Understanding the state of the DSP market

[3ll3d00d]

Yes that is the one, thank you! The measurements more or less agree with what I have. So how would you '"time align" something like this? Is it -2ms or -6ms or?

Generally, if using rew and min phase EQ, i would align the start of the sw ir with the main channel then use alignment tool to move it +/- 5-7ms and examine the phase alignment (and hence sum through the crossover) as you adjust delay. Whatever works best is what I would then use.

 

[3ll3d00d]

Sounds that REW may not provide reliable measurement results, if I understand correctly?

Plenty of people use this fine, there's a whole faq section on it https://www.andyc.diy-audio-enginee...ence-manual/measuring.html#general-rew-timing

 

[markstein]

Plenty of people use this fine, there's a whole faq section on it https://www.andyc.diy-audio-enginee...ence-manual/measuring.html#general-rew-timing

Yes, exactly, that's what I did so far as well. However the question is whether there is a more reliable way of doing the measurements? Maybe I am overcomplicated things...

 

[Chr1]

Have to say that for me this thread has made me realise that there's a fair bit of variation in both approach and results, even amongst those people here with a lot of knowledge and experience using DSP.
Even REWs reliability for measurements is being brought into question.
I continue reading with interest, but I find it curious there appears to be no real consensus on what the best approach is.

 

[Sokel]

Have you considered these settings? (see attached image)

I think I have the settings right so lets try and see:

Distances while measuring left speaker and mic positioned as follows:

- Mic 1.235 m from Right tweeter
- Mic 100 cm from left mid upper ring
- Mic 125 cm from left port

Results:


other speaker ac. ref



same speaker ac ref



both

Now, REW reports -1.340 m for the "other" speaker ac ref which is close to the 1.235 m. No EQ or any DSP so no delay there, the electronics should be some negligible some.
Is there a chance to calculate by triangulation?

What's interesting is that both generated minimum phase are referring to 0 despite all other traces.

 

[3ll3d00d]

Yes, exactly, that's what I did so far as well. However the question is whether there is a more reliable way of doing the measurements? Maybe I am overcomplicated things...

If you get working results out of mso consistently then it's reliable for you and changing to a hardware loopback isn't going to change anything

 

[Music1969]

I used REW and Acoustic Timing Reference for one thing only - to verify my Acourate DSP correction.

I play REW generated .WAV files with the acoustic timing reference, in my convolver and use a separate laptop running REW to 'listen' for the chirp.

My verification measurements very closely match my Acourate simulation, so I have to assume the acoustic timing reference works just fine.

What happens when you do this @Keith_W ?

 

[Keith_W]

“I would not use the value given there since the woofer's bandwidth is limited (i.e. there is no tweeter). It only works if you are measuring speakers with HF drivers with the range the chirp uses.” Quoting myself.

I can believe that. I noticed that the tweeter chirp is almost useless for measuring subwoofer timing. I took perhaps a dozen sweeps of my subs and examined them all for consistency, and they were hopelessly off with no consistency whatsoever. So I switched to loopback and did a few measurements. I did half a dozen, and all of them were reasonably close. Take a look:

Mind you, I don't believe those subwoofer timings one bit. First, it is -3.75 seconds which would appear to violate the principle of causality. And the total discrepancy between the subwoofer and the mains is 5.8 seconds. I was aghast when I saw this so I rechecked the measurement setup and cross-referenced it with Acourate's loopback timing measurement. The good thing about Acourate is that it allows you to zoom in to the impulse peak and verify the reading yourself. The bad thing is that the excitation signal is a Dirac pulse (a click, not a sweep), meaning it is very low energy.

As you can see, the measurement is heavily contaminated with noise. Regardless, Acourate's loopback measurement gave me roughly the correct value of about 6.5ms (you need to subtract the tweeter loopback measurement which I did not show).

So REW's loopback measurement is consistent, but it is also consistently wrong.

And it's not the left side either. REW thinks my R sub is 5.9 seconds delayed to the R speaker using loopback.

Using the delayed tweeter method that I described earlier, I know that my subs are delayed by about 6.5ms. That seems about right.

