Excess Group delay show in overlay possible or maybe can add in future ?

[user44455555]

Hello

I do ramped levels at nearfield with following method.
make microphone distance 5 cm to Woofer. set volume of each Speaker you want test with REW check Levels in measure tab sp that Level - 6 have 95 db . this is at 1 meter 69 db, and Level 3(L3) so it is not so loud but loud enough to show how much group delay change depend on level.

set sweep length to 1 mb set ramped Levels in measure mode 3 measurement -18 db .. -6 db frequency from 35 to 300 hz
After measure switch to GD choose with mouse right click menu generate minimum phase

L2 is 6 db less (63 db)
L1 is 12 db less (57 db)

the short distance is need, to avoid room influence. Keep in mind you can not use much higher level, because the microphone is so near you can get clipping or damage microphone .
for volume of 84 db at 1 meter you get at 5cm 110 db. thats too high for microphone. so keep silent

I think this is correct and no a problem of room influence. there can see in the results that excess Group delay get larger depend on volume in the bass range and i think everybody can do such tests to see if speaker is good. because in stereo left and right speaker play diffrent waves, the group delay jitter on worse speaker between left and right speaker at same frequency. because interaural time delay is used for audio location in brain this give bad stereo image and mud on low bass(kick drum reverb)

the teststeps

in the mdat are also measures from Kali and HS5. the kali have on low volume much more random changes, i did not think it is problem from room. I always notice when hear kali lp6 very loud sound ok. but i want not hear loud.

Edit: remove attachment, because there is newer one in later posts with high pass filters 50 and 100 hz and adam audio d3v 5 cm measures.

 

[Keith_W]

I decided some time ago not to respond to any more of your posts because my patience ran out much earlier than John's. You don't seem to learn and you are asking the same question over and over, but in a different form.

1. Please read John's post here as many times as you need until you understand what he said. In particular, do what he suggested in that thread. Create a simulation of two loudspeakers and examine the phase and group delay characteristics of each simulation. Do this:

- Press the "EQ" button
- In the popup window, press the "EQ filters" button

- Scroll all the way to the right. Create a bandpass filter. Here is a speaker simulation with simulated bass and treble roll-off. Do this twice with different values to simulate two different speakers.
- Under "Filter tasks", click "Generate measurement from filters".

- Result: two loudspeaker simulations for you to play with and compare. Try shifting the position of the impulse (t=0) and see what effect that has on the phase graph and group delay.

2. Group delay is the negative derivative of phase, see REW manual. You can think of it as another way to view the phase graph.

3. You need to understand group delay audibility thresholds. see Liski, Makivirta, Valimaki's paper on Audibility of Loudspeaker Group Delay Characteristics or read Toole.
4. Ported speakers have different group delay characteristics compared to sealed speakers. Without googling exactly what speakers you have, it is impossible to comment.

5. There will be minimal ITD (interaural time difference) if the same speaker is used for left and right channels, unless one speaker is defective.

 

[user44455555]

.1. Please read John's post here as many times as you need until you understand what he said. In particular, do what he suggested in that thread. Create a simulation of two loudspeakers and examine the phase and group delay characteristics of each simulation. Do this:

But the frequency response of 1 speaker is same only it play at diffrent levels L1 L2 L3 and get diffrent excess group delay over 1 ms (1000 µsec) diffrence. how can you explain this ?. you can download the mdat and choose spl align to see the FR of L1 L2 L3 levels measure

EDIT: i remove that part about predictat phase, because it use phase in EQ settings of the measure. so ok i accept that 2 speakers can not easy compare, but to choose the best speaker can use clarity or excess group delay and the best speaker is the speaker which have between L1 L2 and L3 the fewest diffrence. that the kali have jitter on low freq i get in mind it is maybe because the kali have lots hiss on woofer. On tweeter not so much. the kali hiss sound dull. So my test show also worse speakers with much hiss when L1 jitter much on Low freq.

and btw when you look at the clarity results with the 5cm measure you can see the ranking too, the most expensive speaker i have the focal is the worst this test and HS5 the small 5 inch is best

There will be minimal ITD (interaural time difference) if the same speaker is used for left and right channels, unless one speaker is defective.

and how many µsec do you mean with minimal ? .

they test it and found

Smallest perceivable interaural time differences - PubMed

It is well-established that the smallest discrimination thresholds for interaural time differences (ITDs) are near 10 μs for normal hearing listeners. However, little is known about the hearing and training status of the test subjects from past studies. Previous studies also did not explicitly...

pubmed.ncbi.nlm.nih.gov

Smallest perceivable interaural time differences​

The average threshold ITD for this condition at the 75% correct level was 6.9 μs for nine trained listeners and 18.1 μs for 52 un-trained listeners.

