• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Crossover design based on datasheet FR of drivers

mjvbl

Member
Joined
Feb 9, 2021
Messages
21
Likes
17
Chances are this has been discussed before on forums. (But searching for it isn't that easy.)

I wonder to what extent is it possible to design a crossover (or a whole speaker actually) just based on FR data on the datasheets of the drivers, and it would be interesting to spell out the issues with this idea.

Probably the most obvious issue with this idea is that datasheets often contain only on-axis FR data.
So the best that could be done for the off-axis of the speaker is to design (guess) the crossover with some assumptions in mind about the off-axis FR of the drivers, based on the drivers' radiating area, measurements of drivers that are considered similar, etc. Although I guess that gets more difficult for waveguided tweeters, and probably next impossible for horn/CD combinations?

But suppose part of this is just to have some fun with the crossover design program, or to get something preliminary going, like quick and dirty style. Or in the first case, maybe we don't even have the drivers, just want to have some fun designing a crossover for some interesting drivers on paper, but again to what extent does this make sense?

Another issue is that the FR on datasheet has been measured on a test baffle. Certainly this test baffle won't have the same dimensions as our speaker's front baffle. But if the dimensions of the test baffle are known we could create a simulated baffle response for such a baffle and subtract it from the digitized datasheet FR and sum that with our speaker's simulated baffle response. This could be done in VituixCAD. Also splicing it with a simulated box low-end response. I guess this could work for woofers, maybe?

For waveguided tweeters it seems more complicated as we don't have simple radiating area value to work with to create our virtual baffle response. Or is there a rule of thumb that would provide an approx. equivalent radiating area in some cases? I guess similarly in case of a (big, Tannoy-style) coaxial driver, the cone is a (big) waveguide for the tweeter, at higher frequencies the off-axis FR will be all inherent in the design of the driver, but below a certain frequency (depending on cone/waveguide size) what would be the equivalent radiation area if one wanted to create a simulated baffle response?
 

alex-z

Active Member
Joined
Feb 19, 2021
Messages
253
Likes
368
Location
Canada
Without off-axis data, it would be impossible to do properly. First rule of assumptions is they make an ass out of you.

Simulating baffle step loss + diffraction with VituixCAD isn't perfect, you certainly cannot translate IEC baffle data in the way you want. Short of using something like Comsol, you cannot fully simulate a speaker in the detail required to get high accuracy data.

Your best bet is always to purchase the drivers, build a test box, and then measure them yourself. In the long run this saves you money not having to do 2-3 crossover revisions to fix the baffle problems.
 

headshake

Active Member
Joined
Apr 20, 2019
Messages
291
Likes
202
Another issue is that the FR on datasheet has been measured on a test baffle. Certainly this test baffle won't have the same dimensions as our speaker's front baffle. But if the dimensions of the test baffle are known we could create a simulated baffle response for such a baffle and subtract it from the digitized datasheet FR and sum that with our speaker's simulated baffle response. This could be done in VituixCAD. Also splicing it with a simulated box low-end response. I guess this could work for woofers, maybe?

I think you are on the right track. Also, you can take the farthest real off-axis trace and run that through the baffle sim as an 90 degree with a little padding. Set the mic angle to 100-120 degrees and save the file as a 90-degree measurement. This is how I made the >60-degree measurements below. The vertical was also made with the baffle sim and a single on-axis measurement for each driver. All of the magic from the graph comes from the vcad sim of the xo and the driver spacing.

Is it a 100% match to reality? No. Is it close enough that I could compare XO's and use that XO for my real speaker? Yes.

I think fuzzy answers are better than spitball guesses.
 

Attachments

  • eton3__t25b (hor).png
    eton3__t25b (hor).png
    90 KB · Views: 28
  • eton3__t25b (ver).png
    eton3__t25b (ver).png
    228.6 KB · Views: 29
OP
M

mjvbl

Member
Joined
Feb 9, 2021
Messages
21
Likes
17
Your best bet is always to purchase the drivers, build a test box, and then measure them yourself. In the long run this saves you money not having to do 2-3 crossover revisions to fix the baffle problems.

For sure, I'm still curious though.

Simulating baffle step loss + diffraction with VituixCAD isn't perfect, you certainly cannot translate IEC baffle data in the way you want. Short of using something like Comsol, you cannot fully simulate a speaker in the detail required to get high accuracy data.

This point sounds to me like the baffle diffraction sim is something you'd hardly ever want to use. Or otherwise why is it worse for creating a simulated response for a big IEC baffle compared to a speaker sized baffle?


I just remembered there's a thread with measurements of individual drivers in a real finished box. I guess I could try and compare the results of my attempt based on datasheets and simulated box to those real measurements.

"Try your hand at a 2 way VituixCAD crossover with full Spin-O-Rama data (108 sweeps)!"
https://www.audiosciencereview.com/...-with-full-spin-o-rama-data-108-sweeps.21583/

Although it looks like there's also a waveguide for the tweeter and that seems tricky to me right now. I wonder if there's a simplistic model for conical waveguides to come up with an equivalent direct radiator size to use in a baffle sim for the frequency range where the speaker's baffle isn't providing half-space. Something along those lines. Would be useful for coaxials as well. Or there's mabat's ATH, maybe it could be used for something like this, and baffle sim as well.
 

Rick Sykora

Major Contributor
Forum Donor
Joined
Jan 14, 2020
Messages
1,652
Likes
2,698
Location
Stow, Ohio USA
Without off-axis data, it would be impossible to do properly. First rule of assumptions is they make an ass out of you.

Simulating baffle step loss + diffraction with VituixCAD isn't perfect, you certainly cannot translate IEC baffle data in the way you want. Short of using something like Comsol, you cannot fully simulate a speaker in the detail required to get high accuracy data.

Your best bet is always to purchase the drivers, build a test box, and then measure them yourself. In the long run this saves you money not having to do 2-3 crossover revisions to fix the baffle problems.

Ofc this assumes every driver measurement you have was done under the same test conditions, but even on-axis is tricky. Will share some experience using an extreme example...

If you pick a tweeter and use the manufacturer‘s on-axis response. For this example, let‘s assume an IEC baffle was used. If I take the same tweeter and mount in a minimal baffle enclosure, the tweeter will not have near as low a low frequency rolloff as with the IEC baffle. As knowing the rolloff is often how you determine the crossover design, you cannot accurately pick the correct crossover frequency. The high frequency rolloff for a woofer in a 2-way has a comparable issue.

So, unless you know you can rely on a baffle sim, as stated you have to buy some drivers and measure them in the cabinet.
 

headshake

Active Member
Joined
Apr 20, 2019
Messages
291
Likes
202
Seems I am in the newbie minority. Have any of you actually made a sim using traces and compared it to a real speaker? My speaker does not use a waveguide so maybe that is why it matches my sim.

I attached a sim of the hificompass baffle and the iec small baffle setup. They don't seem to muck up 1khz or above? This is why I think 3rd party and manufacturer data is fine for a tweeter xo above 1khz.
you cannot accurately pick the correct crossover frequency
Isn't the whole point of the baffle sim is to predict how things roll off? Here is what Vcad uses:
The Distributed Edge Dipole (DED) Model for Cabinet Diffraction Effects
https://www.aes.org/e-lib/browse.cfm?elib=13024 It is free to read. A bunch of people with PhD's and master's degrees wrote it. If you dig in, you can see it is more accurate for how things roll off than the power just off-axis. This is why I think feeding manufacturer data helps.
 

Attachments

  • iec_vs_hificompass.png
    iec_vs_hificompass.png
    50.3 KB · Views: 23
Last edited:

Rick Sykora

Major Contributor
Forum Donor
Joined
Jan 14, 2020
Messages
1,652
Likes
2,698
Location
Stow, Ohio USA
Seems I am in the newbie minority. Have any of you actually made a sim using traces and compared it to a real speaker? My speaker does not use a waveguide so maybe that is why it matches my sim.

I attached a sim of the hificompass baffle and the iec small baffle setup. They don't seem to muck up 1khz or above? This is why I think 3rd party and manufacturer data is fine for a tweeter xo above 1khz.

Isn't the whole point of the baffle sim is to predict how things roll off? Here is what Vcad uses:
The Distributed Edge Dipole (DED) Model for Cabinet Diffraction Effects
https://www.aes.org/e-lib/browse.cfm?elib=13024 It is free to read. A bunch of people with PhD's and master's degrees wrote it. If you dig in, you can see it is more accurate for how things roll off than the power just off-axis. This is why I think feeding manufacturer data helps.

Have not used, but agree in the approach. Your measurement has quite a bit of response rippling. In my case was caused by the mic clip. Might want to try wrapping the mount in felt or some damping material to improve. :)
 

headshake

Active Member
Joined
Apr 20, 2019
Messages
291
Likes
202
Have not used, but agree in the approach. Your measurement has quite a bit of response rippling. In my case was caused by the mic clip. Might want to try wrapping the mount in felt or some damping material to improve. :)
Thanks for the tip. I saw your post about the clip and that did have me thinking about confirming my mic. I for sure need to test out diff mic positions/covers or just stick it in a pole like Troels/hardjs. I'm still trying to get at the root of that wobble... my newbie pet theory: some of the wobble is from the edge radius. Testing out the mic will answer that for sure.

When I tested an early foam baffle w/ and w/o a 44mm edge radius, the measurements with the edge had all kinds of little humps that the w/o did not. The sim said it would be flat! ;P If you check out the first image, I measured 0 and 90 degrees without (blue) and with (red) an edge with the eq meant for the one with the edge. The measurements were taken at diff. SPL's to help compare them. More measurements will hopefully answer all the questions.

http://www.troelsgravesen.dk/measurements.htm
 

Attachments

  • 0_90_degree_edge_VS_no_edge.png
    0_90_degree_edge_VS_no_edge.png
    60.5 KB · Views: 19
Top Bottom