What's Delay For?

Isn't delay so we hear sounds from various speakers at the same time?

Not exactly. No speakers in a car are far away enough to be heard as multiple events - they always sound like a single sound. The gap between two sounds needs to be around 40mS before the second of the two sounds is apparent as a separate event.

So what's it doing?

Sound travels pretty slowly. While light travels 186 miles, sound travels about 13 inches. When the same sound arrives from two speakers in different locations, it often has to travel two different paths of different lengths, covering different distances. That means two different arrival times.

Every frequency of sound is a wave, and that wave has a physical length. Those lengths range from 675 inches at 20 Hz, to 0.675 inches at 20,000 Hz.

When the path length difference between the speakers is the same as the wave length of the frequency being played (or a whole multiple - 2x, 3x, 4x, etc), the two waves reinforce each other and the sound gets louder.

When the path length difference between the speakers is half of the wave length of the frequency being played (or an odd multiple of half, such as 1.5x, 2.5x, etc.), the two waves are 180 degrees out of phase, and they cancel each other out almost completely.

So, let's say the two speakers have a 27" inch path length difference from the driver's listening position. 27" is the wave length of a 500 Hz wave, but it's half the wave length for a 250 Hz wave. So the sound from these speakers is louder at 500, but almost disappears at 250 (as the waves cancel each other out).

The graph below shows three traces. The blue and red trace are the left and right sides, with the effects of reflections eliminated for clarity. One of these is delayed, but they still measure the same individually.

The green trace shows the result when both speakers play at the same time, with that 27" path length difference applied to one and not the other. We see a cancellation at 250 Hz. We also see one at 750 Hz, and 1250 Hz, etc. These are because it occurs at the half-wavelength mark, and at the odd multiples of that value. If we have a cancellation at 250, we have one at 750, and we have one at 1250, etc. It get louder at 500, 1000, 2000, etc.

Those cancellations are phase cancellations. Everywhere you see a dip, the two speakers' outputs are not in phase. At the deepest point of the dip, the phase mismatch is 180 degrees. Where the shoulder of the dip is only -3dB lower than the sum, that's 90 degrees out of phase. Where the dip has barely started - when the sum is 0.35dB down, or 0.65dB down, the phase difference is 30 degrees and 45 degrees, respectively.

So, we in the aftermarket usually use delay to correct this for the driver's seat. We delay the nearer speaker's signal, so that the output of the two speakers arrive nearly at the same instant, and the two sounds are simultaneous again. We don't really care about the simultaneity, but we want a linear frequency response - and we can't EQ out a phase cancellation.

Usually, we enter in "27 inches" and our DSP software converts that to about 2.01 milliseconds, and applies 2.01 milliseconds of delay.

And that eliminates the phase cancellation effects of multiple arrivals when speakers are different distances from the listener.

What else does delay do for us?

That's all it does. If there were phase cancellations present BEFORE the path lengths happened to the sound, they are still present.

It only works when the left and right channels start out aligned in time and phase. If the signals don't start out aligned, then delay can't get them all back to alignment.

If there are reflections which are also creating additional multiple arrivals (and there always are), then they have their own effects on the phase.

So, delay has a very specific purpose - to eliminate multiple arrivals and the phase cancellations caused by them. If you want to eliminate other phase cancellations, delay may not be the optimal tool. If you want to solve phase misalignments which have effects of less than 1dB, perhaps that isn't the most remunerative way to spend your time?

Why do OEMs manipulate the phase using all-pass filters?

So let's keep working with our imaginary car with a 27" path-length difference between the left speaker and the right speaker, as discussed above.

So in the graph in that post, the sum shows a bad cancellation at 250 Hz. On the chart, it's the worst of all of them.

When we in the aftermarket apply delay to the near speaker, we solve all the cancellations on this graph - that is, all the cancellations caused by path-length differences. But we only solve them at the listening position which we used in our measurements. The other seats are out of luck - we make the problem worse in the other seats.

OEMs prefer to have partial solutions which are symmetrical over more comprehensive solutions which are not symmetrical (that is, which only apply to one seat).

One such solution is phase manipulation using all-pass filters.

In this graph, you can see the 250 Hz cancellation is gone. That's because one signal isn't linear in phase. In this example, we used an all pass filter on one channel to invert the phase 180 degrees at 250 Hz. So now, the left and right sides are out of phase anywhere the path lengths are equal (such as sitting on the center console, or in the middle seat in the back). However, any place the path length difference was 27" - and the sounds would have been out of phase at 250 Hz - they're now back in phase!

So this method is literally a partial solution for more than one seating position. This technique has been detected in use in some GM, all Toyota for years, all Hyundai and Kia for years, and some Subaru head units.

And if all you're doing in this situation is swapping in new speakers, this processing helps you (it's not speaker dependent, it's speaker-location dependent). If you're adding an amp and speakers, this processing also helps you (as long as you don't sum any channels together in an attempt to make a full-range signal!)

But if we are adding a DSP processor into the system to get better sound and better stereo performance, we may have a problem.

Unfortunately, the way we usually use delay in the aftermarket doesn't work with this signal - because our typical use of delay presumes that left and right signals match in phase at every frequency! Once the phase has been manipulated, we can't use delay the way we normally do. There are a few workarounds which are possible with the right speaker driver component and the right crossover point, but in general, the presence of phase manipulation will foil the use of delay for driver's seat imaging.

So how do we know it's there?

Well, I used to sell a device called the Phase Summer to sum 2 channels together into a single-channel RTA so that phase cancellations could be detected (and thanks to all 14 of you who bought one). Some of you have purchased external devices which can perform the same function. Or you can use a dual-channel RTA with summing capability. This describes the Audison bit Tune using V 3.1 softtware, as well as the electrical input RTA capability in all Audison AF Forza bit amplifiers.? If there's a cancellation on the RTA graph when you sum the two channels together, you have a phase problem.

To learn what Audison suggests you do about it, see my article A New Framework for OEM Upgrades using AF Forza!