Home Theater Forum and Systems banner

1 - 20 of 29 Posts

·
Registered
Joined
·
678 Posts
Discussion Starter #1 (Edited)
OVERVIEW

I was looking forward to using the new features of REW v5 to evaluate the distance calculation of my sub. The distance calculation made by my Audyssey-enabled Denon receiver always left questions as to why its times differed from the physical distance, particularly why the calculated sub distance was less than the physical distance.

The best article I’ve seen so far on how to align the sub is “Subwoofer Alignment with a Full Range System”, by Charlie Hughes and published by Syn-Aud-Con, kindly cited here on the REW Forum by Dr Who. This article describes the problems with using the Impulse plot to align the sub, because of the low resolution provided by the low frequencies, and recommends using a Group Delay plot.

REW v5 provides two new features very helpful in aligning the sub distance with the main speakers. There is the Group Delay plot, that shows the relative or absolute arrival times of the impulse at various frequencies. There is the ability to use the Left Channel for timing, which gives an absolute time reference to the Impulse and ETC plots, so that one can compare these across measures of multiple speakers. In my situation, I needed to use both of these, as the alignment exercise was trickier than normal. For my front speakers I have classic Klipsch floorstanding all-horn speakers (LaScalas) so there is a significant delay between the woofers and the tweeters. I chose to align the sub sound with the woofer, as it seemed most important for these two to work together.
 

·
Registered
Joined
·
678 Posts
Discussion Starter #2 (Edited)
SETUP

For the phase and group delay information in the new plots to be accurate, the soundcard calibration files must be rebuilt with REW v5. V5 now includes phase information in the soundcard file, which was not present in earlier versions.

In the Settings -> Analysis screen, you need to check the Use Left as Timing Reference box to use the Impulse plot to compare multiple speakers. You will also need to loopback the left channel on your soundcard from its output back to its input. This is similar to the loopback needed to build the soundcard calibration file, but now it is on the left channel instead of the right and you need to leave it connected all the time. As described in the REW helpfile, this involves one cable in addition to those shown on the REW Interconnect Diagram, looping left to left. The helpfile shows this optional cable in the pictures illustrating the connection to a BFD Pro equalizer.

You can examine the Group Delay plot without using the left channel as timing. If you are using the left channel as timing for impulse curves and are using the first beta build of REW v5, you will need to uncheck the Allow 96 PPO Log Spacing option or update to a later build.

SettingsPPO.jpg
 

·
Registered
Joined
·
678 Posts
Discussion Starter #3 (Edited)
GROUP DELAY PLOT

Unlike the nice smooth plots in Charlie Hughes’s article, when I measured mine the data was not nearly as easy to interpret. The following plot shows the Group Delay view of two measurements of the left front speaker plus sub before this exercise. The red curve represents the sub with the Audyssey calculated distance, 12.6’, and the green measures the result after increasing the distance in the AVR to match the physical distance to the sub at that time, 14.8’.

groupdelaybefore16.jpg

One can see in this curve that the midrange and bass frequencies are delayed about 2 msec relative to the tweeter. The sub might be delayed about 3.8 msec relative to the woofer. More significant for me was the noise in the measurement, making it hard to judge the group delay over the range of each component.

groupdelaybefore12.jpg

Changing the smoothing to ½ octave made it easier to see an edge, but it suggests that the sub is too late by 7 msec relative to the woofer, which is probably excessive. It even indicates at one point that the smaller AVR distance value (green) is appearing earlier than the greater AVR distance value (red).

This noise is to be expected in a complex room environment. Essentially peaks and nulls from reflections and modal resonances cause phase changes that appear in the group delay view, making it hard to isolate differences in distance. For comparison, I took a near field measurement of the front speaker, still indoors, and the 1.7 msec delay from tweeter to midrange is obvious; at the low end room resonances still have a significant effect.

leftfrontnearfield.jpg
 

·
Registered
Joined
·
678 Posts
Discussion Starter #4 (Edited)
BAND-LIMITED IMPULSE PLOT

As described in the article, the problem with using the Impulse plot on the sub is that the limited frequency range of the sub acts as a filter and limits the resolution of the information. To resolve this problem, I came up with the idea to measure the sub and the front speaker over the same frequency range. By using the same frequency range, I would be looking at the data through filters of the same bandwidth, and so the results should look similar if they are aligned in time/distance.

