Group Excess Delay plot interpretation?

[andyr]

My Excess Group Delay plot is shown below (no smoothing).

I'd appreciate someone telling me whether or not I'm interpreting it correctly. :

* I have a 24dB LP/HP XO between subs and woofers at 100Hz; the (shallower) XO between woofer and mid is 300Hz. There's another XO between mids and tweeters at ~4kHz.
* the graph shows that either side of 100Hz, the bottom of the plot is at the same level - I'm assuming this signifies that I have applied the correct delay, in my miniDSP plug-in, to the mains ... vs. the subs.
* likewise the graph either side of the mid-tweeter XO (~4kHz) is at the same level (0ms delay).
* however, the graph shows that the subs & woofers are 5ms delayed, compared to the mids & tweeters. (There's a discontinuity at 300Hz.)

Does this discontinuity mean that I should apply a 5ms delay to the mids and tweeters, to "line them up" with the woofers and subs?


Thanks,
Andy

 

[John Mulcahy]

Does this discontinuity mean that I should apply a 5ms delay to the mids and tweeters, to "line them up" with the woofers and subs?

That's what it looks like, yes. You can apply a frequency-dependent window to remove the high frequency reflections by the way, and clean up the plot. Another useful view for time alignment is the wavelet spectrogram.

 

[andyr]

That's what it looks like, yes. You can apply a frequency-dependent window to remove the high frequency reflections by the way, and clean up the plot. Another useful view for time alignment is the wavelet spectrogram.

Great - thanks for the confirmation, John.

I'll come back to you about:
* a 'frequency-dependent window', and
* the 'wavelet spectrogram'

... when I've researched them.


Andy

 

[andyr]

OK, John - have had several days of playing around with my miniDSP settings and REW.

The first attachment is the EGD plot with the delays I had on Jul 26 - with frequency dependent windows applied. It's not exactly the same as the EGD plot from the other day - but I have changed some miniDSP PEQ settings in the last few days!

But it does still show that the subs (whose LP filter is 24dB @ 100Hz) are delayed, compared to the main speakers:

I was able to add an additional 6ms of delay to the 3 drivers in the main speakers, to see whether this would line up the subs with the main spkrs. The 2nd attachment is the EGD plot resulting from this:

Looking at this graph:
* yes, it has lined up the main spkrs with the subs - except I should probably reduce the delay on the woofers by, say, 2ms.
* however, I now have:
- a large positive spike in group delay centred on 80Hz (there is no EQ applied at this frequency), and
- 2x very narrow negative spikes at about 180Hz & 390 Hz.(there is +ve EQ applied at 175Hz & 400Hz).

Do these spikes in the EGD plot show that it is a worse configuration than my first one?

Andy

 

[John Mulcahy]

No, those will usually be effects from room resonances and reflections, though the spikes would normally be positive. Where you get sharp changes there is typically a notch in the response due to a partial cancellation. It is not a good idea to try and EQ those, they are very dependent on location so the response can be very different a short distance away. EQ is best kept below about 200 Hz unless it is very broad (i.e. more of a tone control). Narrow boost filters should be avoided.

 

[andyr]

No, those will usually be effects from room resonances and reflections, though the spikes would normally be positive. Where you get sharp changes there is typically a notch in the response due to a partial cancellation. It is not a good idea to try and EQ those, they are very dependent on location so the response can be very different a short distance away. EQ is best kept below about 200 Hz unless it is very broad (i.e. more of a tone control). Narrow boost filters should be avoided.

Hi John,

The sharp negative spikes I can understand are merely artefacts; I have not used EQ at those frequencies.

However, it's the spike @ 80Hz which concerns me (in terms of "is this bad for the actual sound?"):
* why should changing the delay between subs and main speakers have caused this?
* surely this spike in EGD @ 80Hz cannot be a good thing?

Andy

 

[John Mulcahy]

The group delay plot is handy for checking time alignment. It shouldn't be your main focus, nor should it be used as a proxy for response quality. Better devoting your attention to the frequency response.

 

[andyr]

The group delay plot is handy for checking time alignment. It shouldn't be your main focus, nor should it be used as a proxy for response quality. Better devoting your attention to the frequency response.

Aah, thanks, John.

Andy

 

[andyr]

An update. I had friends around on Wednesday who know my system well, for a listening session.

We listened to:
1. the config with the delays I had been listening to (with the Maggie drivers delayed by 9ms,whose EGD plot is the 1st one shown in post #4)
2. and then with the Maggie drivers delayed by 15ms - with the EGD plot which is the 2nd one shown in post #4.

Interestingly enough, we all thought #1 sounded better. The sound became 'flat and lifeless' with config #2.

So I will leave the delays as they were!

Andy