• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Subwoofer Time Alignment

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
I did a measurement of the subwoofer alone (mic at my listening position), before and after EQ. After eq, when I click on estimate IR delay, I got ~-17ms. Before eq, I got around -0.04ms. The subwoofer is right below my desk, ~15cm away from me.

What does the -17ms value mean? How do I use this value?
This is my rew measurement (The measurements are done at 2 different Windows volume, because it's too loud unequalized): https://app.box.com/s/6p5mmci1xgnjtf3cox7sratszhqq6mi9

These are my subwoofer eq filters:
MzXmIli.png


Subwoofer before IR delay shift
BJchj2l.png



After Estimation and IR delay shift. So what's the significance of this?
CBy8BbQ.png


Sorry, you actually meant this:

1596570468240.png


The estimate is not going to be correct. It will add ~17ms delay as a timing offset to try again, but when you remeasure with said offset, REW then tells you need to subtract 5ms again... agh... so you go back to the original ~11ms delay suggestion. There is still much that I don't know about REW, but this does not seem to be of much practical help.
 

whazzup

Addicted to Fun and Learning
Joined
Feb 19, 2020
Messages
575
Likes
486
Sorry, you actually meant this:

View attachment 76640

The estimate is not going to be correct. It will add ~17ms delay as a timing offset to try again, but when you remeasure with said offset, REW then tells you need to subtract 5ms again... agh... so you go back to the original ~11ms delay suggestion. There is still much that I don't know about REW, but this does not seem to be of much practical help.

Thanks for trying! Yes, it doesn't seem to work in the manner that I thought it did. Or maybe it serves a different purpose.
 

whazzup

Addicted to Fun and Learning
Joined
Feb 19, 2020
Messages
575
Likes
486
Some interesting and perculiar results.
REW file: https://app.box.com/s/1ibl0bmfnx6f1ko007ex88yvwthkbtj9

First I tried applying a few different delays. This is R + sub channels (delay applied to R only). I noticed that as I increased the delays to the speaker channel, nothing much happens to frequencies above 100hz (my crossover pt), but sub 100hz gets pushed closer to 0, and then negative.
Based on these 3 graphs alone, is 10ms 'better' because it's closer to 0?
FanDbUn.png



Then I looked at step response for the same R+sub. I noticed as delay increases, the peak gets closer to 0m, then goes negative? And the 'truncated' fuzzy peaks? Not quite sure the significance...
ryudPtg.png



Next I tried manually adjusting the delay to get the peak to 'line up' with 0. Note that now it's LR+sub channels. I got to 12.7ms.
QvVY4Hu.png



Now when I went back to SPL, turns out 12.7ms delay gave a deep suck out at roughly the crossover point??!! And I can hear the huge volume drop when I generate single tones in REW and slowly step through the frequencies.
Does that mean I'm better off not adding any delays?
Anyway, although I do not fully understand them, the results are still very fascinating...
RhO735R.png
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
Some interesting and perculiar results.
REW file: https://app.box.com/s/1ibl0bmfnx6f1ko007ex88yvwthkbtj9

First I tried applying a few different delays. This is R + sub channels (delay applied to R only). I noticed that as I increased the delays to the speaker channel, nothing much happens to frequencies above 100hz (my crossover pt), but sub 100hz gets pushed closer to 0, and then negative.
Based on these 3 graphs alone, is 10ms 'better' because it's closer to 0?
FanDbUn.png



Then I looked at step response for the same R+sub. I noticed as delay increases, the peak gets closer to 0m, then goes negative? And the 'truncated' fuzzy peaks? Not quite sure the significance...
ryudPtg.png



Next I tried manually adjusting the delay to get the peak to 'line up' with 0. Note that now it's LR+sub channels. I got to 12.7ms.
QvVY4Hu.png



Now when I went back to SPL, turns out 12.7ms delay gave a deep suck out at roughly the crossover point??!! And I can hear the huge volume drop when I generate single tones in REW and slowly step through the frequencies.
Does that mean I'm better off not adding any delays?
Anyway, although I do not fully understand them, the results are still very fascinating...
RhO735R.png

Not sure why your mains aren't aligned to the 0 time mark. What type of filters or DSP and settings are you using for the mains?

