Group delay simulation goals

[2234rew]

Hi @juicehifi

I'm very happy with SQ but I've been playing around with correction procedure design to see if I can see any improved simulations.

My limited understanding is that we want to convert group delay 'peaks' into 'dips' ?

Is that a good understanding of the general goal when looking at group delay simulation results?

So I've managed to convert this (peak at 300 Hz) :




To this (dip at 300Hz):



Is this an example of a good outcome for group delay?

The large size of the dip is not a concern?

 

[juicehifi]

We are looking at a time domain windowed chart here, and the group delay peak is most likely still present in both ... on an unsmoothed simulation.

A group delay peak may change to a grup delay dip if the time window is shortened, because less reflections will be included. That's most likely what you did here.

The goal is to correct the time domain effectively, but without ending up with potentially audible pre-ringing in the simulation, and ideally also with correction filters that do not show low frequency oscillating behavior left of the peak. The latter sometimes produces audible pre-ringing even though what's on the simulation shouldn't be audible.

The "selective preringing prevention" usually takes care of all of this. Sometimes, increasing or reducing the TTD window is needed for the best possible result. On rare occations it may work to disable the "selective preringing prevention" ... and correct these group delay peaks without a sonic penalty.

 

[2234rew]

A group delay peak may change to a grup delay dip if the time window is shortened, because less reflections will be included. That's most likely what you did here.

Thanks Bernt.

In this example above I actually increased time window from 5 cycles to 5.1 cycles at 10Hz And that particular peak is also gone at 300 Hz in unsmoothed (below).

And pre-ringing reduced also.

See attached zip of new measurement file (working nicely with Denon AVR now) and my particular favourite target file. I've learn that different targets will require different TTD subwindows.

300 Hz peak from right speaker removed but there's still a lot of bad stuff remaining in un-smoothed I can now see.

So if group delay peaks are removed, step response is improved and pre-ringing reduced, then one can be happy that overall time domain correction has been imrpoved?

For a normal living room system (no hifi specific room treatments) have you ever seen zero group delay, unsmoothed?

 

[juicehifi]

That sounds like a very successful tweak. Good job!

 

[2234rew]

That sounds like a very successful tweak. Good job!

Thanks Bernt! I think I updated my post while you replied but I added a query:

For a normal living room system (no hifi specific room treatments) have you ever seen zero group delay, unsmoothed?

 

[juicehifi]

PS: The GD peaks and dips thend to be narrower and higher/lower as we get closer to a perfect cancellation ... for the subject time window.

 

[2234rew]

PS: The GD peaks and dips thend to be narrower and higher/lower as we get closer to a perfect cancellation ... for the subject time window.

Thanks. When you say "closer to a perfect cancellation" , this means closer to "destructive interference" ?

 

[juicehifi]

Thanks Bernt! I think I updated my post while you replied but I added a query:

For a normal living room system (no hifi specific room treatments) have you ever seen zero group delay, unsmoothed?

I don't recall. I mostly use the group delay to examine those cancelling frequencies.

Most people will see a rise down towards 20 Hz due to a minimum phase target and limited bandwidth. Some of the best rigs in some of the most treated acoustics are probably close enough ... from 20 Hz and up.

You will get close if you apply a linear phase target, disable the selective pre-ringing prevention and just apply enough correction. It may not look very good in the time domain, though.

 

[juicehifi]

Thanks. When you say "closer to a perfect cancellation" , this means closer to "destructive interference" ?

Close to where the direct sound and the reflection cancel each other out completely. If the direct sound is a bit stronger, the group delay will have a steep valley, if the reflections are a bit stronger, it will have a steep peak.

 

[2234rew]

Close to where the direct sound and the reflection cancel each other out completely. If the direct sound is a bit stronger, the group delay will have a steep valley, if the reflections are a bit stronger, it will have a steep peak.

Thanks. This is great info. Will continue to tweak to see what I can do.

(while I continue to enjoy the music and incredible results AL has brought)

 

[2234rew]

Close to where the direct sound and the reflection cancel each other out completely. If the direct sound is a bit stronger, the group delay will have a steep valley, if the reflections are a bit stronger, it will have a steep peak.

Hi Bernt, is the group delay simulation in AL an "excess group delay plot" or "overall group delay plot"

I'm picking those terms excess vs overall from here: https://www.roomeqwizard.com/help/help_en-GB/html/minimumphase.html

 

[juicehifi]

It's an overall group delay plot.