REW EQ question--correcting peaks creates a new dip

[lawnchair04]

Hello there. Been browsing the site a lot lately This is my first post, to see if I might get any feedback on a question I have about some REW measurements and EQ adjustments I made.

My question is about the effects of EQ filters generated by REW after performing MMM measurements. A couple of bass peaks were improved, but there's a new dip in the bass that seems to have been caused by the filters. Should I lower the values of the EQ filters around that dip, or add another filter to bring it up and compensate for it, or...? Any suggestions or input would be greatly appreciated! I am just starting to learn about acoustics, etc., so I'm just trying to get a better sense of what's going on with my speakers and in my room and how these measurements and adjustments are affecting things.

I have read the REW Help thread and the MMM for dummies thread. I used the MMM method for this measurement using a Umik-1. I've made several measurements over the past few days and they've been quite consistent, so I don't think I'm doing anything wrong with the measurement.

The speakers are a pair of KEF LSX II. No subwoofer. The listening/measuring position and speakers are set up in an equilateral triangle. The speakers are at the front edge of a low table, about 25cm from the wall behind them. Here is a quick sketch of the room and setup:

Measuring using the MMM method, I got these results:

REW created these filters:

Then I used the "Measure with these filters" button in the EQ settings, which yielded this result:

The peaks in the bass seem to be improved! But there's quite a big dip around 60Hz that wasn't there before. Any thoughts or suggestions would be greatly appreciated!

 

[thewas]

Welcome to ASR , the measurement with EQ is done as a sweep which is different than the MMM one you did first (which was a good choice), just create the filters and do the measurement with MMM again and I don't think you will see that narrow dip there.
Personally correcting the bass region works for me better by correcting L and R separately if you later want to experiment more.

 

[LTig]

Welcome to ASR , the measurement with EQ is done as a sweep which is different than the MMM one you did first (which was a good choice),

Yep. The sweep is more precise but also misleading because it is measured at a single position in room.

just create the filters and do the measurement with MMM again and I don't think you will see that narrow dip there.

Deep dips are usually the result of cancellation and cannot be fixed by EQ. You need to change the speaker position.

Personally correcting the bass region works for me better by correcting L and R separately if you later want to experiment more.

One needs to try both, since deep bass is usually mono.

 

[thewas]

Deep dips are usually the result of cancellation and cannot be fixed by EQ. You need to change the speaker position.

Exactly, but like you say since it doesn't appear in his MMM it seems to be quite a local issue and so probably not really audible.

One needs to try both, since deep bass is usually mono.

I am talking though about the whole region where EQ usually can help (so around 300-500 Hz depending on the room size) and also correcting both bass sources at the same time can lead sometimes to weird sounding constellations, but like you say, everyone should test what work best for them.

 

[LTig]

Exactly, but like you say since it doesn't appear in his MMM it seems to be quite a local issue and so probably not really audible.

Yep, but sharp dips usually aren't audible anyway.

 

[Keith_W]

PEQ's are minimum-phase which means that each PEQ you use modifies the phase as well as the amplitude. Because the relationship between phase and amplitude are inextricably linked, attempting to correct one problem introduces another problem somewhere else. So you add another PEQ, get another problem, add another PEQ to fix that, and get another problem ... and pretty soon you end up with dozens of PEQ's. The problem is, many DSP platforms have a limited number of PEQ's, for e.g. MiniDSP has 10 biquads (PEQ's) per channel, assuming that you aren't using them for something else. This is why you need programs like MSO that juggle PEQ's and all-pass filters to obtain the best result with as few biquads as possible.

Solutions:
1. If your DSP platform supports it, try enabling as many PEQ's as you can and see whether REW can obtain a better result.
2. Try MSO. Yes, it has "subwoofer optimiser" in its name, but it may be able to help.

 

[JohnPM]

PEQ's are minimum-phase

As are the response peaks due to room modal resonances.

 

[lawnchair04]

Welcome to ASR ,

Thank you!

the measurement with EQ is done as a sweep which is different than the MMM one you did first (which was a good choice), just create the filters and do the measurement with MMM again and I don't think you will see that narrow dip there.

Great tip. I redid the measurement with the EQ, this time using the same MMM method, and this is what I got (yesterday's sweep measurement in green, today's MMM measurement in purple):

Personally correcting the bass region works for me better by correcting L and R separately if you later want to experiment more.

Good tip. I went ahead and took measurements, and indeed the bass is quite different in the lower frequencies between the two channels (R in green, L in blue):

Interestingly those two peaks at around 85Hz and 117Hz are about the same, but the troughs above and below that area are different and even inverse in some places. Is this telling in any way?

 

[lawnchair04]

PEQ's are minimum-phase which means that each PEQ you use modifies the phase as well as the amplitude. Because the relationship between phase and amplitude are inextricably linked, attempting to correct one problem introduces another problem somewhere else. So you add another PEQ, get another problem, add another PEQ to fix that, and get another problem ... and pretty soon you end up with dozens of PEQ's.

That's kind of what I suspected.

Solutions:
1. If your DSP platform supports it, try enabling as many PEQ's as you can and see whether REW can obtain a better result.
2. Try MSO. Yes, it has "subwoofer optimiser" in its name, but it may be able to help.

1. My main music source is a Wiim Pro, which allows 10 channels, either Stereo or L/R separately. I've generated filters for both options as per @thewas's suggestion, so I'll test those out.
2. Based on MSO's homepage using it requires something like a miniDSP 2x4 or equivalent? It's also Windows-only, which complicates things given my recent conversion to Mac. Thank you for bringing this to my attention though; I will keep it in mind as another option.

 

[Wolf11Man]

Hello, nice to see you here.
Your dip may be gone if you use the same smoothing like the first curve you show.
This dip may be inaudible, even more at your listening position. Don't get mad about it, any room is a compromise for music reproduction.
Try per channel correction and compare it to the mono version. Pick your poison.
Keep doors and windows closed during measurements, or just like you would usualy listen.

 

[thewas]

Thank you!


Great tip. I redid the measurement with the EQ, this time using the same MMM method, and this is what I got (yesterday's sweep measurement in green, today's MMM measurement in purple):
View attachment 456784

You are welcome, as expected no dip as some mono filters would not create such.

Good tip. I went ahead and took measurements, and indeed the bass is quite different in the lower frequencies between the two channels (R in green, L in blue):
View attachment 456792

Interestingly those two peaks at around 85Hz and 117Hz are about the same, but the troughs above and below that area are different and even inverse in some places. Is this telling in any way?

This is very normal as rooms and placements are very rarely really symmetric, some modes can be seen on both though.

 

[lawnchair04]

Hello, nice to see you here.
Your dip may be gone if you use the same smoothing like the first curve you show.
This dip may be inaudible, even more at your listening position. Don't get mad about it, any room is a compromise for music reproduction.
Try per channel correction and compare it to the mono version. Pick your poison.
Keep doors and windows closed during measurements, or just like you would usualy listen.

Thank you!
I was using the same smoothing (1/48) for all measurements. And I suspected as you say that the dip would probably be inaudible…that said, taking the MMM measurement with the EQ applied erased the dip, so it might have just been a problem of using different measurement techniques.

I have been making measurements with doors, windows, and curtains closed. (Though I have tried using different combinations of open and closed and interestingly they don’t seem to make significant difference…)