REW Alignment tool - Guides or Manual?

[jtalden]

I analyzed both 80 and 100 Hz XOs. The SW XO timing of both was already at the most favorable setting. It would be the conventional target for this setup. There was no indication that FL+FR sum was showing that large dip around 65-100 Hz that we were trying to avoid. Even though the SPL is now starting to rolloff in that range, I expected to again see a clear evidence of the dip in the FL+FR sum. This makes me wonder if there may have been a problem with the measurements analyzed in Post-11 that impacted the phase in that area. Possibly a measurement error created a red herring in the analysis.

Since there was no advantage in using 100 Hz XO the following with respect to the 80 Hz XO data. Since the current timing is already the most favorable for SPL and you commented on hearing differences in the bass with different setups, I identified 2 additional options that also provide favorable SPL.

1. Use the current conventional timing (Trace 6)
2. Reduce the SW delay by 13 ms (Trace 9)
3. Reduce the SW delay by 7.5 ms and invert the polarity of the SW (Trace 12)

There may be a noticeable difference in these even though the SPL is very close - Just options in case you like to experiment. Option 3 with the inverted SW has made a noticeable difference in other systems. The chart below shows the relative SPL difference with an expanded dB scale so the differences can be more easily seen. I did not show other charts as they are not easily interpreted and will just create more confusion. I did not detail the alignment tool methodology used with this data. It is one of the more difficult ones to interpret and would only cause frustration. I will follow next with a simple case. That is the place to start.

 

[jtalden]

Here are some practice files for use with the REW alignment tool.
1. The 'ideal...' file has an idealized 80 Hz high pass and lowpass filter. The 2 impulses are already properly time aligned. The alignment tool was used to create the 'aligned sum'. It is good to see the impact on the SPL, Impulse, phase, group delay, step response and other charts with an idealized LR-24 XO.
2. The 'Offset...' file has the same 2 original measurements, but the relative timing is not ideal so you can practice using the alignment tool controls to recreate the ideal timing as was already generated in the 1st file.
3. The 'Real...' file has a more typical situation using real measurement data. See if you can find the best phase tracking situation in this relatively easy case.

Depending on the measurement data it can be very difficult to analyze some setups. Using some smoothing can often help. Applying an FDW can help understand the phase trend of direct sound. I like to stay as close to that timing as possible, but still optimize the SPL without the FDW applied. A timing deviation of greater 1 wavelength would best be avoided.

 

[JStewart]

I have been working on the timing. There is unusual timing problem between the FR and LF speakers. It most easily shows up in the measurement notes with the FL delay = -0.0089 ms and the FR delay = -5.0457 ms. A 5 ms difference makes no sense. The Lexicon chart in Post-12 shows more expected delay setting differences for a mic at the LP (2.88 and 2.74 ms). This also makes the accuracy SW timing suspect.

Did you get to the bottom of this? If not the bottom screen shot in post 12 may hold a clue. It appears that manual delays can be input on this device even with Dirac delays set. Not sure of the order of operations in the DSP but the difference between the FL & FR manual delays is 5ms. Coincidence?

 

[lexicon]

I analyzed both 80 and 100 Hz XOs. The SW XO timing of both was already at the most favorable setting. It would be the conventional target for this setup. There was no indication that FL+FR sum was showing that large dip around 65-100 Hz that we were trying to avoid. Even though the SPL is now starting to rolloff in that range, I expected to again see a clear evidence of the dip in the FL+FR sum. This makes me wonder if there may have been a problem with the measurements analyzed in Post-11 that impacted the phase in that area. Possibly a measurement error created a red herring in the analysis.

Since there was no advantage in using 100 Hz XO the following with respect to the 80 Hz XO data. Since the current timing is already the most favorable for SPL and you commented on hearing differences in the bass with different setups, I identified 2 additional options that also provide favorable SPL.

1. Use the current conventional timing (Trace 6)
2. Reduce the SW delay by 13 ms (Trace 9)
3. Reduce the SW delay by 7.5 ms and invert the polarity of the SW (Trace 12)

There may be a noticeable difference in these even though the SPL is very close - Just options in case you like to experiment. Option 3 with the inverted SW has made a noticeable difference in other systems. The chart below shows the relative SPL difference with an expanded dB scale so the differences can be more easily seen. I did not show other charts as they are not easily interpreted and will just create more confusion. I did not detail the alignment tool methodology used with this data. It is one of the more difficult ones to interpret and would only cause frustration. I will follow next with a simple case. That is the place to start.
View attachment 31180

Many thanks again for this. I assume that the 5ms error was not present this time?

As regards the 40hz measurement with the dip, I'm not sure that was with the new BM function applied. I'm wondering if that's exactly what problem that system is trying to address as of course both the 80+100 were with that switched on. To prove that theory, I have done a 40 (mains) + 80 (remainder) profile which I will load and measure tomorrow. I can then just about manage to run the trace arithmatic and process to see if that dip remains. If it is gone I will give that profile another try.

At the moment the sub is set at 0ms delay by Dirac (usually now as default) with all other speakers delayed from there. Are you suggesting I increase the sub delay to more than the L&R? That would confound all my assumptions that the speakers have to be delayed most to wait for the sub! My assumption was always that we would need to add more delay to the mains?

My thinking may need to enter a 3rd dimension is that is the case?!

Jon

 

[lexicon]

Did you get to the bottom of this? If not the bottom screen shot in post 12 may hold a clue. It appears that manual delays can be input on this device even with Dirac delays set. Not sure of the order of operations in the DSP but the difference between the FL & FR manual delays is 5ms. Coincidence?

Well spotted! Yes that was indeed the issue. You need to start all new Dirac calibrations from a profile with zero settings for delays and gains or it treats anything set as offsets and adds them in. That one had escaped me as I had just expected the processor to always calibrate from a zero position automatically.

 

[lexicon]

Here are some practice files for use with the REW alignment tool.
1. The 'ideal...' file has an idealized 80 Hz high pass and lowpass filter. The 2 impulses are already properly time aligned. The alignment tool was used to create the 'aligned sum'. It is good to see the impact on the SPL, Impulse, phase, group delay, step response and other charts with an idealized LR-24 XO.
2. The 'Offset...' file has the same 2 original measurements, but the relative timing is not ideal so you can practice using the alignment tool controls to recreate the ideal timing as was already generated in the 1st file.
3. The 'Real...' file has a more typical situation using real measurement data. See if you can find the best phase tracking situation in this relatively easy case.

Depending on the measurement data it can be very difficult to analyze some setups. Using some smoothing can often help. Applying an FDW can help understand the phase trend of direct sound. I like to stay as close to that timing as possible, but still optimize the SPL without the FDW applied. A timing deviation of greater 1 wavelength would best be avoided.

I'm worried I may get lost fairly quickly but I'll give this a go and see how I get on! Thank you again.

Jon

 

[jtalden]

At the moment the sub is set at 0ms delay by Dirac (usually now as default) with all other speakers delayed from there. Are you suggesting I increase the sub delay to more than the L&R? That would confound all my assumptions that the speakers have to be delayed most to wait for the sub! My assumption was always that we would need to add more delay to the mains?

Yes, reducing hte delay on hte SW is the same as increasing the delay on the mains. When the SW is at (or too close) to 0 ms delay and a reduction in SW delay is needed, then all the other main speaker delays can be increased instead. So if you decide to test the impact of Option-3, you could increase the FL and FR by 13 ms and evaluate it in stereo mode. If you decide to adopt that option then the other main speakers need to be increased as well.

Thanks to you and JStewart for letting us know what the issue was with the 5 mm delay. Someone else may run across that same problem.

The dip we saw in the FL+FR average couldn't be caused by this extra 5 ms of delay, but maybe there was another issue with that first measurement file.

 

[lexicon]

Yes, reducing hte delay on hte SW is the same as increasing the delay on the mains. When the SW is at (or too close) to 0 ms delay and a reduction in SW delay is needed, then all the other main speaker delays can be increased instead. So if you decide to test the impact of Option-3, you could increase the FL and FR by 13 ms and evaluate it in stereo mode. If you decide to adopt that option then the other main speakers need to be increased as well.

Thanks to you and JStewart for letting us know what the issue was with the 5 mm delay. Someone else may run across that same problem.

The dip we saw in the FL+FR average couldn't be caused by this extra 5 ms of delay, but maybe there was another issue with that first measurement file.

Of course! - I always find that confusing! Ok so tomorrow I will try both options and see how that goes - many thanks.

