I can see this has taken it's own turn, and that's fine but I wanted to be clear: I'm not trying to create a speaker with a perfectly triangular impulse response. Just excellent driver to driver phase and amplitude matching on axis and best as possible off.
Just keep doing what you've always been doing. It's easy enough to teach yourself to do it using DSP tools. They may be a little more rigid than you're used to but as long as you set yourself up to be able to see the changes you're making, you'll get there. (By the way, don't expect DSP to do anything off-axis that you haven't already been doing.)
If you get the excess group delay plot minimized through the two crossover interference bands, you will also achieve the best impulse or time-based response of the loudspeaker--on-axis or off. It will also sound clearer with much more neutral timbre, which is not really an "on-axis" or "off-axis"-only proposition, but something that will always be audible in-room.
I'm not sure how you can say "all you get is better off-axis improvement". To say that means completely discounting the excess phase and excess group delay effects (which are large and quite visible--as well as audible at some small threshold), and also somehow ignoring the SPL and phase plot disruptions through each crossover band where the SPL cancellations occur and phase jumps are also seen.
Also, the notion of stacking the lower frequency drivers side-by-side in a center loudspeaker--whose listening room requirements require the best off-axis response possible--I think you might see that turning that horizontal set of drivers vertically is a much better arrangement. But the example you showed has drivers arranged in both the vertical and horizontal directions--the worst of both worlds.
Chris
I'm not sure how you can say "all you get is better off-axis improvement". To say that means completely discounting the excess phase and excess group delay effects (which are large and quite visible--as well as audible at some small threshold), and also somehow ignoring the SPL and phase plot disruptions through each crossover band where the SPL cancellations occur and phase jumps are also seen.
Also, the notion of stacking the lower frequency drivers side-by-side in a center loudspeaker--whose listening room requirements require the best off-axis response possible--I think you might see that turning that horizontal set of drivers vertically is a much better arrangement. But the example you showed has drivers arranged in both the vertical and horizontal directions--the worst of both worlds.
Chris
Hi Chris
Whether the design shown in the OP belongs to the "worst of both worlds" is strongly depending on the woofer to mid crossover frequency.
Regards
Charles
Whether the design shown in the OP belongs to the "worst of both worlds" is strongly depending on the woofer to mid crossover frequency.
Regards
Charles
A known compromise, and a required one in my case due to location and aesthetics. I'm not building for a theater or for behind a perforated screen here. There are size constraints I must adhere to. Still, this compromise is much better with a 3-way design than with my current 2-way, MTM center. With an expected crossover location around 200 Hz, and a 5.6' wavelength I should be good for my narrow room. My current center uses a 2kHz crossover point, so you know this has to be better.
No it isnt. Period. If you add a reactive component in series with a driver, you get a continuous increasing phase shift as frequency changes i.e. a phase "error". But when you add delay with a DSP, you get a constant delay irrespectively of frequency i.e. no change in phase - it all just comes later with whatever phase situation that was at hand before the introduced delay. Big difference.
//
Apologies for my part in diving deeper than fits the thread.
Good time alignment between driver sections has proved to be a cornerstone of what i feel is great success with multi-way tuning.
Timbre, vocal clarity, dynamics, solid imaging...... all seem to improve together, when phase traces overlay (which is the sign of good timing)
A good timing/tuning test for me, is i have a LCR setup with all three speakers being the exact same.
Running L&R both in mono, and comparing how they project a central image versus just the center speaker in mono is a good test.
They are often indistinguishable.
There's very little need for a center speaker to go with L&R in stereo, as the L&R image to center so solidly and clearly, on their own.
When I try tunings that are less time coherent thru the entire xover regions, like some lower order xoxer slopes that inherently have a wider frequency range overlap and increased lobing/combing, a center speaker then helps tighten up a sense of a solid center with great vocal clarity etc.
It's very difficult to maintain full phase trace overlay thru wide xover ranges. Any phase trace slippage is just another source of timing error, like having fixed delays off some. Tight timing really seems to matter, ime.
@TNT , you've forked the question there, so I'll address both.
The point I was answering with YSDR is that the "jimmying" Erik refers to is compensating for the driver being non-flat to begin with. This is the same as what YSDR does only in a different order.
The point I was answering with YSDR is that the "jimmying" Erik refers to is compensating for the driver being non-flat to begin with. This is the same as what YSDR does only in a different order.
Not if you change phase independently of response.
