So what is "Group Delay" anyway?

[Raider]

I have heard this explained before, but since it didn't take, I think I may not have really understood it. :|

A concise layman's definition, as much as possible, would be appreciated. As with so many other things, what variables affect group delay? Is it primarily influenced by enclosure design, or do room acoustics come into play? What are the tradeoffs? What is considered optimum, and what does it sound like when it is optimized, or when it is compromised? What enclosure types optimize it, and which types tend to compromise it?

Thanks.

[DrWho]

You're opening a can of worms with that question!

As I understand it, for a fixed diameter piston the lower it plays, the

further it needs to travel in order to "create" the frequency. You

could almost think of group delay as the time it takes for the driver

to produce that note. I hesitate saying that because I know it's wrong,

but I don't know how else to describe it (which probably means I don't

understand it correctly).

I do know that a google search brings up crap loads of debates on the issue. The "audibility of group delay - round 5, start!"

I too would be intersted in a good explanation.

[bodcaw boy]

the delay in time caused by traveling in groups......

[colterphoto1]

that's so poetic Roy!

[bodcaw boy]

hey michael,

bought any good speakers lately.........

[colterphoto1]

didn't you hear? I copped a pair of HIP's just this am. 3 bills

Does much time go by that I don't buy any good Klipsch speakers? I'm a madman!

[colterphoto1]

PM to ya!

[TheEAR]

You're opening a can of worms with that question!

As I understand it, for a fixed diameter piston the lower it plays, the

further it needs to travel in order to "create" the frequency. You

could almost think of group delay as the time it takes for the driver

to produce that note. I hesitate saying that because I know it's wrong,

but I don't know how else to describe it (which probably means I don't

understand it correctly).

I do know that a google search brings up crap loads of debates on the issue. The "audibility of group delay - round 5, start!"

I too would be intersted in a good explanation.

This is a very good,simple description. When a cone/piston has to travel say 4" to produce a given note a say X dB the group delay will be much higher than say the same note produced with a 4mm travel and Y dB. In a perfect world there would be no group delay and full range,point source would be the summum of perfection.In the real world the engineers have to work to minimize group delay or at least keep it in check,if I may say.

Also more travel means recuperation time,the woofer's cone/piston does not reach its central position as fast when under full excursion as it does at moderate levels where the cone has little travel.

[kramskoi]

You're opening a can of worms with that question!

As I understand it, for a fixed diameter piston the lower it plays, the further it needs to travel in order to "create" the frequency. You could almost think of group delay as the time it takes for the driver to produce that note. I hesitate saying that because I know it's wrong, but I don't know how else to describe it (which probably means I don't understand it correctly).

I do know that a google search brings up crap loads of debates on the issue. The "audibility of group delay - round 5, start!"

I too would be intersted in a good explanation.

This is a very good,simple description. When a cone/piston has to travel say 4" to produce a given note a say X dB the group delay will be much higher than say the same note produced with a 4mm travel and Y dB. In a perfect world there would be no group delay and full range,point source would be the summum of perfection.In the real world the engineers have to work to minimize group delay or at least keep it in check,if I may say.

Also more travel means recuperation time,the woofer's cone/piston does not reach its central position as fast when under full excursion as it does at moderate levels where the cone has little travel.

basically, a driver in an enclosure is a high pass filter...allowing the highs to pass and rolling off the lows...with any highpass filter, decreasing frequency brings increasing time delay/group delay (the delay between two differing frequencies, i.e. 10 and 100 Hz)...past a certain point it becomes audible...it is "generally" accepted that group delay becomes noticeable between 1 and 2 cycles of the frequency being reproduced...the allowed limit gets higher as frequency decreases...for example, 20 Hz = 20 cycles per second, or one cycle every .05 second, or 50 ms...so the onset of audible delay @ 20 Hz would be between 50 and 100 ms, or 1 - 2 cycles...at 10 Hz it would increase to 100 and 200 ms...

If we use the phase adjustment on subwoofer plate amps as an example...why do two frequencies canel each other? It is because they are both arriving at different times (out of phase) to the listening position, cancelling rather than reinforcing one another...by adding or reducing delay, one can cause the frequencies to sum together...with this example we can see that reducing the amount of time delay between frequencies allows a closer approximation of the intended sound signature...the different frequencies arrive virtually simultaneously, reducing the "time smearing" effect.