To compare measurements of multiple speakers taken separately, one must use the Left Channel as Timing described above under Setup. As the upper limit of the sweep spectrum is raised, a finer resolution of the curve is obtained. So the first step is to determine how high a frequency can be used. First, I raised the crossover in my receiver to its highest possible value, 250Hz. When I measured the sub, I observed that its frequency response with Audyssey On was flat to about 200Hz. I then changed the AVR to the lowest possible crossover, 40Hz, physically turned off the sub, and measured the frequency response of the left front over 10Hz-200Hz.

subfrontresp.jpg

In this graph, the sub response (green) is even up to 200Hz, so this is a good upper bound to the sweep. Any higher, and the impulse curve of the main will show too much resolution and may have a different character than the sub impulse. The main curve, with a 40Hz crossover, is even from about 35Hz on up to 200Hz. To cover the range shared by these speakers, I chose to examine 50Hz-200Hz.

leftfrontimpulse2.jpg

In this graph, an Impulse plot showing % FS on the vertical axis, the teal is the impulse curve of the left front over its full range, and the green shows what happens when I limited the measurement range to 10Hz-200Hz. The red line shows the impulse response of the sub, located physically very near the left front so it should see roughly the same room response. Its current peak appears 3.06 msec (3.4 ft) after the left front woofer. After increasing the distance value in the AVR by 3.3 ft, the sub’s new impulse curve (red) shows its earlier arrival, now aligned with the left front woofer.

leftfrontimpulse3.jpg

Note that I was able to examine the effect of changing the AVR’s distance to the sub by re-measuring the sub only because it was not the farthest away speaker. The receiver uses the speaker the farthest away as its point of reference, and delays the others to match. Had my sub been farther away than the front speakers, changes to its distance would have been seen as moving the impulse response curves of the front speakers.

Looking at the leading edge of the impulse curves instead of the peak, one calculates almost the same difference in distance. From reading Charlie Hughes’s article and the experiments described below on the effect of the crossover’s low-pass filter on the sub, there is a sound basis for using the leading edge as the point defining the location of the sub. The tricky part is deciding which part of the curve to use as the leading edge, when there is a negative pulse in front of the curve. In the following graph, I measured from the 2% point on the sub curve forward to the center of the leading negative peak of the front woofer and found a difference in distance of 3.2’, very close to the 3.4’ found above measuring peak-to-peak.

leftfrontimpulse2b.jpg

To get the curves to line up like those above, depending on the polarity of the sub when you start, you may need to check the Invert Impulse box in the Controls for the sub’s individual impulse plot. A nice feature in REW v5 is that changes you make in the individual plot are immediately reflected in any Overlay plot displaying the same measure.

The above technique seeks to obtain a reliable measure of the time/distance to the front speaker and the sub by avoiding the effects of the crossover. It does so by using a very low crossover when measuring the front speaker, and a very high crossover when measuring the sub.

I looked at whether a similar result could be obtained measuring narrowly at a selected crossover frequency. As discussed in the next section on the Crossover Filter, its high-pass and low-pass filters have different effects on the two speakers. These make it a challenge to define the correct measurement points for their impulse responses.

leftsub12.2.95-105.jpg

In this graph of the system with the crossover set to 100Hz, the teal is still the left as full-range, the green is the left woofer measured over 95-105Hz, the sub’s response at the original 12.2' distance over 95-105Hz is in red, and the broad sub response, 50-200Hz is in light red. Comparing the sub’s impulse response measured narrowly to the left front, I can only observe that the distance from the sub’s peak to the large negative peak in the left front, 3.3’, tracks well with the distance determined above. That these two peaks are in opposite directions may explain why I needed to invert the sub’s polarity for the best frequency response. That the sub’s peak is very close to the second positive peak in the left front’s response may explain why the 12.2’ distance calculated by Audyssey gave a frequency response about as good as that seen at an adjusted 15.2’.
 

·
Registered
Joined
·
678 Posts
Discussion Starter #5 (Edited)
EFFECT OF CROSSOVER FILTER ON IMPULSE RESPONSES

To interpret what will be seen in the impulse curves, it is useful to experiment and see the results of running sweeps of different ranges through the various components of the system.

First let’s look at simple case of the front speaker configured as Large, so we can later isolate the effect of the crossover on the test spectra.

leftprevvaryingfreq.jpg

Measuring the AVR preout as Large, with no crossover, shows that the impulse response is symmetric around the zero point. As the frequency lowers, the resolution becomes less sharp and the curves broader, from full range (red), through 200-400Hz (orange), 150-200Hz (yellow), 95-105Hz (green), 60-80Hz (blue), and 10-40Hz (purple). There is, perhaps, a slight advance in time in the 95-105Hz curve, of 200 μsec.

