Time align with Wiim amp using Rew

[Arianoxx]

Hey! I'm trying to set the delay between the mains and the sub. I followed this guide, but I'm getting significant cancellations, same result when using the tool in the WiiM app. I don't have these one to show, but it was really bad.

So I started increasing the delay on the mains by 2 ms and measure and got slightly better results, but it still feels far from perfect. The red line is with a 12 ms delay and phase inverted; the blue line is with a 5 ms delay and phase untouched; the black line is just the mains without the sub. The crossover is set at 80 Hz.

Below 60 Hz it's acceptable, but between 60 and 120 Hz it's actually worse than using the mains alone. Unfortunately, the sub and mains placement is far from ideal and I can't change it, so I'm just trying to get the best result possible with what I have.

Speakers: Elac DBR62
Sub: B&W ASW610
Amp: WiiM Amp

Does anyone have any suggestions or experience that might help me improve the integration? I'd really appreciate any advice!
I'm also attaching the measurement files, including those with time reference enabled, in case it helps.

 

[kyuu]

I'm not one of the measurement experts here, but given that those peaks/dips appear essentially identical on the mains only measurement they may not be cancellation issues due to phase mismatch but rather room modes. Messing with the delays won't have any impact on that. Only can address that with placement (which you state isn't flexible) and EQ.

 

[Keith_W]

Firstly, your measurements are a complete mess.

I don't know what the difference is between "sub right old", "solo sub right old", and so on. And if you say the XO point is 80Hz, why is it that your sub measurement goes all the way to 500Hz?

Your subwoofer and mains XO need to be in place before you attempt to time/phase align them. This is because the LPF/HPF are minimum phase, and therefore rotate phase. It makes no sense to phase align the raw subwoofer and speaker measurement first, and then mess it up again with the XO.

This is what you need to do:

1. Put the LPF/HPF in place
2. Measurement of ONLY the sub with the LPF and the timing chirp
3. Measurement of each main speaker INDIVIDUALLY with the HPF and the timing chirp

 

[alankila]

Based on flatness of excess group delay, and the trend from 6 ms to 5 ms measurement, 4 ms delay might be the ticket. This is 5 ms, and it is already nearly a straight line near crossover at 80 Hz and the surrounding area. I don't see any frequency being cancelled near 80 Hz and GD is not spiking anywhere nearby, so I think it's mostly in phase, too.

 

[Arianoxx]

First things first, thanks everyone for the help!

I'm not one of the measurement experts here, but given that those peaks/dips appear essentially identical on the mains only measurement they may not be cancellation issues due to phase mismatch but rather room modes. Messing with the delays won't have any impact on that. Only can address that with placement (which you state isn't flexible) and EQ.

Yeah, I know, unfortunately most of these dips are probably going to stay no matter what. But what concerns me a bit is that these measurements are actually the “better” ones. The others were so bad I deleted them. There doesn’t seem to be much cancellation in these one, but in some areas, the response with the sub turned on is actually lower than without the sub. That’s why I was skeptical that I’m really getting the best result just by trial and error.

Firstly, your measurements are a complete mess.

View attachment 444659

I don't know what the difference is between "sub right old", "solo sub right old", and so on. And if you say the XO point is 80Hz, why is it that your sub measurement goes all the way to 500Hz?

Your subwoofer and mains XO need to be in place before you attempt to time/phase align them. This is because the LPF/HPF are minimum phase, and therefore rotate phase. It makes no sense to phase align the raw subwoofer and speaker measurement first, and then mess it up again with the XO.

This is what you need to do:

1. Put the LPF/HPF in place
2. Measurement of ONLY the sub with the LPF and the timing chirp
3. Measurement of each main speaker INDIVIDUALLY with the HPF and the timing chirp

I’m pretty new to all this and i'm trying to learn as fast as I can, but the results clearly reflect my lack of knowledge, lol. I forgot to remove “sub right old”, that was just a comparison between the sub's old and new position.

Also, for the “timing” measurements, I disconnected one main speaker (the one furthest from the sub, in this case, the left one), on the other channel I've played the chirp (right one). Then i've played the sweep only on the sub (with the left main disconnected). Then I reconnected the left main and played the sweep again, this time with the sub off.

If I understand correctly, I should have followed what you described, the measurements should be “solo sub timing” and “solo main timing.” I’ll redo everything tomorrow just to be sure!