From the look of things, better stick with no delays.

There's quite a strange look in some of the graphs -- almost like ringing...

1596613876056.png


1596613885071.png


Normalized step to IR peak
1596615441776.png

Looks like there's quite a large offset from zero, and the sub's step ain't too clean.

1596615757172.png


Some of the filtered IRs also don't look clean to me... which (probably) partially explains the really low clarity curves -- I've never seen anything go below zero % before:

1596615652723.png


Could be just something inherent in the speakers and/or subwoofer as I also see prolonged ringing in the sub and speaker's bass around 120Hz...
1596617197058.png


1596616407305.png


Sub level looks a little bit hot relative to just the speaker.
1596617690566.png

Time offset of the speakers seen here...

Sorry to veer off a bit from just the sub-speakers alignment subject.

But, anyways, yeah, all of that aside, it definitely doesn't seem like you need to add any more delays for the speakers.
 

whazzup

Addicted to Fun and Learning
Joined
Feb 19, 2020
Messages
575
Likes
486
Not sure why your mains aren't aligned to the 0 time mark. What type of filters or DSP and settings are you using for the mains?

From the look of things, better stick with no delays.

There's quite a strange look in some of the graphs -- almost like ringing...

View attachment 76713

View attachment 76714

Normalized step to IR peak
View attachment 76721
Looks like there's quite a large offset from zero, and the sub's step ain't too clean.

View attachment 76723

Some of the filtered IRs also don't look clean to me... which (probably) partially explains the really low clarity curves -- I've never seen anything go below zero % before:

View attachment 76722

Could be just something inherent in the speakers and/or subwoofer as I also see prolonged ringing in the sub and speaker's bass around 120Hz...
View attachment 76725

View attachment 76724

Sub level looks a little bit hot relative to just the speaker.
View attachment 76726
Time offset of the speakers seen here...

Sorry to veer off a bit from just the sub-speakers alignment subject.

But, anyways, yeah, all of that aside, it definitely doesn't seem like you need to add any more delays for the speakers.

Forgot to mention, I have my speakers flipped (sub still facing me).
Wanted to add the link in the previous post but got distracted earlier:
https://www.audiosciencereview.com/...all-rooms-fire-your-speakers-backwards.14968/

Thanks for highlighting the 120hz, didn't notice that. But it's not in the FR....??
The ringing you mentioned, could be my computer casing? Or other objects in my room.
I'm still not very sure about the other types of plots you mentioned and how to read them, like filtered impulse.
For group delay, the rough idea I have is that the ideal is for it to go to 0, correct?
And how did you get the spectrogram to look like that, haha....mine's just a few blobs along a horizontal straight line...


Sub filters
AtOzr82.png



L and R filters
asVYL6b.png


kuSuysU.png


If this is getting off-topic for this thread, then will stop here.
 

dasdoing

Major Contributor
Joined
May 20, 2020
Messages
4,209
Likes
2,674
Location
Salvador-Bahia-Brasil
oh yea, the @whazzup experiment.
your excess group delay data is hard to analyse cause of all the variations, so I put psychoacoustic smoothing on it. I also put the offset (0 point) on that huge negative spike on the step reponse.
So I got this for your L+R zero delay meassurement:

ffg.jpg


as expected (for backfireing) it's actualy the directional highs that are too late (since they only arrive as reflections). I think the sub (not the mains) needs 8-10ms of delay. then, somehow you would need to delay those mids by about 8-10ms, too. you would need DSP (crossover the mids and delay them)
 
Last edited:

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
No wonder... I was trying to figure out why I could not see a straight peak energy trace for the higher frequencies. o_O All those reflections made for confusing graphs.

I have no experience with the backfiring speakers thing, though.

