Setting subwoofer delay

[ppataki]

I' NOT a room-correction expert...



Phase differences will ONLY be an issue in the narrow band around 80Hz where the regular woofer and sub are working together (where the soundwaves might be partially or completely out-of-phase relative to each other).

Assuming there is no "different DSP processing" to the high-pass or low-pass signal paths, and assuming neither one is inverted relative to the other, the ONLY phase-timing issues will be from difference in distance between the sub and the main speaker to your listening position.

The wavelength at 80Hz is 14 feet. So a half-wavelength is 7 feet and if the sub or main speaker is 7 feet closer than the other, you'll have a 180 degree phase-shift (at 80Hz) and the soundwaves will cancel. That's the worse case. (At 1.5 wavelengths you'll get 540 degrees and cancelation.) It wouldn't be unusual to have the main & sub 7-feet apart but a 7 foot difference would be unusual at home.

And since sound travels about 1 foot per ms, 7 feet is about 7 ms of time difference. (If you delay the nearest speaker by 7ms they will be back in- phase at 80Hz).

I have always got better results when using linear phase crossover vs minimum phase crossover

This post is about my main system:
Post in thread 'Lavoce SSF153.00 15" based dual opposed subwoofer project' https://audiosciencereview.com/foru...-opposed-subwoofer-project.51178/post-1852855

And this one is about my home office system:
Post in thread 'Another 8" dual opposed subwoofer project' https://audiosciencereview.com/foru...-opposed-subwoofer-project.55078/post-2033114

 

[Keith_W]

If you are using a UMIK-1 it is not accurate enough for timing measurements like this due to USB bus issues. You will never know what is correct and what is wrong. A UMIK-2 or an analog Mic using loopback is the best way to get timing measurements. See link to developer of REW showing the issue https://www.avnirvana.com/threads/acoustic-phase-measurements.13300/#post-99941

If you look closely at the horizontal scale, you will see the timing variance to be about +/- 50us (microseconds). That's 0.05 milliseconds.

If Paul is using a 48kHz sampling rate, then each sample is 0.02083 milliseconds. So the variance of the UMIK-1 is +/- 2.5 samples.

Next question to ask is what are the thresholds of audibility for bass group delay? The answer is: nobody seems to know. Some people think you need "several cycles" of bass before you even hear it. If you believe them, then the period of one 50Hz cycle is 1/50 sec (20ms) and "several cycles" may be 40-60ms.

In theory then, a UMIK-1 should be able to correct to below the threshold of audibility if you are using the UMIK-1 to time align a subwoofer. Where it might make more of a difference is with higher frequencies.

That said, there are a lot of other reasons to hate USB microphones:

The guy in the video makes a great point about how interfaces are so versatile. When I bought my interface, I never imagined all the different things I would use it for. When I am not using it for measuring, it is my headphone DAC/amp. It has digital inputs, so I can connect external digital sources like a CD player, a Wiim, my DAP, and so on. It has an ADC, so if I wanted to I could connect an analog source. I can use the interface to route signal from one input to another, for e.g. REW has no built-in convolver. I can send REW signal to a software digital input on the interface, and loopback that input to an output. My convolver then takes signal from that output and sends it back to the interface's DAC's.

In fact those interfaces are so versatile that I have 3 of them. Only one of them (the RME) is expensive. The other 2 (Presonus Audiobox USB and Focusrite 2i2) are downright cheap. But they don't do half the things that the RME can do.

 

[levimax]

Next question to ask is what are the thresholds of audibility for bass group delay? The answer is: nobody seems to know. Some people think you need "several cycles" of bass before you even hear it. If you believe them, then the period of one 50Hz cycle is 1/50 sec (20ms) and "several cycles" may be 40-60ms.

In theory then, a UMIK-1 should be able to correct to below the threshold of audibility if you are using the UMIK-1 to time align a subwoofer. Where it might make more of a difference is with higher frequencies.

20 or 40 or 60 ms is a lot. Hard to reliably ABX but having made some "50 ms mistakes" I think that is audible. I agree with your points in general but if 20 to 60 ms is "good enough" there is really no need to measure at all and you would be better off with a tape measure.

 

[Keith_W]

20 or 40 or 60 ms is a lot. Hard to reliably ABX but having made some "50 ms mistakes" I think that is audible. I agree with your points in general but if 20 to 60 ms is "good enough" there is really no need to measure at all and you would be better off with a tape measure.

Believe me, I would love to be wrong. I time align my subwoofers down to <0.5ms and I take a lot of care doing it. I think I hear a difference, it sounds "tight" and "punchy" to me, but that could easily be expectation bias given how much time I spend on it.

But the reality is that I am very likely correcting it to be way below audible. Toole in Chapter 4.8.2 talks about phase distortion in subwoofers. He notes that minimum phase high pass filters are used at multiple stages of the recording process, starting at the microphone, through various mixers, and finally through speakers and subwoofers. He does not know how much group delay has been added even before we attempt playback but it must be "enormous". So he asks: is there any point preserving the phase/time relationship at the last part of the signal chain, when so much harm has been done upstream? And more importantly, does it even make sense doing it if the room is going to distort the phase even more?