Based on flatness of excess group delay, and the trend from 6 ms to 5 ms measurement, 4 ms delay might be the ticket. This is 5 ms, and it is already nearly a straight line near crossover at 80 Hz and the surrounding area. I don't see any frequency being cancelled near 80 Hz and GD is not spiking anywhere nearby, so I think it's mostly in phase, too.

View attachment 444673

Nice! Sounds great. I’ll also take a measurement at 4 ms tomorrow and see if anything improves.

 

[kyuu]

There doesn’t seem to be much cancellation in these one, but in some areas, the response with the sub turned on is actually lower than without the sub. That’s why I was skeptical that I’m really getting the best result just by trial and error.

Yeah the thing is that the room mode-induced dips will get worse along with the peaks when you have the sub injecting more energy in those frequencies. But honestly with those high Q dips a few dB one way or the other probably isn't going to be audible.

 

[alankila]

If Wiim uses 2nd order crossovers they will probably all by themselves cause about 3-4 ms of group delay that rises near 80 Hz and continues below. This is probably the difference you should be shooting for between the sub and what is in the mains passband in the bass. Group delay should not show a hump around 80 Hz, nor a notch.

Wiim's sub out features a 2nd order crossover, and these are supposed to flip the phase between the low- and high-passed signal, but Wiim might already have flipped it back for you, so you don't have to flip it yourself. It may or may not be in phase once the timing is right, and based on what it looks like, it is in phase when not being inverted by the setting.

You will go mad trying to figure out which frequency response is better. You're going to equalize the frequency response either way, and acoustics is a huge factor in what it looks like. I only looked at the excess GD, because I can read a notch as a big spike in excess GD, and I didn't see any. This "excess group delay", requires some explaining, I guess. It is available after computing the minimum phase version of the impulse response, and is the difference between measured group delay and equivalent minimum phase version of the frequency response's group delay. In this context, we can considered it to show what the group delay of your system would be, if it were perfectly equalized.

 

[Chrise36]

First things first, thanks everyone for the help!

Yeah, I know, unfortunately most of these dips are probably going to stay no matter what. But what concerns me a bit is that these measurements are actually the “better” ones. The others were so bad I deleted them. There doesn’t seem to be much cancellation in these one, but in some areas, the response with the sub turned on is actually lower than without the sub. That’s why I was skeptical that I’m really getting the best result just by trial and error.


I’m pretty new to all this and i'm trying to learn as fast as I can, but the results clearly reflect my lack of knowledge, lol. I forgot to remove “sub right old”, that was just a comparison between the sub's old and new position.

Also, for the “timing” measurements, I disconnected one main speaker (the one furthest from the sub, in this case, the left one), on the other channel I've played the chirp (right one). Then i've played the sweep only on the sub (with the left main disconnected). Then I reconnected the left main and played the sweep again, this time with the sub off.

If I understand correctly, I should have followed what you described, the measurements should be “solo sub timing” and “solo main timing.” I’ll redo everything tomorrow just to be sure!

Nice! Sounds great. I’ll also take a measurement at 4 ms tomorrow and see if anything improves.

Time allign until you get the max spl and then correct FR of the system don't worry too much you need multiple subs to flatten the FR.

 

[Arianoxx]

Today I finally had a bit of time and did some more measurements, including redoing the timing reference like @Keith_W suggested. This time I kept even the really bad-looking measurements in the file, I hope that helps with understanding what's going on.

Unfortunately, the delay suggested by REW’s alignment tool is still quite different from the one that gives me the highest SPL, again. It’s probably something like what @alankila mentioned, actually, when I flip the polarity, the alignment tool's suggestion tends to make more sense.

Anyway, 5 ms at 0° and 12 ms at 180° still give me the highest SPL overall.

I really hope I can get a second sub in the future, but damn, here in Italy they’re so much more expensive compared to the US or other countries. And it's not easy to find a great deal on the used market either.

 

[Keith_W]

All right, here are a few things to see.

First, please stop giving your curves confusing names like this. A casual reader would think you have a left and a right sub. But I know it's only one sub because the frequency and phase response is exactly the same.

This is a good demonstration as to why I don't like REW's automatic timing with the chirp. You took a measurement of your left speaker with the right speaker as a timing reference, and then you took a measurement of your left speaker with the right speaker as a timing reference.

