Struggling to make sense of REW delay measurements

[klettermann]

Per the title. I'm trying to do the speaker delay process described by Deer Creek Audio's method HERE. On the face of it it seems easy enough, but either I'm doing something wrong or I can't make sense of it. Hopefully The Learned Ones here can set me straight.

Room setup is at the bottom.. Main speakers are in front, Sub1 in front right corner, Sub 2 in left rear corner. The respective REW delay measurements from respective speakers to MSP as follows:
o
Left: -0.122ms
Right: -0.127
Front sub1: 3.225
Rear sub 2: -2.980

This the raw data, no further arithmetic. I repeated it a few times and got the same result. It's sensible that the mains would be the same, but how are they at virtually the same place as the MLP? And what's up with the subs? Clearly I'm doing something very wrong. Any help greatly appreciated. Thanks and cheers,

 

[NTK]

I haven't done any of these kinds of measurements myself, so my observations may not worth anything, but based on my layman understanding ...

• The left and right delay numbers are indeed suspiciously close. They are off by 0.005 ms, equivalent to a difference in path distances of ~0.07 in (= 0.005*1.11*12).
• The front sub (sub1) is 3.347 ms behind the left, meaning its "effective" path distance to the MLP is 3.7 ft further.
• The rear sub (sub2) is 2.858 ms ahead of the left, meaning its "effective" path distance to the MLP is 3.2 ft closer.

The sub numbers seem reasonable to me based on the figure.

 

[Keith_W]

A few points.

1. Phase alignment is far more important than time alignment. This is because the audible group delay thresholds are large - approximately 1/2 a period at the frequency in question, so an 80Hz wave will have an audible threshold of 6.25ms. But phase misalignment will create cancellation at the XO frequency which is far more audible. I recommend "ballpark" time alignment, then meticulous phase alignment. If you are using an XO freq of 80Hz, you only need to get it time aligned somewhere around the vicinity of +/- 6.25ms. There is no harm in going for more precise alignment, but time alignment should not be sacrificed in favour of phase alignment.

2. REW's automatic time delay readout measures the delay of the DUT relative to the tweeter reference. So if you measuring (say) your left speaker, and you are using your left tweeter as the reference, then the delay is between the left woofer and the left tweeter. It is NOT the delay between the speaker position and the mic, which I suspect is what you are thinking. So, -0.122ms for the left speaker is perfectly reasonable if the time reference and DUT are on the same side. But the -0.122ms and -0.127ms discrepancy for left/right (0.005ms) either implies you used L tweeter/L woofer and R tweeter/R woofer, or you somehow managed to place your microphone extremely accurately between the speakers if you were using Ltweeter/Lwoofer and Ltweeter/Rwoofer.

If you want to confirm a centred mic position, sweep both speakers together (mono measurement) and examine the ETC:

You should see two peaks. One will be the left speaker, the other the right. There is no way to tell. If the peaks are coincident, it means the mic is perfectly centred. If you really want to centre your microphone, shift it and repeat the sweep. The gap between the peaks will either narrow or widen. Keep shifting the mic until the peaks are coincident (this is a method I invented! I have not seen it described anywhere else?).

Do you need to centre your mic? Not really. But this method will tell you if your mic is egregiously off centre, and that WILL affect your time alignment because of parallax error.

3. You did not indicate the distance of your MSP relative to the subs. Measure the actual distance with a tape measure, then calculate what the delay should be using this formula: t = d/1000c with t = time in ms, d = distance in ft, c = speed of sound 1125ft/s. This will tell you if your measurement is ballpark accurate or not. Bear in mind that you are measuring the delay with respect to the time arrival of the tweeter impulse at the MSP. So the calculations will look something like this:

- measure distance of the tweeter reference to MSP, then calculate the delay.
- measure distance of both subwoofers to the MSP, then calculate the delays.
- Estimated delay of front subwoofer = Delay(tweeter) - Delay(front sub)
- Estimated delay of rear subwoofer = Delay(tweeter) - Delay(rear sub)

Heed the warning on that page - if you have a wireless sub, or if the sub has DSP, it will increase the delay - sometimes by up to 100ms! The reason for doing this calculation is a "sanity check" to see if your measured delay is in the ballpark of what it should be.

