Group delay is valid as a representation of time when it is flat, and constant. Flat areas of group delay, one frequency range vs another is/are usable data.
The constant time difference between the flat regions of group delay equal valid fixed time delays.
And simply reflect fixed distances between acoustic centers, relative to mic placement. (see REW's help Minimum Phase)
Group delay as a representation of time is totally invalid when it is not flat. (over a significant freq range)
If group delay has slope, it's unusable data imo. Phase is the valid data then.
Time to peak energy? ..... not what we want to use for establishing fixed delays
Time to peak energy and group delay are closely related. Group delay is pretty fixed though I'm pretty sure how hard the driver is driven changes things, just like Q... the time to peak energy seems to follow the same trend as group delay except when it is negative, something different happens.
Either it must over my head because its all apart of the time domain, filters have knees, and filters needs have peaks group delay.
Camplo, in my strong but hopefully humble opinion, I advocate erasing the idea of group delay from our audio measurement vocabulary.
Group delay is neither a fixed constant time across all frequencies, nor is it relative time across all frequencies.
For me, it's a techno-babble term that when used, most often betrays a (big) lack of understanding of the two basic fundamentals involved in crossovers .....fixed time, and relative phase.
But that's just me, the way I think....
If group delay works for ya, go for it ! Either way you go, I'd advise playing with electrical transfer measurements (or sims) to get time and phase sorted out, as a build base for acoustical measurements. Also, if you start with studying electrical linear phase crossovers first, before IIR, IIR will be probably be much easier to make sense of. (Sims work for understanding/comparing lin phase vs IIR (if you don't have lin phase processing capability. )
Group delay is neither a fixed constant time across all frequencies, nor is it relative time across all frequencies.
For me, it's a techno-babble term that when used, most often betrays a (big) lack of understanding of the two basic fundamentals involved in crossovers .....fixed time, and relative phase.
But that's just me, the way I think....
If group delay works for ya, go for it ! Either way you go, I'd advise playing with electrical transfer measurements (or sims) to get time and phase sorted out, as a build base for acoustical measurements. Also, if you start with studying electrical linear phase crossovers first, before IIR, IIR will be probably be much easier to make sense of. (Sims work for understanding/comparing lin phase vs IIR (if you don't have lin phase processing capability. )
GD is related to Time to peak energy.... Furthermore, whatever you choose as your starting point in time, I can find it on the slope of Group Delay so its definitely all related. I'd say that some characteristic of GD is chosen as the designated place of the beginning of "time" and if you knew the criteria you can look at GD and know where "time" began.... my point is.... its all related....
Delay is derived from, Group Delay...
Calling it Techno Babble, I don't know about all that lol.
Yes, they are related mathematically,
as Group Delay is defined as the negative derivative (or slope) of the phase response vs frequency.
So sure, we can study group delay for all the ways it might give us useable data we can act on.
Sorry about the techno-babble comment.....but my point is why bother with group delay at all,
when its only two factors, constant time and relative phase vs frequency,
are plain to see without having to ferret out how group delay provides useable data..
If I didn't see so many mistakes based on the false idea that time and phase can be substituted for each other, I wouldn't be so hard on group delay.
But group delay appears to me, to be the culprit behind that false idea.
Anyway....
Hey, without simply quoting the mathematical definition,
please try define what Group Delay pragmatically means/shows .........and in '30 words or less' !
as Group Delay is defined as the negative derivative (or slope) of the phase response vs frequency.
So sure, we can study group delay for all the ways it might give us useable data we can act on.
Sorry about the techno-babble comment.....but my point is why bother with group delay at all,
when its only two factors, constant time and relative phase vs frequency,
are plain to see without having to ferret out how group delay provides useable data..
If I didn't see so many mistakes based on the false idea that time and phase can be substituted for each other, I wouldn't be so hard on group delay.
But group delay appears to me, to be the culprit behind that false idea.
Anyway....
Hey, without simply quoting the mathematical definition,
please try define what Group Delay pragmatically means/shows .........and in '30 words or less' !
And, I'd add to that;
how to effectively and definitively take stock of what you have and resolve alignment to achieve the best possible audible results.