 

[lexicon]

Here are some practice files for use with the REW alignment tool.
1. The 'ideal...' file has an idealized 80 Hz high pass and lowpass filter. The 2 impulses are already properly time aligned. The alignment tool was used to create the 'aligned sum'. It is good to see the impact on the SPL, Impulse, phase, group delay, step response and other charts with an idealized LR-24 XO.
2. The 'Offset...' file has the same 2 original measurements, but the relative timing is not ideal so you can practice using the alignment tool controls to recreate the ideal timing as was already generated in the 1st file.
3. The 'Real...' file has a more typical situation using real measurement data. See if you can find the best phase tracking situation in this relatively easy case.

Depending on the measurement data it can be very difficult to analyze some setups. Using some smoothing can often help. Applying an FDW can help understand the phase trend of direct sound. I like to stay as close to that timing as possible, but still optimize the SPL without the FDW applied. A timing deviation of greater 1 wavelength would best be avoided.

First attempts attached! Ok so I just need to use the alignment tool on the "All SPL" tab.

So - idiots guide:-

0. I click on settings top right.
1. I calculate the average of the L&R with ‘trace arithmetic’ using (A + B /2) and save that “average’ trace
2. I click on ‘Alignment Tool’
3. I compare the ‘average‘ trace (top slot) with the sub (sub at bottom slot)
4. I place cursor on Xover used
5. I click ‘align phase at cursor’
6. If that gives me a positive delay on the sub on bottom trace I need to increase the ms delay on the sub or decrease the ms delay on all of the mains. Conversely, if I get a negative delay showing for sub I decrease ms delay of sub or increase ms delay for all the mains by that amount.
7. The three lines on the phase graph should now cross at the cursor (xover) line but that cross won't likely be at 0 or 180 angle, it could be anywhere, at any angle?
8. I click ‘aligned sum’ to get the summed result - that creates the ‘aligned‘ trace.

In this particular case (Aligned Real) I would need to increase the sub delay by the 6.5ms shown or decrease the delay of all main speakers by that amount?

Fascinating stuff. Does that give you the initial "conventionally" optimised timing?

 

[lexicon]

Attached are two MDAT files - 40Hz mains Xover + 80Hz Xover for all other speakers. These are both with the identical corrected levels. One with BM and the other without.

You will see that the dip is back for the profile with no BM applied and eliminated by BM.

The conclusion (I hope you agree) is that the new BM module does as advertised and corrects the phase in such a way as to eliminate these dips.

We still have two relatively narrow dips created by the room/speaker locations at 127hz and 227hz. If I were to add a second identical sub placed to the left rear of my seating position and re-run BM do you think those dips would be improved as they are both above the Xover frequencies? Correcting phase for multiple subs is what BM was really designed to achieve as far as I understand.

Based on the profiles we have created to date which Xover or Xover combination would you choose to use from a technical point of view?

Will your suggested timing adjustments for the 80hx profile still be the same for the BM 40+80 profile do you think?

We are making real progress here I feel, all thanks to you.

 

[jtalden]

In this particular case (Aligned Real) I would need to increase the sub delay by the 6.5ms shown or decrease the delay of all main speakers by that amount?

Yes, well done. You found the correct answer and are correct in the how the change must be implemented in a DSP.

Fascinating stuff. Does that give you the initial "conventionally" optimised timing?

It did in this case and we can confirm this with a few extra steps. For a conventional timing alignment we are looking to achieve the closest direct sound phase tracking throughout the XO range. We can use REW FDW to remove most of the late arriving sound such the we can see the the phase of the direct sound better. Below I changed the scaling to better see the SPL and XO range and applied a FDW of 5 cycles (adjust that as needed to clean up the traces).

First- ID the XO range: Here it extends from about 60-200 Hz to provide about 30 dB of SPL rolloff.

Then: Scale and view that range with FDW applied.

We can now see there is very close phase tracking with that setting. No other timing setting will result in this although different FDW settings will often shift results somewhat so it is worth exploring that.

The FDW setting can be done first to get a good starting point for the timing. The FDW can then be removed to see if the SPL support still looks good. Try then changing the timing the smallest amount necessary to see if a significant SPL improvement can be made. It is best stay within ±1/4 wavelength (1/100 Hz / 4 = ±0.0025 s or ±2.5 ms in this case).
No adjustment was needed in this relatively easy example:

Tip: For the minor SPL adjustment without the FDW it is often helpful to apply smoothing. I have used 1/3 to 1/12 with very good success (not shown).