 

[klettermann]

A few points.

1. Phase alignment is far more important than time alignment. This is because the audible group delay thresholds are large - approximately 1/2 a period at the frequency in question, so an 80Hz wave will have an audible threshold of 6.25ms. But phase misalignment will create cancellation at the XO frequency which is far more audible. I recommend "ballpark" time alignment, then meticulous phase alignment. If you are using an XO freq of 80Hz, you only need to get it time aligned somewhere around the vicinity of +/- 6.25ms. There is no harm in going for more precise alignment, but time alignment should not be sacrificed in favour of phase alignment.

2. REW's automatic time delay readout measures the delay of the DUT relative to the tweeter reference. So if you measuring (say) your left speaker, and you are using your left tweeter as the reference, then the delay is between the left woofer and the left tweeter. It is NOT the delay between the speaker position and the mic, which I suspect is what you are thinking. So, -0.122ms for the left speaker is perfectly reasonable if the time reference and DUT are on the same side. But the -0.122ms and -0.127ms discrepancy for left/right (0.005ms) either implies you used L tweeter/L woofer and R tweeter/R woofer, or you somehow managed to place your microphone extremely accurately between the speakers if you were using Ltweeter/Lwoofer and Ltweeter/Rwoofer.

If you want to confirm a centred mic position, sweep both speakers together (mono measurement) and examine the ETC:

View attachment 432623

You should see two peaks. One will be the left speaker, the other the right. There is no way to tell. If the peaks are coincident, it means the mic is perfectly centred. If you really want to centre your microphone, shift it and repeat the sweep. The gap between the peaks will either narrow or widen. Keep shifting the mic until the peaks are coincident (this is a method I invented! I have not seen it described anywhere else?).

Do you need to centre your mic? Not really. But this method will tell you if your mic is egregiously off centre, and that WILL affect your time alignment because of parallax error.

3. You did not indicate the distance of your MSP relative to the subs. Measure the actual distance with a tape measure, then calculate what the delay should be using this formula: t = d/1000c with t = time in ms, d = distance in ft, c = speed of sound 1125ft/s. This will tell you if your measurement is ballpark accurate or not. Bear in mind that you are measuring the delay with respect to the time arrival of the tweeter impulse at the MSP. So the calculations will look something like this:

- measure distance of the tweeter reference to MSP, then calculate the delay.
- measure distance of both subwoofers to the MSP, then calculate the delays.
- Estimated delay of front subwoofer = Delay(tweeter) - Delay(front sub)
- Estimated delay of rear subwoofer = Delay(tweeter) - Delay(rear sub)

Heed the warning on that page - if you have a wireless sub, or if the sub has DSP, it will increase the delay - sometimes by up to 100ms! The reason for doing this calculation is a "sanity check" to see if your measured delay is in the ballpark of what it should be.

Great, thanks, I'm trying work through all this. Per the pic, the mains and MLP are very precisely placed and measured. Up to this point I hadn't calculated delay based on the manual measurement calculation. Yesterday I did used those delays in the MiniDSP. They still don't agree with the REW reference chirp results for the mains. Moreover, the right main ought to be just slightly less than the front sub. It's almost directly in front of the sub, so just a few feet closer to MLP. But the REW doesn't indicate this. Still working on it.... Cheers,

 

[Keith_W]

Don't forget that subs have group delay on their own. Furthermore, the subwoofer impulse is time stretched. As an example, the period of a 50Hz sine wave is 20ms. Since the impulse peak occurs at 1/4 impulse (90deg out of 360deg), the peak will occur at 5ms. So yes, it is conceivable that calculating the delay using the distance + speed of sound will give you one number, but acoustic measurement gives you another.

