Sorry for coming back to group delay. I undug my excel spreadsheet that I once made for the the dimensioning of group-delay equalisers in order to show where the comparision between loudspeaker group-delay and most group-delay studies falls short.
The green curve shows the group delay of an LR4 (basically the GD of a 2nd order allpass with a Q of 0.71). The white curve is showing the GD of a high-Q 2nd order allpass like they are used for most GD audibility tests. The main difference is that these high-Q allpass filters have high GD values over a small bandwith but very low GD at low frequencies and falling towards zero at very high frequencies.
So the perceptual studies are mostly about peaking group-delay within a narrow frequency band while the typical loudspeaker crossover has high (and quite constant) group-delay at low frequencies and low group-delay at high frequencies. The only paper I have seen that is dealing with the typical crossover-cause GD is a recent one from Finland. I willl cite it when I have found it again.
Regards
Charles
P.S.: The dark blue curve is from the 2nd/3rd order crossover that I personally use when it is dimesioned for the same crossover frequency as the LR4. Its GD distortion (difference between the highest and lowest GD value is only one third of that form the LR4).
The green curve shows the group delay of an LR4 (basically the GD of a 2nd order allpass with a Q of 0.71). The white curve is showing the GD of a high-Q 2nd order allpass like they are used for most GD audibility tests. The main difference is that these high-Q allpass filters have high GD values over a small bandwith but very low GD at low frequencies and falling towards zero at very high frequencies.
So the perceptual studies are mostly about peaking group-delay within a narrow frequency band while the typical loudspeaker crossover has high (and quite constant) group-delay at low frequencies and low group-delay at high frequencies. The only paper I have seen that is dealing with the typical crossover-cause GD is a recent one from Finland. I willl cite it when I have found it again.
Regards
Charles
P.S.: The dark blue curve is from the 2nd/3rd order crossover that I personally use when it is dimesioned for the same crossover frequency as the LR4. Its GD distortion (difference between the highest and lowest GD value is only one third of that form the LR4).
Attachments
I've recently noticed crossover region can also have peaking GD at off-axis angles due to transducer size and resulting beaming or self interference or how'd ya call it. It depends crossover frequency and slope how attenuated signal is at these GD peaks and it seems LR4 at wavelength ~= driver diameter takes these about ~30-40db down inside listening window. GD peaks move lower in frequency the more off axis one goes until 90 degrees.
Attached is ideal driver simulation but some smaller GD peaks were in my measurements as well. I've got no idea if these are of any concern. Just an observation.
First attachment is simulation of ~3" driver at 30 degrees off-axis and some 1ms GD peaks at high frequencies are seen at interference nulls.
Second attachment is ~15" driver at 70 degrees, 3ms group delay peak at 1.5kHz wowza.
Third is a filter test.
Listening window SPL is roughly 40db down at GD peak frequency with the filter. Listening window in the simulations is +-30 degrees from the reference on horizontal axis and +-10 on vertical axis. Listening window is the light green trace on the SPL and Power&DI graphs.
Attached is ideal driver simulation but some smaller GD peaks were in my measurements as well. I've got no idea if these are of any concern. Just an observation.
First attachment is simulation of ~3" driver at 30 degrees off-axis and some 1ms GD peaks at high frequencies are seen at interference nulls.
Second attachment is ~15" driver at 70 degrees, 3ms group delay peak at 1.5kHz wowza.
Third is a filter test.
Listening window SPL is roughly 40db down at GD peak frequency with the filter. Listening window in the simulations is +-30 degrees from the reference on horizontal axis and +-10 on vertical axis. Listening window is the light green trace on the SPL and Power&DI graphs.
Attachments
I think you are confusing phase as it relates to a wave and measured phase of a speaker. Pattern flip is not a good term to use as it means something totally different in speakers.
Wave phase image
The change in measured phase of speaker is either due to bandwidth limitation or crossover.
