Concept of time delay?

[Deang]

"After a few years listening to these speakers, I realized that I couldn't image stereo with the speakers very well...

If you put them in false corners and turn them into the listening area, they image pretty good -- not like a good monitor, but they do 'image' -- and do something a pair of monitors, or even good size floorstanders can never do -- throw a huge, expansive, and all enveloping soundfield.

"Still later, when working with the Grateful Dead and becoming familiar with the Jerry Gracia's voice, I realized that the K-horn would NOT reproduce Jerry's voice accurately, in fact it sounded like someone else."

They should have epoxied a K-400 to Garcia's mouth and performed some A/B testing -- they probably would have discovered it didn't have much to do with time delay.

[sunnysal]

dean, my khorns image VERY well, even without toeing them in. I get a vivid stereo image in my listening room, not much depth, but L/R stereo image is great! adding the time delay seems to clarify the mid and bass a bit though, nothing scientific, just my impression ya'know. regards, tony

[PrestonTom]

The issue of audibility of time delay is an interesting one.

I am going to give a rule of thumb, since the details would require a deeper explaination. You are probably insensitive to a time delay (across the spectrum) if it less than a couple of msec (or about two feet of extra path length). In the real world case (misalignment across dirvers), there are other cues that result from reverberation and and comb filtering. That is a seperate issue, although it can also be "introduced" with the time delay.

This is seen with measures using signals with certain properties that allow you to rule out other cues and measured in a lab setting that allows the minimization of problems due to auditory memory, attention and practice & fatigue effects. Whether you would get comparable results using music and casual listening is debatable. Although, I suspect that it would not be very noticeable at delays of less than a few (to several) msec.

I am being very careful to specify this as a time (about 2 msec) where the minimal difference is detectable. It does not mean it is detectable in every instance (although the listeners are above a "guessing" rate). It does not mean the difference is clearly audible. Nor does it mean the difference is "heard" as being clearly objectionable.

Good Luck,

-Tom

[DrWho]

How many ms is it before the human ear perceives the sounds as coming

from different sources or that it interferes with phasing or musicality?

That's actually a loaded/complicated question to answer. There are two

concepts at play here as well: "different sources" and "interference"

(to use your words).

I believe Bell Labs conducted some experiments with tap dancers to

indicate that the same tap must be delayed by 10ms before it is

percieved as a different tap. Anything shorter is percieved as a single

longer tap. A tap is a pretty short sound (probably on the order of

10ms actually) so sounds that are much longer - say organ notes in a

reverbrant church - might mask this effect a bit better. I suppose it

wouldn't be too hard to construct some wave forms to demonstrate the

effect.

However, the argument driving the importance of time-delay has nothing

to do with percieving the same sound as happening twice and has been

pretty well explained already. It is something that becomes very

frequency and phase dependant and doesn't only have to be an issue in

the passband of the crossover overlap. A lot of people talk about the

importance of having only one driver reproduce all the notes of a

particular instrument and this would be the technical reason for the

issue.

My favorite example is that of the kick drum - it's super dynamic and

has an important passband from 40Hz-3kHz and an extreme of 10Hz-10kHz

if you want to get picky. And then there are two sounds associated with

the kick: the initial thwack (200Hz and up) and the following boom

(200Hz and below). On a speaker like the khorn you are effectively

increasing the time between the thwack and boom which makes the drums

sound even larger - in fact, this is a trick implemented all the time

in the studio. Combine that with the huge dynamic capability of the

fully horn loaded khorn and you have one awesome drum speaker. (If it

weren't for the crazy peaks in the 100Hz region I'm sure the khorn

would sound a bit on the boomy side).There are plenty of other

instruments where the meat slowly builds after the attack and lots of

stereo pair recordings where the phase response is just smeared in the

lower frequencies (due to reverberation of the room) that work to mask

the effects of time-misalignment.

But take something like a guitar or male voice and you suddenly run

into very audible issues due to the complex harmonic structure of the

sound...especially when you start getting into thicker material (like

heavy strumming or choral peices). You know that lobing you get when a

stringed instrument is perfectly in tune and chords are played? that

goes away when you've got time issues.

Anyways, that's my slighly educated attempt at trying to explain the

audibility which is ultimately what matters. One of these days I'll see

if I can't find some source material to demonstrate with.

[Guest " "]

Now would any of this apply to my current setup?

Since I'm running the Cornwalls as bass bins, the K33 driver is right on the front baffle (motorboard). However, with the 511B horn, the 902 driver is over a foot back behind the motorboard, so wouldn't that already naturally correct a little bit for the time alignment?