When a 100Hz crossover is introduced, the behavior of the impulse response curve of the left preout changes.

leftprevaryingfreqcrossover.jpg

The 100Hz crossover provokes a slight advance in the time of the curves above the crossover frequency, 400-800Hz (orange), 200Hz-400Hz (yellow), 150Hz-200Hz (green), then the curves become more significantly advanced and broader at and below the crossover, 95-105Hz (blue), 60-80Hz (indigo), 10-40Hz (violet). We seem to be seeing two effects here. There is still the broadening of the curves from the lower resolution of the lower frequencies, as seen above when the speaker was Large. There appears a new effect from the crossover, that it is introducing a phase shift and advancing the time, the more the frequency is reduced below the crossover frequency of the high pass filter.

lfevaryingcrossover.jpg

Looking at similar tests of the LFE output shows different behavior below the crossover. In this test run with Audyssey Off giving smooth curves, sub range sweeps, 10-400Hz, were taken varying the crossover, 250Hz (red), 120Hz (orange), 100Hz (yellow), 90Hz (green), 80Hz (blue), 60Hz (indigo), 40Hz (violet). The gray pulse shows a full range measure of the left preout, to establish a point for the starting time. As described in Charlie Hughes’s paper and already seen above, with the lower resolution of the lower frequencies, the curves become broader. The beginnings of all the curves appear to start at almost the same instant, 1.0-1.5msec before the pulse from the left channel preout.

lfetimevaryingfreq.jpg

This graph shows the impulse curves of the LFE output from the AVR of a sub configured 1.7’ (1.5 msec) closer than the front speaker. If we look at a typical crossover frequency, 100Hz, it is interesting to compare the curves as the sweep range is varied, 10-400Hz (red), 150-200Hz (orange), 95-105Hz (yellow), 50-80Hz (green). Here one sees that the 50-80Hz curve is delayed later, the 95-105Hz yellow curve has a similar peak to the full range, but has a leading negative peak, and the 150-200Hz range (orange) above the crossover looks as if its phase is inverted, with a full negative peak before the positive peak. The center of the negative peak of the narrow range around the crossover, 95Hz-105Hz, matches very closely the initial time of the full range curve.

groupdelaylfevsleftpre.jpg

Comparing the group delay curves of the left preout to the LFEout with both speakers configured to the same distance in the receiver confirms the impression from the impulse curves above, that the lower frequencies are delayed relative to the higher frequencies. Here the blue curves are the overall LFEOut and the narrow ranges used above, and the red curves are the left preout, full range and narrow sweeps. Well above the 100Hz crossover, the curves show the same 4.7 msec processing delay in the receiver. But as the frequency decreases near, through, and below the crossover, the crossover introduces delay. Because the Denon crossover is 24dB/octave on the low-pass side to the sub, and 12dB/octave on the high-pass side, the LFEout shows twice as much delay as does the left preout. There are two important implications for setting the sub distance: (1) the front speaker itself appears at a different distance at the crossover than it does at the other frequencies, and (2) even if the sub distance is set to match perfectly at the crossover, its distance will only be close at the frequencies above and below the crossover.

leftprevsleft.jpg

In this graph, I compare the Left preout to the left speaker measured in the room. Particularly interesting is how the narrow measure 95-105Hz at the crossover changes. The leading positive peak does not appear as large, the negative peak is just as large, while the second positive peak now appears larger than the first. This illustrates why narrow measures at the crossover are so hard to interpret.
 

·
Registered
Joined
·
678 Posts
Discussion Starter #6 (Edited)
CONCLUSION

The Group Delay plot after aligning the sub appears to show the sub frequency range more-or-less even with the woofer/midrange. This confirms that the band-limited impulse approach used above had the intended result.

groupdelayafter16.jpg

After adjusting the sub time/distance setting in the receiver, one must go back and review the polarity of the sub and the choice of crossover frequency. I looked at the frequency response curve of the (left +sub) and (right + sub) pairs separately, seeing which polarity and crossover frequencies worked with the left, then choosing among these which worked best with the right. Only after checking each front speaker separately with the sub, did I look at the overall response of both driven together.

After swapping the polarity of the sub, one can see from the phase curves why the corrected 15.2’ distance works a little better than the original 12.2’ distance. Looking at the phase curves over the octave either side of the 100Hz crossover, one can see that the new distance (blue) stays between the green and red curves over most of the range. The sub curve at the original distance (teal) gets more out of phase sooner below 100Hz; its phase is changing more rapidly because its audio is delayed and seems farther away.

phasefrontvssub.jpg

I can't say I can hear a difference, but the total adjustment I made relative to the Audyssey calculated distance, +3.1', is below the threshold of what is considered audible. Correcting the sub distance gave me a clearer choice of sub polarity and crossover frequency, and a better overall response curve. That is audible, and I'm happy about that.
 

