Pretty concise description only doesn't phase alignment simply mean that two or more waveforms are in phase, the frequency doesn't change? I think this is what you meant but it didn't quite come out like that 
. Having re-read your post with a different emphasis I see this is what you are saying more clearly. I was tempted to delete this post, but, what the hell, someone else may have made the same mistake as me
. Having re-read your post with a different emphasis I see this is what you are saying more clearly. I was tempted to delete this post, but, what the hell, someone else may have made the same mistake as me
Last edited:
Yes, that too. We have several possibilities to make quite close to timing properties of acoustical 1st order. For example with Bessel style low pass combined to lower order high pass. Simulator with optimizer makes adapting by f, Q and delay in few seconds without extreme math skills, design tables or white papers. Directivity and power response problems could increase, but that's one thing to control while designing crossover. Timing is not everything.
I appreciate everyone's input.
Hi wesayso
It was never my intention to work on the transient response. My take on that is without DSP, there's very little one can do about it. Transducers are not like transistors where the reaction is instanteous. Perhaps one can work on the crossover to improve the transient but that's likely at the cost of something else.
I was focusing solely on the concept of time alignment. Before the age of DSP, time align means the sound from the woofer and the tweeter reach the mic at the same time. This happens at the acoustic crossover point where it is not only phase aligned but no time delay between the woofer and the tweeter. This crossover can be any freq. Unless I'm mistaken, that's what time-alignment means.
Conventional methods of achieving that is by vertically aligning the acoustic centers of the drivers, using a step to displace the woofer or tilting the front baffle. I did not use any of these methods but I believe I achieved the same objective.
Build a WAW (formerly a FAST). Fairly easy to choose drivers with extension well beyond the XO, arrangement so that no centre-to-centre is greater than a quarter wave-length, and use a 1st order XO. XO is fairly low (150-400 Hz typically), with a more equable distribution of octave-to-octave duties.
dave
I think the conventional method is by including a phase shift at the crossover frequency
FR of Speaker
The above FR plot shows a lot of room influence. This would make it pretty hard to show the phase result without using some form of gating.
I'd love to see this same plot with a frequency dependent window of about 6 cycles on it, which also shows the phase. Not to look at the FR, but to see trends in the phase.
Another (more traditional) way of viewing what's happening could be to set gating at 3 ms, or just before the first clear reflection (would be close to that 3 ms, judging some earlier zoomed out plots).
Impulse of Speaker
The impulse shows a negative peak before moving to the positive side. It could indicate the high frequency part is inverted, could you tell us the polarity of the drivers? If it isn't inverted we would need a harder look to see what that's causing that negative spike. All I can tell you, it isn't REW that's causing it. First arrival of energy is inverted. Sometimes something else in the signal chain can cause weird looking IR's, even if it is only affecting frequencies up high. Knowing the true polarity of the drivers would help.
We also see some pretty early secondary spikes after the positive peak scaled to 100% by REW.
Step of Speaker
The STEP supports what we see in the IR. I've never looked at this combination of filters before to know what an ideal IR and STEP should look like. I'd first like to know the polarity of the drivers used, maybe see the IR from each driver separately (with filters on) would help too.
Especially the tweeter with filters on.
We will need to get this graph in here too, showing the acoustical slopes of the separate drivers.
The tweeter shows a dip, at 10K in every plot we've seen so far. That's why I'd like to see it separately, with phase and an IR.
Last edited:
I think more speakers have been created without a stepped or slanted baffle that with one of these features. That does not mean their developer did not watch the hand over between the drivers. Looking for phase alignment between the drivers at it's crossover point. This is where most "named" crossovers originate from.
As kimmosto said, if we take the response of each driver (with proper timing) into a capable simulation package we can choose the best crossover point and slope and make sure there is phase overlap between the drivers at the crossover. Sometimes this takes inverting one of the drivers, it all depends on the choices one makes.
The frequency response is the first factor, responsible for what we hear by far. However the power response does play a big part to, as we won't hear direct sound only, at our listening spot in the most common places they are used. (like a living room)
Each choice we make in the crossover will determine several factors. For instance, I've already posted a picture showing the difference of a speaker, measured at tweeter height at half a meter compared to a hypothetical listening distance of 2.5 meter.
If we would pick our design axis between the tweeter and woofer, concentrate on getting it right on that axis, it would still hold up better at larger distances. Why? because we wouldn't change the geometry between those two positions as much as we do by keeping that design axis on the tweeter level. So the closer the tweeter and woofer are, the less difference we would get between ear to tweeter and ear to woofer, makes sense?
So even little details like this has an influence on the end result we get at our listening spot.
Last edited:
Is this what you meant. It's at 3ms with Minimum Phase.
We will need to get this graph in here too, showing the acoustical slopes of the separate drivers.
