.png

 

[ernestcarl]

Im using essentially the same settings, except the length.

View attachment 275422


More stable how? How are you generating the pre-recorded tracks? When I generate a sweep in REW I don't get that stuff thats on both L/R in the beginning and cannot find a checkbox for it.

Thanks.

Stable in the sense that one can measure longer without or with less clock drift glitches and random truncations… I haven’t tried 4M, though.

I simply inserted full range pink noise generated tracks that came from REW via Audacity.

I click to “start” measurement after the PN (from the pre-recorded track) has played out (maybe 8-10 seconds worth). However, I use a “dummy” output device in REW itself so no tones are actually generated by my speakers at the same time.

 

[datapata]

At that point, just do MMM.

Oh, interesting! So if I understand correctly, MMM is useful for EQ whereas sweeps are necessary for time/phase correction. Thanks!

Stable in the sense that one can measure longer without or with less clock drift glitches and random truncations… I haven’t tried 4M, though.

I simply inserted full range pink noise generated tracks that came from REW via Audacity.

Great, thanks for explaining

 

[asoka]

Oh, interesting! So if I understand correctly, MMM is useful for EQ whereas sweeps are necessary for time/phase correction. Thanks!


Great, thanks for explaining

I have a similar problem that drives me nuts, spending countless hours to try solving but to no avail. You find a solution for this?
Independent speakers measure correct when both playing disaster in the highs

 

[datapata]

@asoka

Sorry for late reply, you have probably already sorted this but if not, try a RTA with periodic pink. Make sure to set the mode to RTA 1/48 octave in the RTA settings.

 

[Duke]

That dip looks to me like it is consistent with one speaker being 1/2 wavelength closer to the microphone at 13.5 kHz, which would put the two signals out-of-phase with each other at the microphone location in that frequency region. This distance just happens to work out to 1/2 inch (a 13.5 kHz wavelength being 1 inch in length).

 

[LIΟN]

* Is it common that L and R have such different phase response? Is it due to the room?

There may, of course, be an imbalance in phase depending on the room imbalance and the position of course.

There are times when the clock is not fixed, such as the USB umik microphone, and sometimes the timing is misaligned, not only in the actual left and right distances, but also in the speakers and microphones.
Even if you proceed with a multi-point message in the REW and time align, if you do vector Avg without the impulse peak being properly aligned, the high notes can be added up strangely (like your current message)

1. Check the distance between each speaker and microphone again. (This also includes toein angle)
- If you're measuring at a long listening distance of more than 2-3 meters, it shouldn't be a big problem. However, near 1m, near field listening is very sensitively reflected in the message.

2. Check L and R in the impulse overlay (% Graph) to make it easier to check. You should also check Step Response.
- If you placed it in the correct microphone position (listening position), and the position of the speaker is also balanced, it should be very overlapping and similar.

 

[ozzy9832001]

I have a similar problem that drives me nuts, spending countless hours to try solving but to no avail. You find a solution for this?
Independent speakers measure correct when both playing disaster in the highs

If the speakers aren't toe'd in properly or are slightly out of alignment it can easily cause this. There will be only 1 sweet spot and that is dead center at the point in which they meet (depending on toe in). I would not worry too much about it since our ears don't meet dead center of our head.

If you have a desk, chair, other furniture, PC or equipment between you and the speakers it can also cause a bad dip or comb filtering. This is why they are measured separately (amongst other reasons). If you correct through toe in to get a good measurement, it may not sound the best. You may get additional comb filtering at the lower frequencies because the sound fields between the 2 speakers are now closer and more directly pointed at one another.