·
Premium Member
Joined
·
6,711 Posts
Bill, the band limited impulse measurements is an interesting idea, looks to work quite well. I'm not clear what you are measuring with the deltas in those group delay plots, though. In-room measurements in a small space (e.g. HT room versus stadium) will result in GD plots with lots of variation, but comparisons between plots need to be made at specific frequencies - for example, in the 12.6' versus 14.8' measurements the GD should show a difference of about 2ms at any frequency where the sub's output dominates, say below 50Hz or so. To use GD to compare the sub's timing with the main you really need separate measurements of sub only and main only and to make comparisons of their GD's around the crossover frequency.
 

·
Registered
Joined
·
678 Posts
Discussion Starter #9 (Edited)
The deltas in the group delay plots were just arbitrary frequencies compared against what seemed to be the leading edge of the woofer.

I was hoping to see a broad trend in the group delay plot, but as you say the issues with a small room dashed that hope. Being able to look at a broad trend, if that were smooth without so much noise in the data, would have been a nice technique to recommend for people whose mains don't have a wide range of overlap with the sub.

I did look at the Group Delay of the sub compared to a measurement of the main just over the range 10Hz-200Hz. Even over that limited range, there was so much variation between different frequencies that there was no obvious point at which to compare them. Depending on which frequency I picked to look at, the sub was significantly early, significantly late, or just on time.

I didn't think to take a sweep just around the crossover to see if that yielded a more stable measure. I may play with that and see what it says. When I started this exercise, I was thinking more of synchronizing the sub with the woofer for any improvements one might hear in music by eliminating the time difference. You may be right, though, that focusing on the crossover could yield a better result, by ensuring no interference between the sub and the woofer and improving the overall response.

Another idea for the statistically inclined would be to export the data, calculate a mean group delay over the interesting range, say 60Hz-120Hz, and see if comparing these would yield the same recommended change in time/distance as does aligning the impulses.
 

·
Premium Member
Joined
·
6,711 Posts
What would you recommend for time aligning a subwoofer to mains?
I like Bill's idea of band limited measurements of sub and mains over the range both operate. Another option would be to make measurements at the line level outputs to sub and mains to accurately characterise the delays in processors etc, free of any room influences, then add in the delays from the relative distances of the speakers/sub to the listening position. An option for fine tuning is to set the sub to opposite polarity and tweak the delay adjustment to get the largest dip at the crossover frequency, then flip the sub back to correct polarity, though that is subject to room influences on the measurement which may make the evaluation difficult.
 

·
Elite Shackster
Joined
·
7,142 Posts
I am wandering how adjusting the time delay in your AVR, after Audyssey is run, affects what Audyssey is trying to do. I'm all for improving your system, and your results sound positive, but if you manually change settings after Audyssey stores its settings, your then basically altering the response Audyssey is trying to base its calculations on.

This is one of those things that would niggle away at me, even if I cant hear the difference, at least for a while. Ive always been an advocate of helping Audyssey as much as possible to get the very best results, so I really wander how this figures into the equation.
 

·
Registered
Joined
·
493 Posts
I like Bill's idea of band limited measurements of sub and mains over the range both operate. Another option would be to make measurements at the line level outputs to sub and mains to accurately characterise the delays in processors etc, free of any room influences, then add in the delays from the relative distances of the speakers/sub to the listening position. An option for fine tuning is to set the sub to opposite polarity and tweak the delay adjustment to get the largest dip at the crossover frequency, then flip the sub back to correct polarity, though that is subject to room influences on the measurement which may make the evaluation difficult.

I've basically been adjusting distance in my prepro to peak out the crossover region also. Inverting the subwoofer signal gave me the same results.

One tweak that I'd like to see in REW is the ability to send out a DD 5.1 signal where the user can select which speakers to turn on/off so we can tweak distances in the other channels. For example, the user could turn off all but the left and right main, run the signal generator and adjust distances to peak spl. Then turn off the left channel and turn on the center and repeat the test. So on and so forth with the other speakers.
 

·
Registered
Joined
·
678 Posts
Discussion Starter #14 (Edited)
I am wandering how adjusting the time delay in your AVR, after Audyssey is run, affects what Audyssey is trying to do. I'm all for improving your system, and your results sound positive, but if you manually change settings after Audyssey stores its settings, your then basically altering the response Audyssey is trying to base its calculations on.

