Hey guys, wanting to experiment a bit with my active 3 way speaker driver integration. I'm currently running EQAPO and utilizing stacked filters to achieve what I believe is 12db, 2nd order high and low pass filters. This seems to give me good integration with in room response (can't measure outside due to equipment limitations).
My filter settings are quite basic, just picked some easy xover points, 300hz for woofer to mid, 2.5khz for mid to tweeter. I've attempted to utilize higher order filters but always run into dips at the xover points when utilizing the same xover points as the 2nd order filters. Am I correct in assuming that one would need to set the xover points for the drivers so the frequencies overlap? What I mean is that instead of 300hz LP and 300hz HP, I would have to something like 350hz LP and 250hz HP?
My current crossover does appear to give me good integration, I've attached an image. I'm just curious to test the benefits of higher order filters.
My filter settings are quite basic, just picked some easy xover points, 300hz for woofer to mid, 2.5khz for mid to tweeter. I've attempted to utilize higher order filters but always run into dips at the xover points when utilizing the same xover points as the 2nd order filters. Am I correct in assuming that one would need to set the xover points for the drivers so the frequencies overlap? What I mean is that instead of 300hz LP and 300hz HP, I would have to something like 350hz LP and 250hz HP?
My current crossover does appear to give me good integration, I've attached an image. I'm just curious to test the benefits of higher order filters.
Ideally you want both high-pass and low-pass filters at the same frequency. The stated frequency is the -6dB point where the two halves of the filter will sum to full level.
Filters affect the phase as well as amplitude, so for example with a symmetrical 300Hz 2nd order (12dB/Oct) high-pass and low-pass, one of the drivers will need the phase inverting beause the filters make them 180degrees out of phase. If you use 4th order (24dB/Oct) both drivers will be in phase.
As well as the filter phase, also the position of the drivers will have an effect on the timing / phase of the sound. One way to check for alignment is to invert the polarity of one driver so you get a dip and then tweak the delay on the closer driver (usually the tweeter / mid) until you get a maximum dip. Then you flip the driver polarity back to normal and it should sum optimally.
Filters affect the phase as well as amplitude, so for example with a symmetrical 300Hz 2nd order (12dB/Oct) high-pass and low-pass, one of the drivers will need the phase inverting beause the filters make them 180degrees out of phase. If you use 4th order (24dB/Oct) both drivers will be in phase.
As well as the filter phase, also the position of the drivers will have an effect on the timing / phase of the sound. One way to check for alignment is to invert the polarity of one driver so you get a dip and then tweak the delay on the closer driver (usually the tweeter / mid) until you get a maximum dip. Then you flip the driver polarity back to normal and it should sum optimally.
Doing a quick a comparison of 12db vs. 24db filters, the 12db seems to sound much better to my ears. It does measure better still, I tried applying some delays to the mid and/or tweeter but couldn't get rid of the dips at the xover points.
This is what I'm getting with some filters. 2.5khz xover point.
This is what I'm getting with some filters. 2.5khz xover point.
Last edited:
Higher order slopes sometimes have a strange tendency to make it easier to create mistakes like this when you do a without measurement style crossover. I'm basing that on the way you're not simulating with the measurements or using them to sculpt the filters.
How far out of phase are they at the cross.. since this will give you some clue as to the problem? and we can go from there..
How far out of phase are they at the cross.. since this will give you some clue as to the problem? and we can go from there..
You said you got a null when inverting the driver polarity - that's good! Do that, then adjust delay until you can maximise the depth of the null. Then flip the polarity back to 'correct' and the dip should have filled in a bit. The delay needed can be very small, like 0.1ms.
Sorry I edited my post, I said I didn't get a null with 12db, just a different response.
I used rta and noise to dial in xover. 3rd order actually sounds better to me than the 2nd or 4th for whatever reason.
I used rta and noise to dial in xover. 3rd order actually sounds better to me than the 2nd or 4th for whatever reason.
If you didn't get a null when polarity was inverted it means for sure you need to do some delay adjustments.
There is another reason things might not sum exactly flat but it get's a bit complex if you want to go this far:
'Text-book' xovers like LR12 work correctly only when the original sound from the driver has a flat response. Your drivers may already show a roll-off or some dips and bumps in the crossover range for whatever reason. You can overcome this by first EQ'ing the drivers to be flat over the xover region for at least one octave. E.g. with 2.5KHz xover the tweeter needs to be flat down to about 1KHz before the xover filter is applied. Then when the xover filter is applied the response will follow the filter as intended. Don't pop your drivers while EQ'ing though!
There is another reason things might not sum exactly flat but it get's a bit complex if you want to go this far:
'Text-book' xovers like LR12 work correctly only when the original sound from the driver has a flat response. Your drivers may already show a roll-off or some dips and bumps in the crossover range for whatever reason. You can overcome this by first EQ'ing the drivers to be flat over the xover region for at least one octave. E.g. with 2.5KHz xover the tweeter needs to be flat down to about 1KHz before the xover filter is applied. Then when the xover filter is applied the response will follow the filter as intended. Don't pop your drivers while EQ'ing though!
You'll find what you're looking for if you measure each driver separately with phase. The null method will be limited in this situation.
Did some measurements and indeed saw the phase wrap with my filters. Stumbled on the program I'm sure you're familiar with, rephase. I generated some linear phase filters there and boy does it sound better and my phase measurement doesn't have any wraps. Of course it seems I've induced a bit latency to my signal with these filters.
Ok, a good result.
For future consideration, it's possible to get standard filters to come together before you concern yourself with FIR manipulation. When I say standard filters I mean the combined response. The filters themselves will have to be special to fit in with the drivers but the end result will look standard and it will work.
For future consideration, it's possible to get standard filters to come together before you concern yourself with FIR manipulation. When I say standard filters I mean the combined response. The filters themselves will have to be special to fit in with the drivers but the end result will look standard and it will work.
I was working on this today so thought I'd share. Here you can see the combind respose of a LR24 xover at 1.7KHz with the tweeter polarity inverted. The top yellow trace is with no delay, then going down you see added delay to the tweeter in steps of about 0.03ms (1 sample steps at 48KHz sample-rate).
Combined response with drivers "in phase" before (purple) and after (red) the delay. The offset between the drivers is rather small here and the effect can be significantly more noticeable in other designs.
Combined response with drivers "in phase" before (purple) and after (red) the delay. The offset between the drivers is rather small here and the effect can be significantly more noticeable in other designs.