Loudspeaker design measurements - problem with loop-back as timing reference

[Mensink]

Hi guys!

I'm considering switching from ARTA to REW for doing measurements for speaker design. I've been using REW for in-room measurements for a few years now, but I'm finally ready to make a full switch. But I'm running into a challenge that I can't get around. I'm probably missing something obvious. Anyway, I believe AVNirvana to be the best forum to ask for help regarding REW, so here goes!

Today I've been playing around, trying to do measurements with a proper time reference. In ARTA I always used a loopback measurement and I wanted to try the same with REW. What I need is a fixed time reference. That enables me to estimate the relative propagation delays of the drivers to the mic, to remove excess phase, and thus to get reliable phase measurements of the drivers in their enclosures.

I'm using a Focusrite 2i2 audio interface. One output goes to the amp, and the other is looped back to one of the two inputs on the front. The second input is fed the microphone signal.

I'm measuring indoors today. The speaker is a standmount with a woofer on the front and one on the back. The speaker is positioned about midway between the floor and the 2.60m ceiling, the mic is about 60 cm from the speaker. I'm measuring on the main axis, i.e. 0 degrees.

Attachment 1 - Make a measurement
First, I measure the front woofer.

Attachment 2 - Gating the measurement
I set the gating such that I get a clean frequency response without reflections. In this particular setup, that's an about 4.5 ms window after the first arrival of the impulse..

Attachment 3 - Estimate IR delay
Then I estimate the Impulse Response delay (IR delay), to remove the excess phase. I shift and update the timing reference, in order for the next measurement to have the same time reference.

Attachment 4 - SPL & Phase
If you remove the excess phase, you end up with something that looks like the minimum-phase response of the system. This is what I need.

Attachment 5 - Make a measurement, back woofer
Then I disconnect the speaker wires from the front woofer, and connect them to the second woofer on the back. I measure with the same settings as before, plus updated timing reference as per attachment 3.

Attachment 6 - Impulse response and gate, back woofer
The woofer on the back of the speaker is about 50 cm ~ 1.5 ms farther away from the mic than the front woofer. What I'd expect, is for the time reference to be fixed at t=0, and for the signal to arrive at the mic about 1.5 ms after T=0. What seems to be happening instead, is that the window jumps to what seems to be the loudest part of the measurement, which presumably is a reflection off the wall far behind the speaker, about 5.5 meters away.

Attachment 7 - SPL & Phase, back woofer
In the frequency domain, I'd expect to see some excess phase as a result of the pure time delay. However, the SPL and phase response (obviously) don't make any sense.

Attachment 8 - Estimate IR delay, back woofer
The estimated delay should be about 1.5 ms, but instead, it's 35.8 ms.


What is going on here? I presume I'm missing some checkboxes somewhere. What should I do in order to get a fixed time reference?

Thanks!

 

[John Mulcahy]

The issue is the reflection that is larger than the direct signal, since REW will place the window reference at the largest peak. If you change the window ref time to the peak of the direct signal the rest should work OK.

 

[Mensink]

The issue is the reflection that is larger than the direct signal, since REW will place the window reference at the largest peak. If you change the window ref time to the peak of the direct signal the rest should work OK.

Hi John

Thanks for chiming in!

I found the "Window Ref Time" under IR windows. I guess what works is if you select the measurement of the front woofer (the one you'd like to use as a time reference), and then click on "Apply Window To All, Keep Ref Time*[EDIT]". Is there a way to apply this window to, and keep the ref time for all subsequent measurements as well?

EDIT: It seems I was wrong, what I need is the "Apply Window To All" button.

 

[John Mulcahy]

The ref time is only a reference for where the window is placed, it has no effect on phase which depends only on where t=0 is located. If you want to use the same window widths for other measurements you can set them as the defaults in the Analysis preferences.

 

[Mensink]

I'm afraid I'm not totally following. I've already got fixed window width settings, as can be seen in attachments 2 and 6 in the OP.

What I'm struggling with is setting Window Reference Time such that it stays the same relative to the loopback. In the measurement from attachment 2 it's 0.085 ms. With that Ref time I get the window at the specific desired time after the reference measurement of the loopback. In the second measurement (attachment 6) the window widths are the same, but the Window Reference Time is 36.7 ms.

Is it correct that what I need to do, is - after all measurements are completed - select the t=0 measurement and then use the "Apply Window To All" button? Then all measurements will have the same time Window Reference Time, correct? For every next measurement, REW will again place the window reference at the largest peak. It is not possible to use a fixed Window Ref Time, correct?

 

[John Mulcahy]

Yes, that's right. The peak of the IR will always be used as the ref time for a new measurement but you can apply window settings including a ref time using Apply windows to all.

 

[dcibel]

"Apply windows to all" is your friend, for keeping same window for the complete set of measurements. Set window and "apply to all" once all measurements are complete. Unlike ARTA, the phase represented will not be shown relative to the window reference, but rather its relative to t=0, so it includes the excess delay from time of travel from speaker to mic, as you can see in the very dense phase plot in your screenshots. One way around this is to use a timing offset when measuring, set timing offset to distance from speaker to mic. Alternatively, you can adjust t=0 after measuring to move the impulse closer to t=0 to remove the excess phase. Same offset needs to be applied to all measurements to maintain relative phase. Easiest way John has shown to do this is using impulse response tab in the overlay window, here you can move t=0 for all measurements shown at the same time.

 

[Mensink]

Thanks John and dcbel! It finally clicked. Until a few minutes ago I thought having a separate time reference and window reference was a nuisance, but I now see it's actually a great feature! In REW I can now move the window around to get the best possible resolution, without having to worry about messing up the time reference.