 

[user44455555]

here is the confirm excess group delay change only very few with diffrent highpass frequency (to simulate bass lost) .And the most important thing, a big speaker with more bass should get less excess group delay than a small speaker. But my measures show the big speakers have more Excess group delay.
Here i have screenshot of Generate Measurement from Filter. A lowpass filter for crossover did not change this

excess group delay at 37.1 hz . L2 level = 63 db at 1 meter distance. L3 level is 69 db at 1 meter distance. so diffrence between L2 and L3 is only 6 db

the 50 hz lowpass 0.112 ms
the 100 hz lowpas 0.224 ms
Kali with L2 Level -10.7 ms
Kali with L3 Level -13.1 ms = 2.4 ms diffrence between L2 L3
Focal wit L2 Level -14.3 ms
Focal with L3 Level -15.8 ms = 1.5 ms diffrence between L2 L3
HS 5 with L2 Level -13.5 ms
HS 5 with L3 Level -15.3 ms = 1.8 ms
D3v with L2 Level -12.3. ms
d3v with L3 Level -12.1 ms = -0.2 ms

Who do disagree that a complete speaker with case is no minimum phase system ?. Cabinet Diffractions are huge. this is what this measures show. only for the driver is tell that it is near minimum phase. you can ask the KI with this "is loudspeaker minimum phase"

the screenshot are do at 35.31 hz so values are slightly diffrent for the filters than above

the mdat attach here is enhanced. the measures of the d3v and filters are include. see also clarity of filters also intresting to know

 

[user44455555]

I notice that for correct excess group delay the begin of measure frequency (and end) need have a good SNR . so currently it is important to not use just 20-20 khz measure range. I notice at 35 hz is a good start point for the 5 inch speakers and larger

And question is, is it possible to limit the excess group delay calculation when use older measures to a choosable start frequency ?
I have not post JBL 104 results because this can not output 35 hz so measure start must be set higher.

excess group delay do EQ the measure with min phase filter so it is 100% linear. this reduce the frequency depend. and when measure range contain frequencies that are only noise because speaker can not emit it with good SNR then excess group fail complete to EQ them

for example when i measure from 20 hz i get such curve that bass is faster(because of positive values). this is a quick measure on desktop so more jitter as my other measures. My other measures are done from stativ in middle of room

 

[user44455555]

thats really strange wy excess group delay output wrong values when start measure at 35 hz and measure have good SNR.
Seem i need verify if group delay code work correct with burst measure and osciloscope to see if group delay show same diffrence as osciloscope measure with a burst at 35 hz and then 400 hz, or at least near it with same speaker

I notice now that in group delay positive values mean delay of output. In phase it is opposite. negative values mean delay .
really strange, wy no speaker tester show timing measure in nearfield. when do bursts at 5 cm distance. the sine wave can clear see. so no room reflections. maybe continues sine measure is not able to do usefull timing results of speakers. In speaker tests many years ago they show burst . for example

https://www.hifi-archiv.info/Test/Lautsprecher%20Test%201979%20Stereoplay/13.jpg

https://www.hifi-archiv.info/Test/Lautsprecher%20Test%201979%20Stereoplay/09.jpg

https://www.hifi-archiv.info/Test/Lautsprecher%20Test%201979%20Stereoplay/04.jpg

EDIT: I notice when i do in generate minimum phase in the 35 hz measure enable Replicate Data Outside range and enable Add LF Tail 37 hz(35 hz setting do REW change to 37 hz) 12 db result is very simular to the 0 hz measure. also when measure start frequency is 20 hz then there need do Add LF Tail 20 hz to get simular results as the 0 hz from beginning measure. when choose diffrent filter 6 db or 18 db, result change. so it seem really need verify group measure results with the osciloscope

but anyway, my quality speaker suggestion measure is mainly to measure at diffrent volume and compare the diffrence between the volume in excess group delay. and the better speaker is the speaker which have fewest diffrence in excess group delay between measure volumes

 

[John Mulcahy]

I notice that for correct excess group delay the begin of measure frequency (and end) need have a good SNR . so currently it is important to not use just 20-20 khz measure range. I notice at 35 hz is a good start point for the 5 inch speakers and larger

Use the tails to correctly generate the minimum phase response, they let REW know how the response behaves at the lowest and highest frequencies.

bass is faster(because of positive values)

It does not mean that.

 

[user44455555]

Use the tails to correctly generate the minimum phase response, they let REW know how the response behaves at the lowest and highest frequencies.

thanks for your answer, please look also at my osciloscope screenshots what crazy stuff speakers do. this explain alot, wy group delay is negativ when the SPL of the begin of measure is 35 hz. the first 2 cycles are much faster as 50 hz in transient. can see clear in screenshot at 50 hz does not reach the expectet levels and have not correct frequency. this is diffrent depend on speaker. they have at begin transient higher frequency. with continuous sine threre is only 1 transient at begin of measure. this mean when REW use the first cycle at beginnung the following cycles are then measured as if they are ahead. I think this is reason wy show with 35 hz start freq negative group delay

original wave

d3v 1. cycle

d3v 2. cycle

d3v 3. cycle . this is correct frequency the speaker output. the 0,3 hz is cursor set not so exact

now focal 1. cycle

focal 2. cycle

there can see the d3v reach are more near correct levels, but have more frequency drift.
so i have no idea how can verify group delay in measure software with osciloscope