So - idiots guide:- Commented/Clarified?

1. Open graph controls on All SPL panel
2. Calculate the average of the L&R main speaker measurements using ‘trace arithmetic’ (A + B / 2)
3. Click on ‘Alignment Tool’
4. Compare the ‘A,B average‘ trace (top slot) with the SW (bottom slot)
5. Place cursor in the XO frequency range avoiding nulls in either L&R or SW
6. Click ‘align phase at cursor’
7. The three lines on the phase graph should now be horizontal at the cursor. [If they are vertically offset from one another by more than 90° that suggest that we will get better phase tracking by inverting polarity of SW in the tool.]
8. Click ‘aligned sum’ to get the summed result - that creates the ‘aligned‘ trace.
9. A positive ms delay indicates an increase of delay on the SW is needed. This would be decrease in SW distance if done in an AVR.

 

[jtalden]

You will see that the dip is back for the profile with no BM applied and eliminated by BM.

The conclusion (I hope you agree) is that the new BM module does as advertised and corrects the phase in such a way as to eliminate these dips.

That's impressive. I didn't actually review these files as I trust you have already correctly analyzed them.

We still have two relatively narrow dips created by the room/speaker locations at 127hz and 227hz. If I were to add a second identical sub placed to the left rear of my seating position and re-run BM do you think those dips would be improved as they are both above the Xover frequencies? Correcting phase for multiple subs is what BM was really designed to achieve as far as I understand.

You can look at the SPL chart and see if the dips are in the mains - I assume they are. If so, Dirac would have corrected these if it was possible. They are possibly SBIR effects that can't be addressed with EQ. Also, remember that Dirac EQ is correctly using an average of mic locations in listening area for its corrections. The minor dips at the main LP can just be the tradeoff needed to improve other mic locations.
If the dips there are in the SW SPL response and that is dragging down the channel response at those frequencies then yes, multiple SWs are helpful in smoothing the SW SPL. They have other advantages as well.

Based on the profiles we have created to date which Xover or Xover combination would you choose to use from a technical point of view?

80 Hz would be my technical preference, but you can either choose a setting you desire or experiment enough to become comfortable with one the options you are considering.

Will your suggested timing adjustments for the 80hx profile still be the same for the BM 40+80 profile do you think?

I am not sure. For DIY I just always confirm the timing and adjust as needed. For Dirac BM you have not yet found a timing issue that I remember. A quick look at the measured SPL chart will indicate if the timing is wrong. We knew from the beginning via the SPL chart that the timing at 40 Hz appeared favorable we just were confirming that to ease any apprehension.

 

[lexicon]

That's impressive. I didn't actually review these files as I trust you have already correctly analyzed them.

You can look at the SPL chart and see if the dips are in the mains - I assume they are. If so, Dirac would have corrected these if it was possible. They are possibly SBIR effects that can't be addressed with EQ. Also, remember that Dirac EQ is correctly using an average of mic locations in listening area for its corrections. The minor dips at the main LP can just be the tradeoff needed to improve other mic locations.
If the dips there are in the SW SPL response and that is dragging down the channel response at those frequencies then yes, multiple SWs are helpful in smoothing the SW SPL. They have other advantages as well.

80 Hz would be my technical preference, but you can either choose a setting you desire or experiment enough to become comfortable with one the options you are considering.

Will your suggested timing adjustments for the 80hx profile still be the same for the BM 40+80 profile do you think?
[/QUOTE]
I am not sure. For DIY I just always confirm the timing and adjust as needed. For Dirac BM you have not yet found a timing issue that I remember. A quick look at the measured SPL chart will indicate if the timing is wrong. We knew from the beginning via the SPL chart that the timing at 40 Hz appeared favorable we just were confirming that to ease any apprehension.
[/QUOTE]

That’s fantastic - thank you so much for all your help.

Listening to the 40+80 BM profile this evening it does seem to offer a very impressive musical and integrated/coherent sound-field. I believe that this may well be because I am hearing more of the complete range from the JBL mains now and less from the SVS which I have come to regard as better at the low effects than music. Before BM, with the 40hz Xover this did not sound as good and I think you have uncovered the main reason why.

It may be that for action films with less music content where more bass aggression is required, the higher 80 crossover will be the better choice with perhaps the addition of a second identical sub.