This is one of those things that would niggle away at me, even if I cant hear the difference, at least for a while. Ive always been an advocate of helping Audyssey as much as possible to get the very best results, so I really wander how this figures into the equation.
Unless you are using an Audyssey Pro system, that actually looks at how the satellites merge with the sub and calculates the results of each choice of crossover, this doesn't impact the Audyssey did it's equalization at all. Audyssey generates filters to equalize each speaker individually to 75dB up to the points where the speaker rolls off. Time/distance alignment of the speakers doesn't affect that, Audyssey still runs, although it might make the little square or green light turn off that indicates you are using the original Audyssey values. Also, especially for the sub, Chris recognizes that Audyssey's distance calculation is often off, hence the advice when it's too low to set it to the physical distance. With REW v5 we now have a measurement tool to make it right whether the Audyssey calculated distance was too low or too high.

I actually took advantage of this in the band limited impulse plot. Because Audyssey equalized the sub and the mains, the impulse curve, generated just over the range where both can respond, should look similar. With Audyssey equalization, the sub's range is extended as high as possible, to give the best available resolution. I intend to add a plot showing how I picked 200Hz for the upper end, and why I should have used a higher lower end, e.g., 30Hz or 40Hz for the sub impulse to match the main's range.

It is important to have Audyssey On when taking any of the measures for estimating distance. With the left channel as timing, one can see the delay from enabling the Audyssey processing. In my Denon, I'm seeing about 9 msec (10') difference between the impulse curves of Audyssey Off and On.
 

·
Registered
Joined
·
678 Posts
Discussion Starter #15
I am wandering how adjusting the time delay in your AVR, after Audyssey is run, affects what Audyssey is trying to do. ...
As promised, I updated the description to include a freq resp curve showing why I stopped the impulse sweep at 200Hz.

Just for you, Dan, I also included a comparison showing the initial Audyssey chosen distance.
 

·
Elite Shackster
Joined
·
7,142 Posts
Cheers Laser :T

Ive just got in from work an hour ago, and I'm just settling down for the night, I'll sink my teeth into into it in a little while and have a think.
 

·
Registered
Joined
·
678 Posts
Discussion Starter #17
I've expanded my description to reflect later experiments I made investigating the effect of the crossover in the AVR.

I've basically been adjusting distance in my prepro to peak out the crossover region also. Inverting the subwoofer signal gave me the same results.
...
Of course you are right, had I started with the physical distance to the sub and stepped its distance upward, I would have eventually found a distance setting that gave the best frequency response.

I've seen comments elsewhere in the forum, perhaps yours, which described how someone adjusted the sub distance over a wide range looking for the best frequency response. I was always reluctant to do this, both because of the time involved and because frequency response is only one criteria, the absolute time relationship matters too. So I did not follow this route until v5 gave us tools to look at the timing.

I did some more experiments to see if a narrow sweep range around the crossover could give a starting distance as good or better than I determined with my first technique. This proved not as satisfactory as using a range 50Hz-200Hz while moving the crossover filter out of the way.

... Another option would be to make measurements at the line level outputs to sub and mains to accurately characterise the delays in processors etc, free of any room influences, then add in the delays from the relative distances of the speakers/sub to the listening position. ...
I have made some more measurements of the line level outputs from the AVR, some of which I've included in the article. For me, the physical distance of the speakers to the listening position is itself problematic when considering an all-horn floor standing speaker, where the woofer faces backward to use the back of the cabinet to extend the horn. Add to that, there is some delay in the custom crossovers in the front speakers.

Bill
 

·
Registered
Joined
·
678 Posts
Discussion Starter #18
After looking at impulse responses over narrow ranges to see the effect of the receiver's crossover at different frequencies, it occurred to me that one should see the same thing in the group delay curve. So I've added that graph, confirming what crossover design would lead one to expect.

Bill
 

·
Registered
Joined
·
2,604 Posts
Somehow I just found the time to read this thread fo rthe first time... thanks for the great work Bill.
But I'm almost thoroughly confused... so please correct me if I'm wrong...
The sweeps you took to generate the two IRs you used to measure your delay were both done from 50Hz-200Hz, right? One was the Left speaker set with a 40Hz Xover in place, and the other was the sub with a 250Hz xover in place.

Right?
 

·
Registered
Joined
·
678 Posts
Discussion Starter #20 (Edited)
Correct, Greg.

Because of how the range of the sweep affects the shape of the impulse curve, essentially the rise time at the start of the curve, looking at both speakers over the same frequency range lets one try an apples-and-apples comparison of the relative timing of the curves.
 
1 - 20 of 29 Posts
Top