 

[klettermann]

Don't forget that subs have group delay on their own. Furthermore, the subwoofer impulse is time stretched. As an example, the period of a 50Hz sine wave is 20ms. Since the impulse peak occurs at 1/4 impulse (90deg out of 360deg), the peak will occur at 5ms. So yes, it is conceivable that calculating the delay using the distance + speed of sound will give you one number, but acoustic measurement gives you another.

OMG. Another wrinkle. It's all too much.

Anyway, the tape-measure delay method did have a significant effect. My process was:
> set the respective delays on the mains and rear sub (front was farthest and hence zero)
> EQ subs together (virtual sub) and respective mains individually.
> Set XO (still needs refinement -needs better level matching between subs and mains)
> Compared scans with delay on and switched off. Smoothing is 1/24. See below.

1. Using "tape measure" delays 2. Same as 1 but delays turned off.

With

The delayed version FR is a little smoother (more noticeable with more smoothing) and transition between mains and subs is a bit nicer. But the proof is in the listening. The difference in soundstage "air" and imaging is immediately apparent. Just more holographic and real sounding. I tested it on my partner, flipping back and forth between the 2. She's quite certain that I've lost my mind. Regardless, she could tell which was which every time and preferred the delayed version. I don't know how to quantify any of that, nor do I know what else there is to do, if anything. I'm thinking nothing except maybe smooth it out around the crossover a little better. Now I tell myself to quit fiddling and just listen. But I doubt I'll be able to resist for very long. There's still Dirac to play with.... Thanks and cheers,

[ps edit:

Comparison between REW-generated delay and "tape measure" delay below. I really don't know what to make of it, if anything.

REW: Tape measure

Left: -0.122ms 3.33ms
Right: -0.127 3.33
Front sub1: 3.225 0
Rear sub 2: -2.980 7.493

 

[Keith_W]

The tape measure delay is a CHECK, it is not meant to be used as your actual delay. It tells you what to expect. Delay must be set with acoustic measurement, always. And I don't think you understood my point about measuring delay with tweeter reference.

When we are talking about delays, there are 3 causes for delay:

1. Driver group delay. It takes longer for 20Hz to exit the driver (1/20 sec, i.e. 50ms) compared to 200Hz (5ms). All drivers have rising group delay at the lower end of the frequency range.
2. Device latency and DSP delay. This can be substantial, especially if the signal chain is long or complex. FYI my Focusrite 2i2 which is connected via VB-Matrix on my system has a delay of 316ms. That's right, almost one third of a second.
3. "Time of flight" delay - how much time it takes the sound to leave the speaker and arrive at the microphone

Note that the tape measure only accounts for time of flight delay. This is why acoustic measurement is superior. All types of delay are accounted for.

And please read this very carefully: this is how the tweeter reference works. You have a device under test (DUT) and the tweeter reference. When REW makes a sweep, both the DUT and the tweeter are swept. When the tweeter reference arrives at the microphone, REW takes this as time 0. It then measures the arrival of the DUT, and it will relative to the tweeter reference. In your case, -0.122ms.

What this means is that the DUT is -0.122ms relative to the tweeter reference (in your case it is arriving slightly early). The time of -0.122ms DOES NOT REFER to the time the sound took for it to launch from the DUT and arrive at the microphone.

If you want, inbox me. I am happy to call you on WhatsApp to explain the difference.

 

[klettermann]

The tape measure delay is a CHECK, it is not meant to be used as your actual delay. It tells you what to expect. Delay must be set with acoustic measurement, always. (1) And I don't think you understood my point about measuring delay with tweeter reference.

Actually I think I did, just not sure what to do about it.

