Manually time-aligning subwoofer(s) to mains - how to

[janbth]

View attachment 550379

Here are 3 measurements of my right sub, all with tweeter timing reference. The measurements were taken back to back, and the mic was not moved.

View attachment 550380

And here are the delays reported by REW. There is a +/- 0.3ms discrepancy between all 3 measurements, despite all of them being correctly aligned on the impulse response.

View attachment 550381

If we drop a marker at the 7ms mark, where REW has indicated that the subwoofer impulse starts, we can see the problem. Each measurement has a tiny discrepancy in SPL, which is throwing off REW's automatic delay reporting by +/- 0.3ms.

Again, I think you are mixing things up. The time delay reported in the measurement panel ("delta t") is the system delay - an estimated number. The three graphs you show seems perfectly time aligned to me, which they should be if you have used the same acoustic timing reference. Differences in SPL is another matter, probably due to normal run-to-run variation, and is probably the reason why the estimated numbers are different.

Another quote from the same REW help page as above:

System delay​

If using a timing reference REW can calculate the delay through the system being measured relative to the reference and show it in the measurement Info panel as System Delay in milliseconds, with the equivalent distance in feet and metres shown in brackets. Any Timing offset is shown below the System Delay. For speakers the delay estimate is based on the location of the peak of the impulse response. Subwoofers have a broad peak and a delayed response due to their limited bandwidth so the delay is instead measured relative to the start of the impulse response. The start of the impulse response cannot be located as precisely as the peak, however, so delay values are less accurate for subwoofer measurements.

 

[OCA]

There is a +/- 0.3ms discrepancy between all 3 measurements,

Have you been using "adjust clock with acoustic timing reference" option with dual ATR chirps?

 

[SoundsGood64]

At the end of my last post (#250), I suggested that there could be an interaction among L,R, and subwoofer when L+R is swept with the crossover active and the subwoofer on. I think this interaction depends on having the sub delay present.

Correct summation is L + Sub + R + Sub, vector average will count sub only once.
Always generate EP of the measurement, Remove IR delays from the EP and only then compare phase graphs
If you use "Align IRs at cursor" instead of "Align phase at cursor", REW's alignment tool will automatically give you correct sub polarity. Phase alignment cannot invert.

Ok, thanks for this post. I will try it. When time-aligning the sub to main speakers at the crossover point (80hz for me), do you recommend separately measuring 1) L alone (XO active, sub off) and R alone (XO active, sub off), OR 2) L + R together (XO active, sub off)? The reason I am asking is because of your suggestion of "Correct summation is L + Sub + R + Sub, vector average will count sub only once" could imply measuring L and R separately (not together). If measuring L and R separately when it comes to verification stage, take the vector average of L+sub (XO active) and R+sub (XO active) ("L + Sub + R + Sub") and compare that with a new measurement of L+R (XO on , sub on).

Before time aligning the sub to L and R speakers, do you recommend phase aligning L (#1 blue) and R (#2 red) to each other at the crossover point using a second order allpass filter? I've already time aligned L and R with a mid-range sweep but they aren't quite phase aligned at 80hz. Also shown in this 'unwrapped phase' overlay is L+R (sub off, XO active, #3 magenta) and subwoofer (#4 green). I used "+360" on the L line to bring it next to R line (I think that is ok but stating here in case it is not):

 

[OCA]

At the end of my last post (#250), I suggested that there could be an interaction among L,R, and subwoofer when L+R is swept with the crossover active and the subwoofer on. I think this interaction depends on having the sub delay present.

Ok, thanks for this post. I will try it. When time-aligning the sub to main speakers at the crossover point (80hz for me), do you recommend separately measuring 1) L alone (XO active, sub off) and R alone (XO active, sub off), OR 2) L + R together (XO active, sub off)? The reason I am asking is because of your suggestion of "Correct summation is L + Sub + R + Sub, vector average will count sub only once" could imply measuring L and R separately (not together). If measuring L and R separately when it comes to verification stage, take the vector average of L+sub (XO active) and R+sub (XO active) ("L + Sub + R + Sub") and compare that with a new measurement of L+R (XO on , sub on).

