Driver time alignment

JDUBS · « **on:** 7 Mar 2011, 02:00 am »

Guys, is there a specific process to follow for time aligning speaker drivers? I have the impulse response as measured by Fuzz Measure but am unsure of how to translate this into the appropriate driver delay(s).

This is for my Yorkvlle Unity 3-way speakers with the tweeter and midrange both located in the horn section and situated above and behind the woofer.

Thanks,
Jim

jtwrace · « **Reply #1 on:** 7 Mar 2011, 02:02 am »

Great question. How about posting the data so we / I can follow along with you?

JohnR · « **Reply #2 on:** 7 Mar 2011, 04:42 am »

I'm just about to try a technique for that today... will report back.

JohnR · « **Reply #3 on:** 7 Mar 2011, 10:48 am »

Let me know what you think of this. I'm assuming you are using some kind of digital crossover?

Disable the crossover filters
Set the "furthest away" driver to a delay of 1 ms
In Fuzzmeasure run a sweep within an overlapping range for both drivers - for my woofer and mid I ran the sweep from 400 to 800 Hz
Zoom in on the impulse response

Heres what I got:

If the woofer and mid were time-aligned physically then the impulse responses would be 1 ms apart. However the woofer impulse is a little more delayed - I figure it's 0.08 ms. So, the time delay on the mid to bring them into line will be 0.08 ms.

Note: this is an open-baffle speaker so the impulse response has a double (positive and negative) peak. For boxed speakers I think you should get a cleaner impulse - I'd be interested to see

JDUBS · « **Reply #4 on:** 9 Mar 2011, 03:49 am »

Hey guys. John, very interesting results and the logic definitely makes sense! I pulled these out of Fuzz, last time I did some measurements:

Frequency response:

And impulse response:

-Jim

JohnR · « **Reply #5 on:** 9 Mar 2011, 03:18 pm »

Hi Jim - you'd need to zoom right in on that impulse response to see anything useful at all for this - the time differences you are looking for are a fraction of a millisecond. (I measured 0.08 ms delay from woofer to mid)

JDUBS · « **Reply #6 on:** 10 Mar 2011, 02:17 am »

Sorry, hopefully this is better:

-Jim

HAL · « **Reply #7 on:** 10 Mar 2011, 02:21 am »

Try looking at the step response instead of the impulse response. Should be easier to see the time offset between the drivers.

JDUBS · « **Reply #8 on:** 10 Mar 2011, 02:34 am »

Here's the step response:

HAL · « **Reply #9 on:** 10 Mar 2011, 02:39 am »

If you expand the time from 25-40 mSec you will see the tweeter, midrange and woofer time offsets pretty clearly.

EDIT: looks like the time scale is much longer now. Is this a really big horn speaker?

JDUBS · « **Reply #10 on:** 10 Mar 2011, 02:43 am »

Here it is with 25-40 on the horizontal:

HAL · « **Reply #11 on:** 10 Mar 2011, 02:51 am »

Looks like 1=tweeter, 2=midrange, 3=woofer. The time between them is the delay needed for the correction.

This looks like a really big horn setup!

JDUBS · « **Reply #12 on:** 10 Mar 2011, 02:52 am »

Quote from: HAL on 10 Mar 2011, 02:39 am

If you expand the time from 25-40 mSec you will see the tweeter, midrange and woofer time offsets pretty clearly.

EDIT: looks like the time scale is much longer now. Is this a really big horn speaker?

Hal, its a unity horn (tweeter at throat surrounded by three mid-range drivers) with the woofer sitting below the horn. Here's a pic:

http://www.yorkville.com/products.asp?type=29&cat=38

Should I adjust the horizontal scale further? Thanks for your help with this!

Jim

HAL · « **Reply #13 on:** 10 Mar 2011, 02:57 am »

The original time axis for the impulse response was about 2.2mSec long. The step response is 15mSec long. Not sure with the software what happened, but the step response should also be about the original length.

JDUBS · « **Reply #14 on:** 10 Mar 2011, 03:13 am »

Here's the impulse response with the same value limits on the horizontal axis:

Does this make more sense?

-Jim

HAL · « **Reply #15 on:** 10 Mar 2011, 03:24 am »

Jim,
The shorter impulse response is the correct one. The longer one shows room reflections coming in at 39mSec.

From the size of the Yorkville cabinet, the shorter impulse response you showed earlier from ~27mSec to 30mSec is good. I do not know the software, but looks like some setting changed when you did the step response. It looks like it is 10x the time it should be.

The reason I say this is that sound travels at ~1130ft/sec. That means that sound travels 1.13ft in 1 mSec. The driver alignment from the picture has them within inches of each other. So the mid at 1.1mSec and woofer at 7mSec after the tweeter are very large travel times.

Something does not make sense.

JohnR · « **Reply #16 on:** 10 Mar 2011, 03:29 am »

Jim, is this with the original crossover still in place?

JDUBS · « **Reply #17 on:** 10 Mar 2011, 03:50 am »

Ok, here are the impulse and step using the same horizontal axis limits...27.5 on the min and 31 on the max.

John, the passive crossover components had been removed when the measurement was taken. I was running everything through my Mac Mini, Pure Music, and TC Konnekt 48 8-channel DAC.

-Jim

JohnR · « **Reply #18 on:** 10 Mar 2011, 04:00 am »

Hi Jim, I'd suggest turning off the tweeter and adding a time delay to the mid of say 2 ms. Then disable the crossover and run a sweep just around the crossover region. That should enable you to see both impulses clearly. (I'd be interested to know

)

The other way to do it is to loopback the measurement signal into the other channel of your measurement interface. That will give you a zero reference so you can get an exact time delay, and then you can measure each driver separately and take the differences. I haven't tried this with Fuzzmeasure but it should be possible.

jimdgoulding · « **Reply #19 on:** 10 Mar 2011, 04:50 am »

Now, I'm not a speaker builder so feel free to tell me to piss off, but why wouldn't one physically align the coils of the drivers in the first place? I'm sure there are other considerations. You'd have to dispence with the single box design, I suppose, and in doing so efficiency.