Toole cites several studies in his book, all of which report a subtle difference, but he notes that none of them were controlled double blind studies.

The only study I was able to find was by Liski et al using test signals and loudspeakers in a room (not mentioned in Toole's book).

Take note that the group delay thresholds start to shoot up like a rocket below 100Hz. But it does suggest that at 40Hz, a threshold of >10ms is potentially audible. Study can be found here: https://acris.aalto.fi/ws/portalfil...udspeaker_Group_Delay_Characteristics_AAM.pdf

This discussion started off by noting that the UMIK-1 has a variation of 0.05ms. I think it is going to be well below the threshold of audibility (if that is your aim). I think for many of us, "below audible" is not good enough ... we also want pretty graphs. No shame in that, that is my goal too. So that's why I persist even though I know it is probably a fool's errand.

There are many reasons to not use a USB microphone, but I don't think this is one of them.

 

[Titurel]

I would like to add a question to this discussion.

If room gain contributes significantly to the bottom octave or so of subwoofer response, shouldn't we considered it (room gain) when setting the subwoofer time adjustment?

I'm not debating, just asking a question.

 

[ppataki]

I am getting my UMIK-2 next week, will share my experience accordingly

 

[ppataki]

So finally today I got my UMIK-2:

Visually, the difference vs UMIK-1 is the colour and the capsule that looks totally different

It is no longer fully Plug & Play (at least in Windows), meaning that you have to install a driver for it to work properly. Also there is a firmware update capability which I performed since my unit was not fully up-to-date. And you will have to download the calibration file too, just like with UMIK-1

I have performed an acoustic timing reference test in REW with this unit vs my UMIK-1
I have made 5 measurements with each microphone exactly at the same position (MLP) to see how the measured delay value will look like using the front left channel as the reference and the subwoofer far behind my back as the other channel

Results:
- UMIK-2 has measured the same delay value with a two decimal precision for all 5 measurements, differences started to appear only from the third decimal (and we are talking milliseconds here)
- UMIK-1 has measured the same delay value up to the integer value only, meaning that there were differences starting with the first decimal already. This is definitely audible, especially in case of full-range channels

Additionally, I have performed a full-range measurement using the front channels of my home office system without any DSP:

Obviously I cannot tell which one is 'better' or 'more correct' but there is quite significant differences throughout the whole frequency domain reaching up to 3dB!
This means that there will be audible differences when you calibrate your system with an UMIK-1 vs UMIK-2

I am yet to try this microphone with Dirac Live 3 - will post my experience later on

So far so good, the UMIK-2 looks to be a keeper for me!
The measurement file is too big to attach, I will upload it to Google Drive and share here later today (I need my other PC to do that)
EDIT: see the measurement file here https://drive.google.com/file/d/1zP-0OmXTlPJGLRQJLkdRq7y7ZSAEKVVF/view?usp=sharing

 

[Paffi]

The following impulse response shows two measurements (each 20-200hz): Subwoofer and Front Left. I´m curious which impulse you would choose for alignment, as one or the other would need completely different delays. Aligning the first valleys around 4ms, choosing the two biggest valleys around 12,6ms. Choosing the biggest peak it would be around 21ms. So what would you choose? Peak or valley, first or strongest?

If I would choose the delay according to the acoustic timing reference (that´s why I choose the frequency range 20-200hz on both), it would be no further delay, as fronts and subs a nearly in sync. This is probably represented in the attached measurement, but it isn´t the first or even stronges peak/valley...

The alignment tool of REW also shows many different values, depending on the frequency choosen for crossover.

e.g.
140hz, impulse alignment: 28 ms
140hz, phase slope alignment: 16,5 ms
140hz, phase alignment: 17,3ms

100hz, impulse alignment: inverting and -10,6 ms
100hz, phase slope alignment: 15,2 ms
100hz, phase alignment: 18,2ms

So as a summary: different suggestions of REW tend from nearly zero delay up 21ms or even inverting.

 

[ppataki]

The following impulse response shows two measurements (each 20-200hz): Subwoofer and Front Left. I´m curious which impulse you would choose for alignment, as one or the other would need completely different delays. Aligning the first valleys around 4ms, choosing the two biggest valleys around 12,6ms. Choosing the biggest peak it would be around 21ms. So what would you choose? Peak or valley, first or strongest?

If I would choose the delay according to the acoustic timing reference (that´s why I choose the frequency range 20-200hz on both), it would be no further delay, as fronts and subs a nearly in sync. This is probably represented in the attached measurement, but it isn´t the first or even stronges peak/valley...