Before time aligning the sub to L and R speakers, do you recommend phase aligning L (#1 blue) and R (#2 red) to each other at the crossover point using a second order allpass filter? I've already time aligned L and R with a mid-range sweep but they aren't quite phase aligned at 80hz. Also shown in this 'unwrapped phase' overlay is L+R (sub off, XO active, #3 magenta) and subwoofer (#4 green). I used "+360" on the L line to bring it next to R line (I think that is ok but stating here in case it is not):
View attachment 550425

With a single sub, you need to work with L+R together and an allpass to a single speaker is also not an option as it will shift phase everywhere else for only one speaker and damage phantom image. Using sub twice in the summation only corrects frequency response but will not effect sub time alignment. If you will use "vector sum" in REW when sub, L and R are selected, increase sub volume by 6.02dB first and the summation magnitude response will look correct.

Similarly, you cannot time align L and R between each other, they need to stay put for centre image to keep central at MLP.

This video explains the process (although you already seem to know it):

 

[SoundsGood64]

Thanks OCA, very nice video which helped a lot. I followed OCA's method for determining crossover point and sub gain. I did a combo of Keith_W post #225 and OCA's method for time aligning sub and main speakers. While OCA appled crossovers in REW to sweeps without the crossovers, I did sweeps with the crossover in place. When 'muting' speakers below I applied a -90dB gain to speakers in CamillaDSP. Here is what I did, please review and critique:

1. In REW, set acoustic timing reference to surround right. This speaker's chirp is
timing anchor.

2. Capture 1 — sub only, driven by "L+R" (select L+R in REW): mute the L and R speakers
themselves, sub on, crossover active.
- Sweep: 10-250Hz, 512k length
- Level: -9dBFS
- AVR: -3dB (Note I have a master gain set at -6dB in CamillaDSP that is applied to each channel)
- Single sweep mode (required when using acoustic timing
reference)
- Name: "sub only, L+R driven"

3. Capture 2 — main speakers only: mute the sub, L and R main
speakers on, same sweep settings (select L+R in REW). Name: "L+R mains, sub off".

NOTE: Do not substitute a sweep sent to the discrete LFE channel
for this capture. LFE does not go through crossover
filter and is a different, unfiltered signal path. Both captures above
must go through the real crossover. More on LFE channel below in #10.

4. Under All SPL, select Actions, then Alignment Tool.

5. In the Alignment Tool, select "L+R mains, sub off" on top and
"sub only, L+R driven" on the bottom. Select Impulse Alignment
and place the cursor at crossover frequency. Press Align IRs
at cursor. If Invert Polarity gets selected, note this. Read the Delay value: if
positive, delay the sub by this amount. If negative, delay the main speakers by
the absolute value of this number. (You can also swap the positions of the sweeps
in the alignment tool which I did). In my case, I ended up delaying main speakers (L, R, center, surround L, surround R) by 17ms and kept normal sub polarity, per the alignment tool.
Here I have overlays from the alignment tool for both impulse and phase, as well as an unwrapped phase overlay. Green is L+R channel with sub only, actual L&R speakers off, and magenta is L+R channel with L&R speakers on with a 17.81ms delay applied to them, and the sub off. In the unwrapped phase overlay, the green and magenta lines are equal at the crossover point of 40hz and their slopes (group delay) are also the same. (Note when doing time alignment, to leave the 17.81ms delay OFF for surround R which is timing reference. Once everything is set though, then surround R also gets 17.81ms delay)

6. In my system, I also have a center speaker (XO 115hz) and surround L and surround R speakers (XO 110hz). I found crossover filters can delay the signal and that delay varies by crossover frequency. So I thought I should also align the low frequency output from those channels that would go to the subwoofer with the sub only (L+R driven) from step 2 above.

7. Capture — sub only, driven by Center: mute the actual Center speaker but choose Center speaker in REW, sub on, crossover active, same sweep settings as before. Name: "sub only, C driven".

8. In the Alignment Tool, compare "sub only, L+R driven" (from step 2 above) against "sub only, C driven". Choose an alignment frequency well below every crossover group's corner — say 25hz.