Here is an example of a theoretical woofer. There is a 2nd order rolloff at 50Hz to simulate a sealed box and then either no lowpass or one added at 1K. When there is no upper bandwidth limitation the phase changes 180 degrees over the bandwidth 0 degrees at HF (no rolloff) to plus 180 at DC
When the upper bandwidth is limited with a 2nd order low pass at 1K there is 360 degrees of change from -180 to plus 180.
This the the difference in the impulse response due to the bandwidth change.
Attachments
I found the paper again that I mentioned in Post 7001:
AES E-Library » Audibility of Loudspeaker Group-Delay Characteristics
Regards
Charles
AES E-Library » Audibility of Loudspeaker Group-Delay Characteristics
Regards
Charles
Group delay becomes more clear. Group delay is an understanding of the phase shift of inductors & capacitors in the time domain.
That's the reason I'm building a 2 way, with 1 inductor series the woofer, and 1 capacitor series the tweeter. Inexpensive crossover. Less complex phase shifts. Plus, neither transducer is awful 1 octave out of band. I'm crossing @ 1000 hz. Eminence Deltapro-15A Woofer plots response to 4 khz, although it probably beams (10 deg horizontal dispersion) 1800 hz up. Woofer should not sound awful 2000 hz up attenuated 3 db. RX22 Tweeter has been used by Peavey down to 800 hz in a 150 W late eighties SP2. Will be resistor series tweeter to match volume, & tungsten bulb series tweeter to suppress excess HF peaks. Phase shift plot should be simple, group delay I hope low. No bomb whistle (decending pitch) drum hits. A drum hit is a musician's term for an impulse.
The Peavey SP2-XT I am copying sort of, had no objectionable pitch shifts on drum hits. No decending or ascending whistles, at 1 W in my living room.
Last edited:
One critical factor that is being ignored in all this GD (group delay) discussion is the fact that the audibility of GD is known to increase with increased SPL. This means several things:
1) any level of GD will be audible at some level of SPL.
2) any source will have some level of SPL where its GD is audible.
3) this audibility threshold SPL is likely to be vastly different between different loudspeakers.
To my knowledge only Lidia and I have done tests of GD audibility with level and we found a statistically significant audibility with level. Hence, when one quotes X ms or less is not audible, then at what SPL is that true? Is it even disclosed? This fact cannot simply be ignored as it is critical to the discussion.
Also, good researchers do not seek to make things more complex, simplicity is always the goal, but the world is not always simple. There is a strong tendency for things to get more and more complex as we get deeper and deeper into the details. That's just the way the world works.
1) any level of GD will be audible at some level of SPL.
2) any source will have some level of SPL where its GD is audible.
3) this audibility threshold SPL is likely to be vastly different between different loudspeakers.
To my knowledge only Lidia and I have done tests of GD audibility with level and we found a statistically significant audibility with level. Hence, when one quotes X ms or less is not audible, then at what SPL is that true? Is it even disclosed? This fact cannot simply be ignored as it is critical to the discussion.
Also, good researchers do not seek to make things more complex, simplicity is always the goal, but the world is not always simple. There is a strong tendency for things to get more and more complex as we get deeper and deeper into the details. That's just the way the world works.
The Klipschhorn dealer was very careful to demonstrate his speakers with very easy to reproduce sound source. Light Jazz, trumpet, string bass, brush snare. That long path from the woofer to the ear, around a plywood fold . . . Could lead to delays of different frequencies. Certainly to time mis-alignment. I'm not going to buy one to find out. Pity, in 1974 my ears still went to 20000 hz. Now I am limited to 14000.
Last edited:
Charles,
A drum makes multiple notes at the same time when struck. The lowest fundamental, next higher note, and higher overtones occur at non-musical intervals above the fundamental, completely normal for drums.
Circular Science – Technology Serving Art
Since most drums have two heads, the tuning can be arranged for either a generally descending or ascending drum pitch.
