Measurement with 20Hz-200Hz sweep results in a very different IR vs 20Hz-20kHz

Krunok

Member
Thread Starter
Joined
Oct 27, 2018
Messages
29
Hello,

my REW version is V5.20 Beta 25 and I measured my left speaker with measurement sweeps 20Hz-200Hz and 20Hz-20kHz. In both cases timing reference signal was set to right speaker.

In the 20Hz-200Hz scenario my left speaker was acting as a SW (which I don't have yet) and in this hypotethical scenario I was trying to measure time delay between my left speaker (with 20Hz-20kHz sweep) and my future SW (with 20Hz-200Hz sweep and my same left speaker acting as SW). I was expecting to get identical time delay figures, but I didn't. Here is what I got:

Freq. response:

25874


Phase response:

25875


.. an Impulse response:

25876


As you can see, while FR is šretty much identical phase response is not, and the one that really differs is IR. From IR graphs I was not able to tell that this is identical speaker playing 2 sweeps with identical time delay.

So, am I doing something wrong, and if so, what is the right way to measure phase and time delay of SW that cannot play above 200Hz so measurement sweep 20Hz-200Hz must be used?
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
2,131
The IR is of course very different, all of the high frequency content is missing from it. Phase response, and hence timing, is almost identical, so everything looks to have worked very well to me.
 

Krunok

Member
Thread Starter
Joined
Oct 27, 2018
Messages
29
The IR is of course very different, all of the high frequency content is missing from it. Phase response, and hence timing, is almost identical, so everything looks to have worked very well to me.
Ok, but as that IR cannot be used can you advise on some other method to measure time delay between main and subwooofer?
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
2,131
I didn't say it couldn't be used, only that it looks different. What is wrong with using the delay figure REW gives you?
 

Krunok

Member
Thread Starter
Joined
Oct 27, 2018
Messages
29
I didn't say it couldn't be used, only that it looks different. What is wrong with using the delay figure REW gives you?
Well, in ideal case both delay figures should be the same as the same speaker was measured with both sweeps, but, as you can see from the screenshot below, they are very different.

25892



I would be able to use reading from the overlay IR graph but there is also some small difference there (circled red), probably the same one that corresponds to the difference in phase responses.

25893
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
2,131
in ideal case both delay figures should be the same as the same speaker was measured with both sweeps
No, they shouldn't. REW has made a good estimate of the timing differences in that system, the phase overlay is evidence of that. It is physically a meaningless comparison to make, however. Measuring a full range system over a limited bandwidth is not the same as measuring a system whose bandwidth is limited. The result of such a measurement has a magnitude response that stops suddenly at the measurement limit with no rolloff and no corresponding phase changes. For such a system to exist the impulse response would have to start far into negative time, as REW's measurement result shows. If you want the speaker to act as a proxy subwoofer you will have to filter the signal fed to it to the bandwidth of the subwoofer and bear in mind that the filtering will have its own group delay which will affect the relative timing. Or you could just get on with measuring the subwoofer, making sure to measure it over a wide enough range to include its response drop off and including any filtering (such as bass management filters or LFE channel filters) that will be applied to it in use.
 

Krunok

Member
Thread Starter
Joined
Oct 27, 2018
Messages
29
No, they shouldn't. REW has made a good estimate of the timing differences in that system, the phase overlay is evidence of that. It is physically a meaningless comparison to make, however. Measuring a full range system over a limited bandwidth is not the same as measuring a system whose bandwidth is limited. The result of such a measurement has a magnitude response that stops suddenly at the measurement limit with no rolloff and no corresponding phase changes. For such a system to exist the impulse response would have to start far into negative time, as REW's measurement result shows. If you want the speaker to act as a proxy subwoofer you will have to filter the signal fed to it to the bandwidth of the subwoofer and bear in mind that the filtering will have its own group delay which will affect the relative timing. Or you could just get on with measuring the subwoofer, making sure to measure it over a wide enough range to include its response drop off and including any filtering (such as bass management filters or LFE channel filters) that will be applied to it in use.

Ok, I see your point. I still don't have a subwoofer, I am trying to establish a method to measure time delay between left main and SW, so I'm using left main speaker to simulate SW.

I can repeat the same measurement with the filter shown below applied to the left main speaker (phase is dotted blue line that is practically all the time at 0deg). That filter pretty colsely matches measured FR of the SW.

Should I use the 20-20kHz sweep with this filter applied?

25914


Is the delay of this filter 185.76ms, as stated with green letters under the "generate" button?

When measuring time delay between actual SW and left main should I do it without filters and then add the the difference in the delay of the main filter and SW filter?
 
Last edited:

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
2,131
Should I use the 20-20kHz sweep with this filter applied?
Typically wouldn't need to measure to 20 kHz, 2 kHz is probably sufficient, but it doesn't hurt to go to 20 kHz. Better to start at 0 to avoid high pass artefacts on the impulse response from starting at 20 Hz.