9. Note the resulting delay for center channel LF, which in my case was 8.08ms.

10. Repeat Steps 7-9 for surround L (needed delay was 7.91ms, to also be applied to surround R channel), and also for LFE channel. The LFE channel in my system does not go through a crossover, so the signal is faster than all the other LF content from main channels (L,R,center, surround L, surround R) which gets passed to the sub through crossover. In this case I needed to delay LFE by 12.07ms.Here is unwrapped phase overlay for 1) L+R (no sub) with 17.81ms delay for L&R speakers(magenta), 2) L+R sub only with L & R speakers off, 3) SL sub only with 7.91ms delay, C sub only with 8.08ms delay, and LFE only with 12.07ms delay:

11. In summary, I time-aligned L+R to subwoofer with a 17ms delay that was applied to L and R, but also to Center, SurroundL, Surround R. With a XO at 40hz, LF content from L and R channels is the slowest in the DSP signal path, so LF content from other channels crossed-over to the sub needs to be delayed to time-match it, and the LFE channel (no crossover) also needs to be delayed. This process aligned L+R to sub and also allows LF content from each channel to arrive at the sub at the same time. Do you agree with the method here?

(I do not think it is best to time align the sub to Center at XO 115hz and surrounds at 110hz, because then the timing of LF content from these channels will arrive at the sub at different times from each other, and at a different time from LF content from L+R.)

One reason for different crossovers is so that a target curve that has some emphasis in bass (Harman 6dB target curve from DecayCore) can be used. I tried 80hz crossover for all speakers and it was difficult to use the Harman 6dB target curve for the center and surround speakers which are physically much smaller than the L and R and need the sub (determined using OCA's great crossover method in video).

edit: correction on gain setting in REW

 

[SoundsGood64]

Correct summation is L + Sub + R + Sub, vector average will count sub only once.
Always generate EP of the measurement, Remove IR delays from the EP and only then compare phase graphs
If you use "Align IRs at cursor" instead of "Align phase at cursor", REW's alignment tool will automatically give you correct sub polarity. Phase alignment cannot invert.

Do you mean that I should generate EP of the measurement and then remove IR delays from EP before I use the Alignment Tool? Or do I do the Alignment Tool without doing this and then do it before I unwrap phase? Doing this versus not doing this before Alignment Tool gives me different results in Alignment Tool. I'm not sure when I should do this so I didn't do it in post 265 above. I don't think I saw this in your video, although perhaps I missed it.

To generate EP of the measurement and then remove IR delays from the EP would this procedure in REW work?
1. All SPL -> IR Windows -> Add 6 cycles FDW to all SPL selections
2. SPL & Phase -> Actions -> Generate Minimum Phase (doing this also generates Excess Phase)
3. All SPL -> Actions -> Remove IR delays

 

[OCA]

Do you mean that I should generate EP of the measurement and then remove IR delays from EP before I use the Alignment Tool? Or do I do the Alignment Tool without doing this and then do it before I unwrap phase? Doing this versus not doing this before Alignment Tool gives me different results in Alignment Tool. I'm not sure when I should do this so I didn't do it in post 265 above. I don't think I saw this in your video, although perhaps I missed it.

To generate EP of the measurement and then remove IR delays from the EP would this procedure in REW work?
1. All SPL -> IR Windows -> Add 6 cycles FDW to all SPL selections
2. SPL & Phase -> Actions -> Generate Minimum Phase (doing this also generates Excess Phase)
3. All SPL -> Actions -> Remove IR delays

Remove IR delays from the summation (aligned sum) to see the result's real EP. Rew's alignment tool has an option to remove IR delays during alignment, too but that's for another purpose.

 

[SoundsGood64]

Hi, I have a series of posts on the DecayCore (FIR generation tool) thread that is about time-aligning subwoofer following generation and execution of FIR for room correction. It seemed to make more sense to put it there but letting you know here for cross-reference in case you are interested:

DecayCore — Free FIR room correction with temporal decay control, automatic optimization and measurement workflow

Hello @VilhoValittu, Thanks so much for putting this together and making it free! Running measurements on Ubuntu 26.04 and getting error / notification "Impulse peak arrived unusually early. Check pre-silence length or latency settings. Detected recording clock drift from chirp spacing (+657...

www.audiosciencereview.com