In the positive, it is time to peak energy. What do I win?
I didn't start by suggesting the use of group delay for time alignment. Sorry for the confusion, though it can be used for that, since its just the attack envelope, per frequency, as far as I'm concerned. I was originally talking about passive filters I believe
Last edited:
Win? Lol
As Charles already said, filters do not necessary have peaks in their group-delay response.
We know GD curves can take on a lot of shapes, with peaks or without.
We know GD varies vs frequency.
Which begs the question....
....at what frequency does GD equal time-to-peak energy?
Gotcha, thx. Since time-to-peak energy is the time constant behind setting delays, it's hard not to take a claim that GD = time-to-peak energy, as a claim for a method for setting delays.
Whether passive or active makes no difference.
Can anyone recommend a good 1.4 inch compression driver and a waveguide.
My priorities are smoothness, and a lack of hardness, over detail, etc. Intended use is from 700/800 Hz up to 15kHZ.
Also, a good open baffle 15 inch woofer that can work up to 700/800. Will use two of them per side.
Thanks.....
My priorities are smoothness, and a lack of hardness, over detail, etc. Intended use is from 700/800 Hz up to 15kHZ.
Also, a good open baffle 15 inch woofer that can work up to 700/800. Will use two of them per side.
Thanks.....
Last edited:
Just as you said, it varies per frequency.
I did some experiments showing GD effects on the start up transient using a speaker, mic, and sine wave. I compared the recorded signal to the source. When the GD is positive, there will be an increase in amplitude going from the very first oscillation to next several then amplitude plateaus. When I was negative, the start up transient had highee amplitude
REW displays the time to peak energy data on the CSD view. It was not exact to the GD but usually within a few ms, in my experience.
If you want smooth and no harshness. Then look at treated TI diaphragms, or polymer/plastic/Carbon fibre, TI with a damped surround like the Beyma might also do good if they are well designed.
HF1460, is one of the cleanest ones to date from the popular pro brands.
HF1440 looks good.
CP755ND is a proven performer, you can find quite a few measurements on here. Got a good motor design, and it's possible to get the AL diaphragms still.
18s ND3SN are excellent. NSD1480N gets a lot of praise too.
HF146 looks decent, but it's got a mode/resonance at 10k, it's in the response, free air impedance, on horns, and visible in the CSDs, whether objectionable or not I can't tell you really, but it's a resonance from the driver. The plastic diaphragm might make it less so.
The Beyma and 18Shas the strongest motor, and is the heaviest ones, also got copper shorting rings.
You will find opinions on most of them, and measurements if you search around.
Might matter more what horn /waveguide you choose than Comp. driver.
As none of them have the same throat exit angles etc. so they don't fit the same horn/waveguide.
https://www.dibirama.altervista.org...-faital-hf146-driver-2-56-8-ohm-160-wmax.html
https://audioxpress.com/article/test-bench-faital-pro-s-new-hf1460-pro-sound-compression-driver-1
A few ms is huge.
Crossing to a sub, being used only as a sub...say with response not above 100Hz..is about the only time being off a few ms can slide.
And that's not really good practice.
I am working on a similar approach, but with the integration of a super tweeter that i already have, here a few pics, more pics are here:
https://www.diyaudio.com/community/threads/system-pictures-description.23208/post-7494338
https://www.diyaudio.com/community/threads/system-pictures-description.23208/post-7494561
i have an adapter to run an 1 inch driver on an 1.4 inch horn, so that i will be able to play back and forth very quick with the following combinations:
HORN
Eighteensound XT1464 horn
https://www.eighteensound.it/media/W1siZiIsIjIwMTkvMDcvMTkvMTBfNTZfMzZfMTY4X1hUMTQ2NC5QREYiXV0
DRIVER
A )
B&C DCM414-16 mid range driver, super tweeter integration required
https://www.bcspeakers.com/en/products/hf-driver/1-4/16/dcm414-16.pdf
B )
Eighteensound NSD1480N mid and high range driver, testing with and without super tweeter integration
https://www.eighteensound.it/en/products/hf-driver/1-4/8/NSD1480N.pdf
C )
Visaton DR45N with adapter mid and high range driver, testing like B )
https://www.visaton.de/sites/default/files/dd_product/DR 45 N_6060.pdf
D )
Eighteensound NSD1095N with adapter high-mid and high range driver, driver cutoff freq. is too high for the large horn, testing anyway for "academic" purposes and curiosity of comparison of the measurements
https://www.eighteensound.it/en/products/hf-driver/1-0/8/NSD1095N.pdf
I will be able to do all measurement with CLIO pocket with SPL on axis, 15 degrees and 30 degrees and CSD and WCD plot
I will try to find the time for the corresponding listening tests.
