Design x-overs for 2way - direct bass/mid + CD horn

This should be a simple one - when designing a passive x-over for a two-way speaker consisting of a bass/mid driver (direct radiating) and a CD-driven horn, are there any particular issues that need to be taken into consideration in the design process, and if so, what method(s) should be used to address them?

I'm in the middle of designing and testing such a build (see Updating my Blastoramas... - Techtalk Speaker Building, Audio, Video Discussion Forum), hence my question.
 
Can you borrow a good quality active crossover to experiment with and set a "baseline"?
Do you have high quality test CDs? Mixed tones and scales across the crossover frequency?
Do you have reasonable measurement equipment?
------------

Time and phase alignment are very important, but very challenging for speaker+horn without electronic crossovers.

Physical alignment pushes the horn mouth 5" -10" in front of the midbass cone.
Passive crossovers which modify M - T frequencies and phase leave auible artifacts. A reduced "optimal soundstage" area often results.

PASSIVE CROSSOVER: Since you are using the midbass near their linear high frequency limits, you will need to explore steep slope crossovers like LR4/LR6 and 6th order Bessel. Midbass -to- Horn_Tweeter time alignment is very important, and it is worth some research to study "quasi-optimal" crossovers, like the LeCleach work, which use 2-different slopes to improve phase and time alignment with long horns. "Quasi-optimal" crossover for high-efficiency loudspeaker system

ACCORDING TO MARCO: Here's a list of "approved" “quasi-optimal” crossovers in order of increasing offset:

3rd order Butterworth Low Pass, -3dB @ Fx*0.87 (+)
3rd order Butterworth High Pass, -3dB @ Fx*1.15 (-)
Offset = 0.22*c/Fx

4th order L-R Low Pass, -6dB @ Fx (+)
3rd order Bessel High Pass, -3dB @ Fx*1.4 (-)
Offset = 0.29*c/Fx

4th ordrer Butt Low Pass, -3dB @ Fx* 0,93 (+)
4th order L-R High Pass, -6dB @ Fx (-)
Offset = 0.31*c/Fx

6th order Bessel Low Pass, -6dB @ Fx*1.25 (+)
2nd order Butterworth High Pass, -3dB @ Fx*1.3 (-)
Offset = 0.40*c/Fx

6th order L-R Low Pass, -6dB @ Fx * 1.06 (+)
3rd order Butterworth High Pass, -3dB @ Fx* 1,13 (-)
Offset = 0.445 c/Fx

6th order L-R Low Pass, -6dB @ Fx * 1.07 (+)
4th order L-R High Pass, -6dB @ Fx * 0.92 (-)
Offset = 0.465 c/Fx

------------------------
 
You'll want to take polar measurements to calculate power, decide the listening angle and match the driver polars as well as smoothing the direct/reflected output through the crossover.

Thanks. I did take some rough measurements of the drivers off-axis, and it *looks* like I might be Ok with a 3kHz acoustic x-over like what I eventually ended up with, at least for my second try (the first one was a bit of a disaster). Fingers crossed....
 
Whoa, lots of info there. Let me see....

Can you borrow a good quality active crossover to experiment with and set a "baseline"?
Do you have high quality test CDs? Mixed tones and scales across the crossover frequency?
Do you have reasonable measurement equipment?

I have an iNuke 3000DSP (which is actually going to be the amp I'm going to use to drive these things. I can set it in two-way mode and experiment with the active x-over settings. It does offer a LOT of settings options, including filters up to 48dB/octave and time delay.

Measuring equipment consists of the Dayton UMM6 (calibrated) and a PC running REW. I also use HolmImpulse for distortion measurements and occasionally TrueRTA. I also use WinPCD for x-over design.

------------


Physical alignment pushes the horn mouth 5" -10" in front of the midbass cone.

Using one of the features built into WinPCD, I've determined the actual offset to be about 3.5 cm, tweeter behind the woofer. The "horn" is an APT150, so this seems about Ok.