The shallower the rolloff (highpass filter), the least amount of phase shift over the passband...and if this also represents time delay the way i think it does, then we should strive to keep this delay as small as possible, in the interest of the most cohesive reproduction of sound.

That said, all this becomes moot, however, once you factor in room acoustics...this is what lots of mfgs. conveniently forget when pushing for sales...a ruler flat frequency response means nothing once placed in a particular room...the loudspeaker is then at the mercy of a myriad of reflections cancelling and reinforcing various frequencies throughout its range...

I'm no EE, but group delay, phase shift, and impulse response are all related to a degree, and all will affect the final presentation in the time domain.

I don't pretend to have this all figured out...this is a "very" tricky and confounding subject...one in which i feel uncomfortable responding to...so take the above with a grain of salt and do your own independent research...the axiom, "rely but verify", definitely applies here...

[WMcD]

Nevermind. I'll have to study up more.

Gil

[bodcaw boy]

deang posted this another thread. it helped me to keep it straight.

[TheEAR]

Very ineteresting. Acoustics as a whole are a complex beast,the one reads to more there is to learn.Very interesting to say the least.

[DrWho]

Gil, I have asked...and never get a solid answer. I just need to find a prof that speaks english and doesn't do research 24/7.

One thing I do know is that group delay does amount to some level of

"time delay" in the sense that the impulse of the wave arrives later

(versus the arrival time of an impulse with less group delay). How much

later is a bit more complicated since it is directly relative to

frequency.

I have also read from several sources about the use of "all-pass"

filters to introduce time-delay. It seems they're all using group delay

to achieve the necessary time delay. Back in the days before digital

processing these kinds of analog filters were the only way to provide

the delay needed for speakers further out into the audience. I once

worked with such a unit and I remember spending years reading the

manual and circuit diagrams trying to figure out how the bloody thing

was able to "slow down electricity". I understand now that electricy

isn't slowing down, but it still doesn't make much sense. The units

were annoying because you couldn't arbitrarily pick any level of delay

- speaker positioning had to be tweaked based on the delay

compensations available in the unit (which made upgrading a royal pain

in the butt). But since I've always been told that these units are taking advantage of group delay - I get leery when people mention that it has nothing to do with time delay (in the sense of measuring when the impulse arrives).

This is one of those things where I'd like to sit down in a lab with good measuring equipment and measure / listen for myself.

[mas]

<?xml:namespace prefix = o ns = "urn:schemas-microsoft-com:office:office" />

Simply put, Group Delay is the rate of change of phase. Put another way, it is the slope (derivative) of the phase response with respect to time relative to an angular frequency.

OK, I guess that should put to rest any questions anyone has! [][]

Rather than wax on (I'll wait a moment for the cheering to subside...), take a look at some correspondence regarding group delay from John Murphy. It describes the Blauert and Laws research* as well as the need for additional measurements addressing the audibility limits of group delay in the LF region where it is most pronounced and most audible.

Generally speaking, it is desirable for the group delay (rate of change of phase) to be constant across all frequencies; otherwise there is temporal smearing of the signal.

It is essential that the total delay through the system is kept to a minimum. This requirement limits the amount of equalization that can be achieved on the phase. A compromise where the phase is corrected to a point that does not increase the overall perceived delay of the signal is necessary, and ensures that the phase is still improved. It is also important to consider that large group-delay variations between frequencies can cause considerable audio distortion [14], and this should also be kept to a minimum.*

This extends both to the use and design of speaker crossovers that divide frequency pass bands into separate 'pieces' with differing phase as well as to the next order of magnitude involving multiple speakers. When correcting the signal arrival times between groups of speakers, uniformity of phase between speakers is important in improving imaging (the perceived direction of the sound), because the time arrival of each frequency will be the same for a listener sitting equidistant from all speakers. **

I am not aware of any additional detailed study of the audibility limits of group delay in the critical lower frequencies other than the table supplied by Blauert and Laws.

Mike, (Doc), this could be a very good subject for a project and a very publishable research paper. Hint Hint!

______________________________________________________________

* Blauert, J. and Laws, P "Group Delay Distortions in Electroacoustical Systems", Journal of the Acoustical Society of <?xml:namespace prefix = u1 /><?xml:namespace prefix = st1 ns = "urn:schemas-microsoft-com:office:smarttags" />America, Volume 63, Number 5, pp. 14781483 (May 1978)

** Richard Heysers works

*** Transfer-Function Measurement with Sweeps, by Swen Muller, http://saturn5.com/~jwb/mueller.pdf (Excellent Paper!!! But too large to attach)

**** "The Differential Time-Delay Distortion and Differential Phase-Shift Distortion as Measures of Phase Linearity" by Marshall Leach, JAES, Vol. 37, No.9, September 1989

LdspkrGroupDelay-JMurphy.pdf

[mas]

Here is:<?xml:namespace prefix = o ns = "urn:schemas-microsoft-com:office:office" />

"The Differential Time-Delay Distortion and Differential Phase-Shift Distortion as Measures of Phase Linearity" by Marshall Leach, JAES, Vol. 37, No.9, September 1989

DxTimeDelay&DxPhaseShiftDistortion.Leach.pdf

[mas]

Maybe I can take a swing at explaining group delay in a way that may help one be able to more clearly visualize what is going on.<?xml:namespace prefix = o ns = "urn:schemas-microsoft-com:office:office" />

We often talk of a signal arrival time. Take for instance, the Direct signal, Ld, in a room. If we envision a broadband impulse (click) of an infinitely short duration (hey, it's my fantasy where such things exist), we can imagine the entire signal arriving at a finish line at the same precise instant. It is this time differential between the time of origin and the time of arrival that is commonly referred to as 'time delay'.

But in fact, the entire impulse signal does not arrive at the same instant. Rather, it's arrival is spread over a small period of time. And it's behavior during this time distribution (referred to as 'phase') is what is focused upon in 'group delay'.

Remember the comic books and Sci-fi shows where a character moves incredibly fast? Remember how the special effects folks created this effect by showing the character stretch and blur into a character wider then if he were standing still? Well, that is how the actual signal travels. The different frequencies that constitute the impulse signal do not all arrive at the same time. There is a stretching a distribution in time regarding the propagation of the frequencies that constitute the impulse. This distribution (stretching) of the impulse over a finite period of time, and whether the stretching is uniform and contiguous or broken into pieces is what is commonly referred to as 'group delay'.

Now, if the stretching is contiguous, and the blur is a well behaved blur with uniform stretching, and all of the adjacent frequencies arrive in order, only just a little out of time (out of phase), then all is well and good as long as the blurring is well behaved and evenly distributed at the same rate over the frequency bandwidth and not stretched too far. Thus our cartoon super-hero arrives in one piece.

But, if the blurring is not uniform, if parts of the frequency composition are stretched in time (the rate of the change in phase) significantly more than other parts (imagine looking at yourself in a fun house mirror), or even more significantly, if the total frequency band pass constituting the impulse is broken into pieces and these various pieces arrive spread over different non-contiguous times, each acting as separate discrete signals (as often happens with a passive speaker crossover & separate drivers), there is a temporal smearing that becomes apparent in the perception if the 'change in the degree of stretching' is great enough or if the differing arrival times of the discrete packets are great enough apart. In this case, our super-hero is in trouble. He has become stretched out of proportion, or worse, he has broken into several pieces, and we enter the realm of Humpty Dumpty - as this uneven behavior becomes audible.

Now the exact limits of the unevenness of the stretching (the phase) and the exact limits of this separation delay for discontiguous 'bandpass packets (pardon the abuse of language here) that are noticeable have yet to be detailed as finely as we might like. But this issue of temporal smearing is a fundamental issue in Dick Heysers work.

I hope that my poor description allows you to have a better idea of the order of magnitude difference between the basic arrive time of a signal considered as a whole (the time delay), and this same signal magnified and looked at as a continuum (the group delay).

[PrestonTom]

Frequently the question of group delay (or deviations from it) is asked in reference to time-alignment of drivers or the audibility of of having different drivers at different distances from the listener (this would be one application). Blauert's paper is frequently cited but there is more recent work by BCJ Moore et al in JAES (within the last year or so) that is a good reference.

-Tom

[kramskoi]

Maybe I can take a swing at explaining group delay in a way that may help one be able to more clearly visualize what is going on.<?xml:namespace prefix = o ns = "urn:schemas-microsoft-com:office:office" />

We often talk of a signal arrival time. Take for instance, the Direct signal, Ld, in a room. If we envision a broadband impulse (click) of an infinitely short duration (hey, it's my fantasy where such things exist), we can imagine the entire signal arriving at a finish line at the same precise instant. It is this time differential between the time of origin and the time of arrival that is commonly referred to as 'time delay'.