Post # 6891 has an example of a kick drum hit waveform descending an octave over it's envelope.
https://www.diyaudio.com/forums/mul...m-spl-low-distortion-2-a-690.html#post6626423
This type of signal could create a "whistle" from a small port with an Fb (Box tuning Frequency) near or above the low fundamental frequency when driven hard.
Art
Last edited:
Hi Art
Thanks for the explanations. I am already aware of that (I have actually played in several bands over the years).
What I meant is whether there are really speakers with such severe group delay distortion that they can decompose transient signals into downward whistles.
Regards
Charles
Thanks for the explanations. I am already aware of that (I have actually played in several bands over the years).
What I meant is whether there are really speakers with such severe group delay distortion that they can decompose transient signals into downward whistles.
Regards
Charles
Last edited:
Ty kindly for the graph Ro808. Just what I needed.
I can relate this to tracking the movement of the woofer, for phase or group delay....this would be the best kind of visual reference, even better if animated...seeing is believing.
So where were we? 200hz and -90 degrees.....is that the 270 mark on the graph here? So the phase response measurement if 200hz -90degrees...what does this mean in relation to when the woofer takes flight.
When in time and were is 0? 0 has to be when the signal arrived....the woofer will be at 0 for this example....the measurement says -90 phase angle for 200hz....so the rate of change at 200hz... (1cycle/200hz) * (-90/360) = -1.25ms...it takes the woofer 1.25ms to reach where on the graph?
I have to be able to visualize this
All I want to know is how long it takes the woofer to start moving at each frequency....I think unwrapped phase is the backdoor...
Regarding Martin Kings response...I'll have to look at the formulas and do the comparison at the same frequency to make it makes sense to me. The period size changes with frequency so that variable throws me off...
I can relate this to tracking the movement of the woofer, for phase or group delay....this would be the best kind of visual reference, even better if animated...seeing is believing.
So where were we? 200hz and -90 degrees.....is that the 270 mark on the graph here? So the phase response measurement if 200hz -90degrees...what does this mean in relation to when the woofer takes flight.
When in time and were is 0? 0 has to be when the signal arrived....the woofer will be at 0 for this example....the measurement says -90 phase angle for 200hz....so the rate of change at 200hz... (1cycle/200hz) * (-90/360) = -1.25ms...it takes the woofer 1.25ms to reach where on the graph?
I have to be able to visualize this
All I want to know is how long it takes the woofer to start moving at each frequency....I think unwrapped phase is the backdoor...
Regarding Martin Kings response...I'll have to look at the formulas and do the comparison at the same frequency to make it makes sense to me. The period size changes with frequency so that variable throws me off...
Attachments
Last edited:
As I said the measured phase of speaker and the phase of a wave are not equivalent which is where I think your confusion is.
It will take 5ms for a single cycle of 200Hz regardless of the measured phase. 1000(ms) / 200(Hz) = 5(ms) to be at the 270 degree point will take 3/4 of 5ms. There can only be a -90 when comparing relative phase or go back in time.
Camplo, you appear to be thinking about phase as it relates to a sine wave going through a single driver. That ain't it at all.
Phase is the relative timing of frequency vs frequency across the spectrum.
Perfect phase is when are all the frequencies rise together from a starting 0 amplitude at the same time.
A sine wave through a driver is simply described by polarity....does it move out, or in, with a + signal.
Phase is the relative timing of frequency vs frequency across the spectrum.
Perfect phase is when are all the frequencies rise together from a starting 0 amplitude at the same time.
A sine wave through a driver is simply described by polarity....does it move out, or in, with a + signal.
This paper was published by the same authors of the equalizer paper.
They are affiliated with Genelec.
Wth...Sorry about that...I meant Fluid, little mix up.
Its all related, if not how can REW calculate GD and phase from a sine sweep, instead of a literal impulse
Its all related, if not how can REW calculate GD and phase from a sine sweep, instead of a literal impulse
Now that makes a lot of sense...I haven't established at what part of the cycle peak spl is reached....I know the formula for phase delay which I think is the same event....max spl is reached before the cycle is finished...I think....
Last edited: