This keeps bugging me after seeing a video (forgot where) about how to optimize active crossovers. In it the presenter deftly uses tools to create absolutely ideal LR4 crossover slopes. My question is, does it really matter if the speaker works correctly?
Here's an example. I overlay an ideal LR4 high pass filter for 3 kHz on top of the tweeter response with a filter. In the context of the total speaker, I have the output response and phase matching I want. Is there much value in fixing the area between 2 kHz and 4 kHz so that it more closely tracks the ideal LR4? My intuition is to leave it alone, less filters is better and that there's no practical benefit in attempting to peen this curve to "ideal." Am I wrong?
Here's an example. I overlay an ideal LR4 high pass filter for 3 kHz on top of the tweeter response with a filter. In the context of the total speaker, I have the output response and phase matching I want. Is there much value in fixing the area between 2 kHz and 4 kHz so that it more closely tracks the ideal LR4? My intuition is to leave it alone, less filters is better and that there's no practical benefit in attempting to peen this curve to "ideal." Am I wrong?
I don't think it matters much. I like when I can match a target though. If I get a decent null when swapping polarity, that's good enough for me.
If this were a matter of simple mods to the HP filter values, then the expense would not suffer, and you could optimize a bit.
I could add some EQ, but it feels like I'd be optimizing for the sake of a textbook result, without audible benefit. Am I wrong??
My final results look right:
and the nulls seem ideal. Is there a practical benefit I could achieve by "cleaning up" the tweeter curve?
Response looks fine without EQ other than the HF peaks either side of 4kHz, a critical band for most peoples hearing.
The practical benefit of a closer match of the mid/tweeter slopes would be a more uniform vertical dispersion.
That said, the direction of the main lobe might be preferred in direction other than perpendicular to the speaker baffle, so what you have might be exactly what you want.
As the various in-depth Vituixcad off axis analyses here on Diya demonstrate: there is a bit more to system optimization than just the on axis summing and textbook acoustic slopes.
See here e.g. :https://www.diyaudio.com/community/threads/a-3-way-design-study.376620/
See here e.g. :https://www.diyaudio.com/community/threads/a-3-way-design-study.376620/
Hi,
I think it's a good practice to always try to come up with a listening test if questions like this come up. Performance in paper is nice and it seems you are curious about it so it is a good opportunity to take a dive bit deeper trying to listen it.
To some people it might be meaningful difference, to some it might not, we all need to figure out ourselves if it's for us. Be aware of expectation bias of course, because often any single thing like this isn't gonna make much of a difference in the context of whole, million things that go into the playback situation including the bias.
So if you cannot notice a meaningful audible difference then the feature you are listening on is not meaningful within your context you are having there at the moment, but doesn't necessarily mean it's not meaningful to others, or to you, in all scenarios. For this reason allow yourself to explore to get some perspective, and change opinion later if some new scenario changes it.
Thinking about it, what could be a difference? Boosting tweeter 3dB below pass band basically means power to it is doubled, so your amplifier runs out headroom bit sooner. It would increase excursion on the tweeter running out headroom sooner as well. Both of these would distort the whole pass band of the tweeter, but you might never run into this within your scenario, and it's never audible in practice. On the other hand, it's on a bandwidth where two sound sources have output at the same time, and on the driver that has lower output of the two, so the sound from the louder dominates anyway if you boosted it or not. From this I'll speculate that audible differences highlight if you run the system loud.
Another, if phases match currently, correcting the tweeter would mean you'd have to correct the woofer as well to maintain the phase match. But what is phase match, exactly on top of each other or within +/-45deg or what ever? What happens to phase match if you move your head 10cm up or down from designed listening axis, at the listening distance, still phase match? How much your head actually moves on your typical listening scenario? Can you hear difference with that movement? What if you deliberately move as much as necessary to make a perceived difference? How big the difference is audibly, how much the phase is off? How easy it is to perceive? how it actually sounds like? does it matter? Now move closer of farther of the speakers to affect how much direct / reflected sound ratio is and repeat the test, still the same result? Can you actually hear phase matching or not at your practical listening spot or do you have to be very close to hear it? Simple listening tests like this, that you can do just by moving yourself, help to give perspective on this stuff how big of a difference it actually makes to you. Can be done with passive speakers as well, no need to modify the system, just change your listening position a bit, actively move to make an AB test.
You could try this as well if you have a DSP: deliberately affect phase, like by adding delay to your tweeter. Test how much of a delay you can add before sound changes? How the sound changes? What about at close or far listening distance, still the same? Lot's of things that could make this stuff matter, or not 🙂 Only way to find out is determine yourself whether it matters to you in your situation. Have fun!🙂
I think it's a good practice to always try to come up with a listening test if questions like this come up. Performance in paper is nice and it seems you are curious about it so it is a good opportunity to take a dive bit deeper trying to listen it.
To some people it might be meaningful difference, to some it might not, we all need to figure out ourselves if it's for us. Be aware of expectation bias of course, because often any single thing like this isn't gonna make much of a difference in the context of whole, million things that go into the playback situation including the bias.
So if you cannot notice a meaningful audible difference then the feature you are listening on is not meaningful within your context you are having there at the moment, but doesn't necessarily mean it's not meaningful to others, or to you, in all scenarios. For this reason allow yourself to explore to get some perspective, and change opinion later if some new scenario changes it.
Thinking about it, what could be a difference? Boosting tweeter 3dB below pass band basically means power to it is doubled, so your amplifier runs out headroom bit sooner. It would increase excursion on the tweeter running out headroom sooner as well. Both of these would distort the whole pass band of the tweeter, but you might never run into this within your scenario, and it's never audible in practice. On the other hand, it's on a bandwidth where two sound sources have output at the same time, and on the driver that has lower output of the two, so the sound from the louder dominates anyway if you boosted it or not. From this I'll speculate that audible differences highlight if you run the system loud.
