Let's assume everything possible has been done to reduce edge diffraction, within the constraints of a project. So we're left with some edge diffraction, and as a result driver frequency response measures different in the box than they do in an "infinite" test baffle. Reducing edge diffraction further is not an option, for whatever reason.
How does one appropriately deal with the edge diffraction when designing a crossover? Would you use the test baffle driver responses, and effectively "ignore" the diffraction? Include it and adjust the crossover to compensate the best you can? Average the test baffle and in-box responses and use that to partially compensate? Something else? What's the crossover design process if you have some edge diffraction worth considering?
I'm not talking about baffle diffraction step, but rather the ripple that results in the high frequencies.
Thank you for your input.
How does one appropriately deal with the edge diffraction when designing a crossover? Would you use the test baffle driver responses, and effectively "ignore" the diffraction? Include it and adjust the crossover to compensate the best you can? Average the test baffle and in-box responses and use that to partially compensate? Something else? What's the crossover design process if you have some edge diffraction worth considering?
I'm not talking about baffle diffraction step, but rather the ripple that results in the high frequencies.
Thank you for your input.
Additional Thoughts
It seems like adjusting for diffraction ripple in the crossover is akin to equalization. This video suggests that in simple cases, equalization can compensate for edge diffraction ripple:
Can a Reflection Be Equalized?
I have 3 concerns about this idea.
1. Generally, ripple is more complex than his example, with a half dozen or more peaks and dips.
2. I'm not sure the fact that the diffraction radiation is delayed WRT the direct radiation would be as neatly compensated for as in his example.
3. By equalizing, the direct radiation response is being altered; the total amount of sound being radiated into the room no longer has a flat response. It seems like this might be a competing psychoacoustic concern to the flatness of response at a particular listening point including diffracted/reflected sound.
Granted, all these are reasons to avoid the issue by minimizing diffraction and reflection in the first place, but I'm interested in what to do given some residual cabinet effects which can't be removed via structural methods and physical treatments.
It seems like adjusting for diffraction ripple in the crossover is akin to equalization. This video suggests that in simple cases, equalization can compensate for edge diffraction ripple:
Can a Reflection Be Equalized?
I have 3 concerns about this idea.
1. Generally, ripple is more complex than his example, with a half dozen or more peaks and dips.
2. I'm not sure the fact that the diffraction radiation is delayed WRT the direct radiation would be as neatly compensated for as in his example.
3. By equalizing, the direct radiation response is being altered; the total amount of sound being radiated into the room no longer has a flat response. It seems like this might be a competing psychoacoustic concern to the flatness of response at a particular listening point including diffracted/reflected sound.
Granted, all these are reasons to avoid the issue by minimizing diffraction and reflection in the first place, but I'm interested in what to do given some residual cabinet effects which can't be removed via structural methods and physical treatments.
Have a look here:
Diffraction from baffle edges
At the end, there is a conclusion that I think will give some answers to your questions.
regards,
George
Diffraction from baffle edges
At the end, there is a conclusion that I think will give some answers to your questions.
regards,
George
I also believe there is a discrepancy, when it is said to measure the speaker with diffraction to make a crossover, but do not measure the room. What is the difference and where to draw the line?
I think it is worth learning about the precedence effect. I still believe short delay diffraction can be a problem.
I think it is worth learning about the precedence effect. I still believe short delay diffraction can be a problem.
In all the years I've been making speakers I've never seen any diffraction effects large enough to warrant attention.
Room reflections are a different matter. Where these can't be sorted then yes, you need to compensate for them in the crossover. After all, isn't that one of the advantages of having your own design - a flat in room response.
Room reflections are a different matter. Where these can't be sorted then yes, you need to compensate for them in the crossover. After all, isn't that one of the advantages of having your own design - a flat in room response.
Diffraction correction in a crossover isn't possible outside of one axis. It's a 3 dimensional effect, which is why it's preferred to solve at-source with rounds/absorption.
The Linkwitz article doesn't really answer my question.
I currently have diffraction ripple that approaches 4 dB, but I have yet to apply felt. I'm not sure how much that will improve things. David Ralph's testing showed an improvement of almost 2 dB. With more ripple I may get more improvement; I don't know.
I just am not sure what to use for driver frequency response for crossover design. It sounds like maybe the test baffle measurements, which is what I suspected.
Room effects are a different subject entirely.