Is the delay of this filter 185.76ms, as stated with green letters under the "generate" button?
No, a 150 Hz LR2 low pass has a group delay of a little over 2 ms at low frequencies. That 185 ms is the overall length of an 8192 sample response at 44.1 kHz.

When measuring time delay between actual SW and left main should I do it without filters and then add the the difference in the delay of the main filter and SW filter?
When measuring with a sub use whatever bass management processing or crossover filtering will normally be in place, don't need any extra filtering.
 

Krunok

Member
Thread Starter
Joined
Oct 27, 2018
Messages
29
Typically wouldn't need to measure to 20 kHz, 2 kHz is probably sufficient, but it doesn't hurt to go to 20 kHz. Better to start at 0 to avoid high pass artefacts on the impulse response from starting at 20 Hz.

No, a 150 Hz LR2 low pass has a group delay of a little over 2 ms at low frequencies. That 185 ms is the overall length of an 8192 sample response at 44.1 kHz.

When measuring with a sub use whatever bass management processing or crossover filtering will normally be in place, don't need any extra filtering.
Thank you for the reply!

Let me explain the scenario I'm targeting in more detail: currently I'm running a pair of Castle Harlech S2 speakers with room correction filters which I designed mannually with rePhase based on the measurements I did with REW. I plan to add a Kef Cube 12b subwoofer. My room can accomodate only 1 SW, close to the left main speaker. Based on the discussion I had with Thomas Durgeon (author of the rePhase) and measurements of my mains and SW, I plan to add a high pass LR filter at 40Hz (48db/octave) to my main filters and create a low pass filter (with room correction) for the SW at 140Hz.The idea is to have mains work together with the SW in the 40-140Hz range to better energize the room in order to have less issues with room modes.

Kef Cube 12b has it's own filtering which can be disabled when switched in "LFE" mode as in that mode it relies on external source to do the filtering, althoug I am not sure if that means that no digital processing will be active in the sub. Also, as filter for the sub that I will create will be shorter than filters for the mains it will probably have less delay than the mains filters. My idea is to measure the time delay between left main and SW so I can configure the convolution engine properly so I can do phase integration between the left main and SW.

So, what would be your advice on the best way to measure time delay between left main and the sub?
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
2,131
There is not necessarily any inherent delay in using an FIR filter to apply an IIR response such as an LR HP beyond that of the LR filter itself. Easiest is probably to measure left+sub together as a full range measurement and look at the spectrogram in wavelet mode (example settings below). The peak energy trace will show you the overall relative time alignment across the band.

1570963765085.png
 

Krunok

Member
Thread Starter
Joined
Oct 27, 2018
Messages
29
There is not necessarily any inherent delay in using an FIR filter to apply an IIR response such as an LR HP beyond that of the LR filter itself. Easiest is probably to measure left+sub together as a full range measurement and look at the spectrogram in wavelet mode (example settings below). The peak energy trace will show you the overall relative time alignment across the band.

View attachment 25916

Here is how that spectrogram currently looks for left main speaker, measured at app 1 meter, on axis:

25917


What do you expect to be different when sub will be playing together with left main?
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
2,131
More delay in the peak energy at the low end, most likely, relative to the high frequencies. You don't need to use a timing reference for those measurements, since it is relative delays within the response that are of interest, or can use the "Estimate IR delay" control to remove the delay so the high end sits at or close to 0 for easier reading of the low end delay.
 

Krunok

Member
Thread Starter
Joined
Oct 27, 2018
Messages
29
More delay in the peak energy at the low end, most likely, relative to the high frequencies. You don't need to use a timing reference for those measurements, since it is relative delays within the response that are of interest, or can use the "Estimate IR delay" control to remove the delay so the high end sits at or close to 0 for easier reading of the low end delay.
I'm using Volumio, which is UPnP renderer, to play music from NAS. I cannot play anything from my notebook so I always have to use timing reference when doing sweeps. My initial idea was to use right speaker for timing reference when measuring IR of left main and subwoofer.

Btw, does this spectrogram graph looks ok to you?
 

Krunok

Member
Thread Starter
Joined
Oct 27, 2018
Messages
29
Some issues around 250 - 800 Hz in that measurement.
Can it be due to side wall/floor/ceiling reflections? My speakers have 2nd woofer firing upwards up to 1800Hz and left speaker is app 30cm from side wall.

25922


This is the same graph measured from 30cm distance:

25921
 

Krunok

Member
Thread Starter
Joined
Oct 27, 2018
Messages
29
It seems you advice was spot on! :)

Here is IR made with 0-20kHz sweep with left main with SW simulation filter vs IR with no filter and delayed by the 185.76ms (length of the SW simulation filter):

25924


So, does this mean that this method would measure time delay between SW due to DSP/internal filter/small distance difference between SW and left main?
 
Top Bottom