It could be a potiential problem for any system in which more than one driver is used, and the mid-bass area, mid-range area, and high range area were not on the same driver.

My reading of the below, assuming we could agree where the mid-bass ends (assume 480hz for now), is that the only way to immune your present configuration from the effects of time-delay in a 2-way system, is to crossover at the begining of the mid-bass area, which could be defined at 240hz. The below defines time delay as being a problem specific to mid-bass and tweeter area. So if it were possible to setup a two way to crossover at 240hz, you would be immue from the effects of time delay. Makes sense, since lower frequencies are omi-directional. Assuming mid-bass actually began at 240hz.

""A time delay is introduced whenever the acoustic centres of any two loudspeakers are different from the perspective of the listener. It is not usually a problem in a three way system between the woofer (or subwoofer) and midrange or mid-bass, but between the mid/ mid-bass and tweeter there is usually considerable room for misalignment in the time domain."

[DrWho]

And if the speaker designer included this

time difference in their original voicing, will using an outboard delay

possibly cause other problems in the soundwave?

never thought about that one before...

I suppose by going to an active crossover you are technically unvoicing

the voicing they accomplished with the crossover and starting that part

over. Now if they modified the design of the individual components to

take this into account, then you'll probably need to replicate those

parts of the previous crossover with the active.

Ultimately, by going with the active you would have one less variable

to worry about in the compromise chain, so arguable you should be able

to arrive at an overall less compromised system (which should imply

sounding better).

[Marvel]

I am trying to understand how to implement delay in my system.

I hope I have worded this correct.

Greg

I thought that was when the wife says "Not tonight......"

[DrWho]

Now would any of this apply to my current setup?

Yes! Choosing different crossover points may shift around the comb-filtering issues in the passband, but it's the rest of the sound you gotta worry about it.

Ah, there's my bus - gotta run.

[Guest " "]

Can everyone be expected to notice the effects, or just a few?

" If you hear nothing at all when changing the delays in 1' increments, you may not be sensitive to such small timing errors, and thus shouldnt worry about changing speakers -- or the timing errors in the speakers are large enough to mask improvements in speaker setup dimensions so you can't really tell if time- and phase-correct speakers would be a further improvement"

[Deang]

http://www.trueaudio.com/post_010.htm

[Guest " "]

http://www.trueaudio.com/post_010.htm

intresting chart at this link. the spec's for the 2k - 4k band is consistent with the notion that humans have more sensitve hearing in these bands.

So a 3 msec delay at 8khz may require only 1 msec of compensation to be below the 2 msec threshold of audibility.

And a 4 msec delay at 500hz could require just .8 msec of compensation to be below the 3.2 msec threashold of audibility.

Frequency Threshold of Audibility

8 kHz 2 msec

4 kHz 1.5 msec

2 kHz 1 msec

1 kHz 2 msec

500 Hz 3.2 msec

[DrWho]

I am fairly certain that group delay is not the same thing as time-arrival differences.

[D-MAN]

The lascala bass bin is about a 4 foot path from the driver to the exit of the horn? So from the isntant the driver starts playing, there will be 4ms before sound makes it out the front. I believe the altec HF section is about 18 inches long? So it takes 1.5ms for the sound to get to the front. You will want to delay the HF section by 2.5ms so that the wavefronts lineup at the motorboard.

The drivers with the shorter propogation distance are the ones getting the delay.

Your premise is completely flawed, being based on an inaccurate assumption.

The LS has an approx. 28" overall pathlength (center of channel), NOT 4 FEET!

(The MOUTH size is 4 square feet).

DM

[DrWho]

The lascala bass bin is

about a 4 foot path from the driver to the exit of the horn? So from

the isntant the driver starts playing, there will be 4ms before sound

makes it out the front. I believe the altec HF section is about 18

inches long? So it takes 1.5ms for the sound to get to the front. You

will want to delay the HF section by 2.5ms so that the wavefronts

lineup at the motorboard.

The drivers with the shorter propogation distance are the ones getting the delay.

Your premise is completely flawed, being based on an inaccurate assumption.

The LS has an approx. 28" overall pathlength (center of channel), NOT 4 FEET!

(The MOUTH size is 4 square feet).

DM

Which is why I put a ? behind my assumption. I

wasn't sure how far back the woofer sat in the cabinet - I just know it

has a 2x2 footprint on the ground. So 2 feet to the first bend and then

at least another foot to the middle along the back. But since the

center of channel pathlength comes out at an angle from the sides I

figured I'd round up to keep the math pretty - making the concept

clearer.

There is more involved with actual propogation delay than just the

length of the path. The actual acoustic center is something that will

also vary with frequency - and vary with the source material as well.