I think it will make sense to open an own thread for this.
So far, so good, Stefano
Last edited:
How to use group delay to setup delay for XO?
Here's a 1st order BW LPF and HPF generated by REW's EQ function. The LPF and HPF did not include any delay at the beginning.
And the green is their sum.
The impulse response of the 2 FR.
Also the impulse response for summation. Perfect delta function after summation due to 1st order filter's summation still a linear phase response.
And their corresponding group delay. The group delay shows 80us higher at LF compare to 20kHz.
Both LPF and HPF shows ~80us delay at 20Hz compare to 20kHz. However they did not have time offset at all.
If I offset tweeter (HPF) by 80us forward/backward and sum the response I got the following:
And their corresponding phase response.
The proper way to sum this demo response is apply 0 offset of the 2 response and it'll form perfect summation.
Is group delay a useful too here to estimate time needed to delay for both drivers?
Or is it excess group delay that's useful?
Here's a 1st order BW LPF and HPF generated by REW's EQ function. The LPF and HPF did not include any delay at the beginning.
And the green is their sum.
The impulse response of the 2 FR.
Also the impulse response for summation. Perfect delta function after summation due to 1st order filter's summation still a linear phase response.
And their corresponding group delay. The group delay shows 80us higher at LF compare to 20kHz.
Both LPF and HPF shows ~80us delay at 20Hz compare to 20kHz. However they did not have time offset at all.
If I offset tweeter (HPF) by 80us forward/backward and sum the response I got the following:
And their corresponding phase response.
The proper way to sum this demo response is apply 0 offset of the 2 response and it'll form perfect summation.
Is group delay a useful too here to estimate time needed to delay for both drivers?
Or is it excess group delay that's useful?
Hi yys310, cool to see folks study how electrical crossovers work. I find it so much easier to understand the factors in involved with 3D acoustical crossovers, when I have how 1D electrical crossovers work, firmly in mind.
The summation of first order butterworths is as simple a starting point as it gets...nice!
I'm sure you noticed the GD traces in your example, for the high pass, the low pass, and summation were all the same.
If you compare the GD trace of whichever side you shifted 80us, to the same side before the time shift, you should see the two GD traces are the same trace, one on top of the other, with a parallel separation of 80us.
In this easiest case posible, we can determine fixed time delay by the parallel distance between curves.
Good luck moving beyond such a simple case.
There are two primary direct ways to determine the constant delay between acoustic sources that I'm aware of.
Both of course require comparing time arrivals of one passband, electrical or acoustic, to another.
In your set of graphs, the second one 'Impulse'...that shows the impulses of both sides independently.....
Notice the peaks aren't at exactly the same time, but that the initial rise is the same for both.
Using impulse responses, whatever constant delay aligns their initial rises together, is the correct fixed delay.
Which gives the same constant delay using the phase trace overlay method.
Where individual phase traces become aligned/overlayed with each other, by varying the amount of constant delay.
These two methods go directly to isolating contant fixed delays needed in acoustical designs.
Try comparing individual side's acoustical GD's....yikes.
But like said in an earlier post, for summations, flat areas of group delay, one frequency range vs another is usable data.
The constant time difference between the major flat frequency regions of group delay can equal valid fixed time delays. (per REW help on Min Phase.)
That sounds good but I mean the trend line is obviously following the GD.... its not unknown that the delay in the group is the start up transient... its not a delay in phase but delay in amplitude.