When we are talking about delays, there are 3 causes for delay:

1. Driver group delay. It takes longer for 20Hz to exit the driver (1/20 sec, i.e. 50ms) compared to 200Hz (5ms). All drivers have rising group delay at the lower end of the frequency range.
2. Device latency and DSP delay. This can be substantial, especially if the signal chain is long or complex. FYI my Focusrite 2i2 which is connected via VB-Matrix on my system has a delay of 316ms. That's right, almost one third of a second.
3. "Time of flight" delay - how much time it takes the sound to leave the speaker and arrive at the microphone
Note that the tape measure only accounts for time of flight delay. This is why acoustic measurement is superior. All types of delay are accounted for.

OK, makes sense. And I'm just looking at (3) with the time-of-flight calculation. So, that might be good as approximation, but still inaccuate.

And please read this very carefully: this is how the tweeter reference works. You have a device under test (DUT) and the tweeter reference. When REW makes a sweep, both the DUT and the tweeter are swept. When the tweeter reference arrives at the microphone, REW takes this as time 0. It then measures the arrival of the DUT, and it will relative to the tweeter reference. In your case, -0.122ms.

What this means is that the DUT is -0.122ms relative to the tweeter reference (in your case it is arriving slightly early). The time of -0.122ms DOES NOT REFER to the time the sound took for it to launch from the DUT and arrive at the microphone.

OK, but maybe we can back up to the subs. Those measurements at least produced sensible results (maybe?). Here's my understanding of how that works using my original data for Sub2:
1. System is set up to send the chirp to the right tweeter (speaker) and sweep to Sub2 in the opposite corner of the room.
2. I hit the REW "start" button. REW measures noise level and front right tweeter chirps.
2. The chirp arrives at the mike and REW hears it. REW's timing clock starts with that time as >zero<.
3. Next the DUT signal arrives at the mike. REW records the time difference between zero and receipt of the DUT arrival time, in this case -2.98ms.

Can you confirm that I have this right or, if not, correct? In the meantime I'm going to recheck the timing setup for the mains. Thanks, I do appreciate taking the time to school me (and whoever else is as confounded as I am). Thanks again and cheers,

 

[Keith_W]

Actually I think I did, just not sure what to do about it.

Cool! You are only concerned about the delay that arrives at the listening position. So if you measure -2.98ms, then that is the delay you use with your DSP.

If you want to confirm that the delay has been set correctly, sweep the subwoofer and the tweeter together (disconnect the woofer and midrange driver if you are able to). Switch to the step response (not impulse response). You should see something like this:

The skinny spike is the tweeter, and the big wave is the subwoofer. The tweeter should be aligned to the steepest part of the subwoofer curve.

OK, but maybe we can back up to the subs. Those measurements at least produced sensible results (maybe?). Here's my understanding of how that works using my original data for Sub2:
1. System is set up to send the chirp to the right tweeter (speaker) and sweep to Sub2 in the opposite corner of the room.
2. I hit the REW "start" button. REW measures noise level and front right tweeter chirps.
2. The chirp arrives at the mike and REW hears it. REW's timing clock starts with that time as >zero<.
3. Next the DUT signal arrives at the mike. REW records the time difference between zero and receipt of the DUT arrival time, in this case -2.98ms.

Can you confirm that I have this right or, if not, correct? In the meantime I'm going to recheck the timing setup for the mains. Thanks, I do appreciate taking the time to school me (and whoever else is as confounded as I am). Thanks again and cheers,

This is correct. REW has a start chirp and an end chirp. I don't know how REW detects the delay, from experience it needs both chirps otherwise the delay measurement is off. I have said many times on other threads that I am far from a REW expert since I only use REW to help people on ASR, I don't actually use it myself.

You may have missed the part in my earlier post about the importance of subwoofer time alignment: IT IS NOT VERY IMPORTANT. Phase alignment is FAR more important. Do an approximate subwoofer time alignment and then meticulous phase alignment. You are going to shift your time alignment when you do your phase alignment anyway, so don't worry about it too much.

 

[klettermann]

Thanks, much to digest. The project continues........

Cool! You are only concerned about the delay that arrives at the listening position. So if you measure -2.98ms, then that is the delay you use with your DSP.

Yes. Done.

If you want to confirm that the delay has been set correctly, sweep the subwoofer and the tweeter together (disconnect the woofer and midrange driver if you are able to). Switch to the step response (not impulse response). You should see something like this:

View attachment 433896

The skinny spike is the tweeter, and the big wave is the subwoofer. The tweeter should be aligned to the steepest part of the subwoofer curve.

Several things here. First, like most of us, I can't do that. Spkr XOs etc are buried inside the speakers. But that's a minor point. More substantive is your graph. It's not REW, right? The closest I know how to get on REW is along the lines of the below. Orange and blue are respective subs and green is the impulse speaker. In the pic nothing is aligned with anythiing but it's easy to slide them back and forth to get them all aligned as desired. On your curve it looks like you're delaying mains as shown. Nowhere, anywhere have I seen this approach. Can you clarify the rationale?

This is correct. REW has a start chirp and an end chirp. I don't know how REW detects the delay, from experience it needs both chirps otherwise the delay measurement is off. I have said many times on other threads that I am far from a REW expert since I only use REW to help people on ASR, I don't actually use it myself.

You may have missed the part in my earlier post about the importance of subwoofer time alignment: IT IS NOT VERY IMPORTANT. Phase alignment is FAR more important. Do an approximate subwoofer time alignment and then meticulous phase alignment. You are going to shift your time alignment when you do your phase alignment anyway, so don't worry about it too much.

OK, so I guess I figured out time alignment. As for phase alignment, I hear you but I wasn't clear. Then I found THIS which I hadn't seen before but looks useful. Stay tuned.... Thanks and cheers,

 

[Keith_W]

Several things here. First, like most of us, I can't do that. Spkr XOs etc are buried inside the speakers. But that's a minor point. More substantive is your graph. It's not REW, right?

Correct, it's not REW. It's Acourate. I have to confess I am all thumbs when it comes to REW. I don't even know if some of the things I suggest can be done in REW, let alone describe the procedure to you. Ideally we would have a REW expert here to help you, but it seems like it's only me.

The closest I know how to get on REW is along the lines of the below. Orange and blue are respective subs and green is the impulse speaker. In the pic nothing is aligned with anythiing but it's easy to slide them back and forth to get them all aligned as desired. On your curve it looks like you're delaying mains as shown. Nowhere, anywhere have I seen this approach. Can you clarify the rationale?

There are many ways to skin a cat. The first thing to realise is that the impulse response and the step response are two views of exactly the same thing. On an impulse response, the peak of the tweeter is aligned to the peak of the subwoofer. The subwoofer can barely be seen because it is very low amplitude and it is time-stretched, so you need very high zoom to see the subwoofer impulse peak. Furthermore, the subwoofer impulse peak will be buried within the tweeter impulse peak making it very difficult to see. With the step response, the steepest part of the tweeter impulse is aligned to the steepest part of the subwoofer impulse, and the subwoofer is much easier to see.

All you need to do is zoom in until you see the subwoofer impulse (on the impulse response) then switch to the step response. Observe how the curve changes. It's just a different view of the same thing. I guess it's a bit like sunglasses vs. no sunglasses. Both can help you see, but using the wrong eyeglasses in the wrong situation will make it harder to see. But whatever you are looking at is exactly the same.

The goal of time alignment is to align the tweeter impulse peak to the subwoofer impulse peak (if you are using linear phase FIR), or the start of the tweeter impulse to the start of the subwoofer impulse (if you are using minimum phase IIR). I'll show you:

For a minimum-phase IIR, the start of the subwoofer impulse is aligned to the start of the tweeter impulse. This means the blue curve needs to be shifted to the right (delayed), and the mustard yellow curve shifted to the left (advanced).

