Home Theater Forum and Systems banner

1 - 13 of 13 Posts

·
Registered
Joined
·
27 Posts
Discussion Starter #1
I always thought the wavelengths for say under 80Hz were to large to measure in order to determine the distance for a subwoofer using REW or most any other program.

Can the REW distance measurements be used to time/phase align subwoofers?
 

·
Premium Member
Joined
·
324 Posts
You can measure the time it takes for the pulse of sound to hit the mic and calculate the distance that way. Wile some frequency are a larger wave they still move the same speed.
 

·
Premium Member
Joined
·
1,562 Posts
A phazewolf indicates, REW can accurately measure the relative phase between SWs and also between SWs and main speakers. It is necessary use the REW loopback feature to determine relative timing between speakers.

Be advised:
> A simple tape measurement is adequate to closely time align multiple identical SWs.
> It is more important to have smooth/flat SPL response through the XO range than closely matched phase response IMO.
> If the the SWs/mains phase is relatively close in the XO range EQ can be implemented to smooth the response and the speakers will not be overly taxed to provide the needed SPL output. A timing error of 2 ft or less for a 100 Hz XO is often adequate for good performance.
 

·
Registered
Joined
·
27 Posts
Discussion Starter #4
Thanks for the replies.

However, if the wavelength is say 14 feet long (80Hz) I do not understand what part of the wavelength REW uses to determine the distance for the sub. The distance could vary widely depending on what part of the signal REW is using.
 

·
Premium Member
Joined
·
1,562 Posts
REW provides the timing/phase of all frequencies a speaker produces relative to the loopback timing signal. The loopback signal is consistent in timing so the differences in timing between 2 or more speaker measurements can be compared.

It's easy enoughto prove the accuracy. Just measure the same speaker over and over again and see the the impulse is always located at exactly the same timing. Move the mic position a known distance and see the impulse move accordingly.
 

·
Premium Member
Joined
·
129 Posts
I'm quite puzzled by my own subwoofer impulse measurement, and I have loopback working and it is working properly. When I measure the full range measurement, or the satellites without the sub, I get a time of flight of 9.52 feet, which corresponds very well to the loudspeaker-microphone distance plus a bit of latency in the interface. But when I take the measurement of the subwoofer alone, the time of flight is 20.14 feet! Which we all know is wrong.

The subs and the mains are currently all crossed over by an analog crossover, there is only one D/A converter so we can't blame latency or software on this. It is strictly either an REW issue, or simply some phenomenon related to subwoofer measurement. I have never been able to measure time of flight accurately from a band-limited sub, and I believe that the impulse is not short enough to provide a proper trigger for the measurement hardware. My method of integrating a subwoofer has always been to feed wideband noise or a sweep, and adjust the woofer's phase control (if there is one) and/or the woofer's position in the room to minimize peaks and dips and/or obvious comb-filter artifacts at the crossover frequency. This is very accurate from my experience.

The RTA in REW, set to very few averages, has enough resolution to make this measurement and adjustment possible in real time. When I convert to JRiver and AudioLense (thanks, Mitch!) I'll see if what that system does about analyzing subwoofer position.
 

·
Premium Member
Joined
·
1,562 Posts
Bob,
There may be an issue or there may not. The "Time of flight" or delay is a function of frequency. The SW to Main timing is not easily evaluated by looking just at the IR peaks. The conventional wisdom is to adjust the timing such that the SW and Main provide close phase tracking throughout the XO range.

If you post a measurement of the SW, Main, and SW+Main I will review it and comment in more detail.
 

·
Premium Member
Joined
·
129 Posts
Bob,
There may be an issue or there may not. The "Time of flight" or delay is a function of frequency. The SW to Main timing is not easily evaluated by looking just at the IR peaks. The conventional wisdom is to adjust the timing such that the SW and Main provide close phase tracking throughout the XO range.

If you post a measurement of the SW, Main, and SW+Main I will review it and comment in more detail.

Well, who can argue with that! Thanks for the offer. Attached are the measurements, Left front full range, Left front without the left sub, and left sub alone. Crossover is about 60 Hz, all done in the analog domain.
 

Attachments

·
Premium Member
Joined
·
1,562 Posts
Bob,
I see your concern is valid.
The SW is excessively delayed according to the files you provided. This is not good timing for any philosophy regarding proper XO timing alignment. Of course if the timing was somehow corrupted by your setup/calibration then all the following comments are not relevant.

[There was a 2 dB SPL offset between the SW and Mains as indicated by the SW+Mains. I just ignored this for these comments as an SPL offset does not impact any comments regarding the delay/phase.]

[To be safe, I normally sweep full range for SW and Main as this provides REW a nice sharp loopback IR signal to set the timing on. I have not seen any error even when reducing the SW sweep range, but it is safer to use the full range measurement. It is pretty safe to assume this was no issue here.]