I currently have diffraction ripple that approaches 4 dB, but I have yet to apply felt. I'm not sure how much that will improve things. David Ralph's testing showed an improvement of almost 2 dB. With more ripple I may get more improvement; I don't know.
I just am not sure what to use for driver frequency response for crossover design. It sounds like maybe the test baffle measurements, which is what I suspected.
Room effects are a different subject entirely.
Measurement or measurements? Certainly one measurement per driver would be a little harsh by showing one side to the diffraction effect. If you were to take several and average them, things would smooth out. You could blend power and response, or you could average a listening window set of measurements, surrounding the listening angle.
With a midrange, at the crossover point, there is likely some "beaming" so the baffle edges may not illuminated by this driver. With a waveguide on the tweeter you can reduce its low frequency dispersion as well, making the baffle diffraction much less of an issue. The Visaton Boxsim free software works well for exploring baffle diffraction as it lets you enter a box shape and position drivers on the baffle. If you are not using Visaton drivers, you can overwrite the driver TS parameters and/or load driver measurements, acoustic and electrical impedance frequency sweeps. The driver diameter can be varied to see what happens. They have a version of the Visaton 25mm tweeter with and without waveguide in the library.
You make some interesting suggestions. Taking several measurements per driver in the cabinet could potentially be a way to take diffraction into account without overemphasizing it in the design.
The test baffle measurements are a single frequency response for each driver; but they don't have the diffraction, being in a 2' x 2' baffle. So that would be simply not taking diffraction into account when designing the crossover. It seems like people are imply that's what I should do, since I'm being told diffraction can't be compensated for in the crossover. That suggests that video I posted on equalizing out diffraction effects was an overly simplistic view on the topic.
There's plenty of diffraction and reflection affecting the midranges, actually. But this question isn't about how to minimize diffraction. It's about how to design a crossover for drivers in a box which has some residual diffraction which can't be eliminated. Specifically, what method of measurement and design is best when there is diffraction if you measure the drivers in the box. Generally what I'm hearing is that I should be using the "pure" driver frequency responses, measured in a way which doesn't include the diffraction from the cabinet - measured in the test baffle. Otherwise, I'd be optimizing the crossover to try to flatten out the ripple, which it sounds like the consensus is that it's a bad idea.
Is that +- 2db or a 4db spike or dip?
Also what frequency is this at? I've seen stuff this big around the baffle step hump and dip after it but not at higher frequencies...
Have you tested the same drivers on an infinite baffle, or are you comparing your response to manufacturers data?
Tony.
+/- 4 db. And that's compared to an "infinite baffle" which is actually 4' x 4' (I think I mistakenly said 2' x 2' earlier).
It isn't at one frequency, so I can't easily answer. It starts below 1 kHz. By around 8 kHz, the ripple is attenuated enough not to worry about. It shows in both midranges and tweeter.
It's a pretty ugly baffle diffraction and reflection wise. I designed it some 25 years ago, not realizing the problems it would cause. It's in a fairly elaborate cabinet, has some intricate routing one it, and the baffle is made of an exotic hardwood with an expensive professional finish, so I really can't easily just ditch it. I hope the David Ralph felt treatment will improve it enough, but realistically I expect some residual ripple so I'm deciding how to deal with it. This seemed like a good time to ask how one deals with what's left over once you've done what you can to eliminate it.
The felt could be a complete solution if suitably applied. Try a specialist felt shop, there are several online, they should have a product that is dense enough and thick enough. Some will even cut to your plans for a low cost.
Following the Ralph article on felt, I have some 2" wide 1/2" thick F13 felt strips on order from Grainger.
You can place the sheet as a sort of coating of the whole box. Then start from there with the addition of pieces: it will work with the pieces stacked on the thin side and protruding 2". Around the tweeter, 1". You can cut those stripes in many triangles or half circles.
OK thanks, for the additional info. Hopefully the felt will help to get things under control. It is quite surprising sometimes how something will cause negative effects diffraction wise. I 45 deg chamfered my MTM baffles on all four sides. I shouldn't have done the bottom. It actually causes a dip in the treble region that wasn't there before I took that final cut with the router. The boxes were already veneered so I just lived with it 🙂
I've been wanting to do some grills with integrated felt in strategic places for years but never get around to it.
Tony.
This is not like the inside of the cabinet. I propose that the felt creates diffraction in order to smooth it out, just as a roundover does.