This comes across as rather absolute and final but it's possible, likely even, to be user error or related to your particular environment/speakers. As a counterpoint, I have been given measurements by someone using it to design a passive crossover for them (more than once in fact, with off axis measurements as well). I designed it, they built it, measured it, it worked as expected. You can't do this if it's unreliable in general. I have also used it in the past with no obvious issues (albeit I do have hardware loopback so don't generally use it)

Maybe you're right, maybe it's just me. Or maybe not many people repeat measurements, checks them for consistency, and cross-checks them with other methods of timing measurements like I do. So their timing measurements could be completely out of whack and they don't notice it.

I have spent enough time experimenting with REW's acoustic and loopback timing measurements to be really sceptical of it. When I saw those weird readings, my first thought is that I did something wrong. So I used the same setup to take loopback measurements with Acourate just to verify the measurement setup, and I got realistic readings. I heard that some USB ports are less reliable than others, so I tried another USB port. No difference. So i'm out of ideas.

 

[Keith_W]

I used REW and Acoustic Timing Reference for one thing only - to verify my Acourate DSP correction.

I play REW generated .WAV files with the acoustic timing reference, in my convolver and use a separate laptop running REW to 'listen' for the chirp.

My verification measurements very closely match my Acourate simulation, so I have to assume the acoustic timing reference works just fine.

What happens when you do this @Keith_W ?

It is not clear whether you are actually taking timing measurements or not. It sounds as if you are verifying Acourate's filters from 20Hz - 20kHz. The timing measurements only come to play when you are comparing one driver to another, or one speaker to another. You did not describe that. And yes, I can do offline measurements with tweeter chirp too. I just don't like doing them.

 

[gnarly]

Have to say that for me this thread has made me realise that there's a fair bit of variation in both approach and results, even amongst those people here with a lot of knowledge and experience using DSP.
Even REWs reliability for measurements is being brought into question.
I continue reading with interest, but I find it curious there appears to be no real consensus on what the best approach is.

Good observations imo.

I think there is no real consensus because precisely determining subwoofer arrival times, is an impossible task indoors.

Even in anechoic conditions, the nature of FFT measurements' linear data sampling vs frequency, makes sub frequency measurements difficult at best.
The number of data points simply isn't there, like it is for higher frequency.

Then there's the issue of, it's not the 'already hard to find sub impulse peak' we want for time alignment, it's the initial impulse rise from zero.
Good luck putting a finger on that consistently, even if you do have a repeatable impulse peak. It's a mathematical stab.

And still worse, indoors sub reflections are often so near the strength of the direct sub arrival, which one does the mic and FFT listen to?

Frankly, I think the whole idea of sub time alignment indoors is a lot of wishful thinking.....
Even the wavelet tone burst method, the most accurate sub distance finder I've found yet, doesn't have great repeatability.
Still hoping though.....

 

[3ll3d00d]

Maybe you're right, maybe it's just me. Or maybe not many people repeat measurements, checks them for consistency, and cross-checks them with other methods of timing measurements like I do. So their timing measurements could be completely out of whack and they don't notice it.

As noted above, many people use the same for mso and it works, I have designed speakers using such measurements. These things would be impossible if it were as unreliable as you say, ie there's plenty of evidence out there to say it can produce repeatable results.

 

[olieb]

Then there's the issue of, it's not the 'already hard to find sub impulse peak' we want for time alignment, it's the initial impulse rise from zero.

That should not be so difficult. Measure in nearfield and then use a meter measure. That will give you time. [But not if one takes the subwoofer impulse response peak of course ]
The point is that the difficult part is phase and often time shifting is used to correct for phase/group delay.
That cannot work as phase is frequency domain.

 

[gnarly]

That should not be so difficult. Measure in nearfield and then use a meter measure. That will give you time.

Yep. That's the best indoor solution I know of for timing reference. A laser range finder makes a good easy & cheap distance meter, to extrapolate from near-field point.
What I prefer to indoor near-field, is to measure outdoors at 2m ground plane, and establish a timing reference relative to the front edge of the cabinet. I always take outdoor ground-plane meas of a new sub anyway, so i figure why not use best data i can get..

