Another dual-sub alignment question (aka Subs for Dummies like me)

[klettermann]

Sonically I'm getting good results, though I'm still not satisfied that I actually understand the process. And rabbit holes being what they are, I can't resist fiddling until I'm confident that I understand what I'm doing. I'm hoping that you kind folks will indulge me again as I run through what I understand to be going on, along with a couple questions along the way. Here 's the scenario and my logic (???):

1. Layout: Sub 1 is in front right corner about 3ft behind right spkr. MLP is about 14ft away from sub 1. Sub 2 is in rear left corner and about 8ft away from MLP.
2. Sub 1 REW timing measurement is -1.44ms. This means that the mains need a 1.44ms delay to be time aligned with the front sub at mike position/MLP.
3. Sub 2 REW timing is 3.19ms. This means that signal arrives at MLP 3.19ms before timing signal from the mains. And that further means that the rear sub needs 3.19ms delay for signal to arrive at MLP at the same time as the mains.
4. Since both subs are now aligned with the mains per above it follows that they're necessarily aligned with each other.

It's worth noting that the REW alignment tool result is dramatically improved by inverting rear sub 2. It also turns out that setting a 3.19ms delay on the second sub in the alignment tool seems to be optimum. At the same time, these delays don't correspond at all with the actual physical distances involved. So, I'm left with the feeling that there's something I don't understand. Or do I? I'm using MiniDSP SHD Studio for all this. As always, thanks for any comments and let's leave MSO out of it for the time being. I'd like to first grasp how to do it manually first. Cheers,

 

[klettermann]

OMG. It just occurred to me that, in the time it took to write the above, I could have just measured and tested.
I will do that shortly......

 

[Sokel]

Sonically I'm getting good results, though I'm still not satisfied that I actually understand the process. And rabbit holes being what they are, I can't resist fiddling until I'm confident that I understand what I'm doing. I'm hoping that you kind folks will indulge me again as I run through what I understand to be going on, along with a couple questions along the way. Here 's the scenario and my logic (???):

1. Layout: Sub 1 is in front right corner about 3ft behind right spkr. MLP is about 14ft away from sub 1. Sub 2 is in rear left corner and about 8ft away from MLP.
2. Sub 1 REW timing measurement is -1.44ms. This means that the mains need a 1.44ms delay to be time aligned with the front sub at mike position/MLP.
3. Sub 2 REW timing is 3.19ms. This means that signal arrives at MLP 3.19ms before timing signal from the mains. And that further means that the rear sub needs 3.19ms delay for signal to arrive at MLP at the same time as the mains.
4. Since both subs are now aligned with the mains per above it follows that they're necessarily aligned with each other.

It's worth noting that the REW alignment tool result is dramatically improved by inverting rear sub 2. It also turns out that setting a 3.19ms delay on the second sub in the alignment tool seems to be optimum. At the same time, these delays don't correspond at all with the actual physical distances involved. So, I'm left with the feeling that there's something I don't understand. Or do I? I'm using MiniDSP SHD Studio for all this. As always, thanks for any comments and let's leave MSO out of it for the time being. I'd like to first grasp how to do it manually first. Cheers,

Is there anything in your setup reporting processing time, latency, etc?
Cause you have to add those too at the final result, physical distance is only one factor.

 

[klettermann]

Is there anything in your setup reporting processing time, latency, etc?
Cause you have to add those too at the final result, physical distance is only one factor.

Hmmm. I guess not, I'm just reporting the numbers from REW. But wouldn't the REW numbers include that stuff or net it out since it's recording the actual timing as REW "perceives" it after going through the MiniDSP, DACs etc? The setup is UMC-1 mike --> PC --> MiniDSP --> DACs --> amps --> speakers.

 

[Sokel]

Hmmm. I guess not, I'm just reporting the numbers from REW. But wouldn't the REW numbers include that stuff or net it out since it's recording the actual timing as REW "perceives" it after going through the MiniDSP, DACs etc? The setup is UMC-1 mike --> PC --> MiniDSP --> DACs --> amps --> speakers.

REW doesn't know the internal processing time subs need for example. It just measures in>out.
Whatever delay in between is added to the arrival time and that's what mic is seeing.

 

[klettermann]

No, REW certainly doesn't know the internal processing time. But that time doesn't change regardless of speaker position, so don't measured times have that baked in and constant, making measured times are still valid as offsets for position differences? Equally confusing are the actual measurements themselves, which all correspond to much shorter distances than the tape rule method would indicate. It seems like the times should at least be = to the calculated tape measure distance if latency = 0, and latency then adding to that. So there's something I'm doing very wrong or I don't understand any of it.

