Subwoofers and Time Alignment in the Car

While working on some measurements for time delay settings in my car audio system, I remembered something I had talked to a buddy about, which was finding the crossover point between your midbass and sub, flipping one driver (set of drivers) out of phase, play a sine tone between those points and delay until you get as far attenuated as possible. Then flip back in phase and you’re done. The reason is pretty simple: I’ve never had good luck measuring the arrival time of a subwoofer or midbass that is band-passed to play low frequencies. If I increase the crossover to make them play higher – in order to get a clean impulse measurement – I still can’t get it “just right” and wind up having to dial it in by ear. So, I figure if I can do this and it works, all the better. It keeps the measurement more “real” because I’m measuring a frequency that the drivers are playing when the system is really running, rather than setting pseudo crossovers for the purpose of measurement that has yet to provide me meaningful results.

I put it to the test tonight.

Below is a plot. Without telling you what my midbass setting was already, it doesn’t really give you a specific number to go by. However, the point is to show that T/A has a maximal affect at the crossover point and there is a point where adding more hurts you. I hope you’re able to discern this from below.
I used a 65hz tone and delayed or reduced delay until the SPL was at it’s lowest. This was the point where it was most out of phase. Then all I had to do was flip the polarity back on the subs and I was the most in phase.

(click the image below to view larger)

So, as you can see, 2.00ms delay is the point at which the out-of-phase signal is most attenuated (most out of phase). Since the subs were flipped out of phase, I just flipped them back and I was done.

Here is the final result comparing the original in-phase measurement (my starting point at the beginning of all of this) vs the final result with 2.00ms delay added to the sub channels.


There’s about 1dB difference. Not a whole lot numbers-wise, but it actually works pretty well. Flipping between presets, where one is the original and the other is the new setting, the sound of the subwoofer is more solidly anchored up front. However, the tactile impact is also more apparent. This is a personal issue, so don’t let it become yours. 😉

I also went through the same procedure on the midbass/midrange transition and found similar results; there is a breaking point where the phase starts to turn the more you delay.

In case you’re wondering…

Why did I flip the phase (0/180) at first instead of just adding delay? No specific reason, really. Initially this was supposed to be done by ear. You can easily hear when something’s wrong. In this case, “wrong” is “not there”; aka: cancellation. It’s harder for me to hear when something is just right with the sub. The wavelength is super long and I’ve honestly just not had the best go at finding when the phase is just perfect in that transition. Drives me nuts. That’s why I started the process this way; the idea was to go until there was little-to-no sound when playing that specific tone. I just used a mic instead. 😉



Addendum: 07/25/12:

Essentially, what you’re trying to do is align phase of the subwoofer/mids at the crossover via delay (time alignment).

When talking about delay, consider the wavelength as well. That’s why you may notice that even minute changes in delay (ie: 0.02ms; one click resolution on most current DSPs) will affect what you hear.  As shown above, it doesn’t take a large amount of delay between drivers to impart in-phase/out-of-phase response at the listening position.  It seems many believe very large amounts of delay are needed for drivers/passbands of longer wavelength.  The above shows this is not the case.

In every iteration of my system in the past couple years, three 0.02ms adjustments is a way I come circle in the phase on my tweeters. One 0.02ms adjustment may be out of phase, the next is a bit more out of phase and the final (3rd) is *snap*… back in phase. I’ve talked to others who have experienced this. It’s interesting to note that ILD is often purported as what drives our hearing of higher frequencies (2khz+) and many will say that time alignment is not necessary here because our ears are sensitive to differences in frequencies only below approximately 1khz. However, time alignment will net you the ability to incrementally align phase in higher frequency(s) and result in a more coherent soundstage.