*REW has a lot of extra options other than defaults to change views... I switched it to the ‘wavelet’ view as you can see higher resolution in both Time and Frequency within a single window.
 

dasdoing

Major Contributor
Joined
May 20, 2020
Messages
4,209
Likes
2,674
Location
Salvador-Bahia-Brasil
I was curious what Align2 (https://www.minidsp.com/applications/advanced-tools/align2-for-drc *seams Minidsp removed the software form their webpage; probably because I mentioned it here before? lol) as frontend for DRC (http://drc-fir.sourceforge.net/) could do to phase correct this extreme case of @whazzup

took the no delay meassurements

this is the result for the right channel

excess group delay before
1.jpg


excess group delay after
2.jpg


if you want to try this filter out: https://drive.google.com/file/d/1Jhr_YKN3cqtwgApuk31R8McFwMOkB4wJ/view?usp=sharing
it's a stereo IR. you need a convolver like the one in APO-EQ

I have disabled any FR correction, so the diference should only be in the phase
(normaly I would take a listen and adjust parameters if artifacts are present)
 
Last edited:

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
@dasdoing @ernestcarl
I did a couple more measurements earlier, one with the flipped position and one with facing me position. Just the right speaker + sub, but I did add filters to the sub to adjust the fr.
https://app.box.com/s/p3yy46bekri6xsm7zxkkgfxnug6hnfkd

Yeah, I see now the persistent 120Hz room noise... I wouldn't really know which sounds better -- flipped or not. The only thing I can tell for certain is that bass alignment should be the least of your worries. Distortion around the vocal range... well, that must be at least audible.
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
Just an update: I take back what I said about REW not giving me the best delay estimate!

I was wrong.

In my room and the way things are placed and situated here, this method actually works correctly.

I was too caught up in worrying about the apparent suck-out around the xo area, that I did not think how it would actually be so easy to EQ the FR in that xo region after things had become as closely phase-matched as possible.

Subsequent re-measurements gave a value around 11.7ms or so...

1596732971752.png


It was important that no EQ was activated around the xo region other than two room mode EQs for the right speaker at the bass xo area -- the roll-off for the speakers and subwoofer had to follow as closely as possible REW's targets for subwoofer only and bass limited speakers -- using my xo cut-off and slope settings.

1596733256076.png


This is the best match since there's no way around that huge and deep cancellation in the sub between 75Hz-120Hz.
1596733300593.png


1596733589942.png


A small improvement in group delay is still worth-it, nevertheless.
1596733616989.png


1596733663730.png


1596733686332.png


Decay plot does not look much worse than before
1596733788008.png


That EQ boost around the xo region did have an effect seen in the waterfall, but nothing too serious.
1596733952692.gif


1596734006737.png

Higher resolution...

Zoomed-in and at the full FR range
1596734087386.png


Besides that extension in delay / gap at 130Hz due to unavoidable SBIR, everytyhing's good.

I am not sure how this sub-speaker alignment method and automatic delay calculation with REW would hold up if one's sub is not equidistant between speakers or the MLP, though. I'm also none the wiser as to how well it would work with multiple subs. But I can at least say with certainty that it has, in fact, worked for me and my couch and ('half') my desk setup.
 

dasdoing

Major Contributor
Joined
May 20, 2020
Messages
4,209
Likes
2,674
Location
Salvador-Bahia-Brasil
I think we should make clear that phase alignment and time alignment are not the same (not at listening position). when you phase align in a way that you you make the max amplitude of two frequency waves incident at the same time you are not garanting that they occure at the same time. if they are 360 degrees apart they are phase aligned but not time aligned
 

QMuse

Major Contributor
Joined
Feb 20, 2020
Messages
3,124
Likes
2,785

Judging from this overlay of step responses it looks your sub is still behind your mains. The only graphs that are relevant for time alignement are impulse and step response of mains and sub. Can you post overlayed impulse responses of measurements #2 and #3?

You should measure main and sub separately with sweep that starts from 0 Hz. Set sweep timing reference to your right channel and then measure your left main (without sub) and sub (without left main), both from your LP. Do not have REW do time aligning as both, left main and sub, will use right main as timining reference.