If you centred your mic perfectly between the two speakers, the delta T should be 0ms or close to it. Here, we see a crazy delta T. There is no way your mic is 0.747m and then 1.245m between the two measurements unless you are moving your mic like a drunken sailor. This already tells me that all your timing measurements are unreliable and should be ignored.

FYI this is what I get when I do MY timing measurements. You can see that the delta-T is exactly the same, it is about 10mm off-centre. Yes, I am very good at placing my microphone. I hope you're impressed! My system is very different to yours, in that it is fully active and every driver has its own DAC and amplifier channel, meaning that I can selectively turn drivers off when I take measurements. So I use a different time alignment method. I delay my tweeter by a known quantity (20ms) and then sweep the DUT together with the tweeter and then examine the impulse response to see where the impulse peak is. I deduct 20ms from the result and voila.

So why has REW's tweeter chirp failed? Let us look at your impulse response:

Red = left, green = right.

We can see that the polarity of both your speakers are inverted (initial main deflection is negative). Don't be alarmed, some speakers do this, and what is important is that the polarity of both speakers are the same. In fact, it is very likely that the designer deliberately inverted the polarity of the woofer so that it will sum properly with the tweeter. This is very common.

But what I wanted to show you are the reflections - the red curve has an early reflection arriving which is higher in amplitude than the main signal, and the green curve has a much later arriving reflection which is also higher in amplitude than the main signal.

REW's time reading is only an algorithm. It locks onto what it thinks is the main impulse. And if you have reflections which are bigger than the main impulse, it locks onto the reflection instead. I don't know if it is smart enough to see the inverted polarity and lock onto that as your main impulse?

The other possible issue is that the tweeter chirp isn't loud enough for REW to hear it. When you are taking your measurement, keep a close eye on this display. Make sure you see the start chirp and the end chirp.

I also checked your waterfall to make sure your measurement is sufficiently high above the noise floor. The delta between signal and noise is about 40dB so it's OK in my books. I also checked distortion, also looks OK so you are not measuring too loud. You can strike those two off your list of things to worry about.

Anyway, right now we should be looking at your measurement technique and why you are getting spurious timing readings from all those early and strong reflections. Have you read this thread? Are you using a proper mic tripod and not one of those mini-tripods? Moved your furniture out of the way? No coffee table in between speakers and mic, and no sofa/listening chair?

You want to make sure that you can take timing measurements with consistency. Sweep your left speaker with right timing reference 10 times, and then right speaker with left timing reference 10 times. Look at the result. When it is consistently 0ms (or close to it), you know that you have centred your mic properly. Then sweep your sub maybe a dozen times with the timing chirp. Again, what you are looking for the reading to be the same between measurements. Once you have consistency, you can now start to take timing measurements.

 

[alankila]

Unfortunately, the delay suggested by REW’s alignment tool is still quite different from the one that gives me the highest SPL, again. It’s probably something like what @alankila mentioned, actually, when I flip the polarity, the alignment tool's suggestion tends to make more sense.

I like the excess GD plot the most at 4 ms delay. The result reads like about flat level until crossover point and then it gradually reduces towards the main. The difference in timing is hard to read, but I'm expecting it to be 3-4 ms which it seems to be, so that checks out as well. I think this timing is as good as it can be. Based on frequency response -- and all that nasty waving the actual group delay does -- you have quite a lot of nulls and similar acoustic problems in your system, though. Stuff around 35 Hz, 65 Hz, 170 Hz, 230 Hz...

Your 10 ms 0 deg file shows notching around 80 Hz by the way, like two notches placed fairly close and a small hump in between. I guess one system is overpowering the other at the exact cancellation frequency near 80 Hz. If we are nearly completely out of phase at 80 Hz, then it follows that we are about half cycle delayed at 10 ms. How long does it take for 80 Hz sound to flip phase? Answer: 6.25 milliseconds, and this argument points to about 3.75 ms being very close to the right value. I wouldn't worry about 0.25 ms though, the cycle times are so long in bass that as long as we are within a ms or two, the integration is probably going to be fine.

 

[Arianoxx]

First things first, I really appreciate all the help you guys are giving me, I'm learning so much, and I can't thank you all enough.

I think I finally figured out why REW timing was off. I was using AirPlay to stream the signal to the WiiM, and that probably messed up the timing reference. Now I’ve redone everything , this time paying close attention to every detail and using a jack-to-RCA cable instead.