ACCORDING TO MARCO: Here's a list of "approved" “quasi-optimal” crossovers in order of increasing offset:

3rd order Butterworth Low Pass, -3dB @ Fx*0.87 (+)
3rd order Butterworth High Pass, -3dB @ Fx*1.15 (-)
Offset = 0.22*c/Fx

4th order L-R Low Pass, -6dB @ Fx (+)
3rd order Bessel High Pass, -3dB @ Fx*1.4 (-)
Offset = 0.29*c/Fx

4th ordrer Butt Low Pass, -3dB @ Fx* 0,93 (+)
4th order L-R High Pass, -6dB @ Fx (-)
Offset = 0.31*c/Fx

6th order Bessel Low Pass, -6dB @ Fx*1.25 (+)
2nd order Butterworth High Pass, -3dB @ Fx*1.3 (-)
Offset = 0.40*c/Fx

6th order L-R Low Pass, -6dB @ Fx * 1.06 (+)
3rd order Butterworth High Pass, -3dB @ Fx* 1,13 (-)
Offset = 0.445 c/Fx

6th order L-R Low Pass, -6dB @ Fx * 1.07 (+)
4th order L-R High Pass, -6dB @ Fx * 0.92 (-)
Offset = 0.465 c/Fx

Good stuff. So, going by the calculated offset and that table, it looks like I'm on track with the 3kHz acoustic HP for the tweeter if I go with option(3) (not much choice - this tweeter can't go much lower), which means that I should consider adjusting the filter on the woofer to achieve an acoustic BW 4th order @ 2.8kHz, if I'm calculating things correctly. I aimed for a 4th order LR at 3kHz, but I think I can adjust a few values to achieve that target.
 
Just following up on this. I tried all different types of x-overs and eventually settled for an acoustic LR @ 3kHz. This is about 500Hz below the recommended x-over point for the APT150 horn, but it seems to work ok, at least for V1 of the x-over. There's a bit of low midrange "bloom" around 500Hz~2kHz and frequency response measurements suggest that the high frequencies can be pulled down another 2dB or so, and there's a strange small suckout around 5kHz (not x-over related), but given the purposes that these speakers are going to be put to, I think that should be Ok. I'm tempted to take another stab at the x-over - I have to keep reminding myself what these are actually going to be used for :).

I've attached the FR of one of the speakers, measured on-axis and I think around 40~45 degrees off axis (the CD's mesh grill is shaded a little bit from this position). I'll bet the deeper notch @ 3kHz has a lot more to do with the Beta 8A driver's response than the APT150 horn. Oh, I'm not using the Eminence drivers I bought with the horns. I'm using Pyle PDS222 neo drivers instead. Yes, Pyle, LOL. They measured flatter for most of the passband (well, from about 4kHz up). Of course, I had to buy four and choose the two with the closest response, but still, at $17 a pop, they get all the way up to 20 kHz without any major issues (unlike the APT150 drivers). Just don't ask them to do 2kHz like they claimed to be able to do....

I may have gone with the JBL 2414Hs instead, but I can't find any performance data for them. About 3 times the price of the Pyles, but probably more consistent performance.
 

Attachments

  • 20160506-blastoramas-fr-axis.png
    20160506-blastoramas-fr-axis.png
    55.4 KB · Views: 169
Brian, your results look good for this low cost product. I assume you used LR2/LR2 with some frequency shift to keep the passive Xover cost low and still reduce some time alignment gap.

Does your passive crossover perform(measure) as well as the DSP experimental setup? This question should help illuminate component limitations vs. limitations in the crossover circuit.

A significantly higher budget is necessary to get a balanced next step-up in quality for each component... CD, horn, Xover. Upgrading just one section might be disappointing.
 
Brian, your results look good for this low cost product. I assume you used LR2/LR2 with some frequency shift to keep the passive Xover cost low and still reduce some time alignment gap.

Acoustic LR4 actually. I kept trying a few different x-over points, and every time it seemed that 3kHz (really just under 3kHz) was the way to go. Going higher increased the part count (and would cause MORE problems in the polar response because the Beta 8A starts beaming @ 3kHz, if not earlier), and going lower was a no-no - the design notes for the APT150 specific 3.5kHz as the lowest frequency they should be used.