Thanks and cheers,

 

[andyc56]

Are you using an acoustic timing reference for these delay measurements? If not, the delay measurements can't be relied upon.

The following discussion only refers to main speakers, as sub delay measurements have an additional set of problems whose discussion should be deferred until the simpler case of main speakers is understood.

Even if you are using an acoustic timing reference, a measured delay from a speaker to a seat only represents the difference between two distances: the distance from the timing reference speaker to the seat and the distance of the speaker being measured to that seat. So if the speaker being measured has a shorter distance to the measurement position than the timing reference speaker, you'll get a negative measured delay for that speaker. That's because the timing reference itself gets defined as zero delay, and the speaker being measured has less delay than that (because it's closer to the measurement position), so its net measured delay is negative.

So how do you make sense of these measured delays then? Well, if one speaker has a measured delay of -1.0 msec and the other +2.0 msec to the main listening position, this says you need to add a delay of 3.0 msec to the speaker with the measured delay of -1.0 msec to make the delays equal.

We don't actually care what the absolute distance or delay is, only how much delay we need to add to a speaker or speakers to equalize the delays. There are ways one can get close to being able to accurately measure absolute distance and delay, but not with a USB mic. It requires an analog mic with a loopback timing reference, along with some other considerations.

 

[klettermann]

Thanks muchly, very clear. You go to the heart of what I was trying to get at.

Are you using an acoustic timing reference for these delay measurements? If not, the delay measurements can't be relied upon.

Yes, using timing reference.

The following discussion only refers to main speakers, as sub delay measurements have an additional set of problems whose discussion should be deferred until the simpler case of main speakers is understood.

Even if you are using an acoustic timing reference, a measured delay from a speaker to a seat only represents the difference between two distances:the distance from the timing reference speaker to the seat and the distance of the speaker being measured to that seat. So if the speaker being measured has a shorter distance to the measurement position than the timing reference speaker, you'll get a negative measured delay for that speaker. That's because the timing reference itself gets defined as zero delay, and the speaker being measured has less delay than that (because it's closer to the measurement position), so its net measured delay is negative.

Yeah, that was my point. And latency is the same for both, so baked-in so to speak.

So how do you make sense of these measured delays then? Well, if one speaker has a measured delay of -1.0 msec and the other +2.0 msec to the main listening position, this says you need to add a delay of 3.0 msec to the speaker with the measured delay of -1.0 msec to make the delays equal.

We don't actually care what the absolute distance or delay is, only how much delay we need to add to a speaker or speakers to equalize the delays. There are ways one can get close to being able to accurately measure absolute distance and delay, but not with a USB mic. It requires an analog mic with a loopback timing reference, along with some other considerations.

OK, I think I got it now. Time to go downstairs a play around a little.... Cheers,

 

[Keith_W]

I am trying to develop a better time alignment procedure for REW based on how Acourate does time alignment. I think a lot of the confusion is because you are not made to read the subwoofer peak yourself, REW automatically reads it and spits it out for you.

This is how it is done in Acourate:

1. The tweeter is delayed by a certain amount of time. Acourate specifies delays in samples, in this case 1000 samples. 1000/48000 = 20.83ms.
2. A sweep is taken of the DUT with the tweeter reference.
3. The impulse is examined and the DUT peak is read, like this:

Zoomed out view of the impulse response. The only peak that is easily visible is the tweeter impulse, which Acourate centres at sample 6000.

If you zoom in, you can see the subwoofer peak. I positioned markers to the left/right of the peak and read the position of the peak. In this case, it is at sample 4715.

Since the tweeter was delayed by 1000, and Acourate centres the tweeter at 6000, we expect the subwoofer peak to occur at sample 5000. But it is at 4715, which means it is 285 samples too early.

I like this method. You can see with your own eyes where the subwoofer peak is, and that gives you certainty. If I were to replicate this method in REW, this is how I would do it:

Delay the timing reference by a known quantity, in this case 50ms. Then take a sweep of the DUT with the tweeter reference. You will see two peaks in the impulse response. In fact you may not see it unless you zoom in (take note of the vertical scale in my Acourate window). The big one is the tweeter, and the small one to the left of it is the DUT. Measure the timing difference between the DUT and the tweeter, then subtract 50ms.