Your XO is very steep (high order) such that the XO range is very narrow (about 55-85 Hz) and thus the phase matching is not overly critical as it impacts the SPL in only a very narrow range. The more important issue is that the delay/timing is reasonable at the XO.
spl xo range.jpg

The current delays provide the following phase tracking through the XO range. It is okay for phase at 60 Hz, but the phase tracking through the XO is not the best and the SW IR delay relative to the Main is excessive.
[Note that I offset both the SW and Main measurement IRs by 8.475 ms (the offset of the peak of the mains). This is necessary to best evaluate the phase timing around the 60 Hz XO.]
original phase.jpg
original timing.jpg


Option 1:
Reducing the delay of the SW (or increasing the delay of the Main) by 14 ms And inverting the polarity of the SW we get:
14 ms Phase.jpg
14 ms inverted timing.jpg


This provides better phase agreement through the entire XO range and brings the SW timing more in line with conventional practice.


Option 2:
Reducing the delay of the SW (or increasing the delay of the Main) by 22 ms we get:
sw phase shifted 22 ms.jpg
sw timing shifted 22 ms 2.jpg


This provides better phase agreement through the entire XO range and brings the SW timing more in line with the Main. With this alignment the GD will be minimized compared to the other 2 alignments.

Because of the very steep XO the impact on SPL through the XO range is minor with any of these alignments.

I can provide more information regading the process used as needed.

I hope this helps.
 

·
Premium Member
Joined
·
129 Posts
Dear John: Yes, your incredibly detailed help is incredibly valuable! Yes it's obvious that because of the steep crossover that I'm using, the error is pretty minimal. However, I personally did not notice a little audible delay but Mike Chaffee (acoustical consultant) felt that there is a bit of an audible discontinuity between the woofers and the mains, and then I did notice it on some material. So there is some room for work there. This is nit-picking of the highest order, as you know, because this system still sounds fantastic and very accurate... I'm just trying to take it to the next level of performance.

The so-called 2 dB SPL offset you notice is in my opinion a misinterpretation on your part. I never use the SPL measurement reported in REW or any analysis tool as the basis for adjusting the sub level; this is just not a useful tool. For one thing, SPL of a narrow bandwidth signal is going to always be different. Anyway, first I crerate a nominal 0 dB line in 1/1 octave smoothing mode setting the midrange to center around that line using REW's vertical offset adjustment. That's partly an art of course because "what is the midrange", but this method gets the bass very accurate and rarely is a further tweak necessary, no more than 1/2 dB of sub level variance. Anyway, that becomes the 0 dB reference. Then using 1/6 octave smoothing I set the subwoofer level so the positive and negative extremes of the subwoofer output are centered on the 0 dB line, and the result is always quite correct. That's my method, at least.

In addition, the splice between the non-filtered satellite speakers and the sub is level-perfect, you can tell that the sub level must not be offset because you can see the accuracy of the splice in the combined measurement.

The main question, however, is how the hell did the subwoofer get so delayed? To repeat, there is only one D/A converter in my current system and all the frequency dividing and summing is performed in the analog domain. Perhaps it's due to the extremely steep slope of the low pass filter, which is done by two 24-dB/octave filters in series? But in my wildest imagination I can't imagine that causing an apparent 10 ms. or more offset between mains and sub... perhaps some technical expert on analog filters can respond here to that conundrum, but let's move on and respond to your wonderful detailed response! Please forgive the excess quoting and try to find my questions below.

Bob,
I see your concern is valid.
The SW is excessively delayed according to the files you provided. This is not good timing for any philosophy regarding proper XO timing alignment. Of course if the timing was somehow corrupted by your setup/calibration then all the following comments are not relevant.

[There was a 2 dB SPL offset between the SW and Mains as indicated by the SW+Mains. I just ignored this for these comments as an SPL offset does not impact any comments regarding the delay/phase.]

[To be safe, I normally sweep full range for SW and Main as this provides REW a nice sharp loopback IR signal to set the timing on. I have not seen any error even when reducing the SW sweep range, but it is safer to use the full range measurement. It is pretty safe to assume this was no issue here.]
It's interesting to note that when I did sweep the sub full range, there was some tiny analog leakage in my system into the other channel (barely audible, could have been 60 dB down), and there was enough high frequency leakage into the other channel so that the impulse measurement of the sub showed up as identical timing to the mains! In other words, REW got a "clue" from the slight high frequency leakage that made it think the sub was timed the same as the mains; I saw a narrow initial impulse in the sub display, which I intuitively knew as wrong. But when I muted the other channel, then the sub time delay measurement returned to its extreme that we see in these measurements.

Interesting about the usefulness of the sharp loopback signal, I have to take another measurement muting the other channel in my analog system, but anyway, I don't think that's relevant to the discussion, I believe I had a good loopback timing reference for these measurements. Even when I sweeped the sub full range and when I had the HF leakage in the other channel, you can clearly see where the sub impulse begins, 10 or more ms. after the initial short impulse.

