DSP newbie here. I've been spending hours trying to dial in a woofer-to-compression-driver crossover for a large-format soundsystem project, and I've concluded that something just isn't clicking for me. I can't figure out how to get the phases of the two drivers to align for anywhere close to a full octave around my target XO frequency. It seems like there must be some basic tools to deploy or tricks for making this work, but everything I try moves me in the opposite direction. Many google searches hasn't yielded much.
Drivers are a Dayton PA460-8 18" woofer ("W"), and a PRV D2200Ph compression driver ("M") on an 18x10 horn. Target XO point is around 700hz-800hz, I was able to come up with a decent result by crossing the CD at 700hz x 12db, and the woofer at 1khz x 12db, and then delaying the woofer by 0.55ms, but my best case outcome still had a cancellation of around 8-12db at 500hz and a similar one at 1.2khz. I've tried steeper slopes, increasing the target XO point slightly, and even fiddling with all-pass filters but haven't made much progress. It seems to me like the phase curves are just too different at the XO point for any of the tools I am using to make enough of a difference - i.e. flatten the CD curve below 1.5khz or steepen the woofer curve above 500hz. Are these things even possible? Am I screwing up something basic? Any 101-level concepts or tricks to deploy?
Raw phase measurement:
12db Filters at 700hz for both drivers:
Drivers are a Dayton PA460-8 18" woofer ("W"), and a PRV D2200Ph compression driver ("M") on an 18x10 horn. Target XO point is around 700hz-800hz, I was able to come up with a decent result by crossing the CD at 700hz x 12db, and the woofer at 1khz x 12db, and then delaying the woofer by 0.55ms, but my best case outcome still had a cancellation of around 8-12db at 500hz and a similar one at 1.2khz. I've tried steeper slopes, increasing the target XO point slightly, and even fiddling with all-pass filters but haven't made much progress. It seems to me like the phase curves are just too different at the XO point for any of the tools I am using to make enough of a difference - i.e. flatten the CD curve below 1.5khz or steepen the woofer curve above 500hz. Are these things even possible? Am I screwing up something basic? Any 101-level concepts or tricks to deploy?
Raw phase measurement:
12db Filters at 700hz for both drivers:
Try this for the target filter:
Stack two high pass filters with a Q or .707 on the tweeter at your crossover freq, and two low pass filters on the woofer at same freq.That should give you a 4th order x-over.
4th order LR should be in phase throughout the x-over, should make things easier.
Then adjust delay like AllenB said.
Stack two high pass filters with a Q or .707 on the tweeter at your crossover freq, and two low pass filters on the woofer at same freq.That should give you a 4th order x-over.
4th order LR should be in phase throughout the x-over, should make things easier.
Then adjust delay like AllenB said.
The raw curves show around 90° phase difference between the two drivers in the neighborhood of 800 Hz. You might try a 3rd-order crossover, which produces 270° phase difference between the two passbands at the crossover frequency, which will result in 0° or 180° overall phase difference. If 180°, it can be cured by reversing the leads on one of the drivers. Worth a try.
Last edited:
Thanks for the suggestions - I have tried all of these but will try again tonight and add some images. I don't think I can do a 3rd order LR filter - my AHM unit only works in 12db increments for pass filters.
That said, maybe I'm missing something... Looking at the base curves, the phase response curve SHAPES relative to their locations along the frequency axis are quite different around the XO region (ignoring the vertical displacement that is easily fixed with delay and/or polarity inversion). Adding pass filters/increasing the order makes this more pronounced, as shown in the 12db filter graph. The woofer curve starts to change from 'descending to ascending' farther to the left, and the mid-horn bottoms out or goes from 'descending to ascending' farther to the right. This ultimately is just shrinking any overlap of the curves., making the curve slopes at the crossover point steeper in opposite directions, rather than aligning them. Is this how it's supposed to work? Or is it possible my measurements are inaccurate?
That said, maybe I'm missing something... Looking at the base curves, the phase response curve SHAPES relative to their locations along the frequency axis are quite different around the XO region (ignoring the vertical displacement that is easily fixed with delay and/or polarity inversion). Adding pass filters/increasing the order makes this more pronounced, as shown in the 12db filter graph. The woofer curve starts to change from 'descending to ascending' farther to the left, and the mid-horn bottoms out or goes from 'descending to ascending' farther to the right. This ultimately is just shrinking any overlap of the curves., making the curve slopes at the crossover point steeper in opposite directions, rather than aligning them. Is this how it's supposed to work? Or is it possible my measurements are inaccurate?
The way they are descending vs descending shouldn't be summarised that way. It should be seen as more or less steep. The delay mixed in there is only making it appear confusing when it is simpler than that.
As far as the gap between them, they may have differing amounts of delay.
This is why you ought to get the responses right first without looking at phase...
As far as the gap between them, they may have differing amounts of delay.
This is why you ought to get the responses right first without looking at phase...
You're right, I don't know why I put "LR". Point remains that I can only do 12db increments. (I do have an 18hz Butterworth option via a separate XO function.)
Please explain this to me because maybe I'm just not expressing this correctly. When I look at the two curves overlaid below, in the region around my target XO of 700 hz, I see that their phases are moving in opposite directions as frequency increases. Don't these curves need to be "relatively" similar, meaning roughly the same steepness and moving in the same direction as frequency increases? This has 24db pass filters applied at 700hz, no delays on either signal. Based on this, I would expect a deep null at about 800 hz highlighted by the cursor, and then normal summing at around 550hz and 1.1khz.
Looking at the SPL response, I added 0.25ms of delay to the woofer to adjust for alignment, which results in the predicted SPL response from the phase graph above.
Now, given the fact that the phase lines are diverging on either side of the XO point, reversing the polarity of the Mid horn should result in summing at the 800hz alignment point, but then cancellations on either side, right? Voila...
Any other ideas here? Or am I still missing something?
6th order LR low pass on woofer @Fx;
2nd order Butterworth high pass on horn @ 1.3*Fx.
Both drivers connected with same polarity.
Adjust delay for phase match @Fx.
That should give you an almost perfect overlap of the two phase curves over an octave around Fx.
2nd order Butterworth high pass on horn @ 1.3*Fx.
Both drivers connected with same polarity.
Adjust delay for phase match @Fx.
That should give you an almost perfect overlap of the two phase curves over an octave around Fx.
Thanks for the tips so far everyone, I'm still working through this. I did find a super helpful article on another thread on Reddit that I wanted to share here in case anyone stumbles across this in the future. I believe this has been shared around here before.
https://www.merlijnvanveen.nl/en/st...gnment-the-foolproof-relative-absolute-method
https://www.merlijnvanveen.nl/en/st...gnment-the-foolproof-relative-absolute-method
As long as you know that article is not meant for use in a room (at least not one small enough to be called a room). The goals are not the same.
I would start by looking at the GATED tweeter raw measurements. See how much filtering is needed for a 4th order acoustic roll-off.
Then ditto on the woofer, and include any BSC desired.