The apparent time shifting from different xover points and orders with their differing phase rotations is illusionary constant time, but I guess inevitable in our FFT measurements.
It's another reason I think sub timing measurements are often bogus.

 

[ernestcarl]

Mind you, I don't believe those subwoofer timings one bit. First, it is -3.75 seconds which would appear to violate the principle of causality. And the total discrepancy between the subwoofer and the mains is 5.8 seconds.

Okay, so we agree the delta listed there for those measurements is not valid for the intended purpose so then skip that.

Did you try to use the Alignment Tool?

 

[levimax]

The apparent time shifting from different xover points and orders with their differing phase rotations is illusionary constant time, but I guess inevitable in our FFT measurements.
It's another reason I think sub timing measurements are often bogus.

Does that imply that if both the mains and sub are passive that you can just "physically measure" to calculate the delays and skip even the nearfield measurements?

 

[gnarly]

I have spent enough time experimenting with REW's acoustic and loopback timing measurements to be really sceptical of it.

Loop back measuring anything with significant frequency response above a few hundred Hz should be repeatedly consistent....within a sample of two.
Loopback with for very low freq work, like 100Hz crossovers and below, has to be suspect....as true for Acourate as for REW.

REW's acoustic timing reference that sweeps from 5kHz to 20kHz is clearly meant only for full-range speakers. I've seen some guys put a piezo tweeter on their sub, to turn it into a distance finder, but other than that acoustic timing is just a set of training wheels for folks to first get first ride on the measurement bicycle, .....imso.

 

[gnarly]

Does that imply that if both the mains and sub are passive that you can just "physically measure" to calculate the delays and skip even the nearfield measurements?

It's not a matter of passive or active, it's a matter of having acoustically complementary crossovers (and having subs and mains co-located.)
When that exists, it means a fixed time delay exists between subs and mains that equals the distance difference between their acoustic centers to measurement location.

That equals best tuning, phase trace overlay thru crossover, correct time alignment, etc
It comes with no fixed delay other than the distance between acoustic centers.
So, the crossover order, provided it remains complementary, like going from LR 12 to LR 48, will not require different delay settings. But FFT measurements will say otherwise.

 

[Keith_W]

Did you try to use the Alignment Tool?

I did use it, and it is a thing of beauty. My jaw dropped the first time I started playing with it and saw how it worked. So elegant! I place t=0 at the IR peak, and then zoom in to confirm that REW has correctly placed t=0 at the IR peak. At least I know where the IR peak is, rather than using the acoustic reference or loopback reference which I have shown - are give different readings.

 

[Keith_W]

REW's acoustic timing reference that sweeps from 5kHz to 20kHz is clearly meant only for full-range speakers. I've seen some guys put a piezo tweeter on their sub, to turn it into a distance finder, but other than that acoustic timing is just a set of training wheels for folks to first get first ride on the measurement bicycle, .....imso.

Putting a piezo tweeter on a sub to determine SW time alignment is an absurd idea. This is one of the misconceptions of time alignment - we are not aligning to the physical location of the subwoofer, we are aligning to the acoustic centre. The acoustic centre does not correspond to a physical location, e.g. voice coil, cone, baffle, etc. On top of that, some subs are ported, that throws even more confusion into the mix. The acoustic centre varies according to group delay, and is therefore frequency-dependent. The only way to determine the acoustic centre is with an acoustic measurement.

Anyway, I am sure you knew all that. It's just a warning for other people who might be tempted to try it

 

[markstein]

Then there's the issue of, it's not the 'already hard to find sub impulse peak' we want for time alignment, it's the initial impulse rise from zero.
Good luck putting a finger on that consistently, even if you do have a repeatable impulse peak. It's a mathematical stab.

I think the better approach is to look at the sub maximum impulse peak and align it correctly, with respect to the tweeter impulse maximum, based on the theoretical ideal x-over step response. This ideal response can be modeled, e.g., using Acourate.