Post REW file and I will have a look at it.
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
Judging from this overlay of step responses it looks your sub is still behind your mains. The only graphs that are relevant for time alignement are impulse and step response of mains and sub. Can you post overlayed impulse responses of measurements #2 and #3?

You should measure main and sub separately with sweep that starts from 0 Hz. Set sweep timing reference to your right channel and then measure your left main (without sub) and sub (without left main), both from your LP. Do not have REW do time aligning as both, left main and sub, will use right main as timining reference.

Post REW file and I will have a look at it.

Sure, I'll try tomorrow as people are sleeping. Log sweeps in previous posts start at 0Hz and end at 20kHz using 512k to 2M lengths already for the sub and speaker only measurements. I do notice that phase changes / is somewhat unstable below 1M.
 

QMuse

Major Contributor
Joined
Feb 20, 2020
Messages
3,124
Likes
2,785
Sure, I'll try tomorrow as people are sleeping. Log sweeps in previous posts start at 0Hz and end at 20kHz using 512k to 2M lengths already for the sub and speaker only measurements. I do notice that phase changes / is somewhat unstable below 1M.

Let's try like this: run 30-150Hz sweep separately on the sub and on the left main with right main set to play timing reference signal with both sweeps. If your sub is closer to your right main than run sweep on sub and right main using left main as timing reference.
You better use 2M lengths so there is more time for the steady state to stabilize. Set sub phase knob to zero and remove any delay for mains in miniDSP. Run signal via miniDSP but without any crossover and HP/LP settings on sub and mains. If you have room EQ PEQs defined for mains and sub in miniDSP leave them on.

Once again, do not use "Estimate IR delay" feature in REW after you do the measurements. Plz attach zipped REW mdat file or post a link to a location from which I can download it.

Did I understand correctly your mains are Presonus Sceptre S8 and sub is Rythmik F12?

Can you also plz draw a sketch so I can see the positions of your mains vs your sub vs your LP.
 
Last edited:

QMuse

Major Contributor
Joined
Feb 20, 2020
Messages
3,124
Likes
2,785
Sure, I'll try tomorrow as people are sleeping. Log sweeps in previous posts start at 0Hz and end at 20kHz using 512k to 2M lengths already for the sub and speaker only measurements. I do notice that phase changes / is somewhat unstable below 1M.

One more thing: as you will be doing XO with miniDSP it is important that you use LFE input with your sub so it's internal XO doesn't come to play.
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
Room and placement for context:
1597069362148.png

BTW, placement is pretty much fixed so no further adjustments on that front possible.

I have Rythmik's older F12 sub so there's no LFE input -- left and right only.
The xo is already set at max and the phase/time delay knob is set with zero delay or at its initial setting.

There is a little time difference between left and right speakers -- but both are more or less equidistant from the 'center-positioned' sub.

I simplified the EQ around the xo region a bit.

sub-speaker xo phase time alignment COMPLETE mdat files

The problem with adding delays in my particular case is that no matter what, a suck-out develops -- see attachment: full-range L+R+SUB summed measurements with current 11.7ms delay vs no delay.
 

Attachments

  • full-range with 11.7ms delay or no delay (256k) sum only.zip
    2.2 MB · Views: 97
Last edited:

QMuse

Major Contributor
Joined
Feb 20, 2020
Messages
3,124
Likes
2,785
Room and placement for context:
View attachment 77430
BTW, placement is pretty much fixed so no further adjustments on that front possible.

I have Rythmik's older F12 sub so there's no LFE input -- left and right only.
The xo is already set at max and the phase/time delay knob is set with zero delay or at its initial setting.

There is a little time difference between left and right speakers -- but both are more or less equidistant from the 'center-positioned' sub.

I simplified the EQ around the xo region a bit.

sub-speaker xo phase time alignment COMPLETE mdat files

The problem with adding delays in my particular case is that no matter what, a suck-out develops -- see attachment: full-range L+R+SUB summed measurements with current 11.7ms delay vs no delay.

Ok, I see. Can you plz repeat these 4 measurements, this time with 50-130Hz sweep and add 4ms delay to the mains.

Capture.JPG
 
Top Bottom