Erik, can you help us all out a little? Can you fill in the values for the application that you're considering, below?
nom. diameter of woofer diaphragm =
nom. diameter of midrange diaphragm =
vertical distance between tweeter and midrange driver centerlines =
HF crossover frequency (nominal) =
HF crossover network type/order =
horizontal distance between midrange and woofer driver centerlines =
LF crossover frequency (nominal) =
LF crossover network type/order =
Chris
nom. diameter of woofer diaphragm =
nom. diameter of midrange diaphragm =
vertical distance between tweeter and midrange driver centerlines =
HF crossover frequency (nominal) =
HF crossover network type/order =
horizontal distance between midrange and woofer driver centerlines =
LF crossover frequency (nominal) =
LF crossover network type/order =
Chris
While I have everyone's attention I want to focus on one more issue.
Let's take a hypothetical 2 way speaker for simplicity. If we physically set the tweeter back from the front plane to co-locate the acoustical planes of the tweeter and mid-woofer, that co-location is the same at all lateral angles. That is, they are still exaclty the same distance as the listener moves left to right.
This is not achieved with a digital delay. At the listener moves off axis the tweeter starts to "appear" further away. This must have some effect on loving and amplitude response, right? Is there a generalized, simple way to think about these two scenarios and how they are different??
Let's take a hypothetical 2 way speaker for simplicity. If we physically set the tweeter back from the front plane to co-locate the acoustical planes of the tweeter and mid-woofer, that co-location is the same at all lateral angles. That is, they are still exaclty the same distance as the listener moves left to right.
This is not achieved with a digital delay. At the listener moves off axis the tweeter starts to "appear" further away. This must have some effect on loving and amplitude response, right? Is there a generalized, simple way to think about these two scenarios and how they are different??
I Agree. Passive and active crossover design is not that much different. I think you are "good to go"...
In theory, this is true. It is simply the math and physics of the situation. However, in the real world I have not seen much effect. In My two systems (both are active DSP systems with IRR filters, meaning the DSP filters mimic an analog-type character) the tweeter and mid are mounted on a flat baffle. In one system I need about 30 mm of delay on the tweeter, and the other I need about 40 mm of delay. The measured responses from 0 to 180 degree match up quite well to the simulations (which assume geometric alignment).
You can test this with a VituixCad simulation. You can add driver time delay and/or adjust the "z dimension" to geometrically move the driver closer or further from the listening position. Try it out. My experience is that it does not make a enough difference to worry about.
j.
Yeah, example: crossover is 2kHz which is 17cm long wavelength. When the system is exactly time aligned the response has most SPL at crossover point as both drivers sum nicely together (thinkin on-axis). The signals sum positively all the way up to 1/4 wavelength path length difference, or 90 degree phase difference. In this example it is about +/- ~4cm for 2kHz crossover, or offset range of 8cm where the frequency response would look about the same especially if the slopes are symmetric and phases track each other nicely. Perhaps there is smaller window if the slopes are asymmetric and phases don't track.
So the rough accuracy in this example is about 4cm here or there from optimal but there might be some audible difference. This kind of things can surprise sometimes, where small adjustment makes the system snap together, for what ever reason. However, I'm not sure if I have heard it either (having DSP adjusting delay is easy), but that doesn't mean it wouldn't on some other system. Perhaps my system is so much off that time aligment isn't audible?😀
Deep waveguide could make the offset so large that some delay is very much needed, this is easily seen with measurements loaded in simulator with real measured phase intact.
So the rough accuracy in this example is about 4cm here or there from optimal but there might be some audible difference. This kind of things can surprise sometimes, where small adjustment makes the system snap together, for what ever reason. However, I'm not sure if I have heard it either (having DSP adjusting delay is easy), but that doesn't mean it wouldn't on some other system. Perhaps my system is so much off that time aligment isn't audible?😀
Deep waveguide could make the offset so large that some delay is very much needed, this is easily seen with measurements loaded in simulator with real measured phase intact.
Last edited:
Great question. And illustrates there is simply no perfect solution other than true geometric coincidence of acoustic centers. (in all planes)
I run into the issue of geometrically stacked acoustic centers vs digitally delayed acoustic centers whenever doing spinorama work.
It can be hard to get acoustic centers to tack directly over the turntables center of rotation. in either case.
When time-of-flights don't vary under rotation, I've found i'm able to produce the most consistent on and off axis response sets.
I do all that because i think the Harman research about directivity smoothness makes sense, and having timing right reduces off axis variations.
How much all that ultimately matters to audibility...who knows. But i've seen that timing has about as large an effect on polar variations as does diffraction type issues, like horn mouth terminations.
perhaps the synergies I prefer to build have an inherent ability to benefit more from good time alignments, given the relative geometric tightness of acoustic centers compared to normal baffled designs. Dunno, but that kind of makes sense to me. Anyway, i so believe in good flat phase traces...
Thanks everyone. I think that, ballparking, I will have a tweeter 20mm ahead of the midrange, and to heck with the mid to woofer distance.. 😀
So it sounds like using digital delay from tweeter to midrange will maintain coherency off axis pretty well.
So it sounds like using digital delay from tweeter to midrange will maintain coherency off axis pretty well.