For linear-phase FIR, the steepest part of the impulse of tweeter and subwoofer need to be aligned. So both curves need to be shifted to the left (delayed), with the yellow curve requiring more delay than the right.

This is the purist "correct" way of doing it, but I did not mention it because I do not think it is very important. Like I said, phase alignment is far more important. So I don't know whether REW looks at impulse start or impulse peak when it gives you the delay. But I think it also does not really matter.

OK, so I guess I figured out time alignment. As for phase alignment, I hear you but I wasn't clear. Then I found THIS which I hadn't seen before but looks useful. Stay tuned.... Thanks and cheers,

Bob McCarthy is excellent and that article is on point. As he mentions, the problem is that you have two moving targets. I can already see that the shape of your front and rear subwoofer impulses are not the same. This suggests to me that the phase will be rotating at a different rate for each sub. But i'll wait and see.

Good luck!

 

[klettermann]

Excellent, thanks again. I think it's starting to gel for me. More comments on comments on comments below.....

>>snip<<<

The goal of time alignment is to align the tweeter impulse peak to the subwoofer impulse peak (if you are using linear phase FIR), or the start of the tweeter impulse to the start of the subwoofer impulse (if you are using minimum phase IIR). I'll show you:

View attachment 434460

Right, got it. That's exactly what I did after getting these sweeps.

For a minimum-phase IIR, the start of the subwoofer impulse is aligned to the start of the tweeter impulse. This means the blue curve needs to be shifted to the right (delayed), and the mustard yellow curve shifted to the left (advanced).

View attachment 434462

For linear-phase FIR, the steepest part of the impulse of tweeter and subwoofer need to be aligned. So both curves need to be shifted to the left (delayed), with the yellow curve requiring more delay than the right.

This is the purist "correct" way of doing it, but I did not mention it because I do not think it is very important. Like I said, phase alignment is far more important. So I don't know whether REW looks at impulse start or impulse peak when it gives you the delay. But I think it also does not really matter.

I really don't understand what FIR and IIR are or pros and cons. Still, I can try it both ways. Such as, try IIR both and then compare the effect on phase. And, conversely, do phase and look at effect on FIR and IIR.

Bob McCarthy is excellent and that article is on point. As he mentions, the problem is that you have two moving targets. I can already see that the shape of your front and rear subwoofer impulses are not the same. This suggests to me that the phase will be rotating at a different rate for each sub. But i'll wait and see. Good luck!

Perfect lead-in to getting more granular with the process. My example above was just that, not any kind of "final" result. So what's the order-of-operations here? My interpretation would that for a 2.2 system:

1. Determine respective sub delays with respect to MLP and set delays accordingly.
2. Rescan delayed subs and find their best phase alignment centered on XO. [how wide a bans should be, or just getting as wide as possible??].
3. Scan combined subs.
4. Scan respective main speakers. Find best phase alignment at the XO.
5. Scan combined mains.
6. Find best phase aligment at XO freq between the combined subs and the combined mains.
7. Now EQ all the respective channels and repeat to make sure it's OK. Adjust as necessary, repeat as necessary.
8. Pass GO, collect $100.

Make sense or did I go off the rails again? Thanks muchly and cheers,

 

[klettermann]

.....and so, a phase alignment along the lines of the below, is what we're after? These are the 2 respective subs. I'm assuming that the XO will be used
at 80Hz and that the wrap at 88Hz is ok.

 

[Keith_W]

I really don't understand what FIR and IIR are or pros and cons. Still, I can try it both ways. Such as, try IIR both and then compare the effect on phase. And, conversely, do phase and look at effect on FIR and IIR.

Whether you are using FIR or IIR depends on the DSP device you are using. FIR is computationally intensive, particularly if you have a lot of taps, so it requires x86 or ARM. IIR is computationally efficient, so if you have anything with a SHARC DSP chip in it (MiniDSP, AVR) it will be IIR. Actually more likely mixed phase (IIR + limited number of FIR taps).

