Edgar

Except that a speaker system is inherently a highpass filter, and all causal, nontrivial filters introduce phase shift. The objective is to manage that phase shift. By, "keeping the speakers in phase with each other" I suspect that you refer to the individual drivers; woofer and midrange, midrange and tweeter, etc. Ignoring individual driver responses for the moment, it is quite possible, and not even all that difficult, to create lowpass/highpass filter pairs that are in-phase with each other. The Linkwitz-Riley pairs exhibit this property, for example. And in digital signal processing, it is almost trivial to design perfect reconstruction filters.

Holy cow, @captainbeefheart, I applaud any attempt to simplify the concept of phase response, but that might be too much simplification. Still, I don't know of any other way to describe it without things getting very deep very fast. Again, perhaps an oversimplification, but there is really no simple explanation that is also accurate.

No, I only change one variable at a time. So perhaps I should say that the lower orders sound best in the current speaker locations.

Continuing with that thought ... For example, I have a system in my living room with vented bass. (Currently playing Linda Ronstadt.) A vented enclosure is inherently a 4th-order system. I can equalize it to any 4th-order system that I desire; Butterworth, Bessel, Chebyshev, etc. I can also equalize it to any 5th- or 6th-order system that I desire (or even higher, if I want to be absurd). The lower orders sound best. No doubt about it.

Ah, but there are so many other factors to be considered! So I do not disagree with you.

Constant phase would be an anomaly. For example, one way to implement it would be to apply a Hilbert Transform to a signal, which will give constant 90° phase, and then scale and recombine that with the original signal to achieve any constant phase angle desired. But the resulting time smear would be horrible. And implementation in a loudspeaker is something that I would not even know how to approach. I am personally in favor of "well behaved" minimum phase systems, because they concentrate as much of the signal energy as possible into the very beginning of the impulse response. Linear phase systems exhibit pre-ring, which can be very annoying.

I have informally concluded that we live in a 2nd-order universe. That is to say, essentially every resonant system in our natural world can be described by a 2nd-order transfer function (or, if you prefer, a 2nd-order differential equation). I can't claim that it's always true, but it certainly seems to show up a lot in the real world. This is part of the reason that I object to extraordinarily steep, high-order crossover filters. I believe (without proof) that they sound increasingly unnatural as the order increases.

@Travis In Austin, when you say "constant phase" do you really mean "constant delay"?

Absolutely! Beyond a certain point it all just becomes "fun with numbers", and bears no resemblance to the natural world.

Well, I've designed filters for minimum phase and for linear phase, and for many phase characteristics in-between. (Loudspeaker model equations are identical to filter equations.) Zero-phase is a form of linear phase, but is noncausal. Constant-phase doesn't make any sense to me -- I cannot think of a single advantage that it might have. And it could be very difficult to implement.

Yeah, I know. I was just in a foul mood. Apologies.

Yes, I am familiar. Obviously none of the descriptions (temporal masking, tap dance, chorus, etc.) fit the situation exactly, hence the apparent contradictions. I agree that time and phase alignment are important in crossover networks. I just don't know how important they are in relation to all of the other things that are important. I have experimented a bit with it myself, and while "feet" seem to matter, "inches" don't ... at least not to my ears.

Dual motorcycle crashes? Sorry, perhaps in poor taste.

It's a confusing lot. On one hand we have Zwicker and Fastl telling us that temporal post-masking lasts 200 ms and pre-masking lasts 25 ms. On the other hand, Kolbrek and Dunker say, "... They thought this would really show off the capabilities of the Fletcher system, and monitored [Eleanor Powell's] tap dance through it. Due to the 8-ft difference in path lengths through the two horns, every tap produced two taps from the speakers. By moving the high frequency unit back 8 ft, and then moving it forward again until an echo could barely be perceived, an acceptable upper limit of about 1 ms time delay was established." [High-Quality Horn Loudspeaker Systems, p. 309]

I'm pretty sure that the YM3436 cannot handle 96 kHz or 192 kHz.

Nothing at all, as long as it is accurately reproduced by your loudspeaker.

Make certain that the DC-91 can handle 96 kHz or 192 kHz. Not all can. Nope. The ability to accurately send bits has never been an issue with any of the digital interfaces. The problem with Toslink is bandwidth. (At least, it used to be, many years ago. The technology has largely caught up with the problem, now that it has been identified.) The bandwidth problem translated into a clock jitter problem. So if you use ST optical in part of the signal path and Toslink in another part of the signal path, the fidelity is only as good as the weakest link. Correct. Of course, you could still connect the DC-91 analog output to a different preamp input, if you wanted to choose your flavors according to your mood. Hard to say. For a long time, analog volume control performed better than digital volume control. That has changed as DACs have improved.

Not all standalone DACs have multiple inputs. As for sonic benefits of the DC-91, if you like its sound quality then you won't be buying a new standalone DAC anyway. I didn't read the article closely, but it appeared to me that the only changes that the DC-91 can make to the digital signal are to change gain and polarity.

The DC-91 has 13 digital inputs. It can be used as a source selector, so whichever of those 13 inputs you choose is passed-through to the two digital outputs. That way, if you ever get tired of the DC-91 DAC section, you can buy a new standalone DAC and use the DC-91 to select the digital data stream to send to it. https://www.stereophile.com/content/accuphase-dp-90-cd-transport-dc-91-da-processor

Nah, they gave up on me years ago. Checked into my background and couldn't believe that anybody's life could be so boring.

Weird. Don't know how that double post occurred.

That is essentially as it was explained to me. The signal "crosses over" from one driver to the next as the frequency changes.

That is essentially as it was explained to me. The signal "crosses over" from one driver to the next as the frequency changes.

Research into temporal masking tends to refute this concept. Separated by less than a certain threshold, the disparate arrival times cannot be discerned at all. That threshold is in tens or hundreds of milliseconds. https://www.sciencedirect.com/topics/medicine-and-dentistry/temporal-masking

The audibility of phase distortion is the subject of much debate, argument, insult, and fisticuff. There are many good-sounding loudspeakers that are not phase-coherent. There are many phase-coherent loudspeakers that do not sound good. Inside a room, where reflections abound, the whole idea of phase coherence may be moot. [*] [*] @Chris A would quickly point out that broadband controlled-directivity loudspeakers will largely eliminate the problems caused by reflections.