I now have three good verified Dirac profiles that can be switched instantly so I need to spend some time comparing all the options.

I feel a lot more confident in verifying any future changes I make and that’s all down to you.

Thank you again.

Jon

 

[jtalden]

Thanks- Enjoy the sounds!

 

[Michael Boeker]

I own a StormAudio ISP and have made recommendations to Dirac regarding improvements to the time alignment process.

Note that the REW alignment tool is based off the impulse response. The impulse response is using the energy that arrives first (highest frequencies) measured from the subwoofer in order to establish arrival time. As such, it is not in any way accurate in aligning the subwoofer at the crossover frequency to the main speakers. It can be off by up to 30 ms depending on the crossover frequency. This can be shown by using a dual channel measurement system. I use SMAART for time and phase alignment and verification. One also can't get accurate bass phase information when measuring from the listening position. The reflections swamp the measurement. The only way to accurately time and phase align is to use a close mic measurement on both the a main speaker and the subwoofer. You can also only time and phase align a single subwoofer to one speaker unless the subwoofer is directly between the mains. Then it can be time aligned to both.

One benefit of the StormAudio ISP is that one can have a Profile with the subwoofer time and phase aligned to L/R for two channel content and a separate Profile with the subwoofer time and phase aligned to the center speaker for best dialogue intelligibility for movie content.

 

[John Mulcahy]

Note that the REW alignment tool is based off the impulse response. The impulse response is using the energy that arrives first (highest frequencies) measured from the subwoofer in order to establish arrival time. As such, it is not in any way accurate in aligning the subwoofer at the crossover frequency to the main speakers. It can be off by up to 30 ms depending on the crossover frequency. This can be shown by using a dual channel measurement system.

That's not an accurate portrayal of REW's use of the impulse response or the behaviour of the alignment tool. The alignment tool is used with measurements that have a common timing reference. That might be a loopback connection, or an acoustic reference from one of the speakers in the system. The timing reference is analogous to the function performed by the reference channel of a dual channel FFT system, the subwoofer and speaker measurements are presented with the correct relative alignment and the delays shown in the alignment tool have exactly the effect shown in the tool.

You may have been referring to trying to use a measurement without a timing reference and attempting to infer the t=0 time from the response. The impulse response peak is sometimes used for that, and typically isn't far off for full range measurements, but is not at all appropriate for a subwoofer measurement due to its limited bandwidth. In that case the start of the impulse response is a better metric to use, or a cross correlation with a minimum phase version of the measurement (as used by REW's Estimate IR Delay feature). Neither is relevant to the Alignment Tool, however, since that has a timing reference and doesn't need to make any inferences.

 

[Michael Boeker]

That's not an accurate portrayal of REW's use of the impulse response or the behaviour of the alignment tool. The alignment tool is used with measurements that have a common timing reference. That might be a loopback connection, or an acoustic reference from one of the speakers in the system. The timing reference is analogous to the function performed by the reference channel of a dual channel FFT system, the subwoofer and speaker measurements are presented with the correct relative alignment and the delays shown in the alignment tool have exactly the effect shown in the tool.

SMAART's automatic delay finder, which relies on a loopback timing reference, can't calculate the delay time for a subwoofer. It says in the SMAART manual, "Smaart’s Delay Finder is an automated routine for finding the delay time for the active transfer function measurement. It works by measuring the impulse response of the SUT and then scanning the IR it to find the highest peak, which will normally represent the initial arrival of the measurement signal at the measurement point. In acoustic measurements, we call this the first arrival or arrival of direct sound." I assumed that REW was doing something similar and I was also not getting consistent, accurate phase alignment using REW's Alignment Tool.

Sorry I mis-represented what REW is doing. I'll do some more testing using both loopback and acoustic timing references and see what happens. I also verify using Signal Alignment using Wavelets in ARTA. It would be cool to do this in REW.

 

[lexicon]

I own a StormAudio ISP and have made recommendations to Dirac regarding improvements to the time alignment process.

One also can't get accurate bass phase information when measuring from the listening position. The reflections swamp the measurement. The only way to accurately time and phase align is to use a close mic measurement on both the a main speaker and the subwoofer. You can also only time and phase align a single subwoofer to one speaker unless the subwoofer is directly between the mains. Then it can be time aligned to both.

I’d certainly be interested to hear the response of both Dirac and Storm to your assertion.