But in fact, the entire impulse signal does not arrive at the same instant. Rather, it's arrival is spread over a small period of time. And it's behavior during this time distribution (referred to as 'phase') is what is focused upon in 'group delay'.

Remember the comic books and Sci-fi shows where a character moves incredibly fast? Remember how the special effects folks created this effect by showing the character stretch and blur into a character wider then if he were standing still? Well, that is how the actual signal travels. The different frequencies that constitute the impulse signal do not all arrive at the same time. There is a stretching a distribution in time regarding the propagation of the frequencies that constitute the impulse. This distribution (stretching) of the impulse over a finite period of time, and whether the stretching is uniform and contiguous or broken into pieces is what is commonly referred to as 'group delay'.

Now, if the stretching is contiguous, and the blur is a well behaved blur with uniform stretching, and all of the adjacent frequencies arrive in order, only just a little out of time (out of phase), then all is well and good as long as the blurring is well behaved and evenly distributed at the same rate over the frequency bandwidth and not stretched too far. Thus our cartoon super-hero arrives in one piece.

But, if the blurring is not uniform, if parts of the frequency composition are stretched in time (the rate of the change in phase) significantly more than other parts (imagine looking at yourself in a fun house mirror), or even more significantly, if the total frequency band pass constituting the impulse is broken into pieces and these various pieces arrive spread over different non-contiguous times, each acting as separate discrete signals (as often happens with a passive speaker crossover), there is a temporal smearing that becomes apparent in the perception if the 'change in the degree of stretching' is great enough or if the differing arrival times of the discrete packets are great enough apart. In this case, our super-hero is in trouble. He has become stretched out of proportion, or worse, he has broken into several pieces, and we enter the realm of Humpty Dumpty - as this uneven behavior becomes audible.

Now the exact limits of the unevenness of the stretching (the phase) and the exact limits of this separation delay for discontiguous 'bandpass packets (pardon the abuse of language here) that are noticeable have yet to be detailed as finely as we might like. But this issue of temporal smearing is a fundamental issue in Dick Heysers work.

I hope that my poor description allows you to have a better idea of the order of magnitude difference between the basic arrive time of a signal considered as a whole (the time delay), and this same signal magnified and looked at as a continuum (the group delay).

good layman's term description, i'm of the same opinion as Doc on the time delay issue...these are time-domain issues, and to say that they are'nt interrelated would be a bit premature IMO...a constant delay through the signal envelope "is" desirable but here we are talking about delay in the 2-4 msec range in the higher frequencies and anywhere from 10 to 100msec in the lower bass range down to 10 Hz...

it is my thinking that the "room" is the great equalizer in all this. If we agree that the basic loudspeaker is a highpass circuit, passing the highs and rolling off the lows, would this "temporal shift" be more noticeable in an anaechoic chamber? How much is the "room" masking the issue? especially larger rooms, with their reverberant sound fields.

Or perhaps our ear/brain learning has compensated for this phase shift between high and low frequencies...over many years of less than perfect acoustic reproduction.

Keeping the phase shift as minimal as possible seems to contradict the practice (by some) of crossing over mains to the subwoofer at frequencies appreciably lower than 80-100 Hz...for example, 40 Hz, which would add twice the group delay of an 80 Hz crossover...cross lower and it becomes even worse...

Of course, the biggest question is always "when does it become audible"?

[Raider]

GREAT discussion so far folks. Thanks for all the positive and constructive contributions. And i am closing in on an understanding of this.

What is Temporal Smearing?

(I had always thought it was the condition of my wife's makeup after she rubs her head wondering how we can enjoy this kind of converstaion. : ] )

I also see it referred to as "temporal shift"

[TheEAR]

Now you are digging deep. Yahoo takes you to...

And they redirect to Transient Smearing!

Means that the transient "has spread" on time domain. This happens often because of gibbs oscillation (pre echo), but not necessary always. Smeared transient sounds softer or it has a pre- and/or post-echo noisy sound.

wiki.hydrogenaudio.org good place for advanced audio definitions.

Bose must have very pronounced transient smearing. []