I was already using a tripod, but this time I positioned the mic very precisely. Now, when I play with the delay values in the alignment tool, they actually match with what I get in the measurements! (Even if I haven’t found the time delta, so idk if that’s improved)

That also allowed me to experiment with different crossover settings, like 60 Hz, which actually doesn’t seem bad at all! (in the file I've tried just two setting because it was late and I cannot play the sweep anymore).

By the way, I haven’t shown the speaker position yet, but yeah... as I mentioned before, the positioning is pretty much the opposite of ideal. There’s not much I can do about it besides moving the sub and speakers around the shelving unit.

 

[DWPress]

Yes, Airplay will totally mess things up as I recently re-discovered myself, still ok for MMM or a sweep that doesn't need to be correlated with others. The 60 & 80Hz XO both have trade offs, probably neither very noticeable, everything else looks as good as it can. I'm afraid you won't get anything much better unless you're willing to move some speakers around.

 

[Keith_W]

Your new measurements still have unreliable timing. Please think very carefully about what is being measured when the timing chirp procedure is done.

1. The time relationship between the tweeter timing chirp and the impulse of the driver under test (DUT) is constant. Airplay will not affect this, unless you are doing a loopback timing measurement (which you are not doing, and is impossible anyway because you are using a UMIK-1). We are able to take offline tweeter chirp timing measurements, and the variable latency with that method is an order of magnitude greater than the variable latency of Airplay. So Airplay is not the culprit.

2. I also examined your impulse to see if I can find the tweeter timing chirp. This is one of my measurements, the tweeter timing chirp is invisible unless you zoom in a bit. But you can clearly see it. Left is a zoomed out view, right is a zoomed in view. Your impulse response has no visible tweeter timing chirp, and I zoomed in all the way to look for it.

This is where I expect to see the timing chirp. There is nothing there.

I will reiterate what I said in my previous post:

1. Make sure you hear and see the tweeter chirp when taking the measurement,
2. Repeat the measurement several times until you are certain that you have consistency.

 

[DWPress]

I can verify if you're using a laptop with a Umik and send sweeps via airplay to the main system the results are undependable in the time domain.

 

[alankila]

By the way, I haven’t shown the speaker position yet, but yeah... as I mentioned before, the positioning is pretty much the opposite of ideal. There’s not much I can do about it besides moving the sub and speakers around the shelving unit.

This picture suggests to me that right channel could be the better to match with (more distance to side wall). That means that I would want to use the "l ref" files, for measuring the right channel and the sub together. I don't know why you don't use the same channel as timing reference, though? Not that it matters as long as the acoustic reference is the same, I guess.

REW autoalign seems to think that 4.1 ms is doing the right thing, anyway, using the "align phase at cursor" function, with cursor around 80 Hz. I adjusted gain a little to make the overall level match (sub is too quiet).

 

[alankila]

Another thing. Measurement shows that sub is being cancelled by something near 80 Hz. This is probably a poor crossover frequency because it is possible that both phase and level are disturbed by some cancellation simultaneously. In this case, I think phase still looks to be close to correct. Regardless, you might avoid some of that null and achieve cleaner crossover around 70 Hz, where the levels of sub and mains can be made to align very well, suggesting that acoustic environment between the two is quite similar near that frequency. Delay estimated for that by REW is 4.3 ms.

For integration, we are concerned with the level and the timing, and we can and should adjust the frequency response via equalizer, because you can never read a whole lot from the response other than maybe the absence of notches. Turn your sub up by about 1.7 dB, and match the phase at crossover with 4.3 ms delay. Phase matching can also be done with the excess group delay of total system, which is what I've used, because it will have a generally flat shape that rises a little due to the crossing over filter when it is correct, but it clearly isn't as precise a method as the alignment tool allows. Still, I'd say it is close enough for integration, either way.

The next step is to perform room equalization. Bring the peaks down, either by manually designing filters, or let something like REW design them for you. You would benefit from taking full 10-20000 Hz sweep for this, and should equalize peaks down between something like 20-300 or perhaps even up to 20-400 Hz if you're feeling particularly adventurous. REW has some fairly reasonable basic room response target shape which could be a good starting point for this. Equalization can't make your system sound perfect, but as long as you don't allow any overall gain, you will avoid boosting the numerous notches in your system, and you should bring your elevated 35-60 Hz region down by something like 4 to 10 dB using filters, and maybe try to approximate the right level between the severely peaking and notching region of 100-300 Hz so that overall tonality would be closer to correct. Room correction can't do wonders in this case due to the poor acoustics and placement, but it can still make the lowest bass octaves quite reasonable, and take care of some severely booming frequencies.