[Guest " "]

So many charts, so little time.

Since 5 meters is quiet a bit of lenth for drivers to be spaced appart, the chart is useless with out some ammendments. My best guess on the amendments (stubby pencil approach)

Metres delay in milliseconds

.87 1.825 msec

1.75 3.65 msec

2.5 7.3 msec

In the below table, both meters and feet are provided

////////////////////////////////////////////////////////

Time Delay Table for loudspeakers

Metres delay in milliseconds Feet delay in milliseconds

5 14.6 20 17.9

10 29.2 30 26.8

15 43.8 40 35.7

20 58.5 50 44.6

25 73.1 60 53.6

30 87.7 70 62.5

35 102.3 80 71.4

40 117.0 90 80.4

45 131.6 100 89.3

50 146.2 110 98.2

55 160.8 120 107.1

60 175.4 130 116.1

65 190.1 140 125.0

70 204.7 150 133.9

75 219.3 160 142.9

80 233.9 170 151.8

85 248.5 180 160.7

90 263.2 190 169.6

95 277.8 200 178.6

[PrestonTom]

I am fairly certain that group delay is not the same thing as time-arrival differences.

Mike,

Folks are frequently confused on this point.

Imagine a plot of relative phase (y-axis) as a function of frequency (x-axis). A pure time delay will be a linear shift in phase. So one can compute the time delay (in msec) by noting the slope of this function. Things get more complicated since the output of a system will frequently not plot out as a straight line in this coordinate system. So we need to do a couple of things. First the function must always intercept the y-axis at at either zero or pi (180 or -180 degrees). There a few other things about this function that are of interest.

If the signal is some sort of modulated carrier signal. Then the "slope" at the frequencies of the carrier is the "carrier delay". The slope extending around the sidebands that created the modulator (plus & minus the around the carrier) is the "envelope delay". Note you can introduce delays / phase shifts so the carrier and envelope have different delays (i.e, the slope changes in this region).

Now to you question about the group delay. The "group delay" is the approximate slope around the carrier and envelope. When written as a spectrum it is the negative of the 1st derivitive wrt to frequency (w). The slope at the asymptotic frequency is sometimes called the "wavefront delay"

Some will cringe at my answer since I have not used any equations and the answer is not complete....

But the quick answer regarding the current topic is that the group delay and propagation delay (time of arrival difference) will be about the same (expept for some extra phase shifts introduced by the crossover and drivers acting like inductors or whatever.

Good Luck,

-Tom

[Deang]

Wow, nice job.

I am fairly certain that group delay is not the same thing as time-arrival differences.

True.

[Guest " "]

Tom and Dean,

Thanks for the detail.

The last sentence in both your replies, in my mind, is a good summary/conclusion.

[sfogg]

Tom hit on an important point that is worth repeating.

"In the real world case (misalignment across dirvers), there are other cues that result from reverberation and and comb filtering. That is a seperate issue, although it can also be "introduced" with the time delay."

You need to look at what was studied and how it was studied to understand if the findings of the study is signifigant. If a study seeks to validate wether time delay and/or group delay is audible did they also try to eliminate all the other 'real world cases' and the other side effects/artifacts that can be brought on?

If so how applicable are the findings to the real world?

For example... is arrival time audible?

I could easily setup a case where it is not.... a single speaker playing vs. that same speaker playing 20ms (full bandwidth) delayed would be inaudible.

And I could just as easily setup a case where it is. Two speakers playing mono material in time alignment vs. two speakers playing mono material with one of them delayed by 20ms.

There was some papers in the AES recently talking about group delay. AFAIK the concensus in them was by itself group delay is pretty much inaudible in most cases.

However, if it (or driver time delay) causes other problems (comb filter...etc..etc) those other problems could be audible and the root cause would be the delay.

As far as if those other problems (comb filter...etc..etc) are audible it is pretty easily shown that it is.

Shawn

[Marvel]

The lascala bass bin is about a 4 foot path from the driver to the exit of the horn? So from the isntant the driver starts playing, there will be 4ms before sound makes it out the front. I believe the altec HF section is about 18 inches long? So it takes 1.5ms for the sound to get to the front. You will want to delay the HF section by 2.5ms so that the wavefronts lineup at the motorboard.

The drivers with the shorter propogation distance are the ones getting the delay.

Your premise is completely flawed, being based on an inaccurate assumption.

The LS has an approx. 28" overall pathlength (center of channel), NOT 4 FEET!

(The MOUTH size is 4 square feet).

DM

36 INCHES would be closer to the truth if you are in the center of the channel. Dennis says 36 to the dustcap.

Bruce