Does your passive crossover perform(measure) as well as the DSP experimental setup?

The DSP was better because I could dial in any slope I wanted to up to 48dB/oct, AND I had several parametric EQ bands to play with. However, the passive x-over, at "Version 1" (I'm working on "Version 2") is good enough for the use these speakers are going to get, and it's pretty cheap to implement. I was able to get an acoustic 4th order LR @ 3kHz by using a 1uF capacitor and a resistor bridge (10 ohms series, 30 ohms parallel) on the horn, and ditto on the woofer using a cap, coil and resistor, for a total of six parts. I used WinPCD to "tune" in the best passband I could get without losing too much efficiency (again, considering what these speakers are going to be used for, LOL). With "Version 2" of the crossover, I'm going to tweak the high-pass to bring the impedance down into the 8 ohm region (right now it flattens out at 20 Ohms, and some class D amps may not be too happy with that). I'm also going to try a bit more BSC on the woofer.

The scoop out at around 5 kHz is puzzling. Seems to be more of a horn or horn + baffle effect rather than an issue with the driver. I close-miked the driver in the horn, and the response was ruler-flat all the way up to about 15kHz, at which point there's a +3~4dB shelf. At the mouth it's fairly flat, and then after that the aberrations start happening. (see image #1)

Oh, and image #1 compares the response between the APT150's original CDs and the Pyle Pro PDS222s. From that, I think you can see why I chose to go with the Pyles :)
 

Attachments

  • 20160505-apt150-all.png
    20160505-apt150-all.png
    86.4 KB · Views: 146
  • 20160411-psd222 vs apt150-2.png
    20160411-psd222 vs apt150-2.png
    19.4 KB · Views: 143
Last edited:

Boden

Member
2010-03-02 9:29 am
Hello all,

The big issue, which i.m.o. is the polar matching at x/o frequencu, is not yet fully adressed here.
I have also been wondering - contemplating a similar project- what off axis curve should be leading in optimizing the x/o filter. I am fully aware of and experienced in the classic "Dickason style" single point measuremen based on axis x/o optimization. I have extensively used Arta plus Calsod 3.0 and LspCad 5.1 for all that.

Should on aim for e.g. be 60 degrees off axis? And what to do if optimizing for 60 degrees leads to a sub optimal 30 degrees curve?

Maybe Earl Geddes or the Finnish gentleman Kimmosto (is that his name?), or Wayne Parham can shed some light on this issue, since they are the among the few that optimize for off axis also.

I do recall having read somewhere that Earl optimizes for 22 degrees, but then again, how then to obtain smooth curves for other off axis positions? Is there such an optimum point?

A lot of questions, I know.

Thanx in advance for any info on a proper design procedure for "x/o optimization 2.0".

Regards,

Eelco
 
All good questions.

FWIW, I took some *very* rough response measurements off-axis for the Beta 8As and the APT150s, as I thought is would help in the choice for the best x-over point if their directivity matched at the x-over frequency (Earl has a long write-up on his website about directivity and what would be best to aim for in that regard- something I ran into *after* I thought this would be a good idea, LOL. See http://gedlee.com/downloads/directivity.pdf).

As it turns out, the Beta 8A starts beaming just before 3kHz. As the APT150 is not supposed to be used below 3.5kHz, that basically 'set' for me in which region the x-over point needed to be.
 

Attachments

  • 20160201-blast2-beta8a-axis.png
    20160201-blast2-beta8a-axis.png
    18.3 KB · Views: 84
  • 20160201-blast2-APT150-axis.png
    20160201-blast2-APT150-axis.png
    18 KB · Views: 81
For domestic use you can probably cross the horn a little below there -- 45Watts RMS at that sensitivity would peel wallpaper, that's pro audio stuff. You won't get within 10dB of that much power in a room without earplugs.

True. However the PDS222s were already starting to roll off below 3kHz. The APT150 drivers could get a little lower, but I'd dropped them from consideration because of the issues at 6kHz and around 11kHz.

Oh, and this build is a small portable PA for external use.