This one?
I think you were the one that clued me in to Danley using BMS at Sound Physics Labs.
Yep. That build turned out particularly well. It was the first passive Synergy crossover I designed. Those crossovers are a tough nut to crack.
The BMS compression drivers sound great and are pretty much bullet proof. I'm constantly amazed how much abuse they can handle.
Would be interesting to an impulse response of a synergy/unity horn. One would think those holes would mess up quite a bit.
Phase and delay are very much linked.
A phase shift of 90 degrees at 1kHz is equal to 0.25ms of delay.
Phase is expressed as an angle, which you can turn into a fraction by dividing by 360. Phase difference is measured as fractions of the wavelength.
Delay is absolute.
90 degrees at 100Hz is 2.5ms, while 2.5ms at 1kHz would be 900 degrees.
Chris
PS - the idea of running a 1" driver down to 400Hz was a bit of an oddball, but never say never... It's just made my list of things to try.
And only at 1kHz, at every other frequency the delay will be different.
Which not enough people understand. This angle means that every frequency has a different delay. With phase you can be aligned at one frequency, and not aligned at every other frequency. (This may not be as tragic as it sounds, as the further apart the time gets, the less interaction there is between the two bands, but still worth noting.)
That is why I thought that the original statement that the filter introduced delay was not accurate enough. Using the word delay implies that absoluteness.
A passive filter introduces delay, but a different delay at every frequency. More accurately expressed as a phase angle. (The phase angle itself is not constant either, but changes with frequency.)
Last edited:
That is pure semantics.
"It's a delay!"
"No, it's a phase angle!"
Sounds like the glass is half empty, not half full.
I looked into this some years ago, because of exactly the situation that some horns are deep and at higher crossover frequencies the depth distance between the acoustic centers of the CD and woofer (Z axis) could end being one wavelength of the crossover frequency. Of course this means 360 degrees of phase rotation but only at the crossover frequency . Above and below that the effective phase rotation is lesser or more, causing an uneven summation through the entire crossover region, excepting at exactly the crossover frequency. The exact same thing would happen when adding electronic delay via DSP or similar. Of course I could only thought experiment this as I don't have a math degree, but Charlie Laub did the sims for me with his software: https://www.diyaudio.com/forums/multi-way/256869-inter-driver-delay-question.html#post3941092
You can imagine that polar responses will also differ for the different conditions. So yes, delay and phase are definitely not simply interchangeable in not only a theoretical sense but in ways that actually impact a loudspeaker design.
I have done dozens of crossover designs between a woofer and a horn and it is certainly possible to get it right. It's not easy, which is the point that I made before, but it is certainly doable. I personally have never considered phase and delay as different things. They do differ in a mathematical sense since delay is the rate-of-change of phase, but you cannot manipulate one without changing the other (in a passive network.) They are intimately connected and best thought about that way.
Yes, phase and delay are surely intertwined, but I do like to think of phase and delay as different entities.
To me 'delay' means a fixed time that applies equally to all frequencies, such as time of flight.
And to me, 'phase' means frequencies timing in relation to each other, timings that vary by frequency.
(Small aside: I think the term 'group delay' is a confusing misnomer.)
You are right, it is semantics. The whole discussion is semantic (fact) based. (But it is not metaphorical, as you write.) You are one of the most respected commenters on here, but sometimes you get lazy to make your point easier. Then when you leave, and are quoted, the misinformation you have sewn comes to fruition.
Your answer implies that the (whole band of) the woofer is time aligned to the tweeter, when in fact you are phase aligned at the crossover point.
For the woofer to be time aligned with the tweeter the phase angle across the system passband, including the crossover, would need to be nearly flat - no group delay anywhere. Correct? That is precisely the case in my system. Hence the woofer and the tweeter are time aligned as well as "phase aligned at the crossover point." So unless I don't understand your point, I do not believe that I mislead anyone since you state that your understanding is precisely what I meant.
This is clearly evident in the impulse response that I posted (somewhere around here). The woofer and tweeter impulses are at precisely the same point (within a few micro-seconds.) WIthout the crossover they are way off.
Last edited:
No, this is not correct. For time alignment as is conventionally termed, the phase angles across the crossover summation region just need to overlay each other, no matter their slope.
Nearly flat phase, with little to no group delay, is not a requirement for conventional time alignment woofer to tweeter. It is only a requirement if linear phase is also specified.
If the alignment is minimum phase, which probably comprises 99+% of all alignments unless FIR is utilized, the phase angle traces can take on nearly any slope, ......providing they overlay.
Most folks prefer phase as flat as possible, but that's not a requirement for time alignment. I mean heck, look at how many folks say group delay doesn't matter...
Just my understanding....
For the phase to be flat through the crossover region doesn't that mean that "the phase angles across the crossover summation ... need to overlay each other, no matter their slope." Or am I missing something.
If the phase is flat everywhere, doesn't that mean that it is flat through the crossover?
If the phase is flat everywhere, doesn't that mean that it is flat through the crossover?
Earl it looks like in your particular application the conditions are such that you make it work just fine. But the generalized case is not the same as your specialized case, I know you know that. In the generalized case there is a distinct difference between phase rotation and the effects of delay. Any novice designer should be aware of this as they apply the principles of the generalized case to their specialized case. Did you look at the plots that I linked to?
That needn't be the case, IF the crossover is designed well.
Here's the measured frequency and phase response of my system, comprising a 15" woofer & a CD + horn, with a passive crossover (there's also a super-tweeter for the last 1.5 octaves, but this isn't shown here for clarity).
The Compression Driver's diaphragm is physically located approximately 20cm behind the Woofer's dustcap.
And yet, as you can see, the two phase traces match, not only at the crossover frequency, but over a wide frequency range around it - in fact, everywhere that matters, really (i.e. they only start diverging when either drivers' output is already attenuated below -12dB).
Attachments
You're crossing at 600hz, which has a wavelength of 56.66cm, while your AC difference is 20cm. Illustrating precisely my distinction between the general and special cases. Move that horn back another 36.66cm and see what happens.
Sure, there are limits to how much AC difference you can compensate for using a passive crossover. And you can always dream up a combination of X-over freq. and AC offset that cannot be made to work.
But, realistically, why would I (or anyone) need to position the CD 56+ cm behind the woofer?
I would think that if you're using a comparatively large - but still realistic - mid horn (as I am), then you'll probably be aiming for a X-over freq. of 600-800Hz, and ~20cm of offset normally work well from a practical standpoint.
If, instead, you're crossing over at, say 1.2kHz, then you'll be using a smaller horn to begin with, typically resulting in a proportionally shorter offset (say, ~10cm, give or take), no?
- Status
- Not open for further replies.