You have your passive crossover in place. That is minimum-phase IIR. So even if you use FIR, you will end up with a mixed-phase system - a FIR-IIR cascade.

ALL IIR is minimum-phase. FIR's can be linear-phase or minimum-phase. The advantage of linear-phase (or "constant delay" as JJ puts it) is that it does not introduce any further phase rotation of its own, meaning that filters are much easier to design.

Perfect lead-in to getting more granular with the process. My example above was just that, not any kind of "final" result. So what's the order-of-operations here? My interpretation would that for a 2.2 system:

1. Determine respective sub delays with respect to MLP and set delays accordingly.
2. Rescan delayed subs and find their best phase alignment centered on XO. [how wide a bans should be, or just getting as wide as possible??].
3. Scan combined subs.
4. Scan respective main speakers. Find best phase alignment at the XO.
5. Scan combined mains.
6. Find best phase aligment at XO freq between the combined subs and the combined mains.
7. Now EQ all the respective channels and repeat to make sure it's OK. Adjust as necessary, repeat as necessary.
8. Pass GO, collect $100.

Make sense or did I go off the rails again? Thanks muchly and cheers,

2: as wide as possible. You want the phase aligned for the entire crossover band, because phase misalignment creates cancellation. If you see phase misalignment, and you see cancellation, one way to deal with it is to design an all-pass filter. But we'll leave that for now.

Rest of the steps: agree, that is one way to do it. In my system, my subwoofers are on either side of my speakers. I align the L sub to the L speaker, and the R sub to the R speaker. Then I align L and R together.

9: In Monopoly, when you pass GO, you collect $200.

 

[Keith_W]

.....and so, a phase alignment along the lines of the below, is what we're after? These are the 2 respective subs. I'm assuming that the XO will be used
at 80Hz and that the wrap at 88Hz is ok.
View attachment 434552

You cut off the vertical scale and legends so I don't know what I am looking at. But you can see the problem: black, pink, and blue are aligned at about 88Hz, but not aligned anywhere else. The alignment between black and pink is reasonably acceptable down to 65-70Hz but will likely produce a huge dip at 55Hz (since it is about 180deg out-of-phase ... I am assuming the vertical scale is +/- 180deg!). Whether this 55Hz dip is a problem depends on the steepness of your XO slope. The alignment between pink-blue is horrible below 90Hz, these two drivers will work against each other and actively cancel the other's output. If pink and blue are your two subs, then you need to adjust the phase of one of them so that the two subs align. If pink and blue is a sub and a speaker, then I would consider a steep XO slope to minimise where the two interact.

This is why steep XO slopes are desirable, but you will run into another problem if you are using minimum-phase IIR's. Every filter order you add will rotate phase at the corner frequency by 90deg. You will have to spend a lot of time experimenting.

BTW, a useful shortcut when trying to align phase is to add the curves together to see the result (Trace Arithmetic A+B). Look closely at the amplitude curve to see where the dip is and whether it shifts. This is much faster than doing an actual acoustic measurement. Once you are satisfied with the result, then do an acoustic measurement to confirm that the sim reflects reality.

You can drive yourself nuts trying to align phase, which is why brute force computation using something like MSO is common. MSO will get you a good result, but I think it is worthwhile going through this exercise because the struggle is what makes you learn.

 

[DHzHolden]

Note that the tape measure only accounts for time of flight delay. This is why acoustic measurement is superior. All types of delay are accounted for.

I puzzled over this just this week. I had done my tape-measure calculations for my 2.1 desktop system, but in my REW delay measurements, the subwoofer delay matched the tape, but not the mains. After a bit I concluded that the mains, identically placed, constitute 0ms delay in REW. Prior to this revelation, I'd been using a tape-measure calculation for the mains, and subtracting it from the REW figure for the sub. And then, dawn broke over Marblehead...

Thanks for your informative response.