A time alignment is best performed after you have settled on the x-over and equalisation of your speakers. As an example I will time-align a 3-way speaker system: woofer–mid–tweeter. If you have a 2-way system you will have less work to do; for a 4-way some more. This will be obvious once you read the recipe.
ingredients
- a multi-way speaker system (heh);
- a working acoustical speaker measurement system, able to show you a magnitude frequency sweep and a pulse response;
- a way to measure each driver individually, through its normal x-over and equalisation;
- a way to temporarily add an extra x-over slope to each driver.
You can see that if you have a speaker system that connects to one amp (per channel) through a passive x-over, you are going to have the problem that when you disconnect the other drivers to measure one of them solo, you have changed the response of that driver. This is because all those disconnected x-overs and/or drivers are no longer loading it down. This happens especially in our regions of interest: the x-over regions, where you have power coming out of two drivers, and thus each must be loading down the other. I have no solution for this problem, sorry.
All acoustical measurements are to be performed at ‘the listening position.’ Yeah that is a fudge (is there such a thing?), but you got the point: no near-field measurements. The time alignment that matters to us is the one at proper listening distances—the ones you use. Measure too close and driver offsets start to distort the time measurement. Measure real close and you ignore the dispersion of the driver; this distorts the magnitude response measurement.
preparation
Measure the magnitude frequency response of each driver individually. From this measurement deduct the effective x-over frequency and slope for each driver. Effective, because you want to know the real response—electrical, mechanical and dispersion combined. For our example we want to know the low-pass of the woofer, the high-pass of the tweeter and high + low-pass of the mid driver.
cooking
We will now time align the woofer with the mid and then the mid with the tweeter. In our case there is no such thing as aligning the woofer with the tweeter.
woofer–mid alignment (one channel):
- that high-pass that you measured for the mid driver, effective frequency and slope: apply it to the woofer too; this is most easily done in the digital domain; leave the woofer’s normal x-over and equalisation intact;
- that low-pass that you measured for the woofer, effective frequency and slope: apply it to the mid driver too; again leave the mid’s normal x-over and equalisation intact;
- measure the magnitude frequency response of the woofer and mid, individually, and compare the two to check that for all practical purposes, their output—in effect the x-over region of these two drivers—is identical (if not, repeat and adjust steps 1 & 2); you will have to turn up your system volume by about 10dB to get some decent volume out of it; if your measuring software allows you to overlay 2 measurements, that helps;
- measure the pulse response of the woofer and mid, individually, and compare the two; if your measuring software allows you to overlay 2 measurements, that helps;
- look for the apex of the first peak of both pulse responses (watch out, the first peak may go down instead of up); the peak of one driver is going to be a bit earlier than of the other driver;
- your job is now to centre these first peaks of the mid and woofer, either by adding delay to the ‘early’ driver (digital, or by means of an analog all-pass filter), or mechanically, by moving either driver forward/backward towards the listening position;
- repeat step 4 to verify your progress, then repeat steps 5 & 6 as necessary;
- first pulse peaks centred? done; remove the high-pass on the woofer and the low-pass on the mid—mind that 10dB extra system volume.
Now repeat for the other channel.
digest
As mentioned before, when you change things in your speaker system, you will have to perform the time alignment again. These changes include: changing driver types, enclosures (including horns), x-over frequency or slope, equalisation, not to mention moving your system to another room, or changing that mythical listening position. Luckily, performing a time alignment for the second time is going to be a lot faster.
Unclear, mistakes, requests? Let me know in the comments below.
Speaker builder magazine covered a lot of this crossover and phase relationship work in the late 80s and early 90s. I got interested in it in Japan in the 1970s when some time-aligned speakers with the tweeters sitting back on the top of the box were produced and marketed. I lived in Japan and went to Akihabara back then.
ReplyDeleteOver the years I've found a few strategies to deal with this, as a speaker designing hobbyist.
Focus on two way speakers, the midrange crossover, and use wide-range drivers (usually).
Two wide range drivers can be made to roll off with first order acoustic characteristics (acoustic rolloff is what counts) for awhile. Hey an octave either side of crossover will do.
I've got a 500 Hz horn crossing over to a 100 Hz horn driven by a 12 inch "full-range" driver.
There is a subwoofer, but you don't hear bass phase in a room, so integrate that later. My 500 Hz horn rolls off at the top end. So do my ears. Altec 288-K drivers are what is in the "World's Fair" straight axis-symmetric horns. I get a first order rolloff on the JBL 123 woofers with a choke. I get a first order rolloff on the HF horns with a series resistor on the horns to bring their level down to match the woofers, then I shunt across the HF horn terminals with a low DCR inductor, which gives a nice first degree rolloff characteristic, despite driver resonance.
The other thing to do is to accept a 4th order Linkwitz Riley crossover, again acoustic characteristics, with something like an 8 inch driver and a horn tweeter, so the acoustic centers are not so hard to align, with the tweeter voice coil set back. A second order networtk on the tweeter usually combines with horn LF rolloff to give 4th order acoustic characteristics. If you choose the right woofer, it will have a 2nd order HF rolloff that you can tweak similarly. I got this to work with a Selenium Slot tweeter (cheap JBL semi-copy) and an 8 inch Fostex banana cone full range with the "wheezer" cone cut off. Your mileage may vary. I've gotten a cone tweeter to work at first order crossover to a couple of 6" woofers in D'Appolito configuration in the past, but a lot of luck was involved in finding parts that actually worked out, and efficiency was lower. That was in my early tube days with a HK Citation-2 PP pentode monster.
Anyway, large format 2 way horn around 1 kHz and smaller format 4th order 2-way around 2kHz or 1st order D-Appolito 2 way around 2 kHz with 6" and cone tweeter can be done, if you want to.
I’d like to point out that the subject of this post is time, not phase. what John describes above is a systems approach (2-way) and the x-over work. like I said in the post: you do all that—to taste and religion—and then you time align.
ReplyDeleteI do not agree that one does not hear phase in the bass region. I have created in the past some amplifier stages with flat magnitude (down to 20Hz), but lots of phase shift (say, already at 40Hz). within a minute of hooking that up I knew something was wrong. very strange bass, e.g. ‘bass only behind the speaker, but nowhere else in the room,’ or ‘bass comes from the next room.’ I look at the phase: ugh
phase and time are intimately connected...
ReplyDeletesure, just like voltage and current. and it really pays to use the right model when working to get things right.
ReplyDeletewhat we are trying to solve here is getting everything in sync out of several drivers, the native, and helpful, model is time-based
Altec Lansing had a great bulletin concerning this issue, might try the Altec Unofficial website and surf it, lot to be learned from those who went before, why reinvent the wheel
ReplyDeleteaaaah, that is why a google search on this topic turns up so many articles that guide us, with the computer tools of today, through this topic… NOT
ReplyDeletePeter, thank you so much for committing this information to print. I have a freshly built four way horn system (OK, 3 way horns plus bass reflex subs.) Like you, I've found a lot of information on time alignment, but none of it usable/reliable/sensible/understandable etc. until now.
ReplyDeleteYour technique is simple to follow AND execute. What's more, it makes sense. Much appreciated!