It seems to me that what you are saying is that their newly launched Bass Control Module V1.0 (which they have spent over two years developing and claim can correct time and phase for multiple speakers and subs wherever they are located in a room) is built upon an entirely false premise and ergo does not work.

I myself don’t possess enough technical knowledge in the field of acoustics to either confirm or dispute what seems to be a very bold statement!

 

[John Mulcahy]

REW's spectrogram has Fourier and Wavelet modes.

Finding delay is equivalent to what I was describing as attempting to infer the t=0 time. When looking at alignments the absolute values of delay don't matter, we just need all the measurements to have the same relative timing. With a dual channel system you can do that by running the delay find on a full range channel and keeping that same same delay for a sub measurement, providing everything else stays the same.

 

[markus]

I own a StormAudio ISP and have made recommendations to Dirac regarding improvements to the time alignment process.

Note that the REW alignment tool is based off the impulse response. The impulse response is using the energy that arrives first (highest frequencies) measured from the subwoofer in order to establish arrival time. As such, it is not in any way accurate in aligning the subwoofer at the crossover frequency to the main speakers. It can be off by up to 30 ms depending on the crossover frequency. This can be shown by using a dual channel measurement system. I use SMAART for time and phase alignment and verification. One also can't get accurate bass phase information when measuring from the listening position. The reflections swamp the measurement. The only way to accurately time and phase align is to use a close mic measurement on both the a main speaker and the subwoofer. You can also only time and phase align a single subwoofer to one speaker unless the subwoofer is directly between the mains. Then it can be time aligned to both.

One benefit of the StormAudio ISP is that one can have a Profile with the subwoofer time and phase aligned to L/R for two channel content and a separate Profile with the subwoofer time and phase aligned to the center speaker for best dialogue intelligibility for movie content.

While it's true that in acoustically small rooms "reflections swamp the measurement", it's also true that this highly correlates with what we hear at low frequencies. Our hearing isn't capable of separating direct sound and reflections the same way as it does at higher frequencies. All sense of directionality is lost. This has been studied for example by Martens et al., 2004 and found to be true up to about 90Hz. So optimizing the steady state magnitude response will indeed optimize what is heard.

As to aligning low frequency sources to the IR peak: this is often done in live event environments where the room is large enough so measurements are less contaminated with reflections. In acoustically small rooms the start of the IR is used instead.

 

[StevieLee]

Good day to you gents. I have recently purchased a mini dsp 2x4hd to integrate 3 subwoofers into my system.

I have recently found the alignment tool within REW, could someone please advise on the following image please.

The phase slopes, are they to be close together?, to get a better response on the top line?

With the subwoofer set to a minus delay, I could not enter this into the mini dsp. Should I add distance in the avr so that it will become a positive delay?

Kind regards
Stevie

 

[lexicon]

They look pretty good to me - 60hz Xover? I think the lines ideally need to be running horizontally at Xover which they are.
If delay on sub is showing a negative figure increase all mains by that delay instead.

But @jtalden is your man ..

 

[JStewart]

The phase slopes, are they to be close together?, to get a better response on the top line?

Yes and put another way, they will be close together when the response on the top line is better.
Close together = in phase and sources sum
Not close together = out of phase (to some extent) and sources interfere (to some extent).

With the subwoofer set to a minus delay, I could not enter this into the mini dsp. Should I add distance in the avr so that it will become a positive delay?

Yes

But @jtalden is your man ..

Absolutely

 

[StevieLee]

They look pretty good to me - 60hz Xover? I think the lines ideally need to be running horizontally at Xover which they are.
If delay on sub is showing a negative figure increase all mains by that delay instead.

But @jtalden is your man ..

Thank you, I think I am slowly getting the hang of it, will continue with some more tests and post back.

Yes and put another way, they will be close together when the response on the top line is better.
Close together = in phase and sources sum
Not close together = out of phase (to some extent) and sources interfere (to some extent).



Yes



Absolutely

Thank you, got it, close together it is, and if negative delay on subs, add to the mains.

Thank you both very much.

 

[jtalden]

To be more specific; if negative ms delay on the subs needed and that can't be done within the DSP then increase the ms delay of all the main speakers in the DSP (if possible).
If not possible (as mains are not controlled within the DSP) then either increase the sub distance or decrease the mains distance in the AVR by the equivalent amount.