This is strange to me. Comb filtering was the first big problem I encountered, decades ago when I started measuring speakers in my living room. SPL meter and graph paper in hand, I didn't understand why the response was so peaky, up and down. That's when I learned about comb filtering and what happens when you measure two speakers at the same time. Ever since, I've been cautious about it.
I don't understand how you could NOT get comb filtering with a single point, fixed measurement.
Well, I was using a 1/3 octave RTA, so the frequency resolution was a bit limited. Still, the thing went up to 16 kHz, and down to 32 Hz, so we could see what the overall FR looked like. Also, this is in a big room (factory floor), but no damping anywhere.
But ... the mike has to be on centerline with a 1/2" (10mm) accuracy, or the HF region will be a mess. And the speakers really do have to phase and amplitude-match pretty precisely, as well as crossover parts (particularly capacitors) that have good matching.
In a speaker as rough as an old-school Altec 604 Duplex, or anything with a simple 1970's-style crossover, no chance this will work. If speakers with rough responses could accurately phase and amplitude-match, sure it would work, but in the real world speakers with rough responses are that way due to bad cabinet reflections, uneven driver response in the crossover region, breakups in the passband, or a whole bunch of other problems.
These problems tend not to phase-match across a speaker pair. And if the phase does not match (within a few degrees), the amplitude summing will be all over the place, along with dubious image quality.
A little story: the first speaker I designed at Audionics in 1975 was the TLM-200, a gigantic 4-way system that was given to me when the previous designer high-tailed it out of town without a forwarding address. The stupid thing had the drivers and cabinets already selected (holes pre-cut in production-quality cabinets that were 5 feet high), but no f*#%ing crossover at all. It was either design the crossover or find a new job; take it or leave it. Some time later, armed with the aforesaid RTA, B&K mike, and large box of resistors, inductors and capacitors, I finished the project. Sound quality was OK for a first effort.
The cabinets were in mirror-image pairs, with the drivers offset but in a vertical line (again, I didn't design it, it was my job to make it work). Once the things were done, the imaging was pretty decent, and the response was flat within +/- 1.5 dB, not bad for 1975.
A few months later, our little band at Audionics arrives at the Chicago Summer CES show, but the half-wits at the factory shipped two LEFT speakers. Faces immediately fell when we unboxed what we thought would be the right speaker. Well, now what? We fired them up, looking a bit strange, but in those days most US-made speakers had drivers casually arranged all over the front face, and were not sold as mirror-image pairs, so maybe we could get away with it.
Until we listened to them. To the dismay of all, but particularly myself, they had no imaging at all. That is, you heard sounds come from hard-left and hard-right, but everything else was a diffused blur. Center sounds were spread all over the place. It was so bad I thought the drivers were out of phase or something, but I pulled them out and checked them and the crossover wiring very, very carefully.
This was the night before the show, and we were really starting to sweat now. The terrible image blurring, it turned out, was solely due to driver placement on the front face; they measured just fine, and all the drivers and crossovers were correctly phased. Not only was I shocked, but this was a real crisis for us; it was the debut of our new speaker line, and we had hours to go before the show opened in the morning. The latest we could hold off the crowds was maybe noon the next day.
So we called the idiots responsible and demanded they ship the correct matching speaker ... open the damn box and check, OK guys? ... and make a rush trip to the Portland Airport freight department for overnight shipment. Not cheap at several hundred dollars, but the exhibit room was several thousand, so this had to happen RIGHT NOW, or we risked a very expensive failed show. Not a good thing in a company with 20 employees.
Bright and early next morning, we rush out to the Chicago Airport freight department in a rented van, collect the missing matching speaker, and worriedly set it up in the demo room. To the enormous relief of all of us, the overnighted Right speaker was a perfect match for the already-good Left speaker. Imaging went from non-existent to spectacularly good. Phew! Disaster very narrowly averted. Audionics would live to fight another day, and the show turned out well for us, with good reviews and the dealers happy with our new product lines.
But that whole experience is seared into my memory. It would
never have occurred to me to intentionally audition a pair of Left speakers; everything we did at Audionics was always mirror-imaged, following UK custom of the time (we were the authorized US agents for Radford at the time, and I owned KEF mirror-imaged speakers at home).
The TLM-200, despite the insane complexity of the crossover (the KEF drivers were not very flat and required tuned notch filters), did have pretty flat response, and there wasn't much bad happening when you measured it 10, 20, and 30 degrees off-axis. In fact it was not easy to see the differences if the speaker was measured off-axis on the left or right side; diffraction artifacts are difficult to see on a 1/3 octave RTA, so you really couldn't see the problem with the measuring gear we had at the time.
My conclusion after this near-disaster was diffraction (in the horizontal plane) has a strong influence on image quality and overall spaciousness, but unfortunately is difficult to measure. So the best approach is to reduce it as much as possible, consistent with reasonable costs of production.
The null measurements mentioned in the previous post, along with the in-phase summation, really only work on speakers that are flat to begin with, and have very accurate phase-matching across the pair. All the speakers I did at Audionics, and afterward, have always been like this, but I've noticed poor phase-matching seems to be pretty common in the US high-end business (I still see it 40 years later), and I guess we were the exceptions. I just kind of took it for granted that qualified designers, like Laurie Fincham at KEF, and the people we met at BBC Research Labs, knew all about this already, since the Brits take image quality seriously.