 

[alankila]

Out of interest, I performed the same phase alignment process myself. This is with some Genelec studio equipment in reasonably well damped room, so the bass runs to way below 20 Hz and the system has some built-in phase correction, though only to a degree. The advantage of using matched equipment from a single reputable manufacturer is that phase alignment between sub and mains is quite good from 40-120 Hz. The level alignment deviates more, but is still reasonable from 30 to 110 Hz, I guess. I was previously estimating the mains delay from the group delay of the summed system, and ended up selecting a delay of 6.5 ms based on what I expected the overall group delay of the summed system to look like, but it seems I made a mistake and there should be 2 ms more group delay with the sub than I thought. The evidence from this measurement clearly indicates that 4.5 ms is the right delay value for my setup. (Note that this makes the summed response worse near 30 Hz, but there is a good reason for that, and when you integrate, it's all about the timing, not about the response -- you can equalize many of these differences away. For instance, I can reduce the level of mains near 30 Hz where the phase is going wrong to reduce the problem.)

Genelec only provides a high order crossover, which I believe to be LR8, which causes a very severe group delay shift when used, up to about +20 ms. I think I heard some phasing artifacts and generally uneven response in bass in the room, so I've here opted to just run the subwoofer without a crossover, relying on its natural 150 Hz cutoff and the fact that multiple sources of bass in room sound more even than single one. This approach makes some of the nicest, tightest bass I've ever heard. The phase error near 30 Hz runs up to 90 degrees, too but there's +3 dB boost near 30 Hz due to spinorama far-field eq and I think I'll just remove that to reduce the problem. The room mode near 120 Hz is also why the mains level is goes so high, and also why I don't care about what the sub is doing there -- it can barely contribute relative to how much the mains are booming. That part of the response is simply destroyed by the room.

 

[Calypso1]

Out of interest, I performed the same phase alignment process myself. This is with some Genelec studio equipment in reasonably well damped room, so the bass runs to way below 20 Hz and the system has some built-in phase correction, though only to a degree. The advantage of using matched equipment from a single reputable manufacturer is that phase alignment between sub and mains is quite good from 40-120 Hz. The level alignment deviates more, but is still reasonable from 30 to 110 Hz, I guess. I was previously estimating the mains delay from the group delay of the summed system, and ended up selecting a delay of 6.5 ms based on what I expected the overall group delay of the summed system to look like, but it seems I made a mistake and there should be 2 ms more group delay with the sub than I thought. The evidence from this measurement clearly indicates that 4.5 ms is the right delay value for my setup. (Note that this makes the summed response worse near 30 Hz, but there is a good reason for that, and when you integrate, it's all about the timing, not about the response -- you can equalize many of these differences away. For instance, I can reduce the level of mains near 30 Hz where the phase is going wrong to reduce the problem.)

Genelec only provides a high order crossover, which I believe to be LR8, which causes a very severe group delay shift when used, up to about +20 ms. I think I heard some phasing artifacts and generally uneven response in bass in the room, so I've here opted to just run the subwoofer without a crossover, relying on its natural 150 Hz cutoff and the fact that multiple sources of bass in room sound more even than single one. This approach makes some of the nicest, tightest bass I've ever heard. The phase error near 30 Hz runs up to 90 degrees, too but there's +3 dB boost near 30 Hz due to spinorama far-field eq and I think I'll just remove that to reduce the problem. The room mode near 120 Hz is also why the mains level is goes so high, and also why I don't care about what the sub is doing there -- it can barely contribute relative to how much the mains are booming. That part of the response is simply destroyed by the room.

View attachment 445290

Just a quick question. I thought the phase alignment would be between L+R and the sub? I see you just used the L and Sub.

 

[alankila]

Yes, I used one channel only. I made the choice without any expectation that it makes much of a difference. Left channel is nearer to the door, so I figured it might have slightly less room reinforcement in the low frequencies, so I preferred that one for that reason, assuming that either gain or phase reading could be slightly cleaner.