Another, if phases match currently, correcting the tweeter would mean you'd have to correct the woofer as well to maintain the phase match. But what is phase match, exactly on top of each other or within +/-45deg or what ever? What happens to phase match if you move your head 10cm up or down from designed listening axis, at the listening distance, still phase match? How much your head actually moves on your typical listening scenario? Can you hear difference with that movement? What if you deliberately move as much as necessary to make a perceived difference? How big the difference is audibly, how much the phase is off? How easy it is to perceive? how it actually sounds like? does it matter? Now move closer of farther of the speakers to affect how much direct / reflected sound ratio is and repeat the test, still the same result? Can you actually hear phase matching or not at your practical listening spot or do you have to be very close to hear it? Simple listening tests like this, that you can do just by moving yourself, help to give perspective on this stuff how big of a difference it actually makes to you. Can be done with passive speakers as well, no need to modify the system, just change your listening position a bit, actively move to make an AB test.
You could try this as well if you have a DSP: deliberately affect phase, like by adding delay to your tweeter. Test how much of a delay you can add before sound changes? How the sound changes? What about at close or far listening distance, still the same? Lot's of things that could make this stuff matter, or not 🙂 Only way to find out is determine yourself whether it matters to you in your situation. Have fun!🙂
Last edited:
Probable. But it depends on your requirements on the result.
The "error" is about 2,5 dB at the crossing (LR4 meet at -6 dB) so this will create a deviation from flat in the resulting FR once an ideal matching woofer (LR4 LP) is matet to this tweeter performance. The result depends also on the phase situation - a deviation in FR results in a phase deviation from ideal.
//
Not important correct.
Something more important would be looking at off axis dips and behavior.
On axis summing is the easy part.
Something that is not wavy gravy in a normalized polar chart to 180 is not
and learning how to fix that, not easy either.
Horizontal off axis being more easy, vertical another story.
Good luck. Even if center to center is close as possible ( most dont )
with Vertical and the typical skinny baffle madness and poor driver locations become extremely obvious off axis
Since the high end is almost 3 to 5 dB of difference with a scoop.
Room response and off axis could possible be a issue.
It is up to your listening preference. If the system lacks highs or bright enough
or has weird dips between sitting and standing
Guess it is nice to have a visual graph, and then hear real world.
Then it is easy to tell when you hear yabba yabba yabba about peoples interpretation of graphs.
Something more important would be looking at off axis dips and behavior.
On axis summing is the easy part.
Something that is not wavy gravy in a normalized polar chart to 180 is not
and learning how to fix that, not easy either.
Horizontal off axis being more easy, vertical another story.
Good luck. Even if center to center is close as possible ( most dont )
with Vertical and the typical skinny baffle madness and poor driver locations become extremely obvious off axis
Since the high end is almost 3 to 5 dB of difference with a scoop.
Room response and off axis could possible be a issue.
It is up to your listening preference. If the system lacks highs or bright enough
or has weird dips between sitting and standing
Guess it is nice to have a visual graph, and then hear real world.
Then it is easy to tell when you hear yabba yabba yabba about peoples interpretation of graphs.
Last edited:
These are 1m, not close mic measurements, and pretty much how I like it. 😀
I cheat and avoid quasi anechoic for 2 way speakers. i also try to kind of follow the B&K curves for in room.
The overall curve is following the existing 2 way with passive x over.
Right but thats not ideal either. at least not ideal to LR4. So the question is if I get very good final FRD and very good phase matching, is there any reason to go further? If this was a purely electical circuit, having ideal slopes is important to having an ideal result, but in speaker land ?
Only you can answer that question, but assuming the above conditions, i.e. you've achieved the desired FR and phase matching with 'near as' slopes, then my short version would be 'no'. Unless it causes you psychological irritation -not said or meant with irony, since these things can be valid for some. But I tend to come from the school that says 'good enough by definition is good enough: perfect is a pain in the posterior & usually isn't worth the effort'.
I.m.o. the answer is yes: there is. Include off axis transfer functions of yer 3 drivers into the design and optimization process as per the Vcad Manual.
Rule of thumb is to optimize response at least one octave beyond xo point. Important for LR2 but not so much for LR4.
I think that I hear differences in matching despite axial summed response shows hardly any. Range 1-4kHz is very sensitive and disfference might be heard as sort of air and clarity of eg. piano's right hand short notes/trills. The error in post ¤1 should be fixed IMO.
With dsp-xo correction is easy, passive might loose overall spl... Remember to finetune delays too!
Another challenge is that one should not look at only axial response with one gating and measurement distance... And then some listening.
I think that I hear differences in matching despite axial summed response shows hardly any. Range 1-4kHz is very sensitive and disfference might be heard as sort of air and clarity of eg. piano's right hand short notes/trills. The error in post ¤1 should be fixed IMO.
With dsp-xo correction is easy, passive might loose overall spl... Remember to finetune delays too!
Another challenge is that one should not look at only axial response with one gating and measurement distance... And then some listening.
The room curves like the B&K are result of an anechoic linear response in combination with boundary reflections at the listening position say 3m away from speakers.
Are you listening at much closer distances?
It might be better than actually matching the curve since you have less energy at the crossover point and there could be bunching at the crossover point in the off-axis response. So having less energy on axis there will help to alleviate the excess energy off-axis and this gives a more consistent in-room response, which is more of what we hear in most situations. I think you know this, but it's much less important to have perfect LR4 curves than it does to have flat on-axis response and consistent downward sloping power in-room. So, forget about matching the "ideal" LR4 curve and get your overall measurements to act correctly, which it seems is already your approach.