The alignment tool of REW also shows many different values, depending on the frequency choosen for crossover.

e.g.
140hz, impulse alignment: 28 ms
140hz, phase slope alignment: 16,5 ms
140hz, phase alignment: 17,3ms

100hz, impulse alignment: inverting and -10,6 ms
100hz, phase slope alignment: 15,2 ms
100hz, phase alignment: 18,2ms

So as a summary: different suggestions of REW tend from nearly zero delay up 21ms or even inverting.

I would take an acoustic timing reference measurement using an UMIK-2 or an XLR analog mic or equivalent (to avoid clock drift issues) like this:
L --> L
L --> R
L --> Sub

Then calculate the correct delay values to apply to each channel as stated in post #8 then check the results in REW (check not only the frequency response but also the Wavelet diagram to see the temporal properties)
Then I would invert the polarity of the sub and verify it with another measurement with the same delay values (again, check the FR and Wavelet curves)
Choose whichever looks better and sounds better
This way I always managed to align subs very well (even multiple subs)

Also it is a good practice to measure the physical distance between the reference channel (L in this case) and the sub and check if the measured delay corresponds to it at least approximately. 21ms equals to like 720cm, is your sub really that far away from the reference channel?

 

[Paffi]

Clock drift / timing issued due to the mic are nearly irrelevant. Several measurements show minimal differences, so let's assume it's fine. Measuring the distance makes no sense, as dsp of the sub causes also delay and it's not only one sub. Let's also assume there are no big differences in spl and GD / specto also show little discrepancies. Your method according #8 is included in the above post whereas I always measure 20-200hz instead fullrange, to give rew the "right" impulses.

As shown above the impulse response method would lead to several competely different values.

 

[ElJaimito]

I know this topic has been discussed before but I would like to provide a new approach (or ask if that makes any sense at all)

I have tried different methods to set sub delay but all of them resulted in a different delay figures.....no consistency whatsoever
Using acoustic timing reference in REW (two different kinds), using Excess Group Delay curve, using the Wavelet Spectrogram - ended up with totally different results and none of them sounded good

An alternative I am trying now is this:
Copying the front channels to the sub then turning off any crossover then start to delay the fronts by an increment of 0.3ms (or 0.1ms) and in the meantime checking the RTA curve in real-time

My assumption is that when the RTA frequency curve looks the best (=highest SPL in the low register) that is where the drivers are aligned

Is that a correct assumption? I would appreciate any comments
Many thanks

I use a tape measure to get the path difference and start with that figure.

 

[levimax]

Clock drift / timing issued due to the mic are nearly irrelevant. Several measurements show minimal differences, so let's assume it's fine. Measuring the distance makes no sense, as dsp of the sub causes also delay and it's not only one sub. Let's also assume there are no big differences in spl and GD / specto also show little discrepancies. Your method according #8 is included in the above post whereas I always measure 20-200hz instead fullrange, to give rew the "right" impulses.

As shown above the impulse response method would lead to several competely different values.

I agree with @ppataki that unless you can get impulse measurements with an accurate timing reference it is difficult to line things up. For instance I have SVS subs which have ~6ms delay but they also have "pre-ringing" which I never even thought about. When I tried to time align without an accurate timing reference it was really hard going by "which peak" to line up. With the accurate timing reference it was "easy".

As a sanity check and maybe even all you need is that if you can get an accurate near field "delay" measurement for your sub(s) then with a tape measure you can easily calculate the delays.

 

[Paffi]

Well, as I said, I already checked how consistent the Umik 1 measures in its time behaviour. The timing difference ist absolutely irrelevant. But I have the impression it doesn´t matter how often I repeat this - the main question won´t get answered, until this "matter" is cleared. For that reason I bought an Umik 2 and compared it to the Umik 1. The .mdat is attached.

2 Measurements each.

20-200hz
Umik1
Delay -4,0321 ms (-1,383 m, -(4 ft 6,4 in))
Delay -4,0392 ms (-1,385 m, -(4 ft 6,5 in))

Umik2
Delay -4,0025 ms (-1,373 m, -(4 ft 6 in))
Delay -4,0001 ms (-1,372 m, -(4 ft 6 in))

20hz-20khz
Umik1
Delay -0,0106 ms (-3,6 mm, -0,14 in)
Delay -0,0105 ms (-3,6 mm, -0,14 in)

Umik2
Delay -0,0120 ms (-4,1 mm, -0,16 in)
Delay -0,0120 ms (-4,1 mm, -0,16 in)

As you can see - nothing to note, the reported delay values are absolutely consistent. I hope I can send the Umik2 back and get full refund.

Unless anybody sees a difference here, I would be grateful If we could return to the question.

 

[Paffi]

And here the mdat to play around.

First measurement: Front left Fullrange
Second measurement: Front left Fullrange with Low Pass 140hz
Third measurement: Front left 20- 200hz
Fourth measurement: Subs with Hp 140hz
Fith measurements: Front and Subs aligned

If you play around with the alignment tool and the impulse response window you will see very different results.