Time align for sub using impulse and spectrogram plots

[Greg Dunn]

I'm trying to set up time parameters for my sub. I've been able to look at the frequency response plots to make sure the phase is correct, but trying to determine time delay is giving me vague results.

First of all, I measured the distances to the sub and mains; due to circumstances beyond my control, the sub is slightly closer and off to one side. The mains are about 12 feet away, 4 feet in front of the wall, and the sub is roughly 7 feet away, closer to the right wall. Changing this layout is physically almost impossible in the room I have, but with some effort I have gotten the frequency reponse to work out well; it is crossed over at 60 Hz so localization should be (and appears to be) a non-issue. But due to the configuration, I need to add delay to the sub instead of the reverse. My measurements told me about 4.5 mS of added delay (5' difference in the distances). The raw data from the REW info panel says about 2.0 mS though (possibly due to differences in the DSP channels?). Audibly, I think the 4.5 mS sounds a bit cleaner - better transients. But I accept that might be confirmation bias, so I decided to do a full FR/impulse plot with REW and see whether it agreed with me.

The problem is that using sub delays of 0, 2, 4.5, and 6 mS seem to generate plots which are only subtly different and none of them look "better" to me. I don't see a clear contribution from the sub, though there is a 'bump' in the step response which moves as I change the delay and it's way beyond the delay parameter - on the order of 15 mS - so it's probably a reflection. Rather than clutter up this thread with 50 ad-hoc different graphs, what should I post to give the most useful data, and with what settings to make the data clearer? I have 4 full sets of data, so I can post pretty much anything. Impulse, filtered IR, Spectrogram, waterfall.... What would you like to see? I'm not busy this weekend so I'll be able to test and re-post if desired.

 

[EarlK]

The member who has the best grip on synching subs is ( IMHO ) jtalden .

My advice is to search out threads where he has helped others synch multiple subs.

 

[Greg Dunn]

Trying to absorb some of the info. Meanwhile, here's an overlay containing impulse responses of L, R, and sub (the blue line). The setup has a 4 mS delay already applied, and interestingly, it puts one of the peaks in fair alignment with the L/R peak. But there are other peaks way out there, and I'm not sure if they're room modes or what. They're certainly too far to be direct responses.

 

[John Mulcahy]

You should be aligning the start of the response, not the peak. Better to post the mdat than screenshots.

 

[Greg Dunn]

I'm having trouble identifying the sub response in the measurement; the capture above was an (obviously failed) attempt to try to measure it separately on the same timebase.

Here's the mdat containing only the full sweep with the sub having 0 delay.

 

[jtalden]

Hmm, offsetting the timing using distances usually works pretty well. Sometimes it is necessary to invert the polarity of the sub. Some use the REW RTA feature while making a small adjustments of distance, or while adjusting the phase control on the sub. Very acceptable results are achieved this way. I use a different method but it requires specific measurements.

If you would like me to confirm your current settings or recommend a distance/timing change, please provide an mdat file containing 3 or 5 measurements:

Setup for all measurements
> Use 20-20k sweeps
> Use acoustic timing (optional- not needed if you don't know how)
> Mic located at the LP
Measure
1 Sub (alone)
2 Left (alone)
3 Right (alone)
4 Sub with Left
5 Sub with Right

Notes:
> Measurements 4 and 5 are not needed if acoustic timing is properly implemented and applied to the first 3 measurements.
> It may be necessary to provide 2 mdat files due to file size? If so measurement 1, 2, 3 and 4, 5 can be in separate files.

 

[Greg Dunn]

Audibly, using measured distances seems to work OK - but the data I'm getting from REW doesn't seem to show a clear 'winner' in delay settings. My hope here was to understand what I'm seeing in REW so that for future use I can dispense with less analytical procedures and get something repeatable.

I am in the middle of repairing a garage door opener at the moment but will run those sweeps today ASAP and post the mdat file(s) here!

 

[jtalden]

I like to align the timing of the direct sound such that the there is a close phase tracking relationship of the sub and mains. Unfortunately, due to room effects that does not necessarily result in a similarly coherent SPL and phase relationship at the LP. There are even often very significant differences between the left and right channels. I don't think it is helpful to depart significantly more than 1/2 wavelength from this 'Direct sound' setting as group delay may become an issue, but if a smoother SPL at the LP is found within this range I see no issue using that setting.

 

[Greg Dunn]

OK, dumb question: How do I set REW to do a 20-20K sweep and only measure the sub? Doesn't the timing reference have to come from one of the mains? Then the test signal will come through the main speaker during the test as well, unless there's a way to disable it after the timing reference signal is output.

 

[jtalden]

If you are not using HDMI for output and thus have no easy way around that problem just measure without acoustic timing. Given all 5 measurements, I still have a way to determine the current relative timing. It doesn't matter if you use 0 ms, 4.5 ms, or something else for sub delay so long as all 5 measurements are the same. Any sub timing change I recommend will be relative to the setting you actually used. You may want use the one you have chosen just to see if it compares well to my suggestion.

 

[Greg Dunn]

OK, here they are - file names + notes should be clear on which measurements are inside. This uses the 4.5 mS sub delay, no delays on the mains at all.

 

[jtalden]

That looks good for analysis. I will work on it.

 

[jtalden]

The delay setting of 4.5 ms as measured looks very good. The LP phase a 60 Hz is aligned and the phase tracking throughout the XO range is reasonably good.

For comparison I changed the delay -4 ms (to 0.5 ms) and found closer delay tracking for the direct sound. With this setting the phase alignment at the LP is improved above and below the 60 Hz XO. It is worse however near the XO due to room effects.
The 0.5 ms delay setting SPL is thus not as good at 60 Hz but is better below and above that. See the charts below.
Note that the SPL differences are minor for both of these settings so it is a tradeoff. In fact, as you apparently found, I would expect there is not much difference in SPL and sound quality with any setting between these 2 values. This doesn't mean you will not hear a difference and have a preference for a setting in this range. It is just difficult to say one is theoretically better than the other.

 

[Greg Dunn]

Fascinating comparison. It "seems" that I'm more sensitive to phase shift at the crossover frequency, at least with the SW off to one side of the LP. That would explain why some delay appears to sound "better" than no delay. It's possibly source material dependent, too, because some music benefits more from the delay than others. With the way modern recordings are mixed, that would be understandable.

I'm happy with this analysis, though I will probably try a listening session with the delay shortened to <1 mS to see if there's any perceptible difference.

 

[Greg Dunn]

Just a quick follow-up: I have adjusted the time delay over a larger range and somewhere in the 4-5 mS region seems to work well for my ears. I'm comfortable that measurement, analysis and listening have converged pretty well for this, and I thank everyone for their contributions!

 

[jtalden]