Your XO is very steep (high order) such that the XO range is very narrow (about 55-85 Hz) and thus the phase matching is not overly critical as it impacts the SPL in only a very narrow range. The more important issue is that the delay/timing is reasonable at the XO.
View attachment 39832

The current delays provide the following phase tracking through the XO range. It is okay for phase at 60 Hz, but the phase tracking through the XO is not the best and the SW IR delay relative to the Main is excessive.
[Note that I offset both the SW and Main measurement IRs by 8.475 ms (the offset of the peak of the mains). This is necessary to best evaluate the phase timing around the 60 Hz XO.]
Did you make the timing offset using the "estimate IR delay/shift IR" settings in REW? I don't see any other way to shift that. I also don't see any way to reverse that "shift IR" setting once you perform it, and there are times when I would want to see the genuine time of flight... ???

Option 1:
Reducing the delay of the SW (or increasing the delay of the Main) by 14 ms And inverting the polarity of the SW we get:

This provides better phase agreement through the entire XO range and brings the SW timing more in line with conventional practice.
John, please help me find where you make that delay change in REW. I can't seem to find it.

Option 2:
Reducing the delay of the SW (or increasing the delay of the Main) by 22 ms we get:

This provides better phase agreement through the entire XO range and brings the SW timing more in line with the Main. With this alignment the GD will be minimized compared to the other 2 alignments.

Because of the very steep XO the impact on SPL through the XO range is minor with any of these alignments.

I can provide more information regading the process used as needed.

I hope this helps.
Thanks, yes, it's a great start. Again, it would be great to find out why the big subwoofer delay. Maybe this weekend I'll bypass both lowpass filters and take another measurement and see if the excessive delay goes away. Well, my plan is to learn how to use AudioLense first on this system, within the limitations of having one DAC. Then to move to multiple DACs, a digital crossover and a dedicated HTPC to drive the whole system. That's going to be a big move, and since this is a mastering studio doing actual production work, I have to do it slowly and carefully!
 

·
Premium Member
Joined
·
1,562 Posts


The so-called 2 dB SPL offset you notice is in my opinion a misinterpretation on your part. I never use the SPL measurement reported in REW or any analysis tool as the basis for adjusting the sub level; this is just not a useful tool. For one thing, SPL of a narrow bandwidth signal is going to always be different. Anyway, first I crerate a nominal 0 dB line in 1/1 octave smoothing mode setting the midrange to center around that line using REW's vertical offset adjustment. That's partly an art of course because "what is the midrange", but this method gets the bass very accurate and rarely is a further tweak necessary, no more than 1/2 dB of sub level variance. Anyway, that becomes the 0 dB reference. Then using 1/6 octave smoothing I set the subwoofer level so the positive and negative extremes of the subwoofer output are centered on the 0 dB line, and the result is always quite correct. That's my method, at least.

In addition, the splice between the non-filtered satellite speakers and the sub is level-perfect, you can tell that the sub level must not be offset because you can see the accuracy of the splice in the combined measurement.
My comment was only to point out that for some reason the level of the SW+Main was somehow shifted -2 dB in the measurements you provided as shown below. Something must have changed the level of that measurement. [I was not commenting on the chosen relative level of the SW and Main.] A shift like this will not happen if the only change is to apply appropriate muting. It was just an observation FYI, one that does not affect the analysis.
spl shift.jpg

The main question, however, is how the hell did the subwoofer get so delayed? To repeat, there is only one D/A converter in my current system and all the frequency dividing and summing is performed in the analog domain. Perhaps it's due to the extremely steep slope of the low pass filter, which is done by two 24-dB/octave filters in series? But in my wildest imagination I can't imagine that causing an apparent 10 ms. or more offset between mains and sub... perhaps some technical expert on analog filters can respond here to that conundrum, but let's move on and respond to your wonderful detailed response! Please forgive the excess quoting and try to find my questions below.
Higher Order minimum phase XO filters cause more total phase rotation and more GD between high and low freq. With this steep XO I am not surprised by the delay shown here, but I am not able to confirm if it is entirely due to the filters.


Did you make the timing offset using the "estimate IR delay/shift IR" settings in REW? I don't see any other way to shift that. I also don't see any way to reverse that "shift IR" setting once you perform it, and there are times when I would want to see the genuine time of flight... ???
See below the “Offset” box. I just measured the 8.475 ms offset of the FL Main IR peak and entered that value into the offset box. I applied that SAME 8.475 ms offset to each of the 3 measurements to keep the same relative timing.
ir offset.jpg
 

·
Premium Member
Joined
·
129 Posts
Silly me, John, I should have seen the horizontal offset in that box, missed it...

Thanks,


Bob
 
1 - 13 of 13 Posts
Top