The tweeter shows a dip, at 10K in every plot we've seen so far. That's why I'd like to see it separately, with phase and an IR.
My tweeter is presently surface mounted. The dip at 10k on the tweeter is due to edge diffraction between the tweeter faceplate and the baffle. It will disappear when I flush mount the tweeter.
I'm afraid I may not be able to respond as quickly as I would like. My internet line is affected by the coming storm. To make matters worse, my pc is still mis-behaving. Nonetheless, I shall try my best. I appreciate your help.
That is the topology that I would call time aligned as well. It is the one that was also called Time Alingn by Long and Wickersham. The difference between them any you is that they used mechanical delay (horn tweters and coaxials for instance) to achieve the delay - just keep in mind that digital delay wasn't easily and cheaply achieved back then.
Kinoshita is using something like this as well although he doesn't use the term "time aligned" anywhere.
Such a topology doesn't give a transient perfect response but one that is very close to transient perfect. One can often see a "slowly" rising ramp just before the steeply rising transient from the tweeter arrives. The Harsch crossover does also belong to this category for instance.
I will also try "time.elign" with my large MTM (apart from "subtractive" etc). That is the benfit of using DSP: One can easily change topologies.
What Michael Chua presents would best be called "time aligned at crossver frequency".
Regards
Charles
Is this what you meant. It's at 3ms with Minimum Phase.
Close, but that minimum phase is a derivative of the frequency response. Could you 'thick' the phase too and repost? I'm still looking at what causes the negative pre-spike. I expect a less flat phase plot as measured.
What we do see is the summing isn't entirely flat, however I do not have a problem with that. A little less energy around 3 kHz might not sound worse, it's almost summing like the famous, though ill named, BBC dip. I can't blame the crossover topology for all of that, as we clearly see the extra energy in both tweeter and woofer graphs taken separately.
The rest of the graph actually looks pretty good.
Take your time, I'm not going anywhere soon
. I suspected the same thing, the tweeter edge diffraction might be responsible for one of those minor peaks after the main spike. It should be worthwhile to get rid of it.What would help you to learn what this crossover topology does is looking at polar measurements. No need to do them immediately, but it would be interesting. First I'd like to find the cause of that (first) negative spike in the IR. Which is why I'd still like to see the tweeter IR as measured with it's filters.
The on design axis FR looks pretty good, the waterfall you showed earlier also had a clean result, if the topology holds up at off axis angles you might have a pleasant sounding speaker here. That first negative spike isn't cause for worries, sound wise. Most of the midrange trough the crossover seems to be in phase which isn't bad at all. Especially if it does not create big dips off axis (which would influence the perceived power response).
I'd still like to see a measurement at listening distance. Raising the enclosure so we get to listen slightly under tweeter level at larger distance might preserve the time alignment.
If you remember this graph:
This shows that at larger distances the geometry between tweeter and woofer distance changes, if we raise the enclosure at larger distances, or slightly tilt it backwards, it would keep this difference to a minimum. We would create time alignment with a slanted baffle though (just kidding).
Attachments
Last edited:
This should read: lower the enclosure or tilt it slightly forward. Keeping the difference between woofer and tweeter to the reference points the same.
Or fitting a very short horn to the front of the treble driver. This physically moves the acoustic centre back towards the same as the mid/bass.
They are actually already thickened. Maybe my screen resolution is too high.
I've screen captured them with a lower resolution monitor. The plots are thicker now.
The crossover frequency is 2.66kHz.
Minimum Phase at 2.66kHz. It cuts the 0 Phase x-axis exactly.
These are the Impulse Responses of the woofer with the LP filter (yellow) and the tweeter with the HP filter (green). Both peaks coincide at the same point, does that mean the sound from the tweeter and the woofer reach the mic at the same time?
Andrew, I'm sorry I can't respond now. Internet bad. I can lose power any time. Big winter storm up in NorthEast. I will reply soonest possible.
Is this what you meant. It's at 3ms with Minimum Phase.
Not what I meant, a bit too fast in typing I guess. Look in this graph. Phase is not selected. Only the generated minimum phase. I'd like to see that phase curve.
I meant tick, not thick
. As in tick a box to show phase.Same for the tweeter Fr curve. With the phase curve shown, not minimum phase but measured phase.
Sorry for the confusion. Hope you'll be allright trough the storm.
These are the Impulse Responses of the woofer with the LP filter (yellow) and the tweeter with the HP filter (green). Both peaks coincide at the same point, does that mean the sound from the tweeter and the woofer reach the mic at the same time?
Do you have a way of applying delay to the tweeter?
If so, try inverting the tweeter, and delay it around 0.4ms.
See if that looks better and if so, fine tune delay till peaks align..
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.