30 Hz. -16 dB
33 -10
36 - 6
40 - 9
50 - 8
60 - 1.5
70 - 4
80 - 1.5
90 0 dbmo
100 0
200 - 2
300 - 1
400 0
500 - 4
600 - 6
700 - 8
800 - 8
900 - 7
1000 -10
1600 - 1
2k - 2
3k + 1
4k + 1.5
5k + 2
6k + 3
7k 0
8k - 3
9k + 1
10k - 2
11k - 2
12k - 4
13k - 5
14k -10
15k - 9
16k - 6
17k -10
18k - 4
19k +10
20k - 4

The 1.0 mH is giving too much attenuation to the DUAL 2226J's, but should be close to what the 2035's will need. The depressed levels between 500 Hz. and 1k were not present without the xover, in prior testing. The 4648A-8's were flat to 1k before...need to swap in 0.8 mH soon. But here's an "honest" in room F10 of 33 Hz. anyway.
From 1.6 kHz. to 11k, pretty happy here. And when measured before, these 35480 alum. diaphrams did not have the top end extension of 34647's I also use, so the drop after there is expected. And up high, the rat meter readings are supposed to be "corrected" for accuracy. This is raw, unmanipulated, in room, whatever.
After 10k, Pepper got pissed off and started barking...hard on the little needle!!! That's what closets are for :)
Driver fore/aft alignment initially was with the HF acoustic center 4.25 inches BEHIND the LF...the drivers relative positions were then slightly adjusted to eliminate any phase difference in the drivers outputs at 1.6 kHz...Sam

When you have two sound sources spaced 1/2λ apart, a notch in response appears. It also presents itself at multiples of this frequency. What is happening is dual slit diffraction, and interference patterns form in the listening area. Diffraction effects are present when there is more than one point source separated by distances of this scale. Some places in the room have deep notches in output at certain frequencies.

There are several ways to deal with this issue:

1. Reduce the number of point sources
2. Reduce distance between point sources
3. Reduce maximum frequency emitted by point sources
4. Reduce overlapping frequency bands of point sources
5. Reduce overlapping coverage of point sources using directional control

OK. But that is somewhat academic. So what to do about it?

If you want to use dual woofers as mains rather than subs, I would suggest running them as a 2.5-way speaker, basically crossing over one woofer much deeper so it doesn't share the midrange band with the second woofer. Both woofers work together in the deepest bass frequencies but only one covers the midrange band. The natural orientation is to run them vertically, with a large coil on the lower woofer to attenuate output above 100Hz. The upper woofer serves as a midwoofer, and is used through the midrange. Crossover the midwoofer to the tweeter using the four π crossover.

The prior testing w/ the 4648a-8 by itself only showed a mutual cancellation between the woofs at 817 Hz. or so. I didn't like knowing that was there in the first place, but like you said, it's the nature of the dual woofer format. The 1000 Hz. dip when both LF and HF are working is probably mother nature telling me to go back to an earlier config. that is less problematical, i.e. crossing lower.

With the 0.8 mH in now, subjectively they sound better. I figured with the 3rd order HF slope, diffraction related problems from co-operating at 1K would not be serious. Even with the anomalies, very listenable, and may still have a overall better presentation than my previous benchmark, the back to back BW's at 600 Hz. Sam

yep, those 2226J's could very well find themselves in 4 Pi Pro boxes, probably less trouble doing that than playing with the 2035's. just a 1.4 mH coil? and the 1.6 kHz. HP xovers. done. nah, that would be too easy...

You're a big help. I looked at my measurement setup, and EUREKA, the path diff. between the lower and upper woofs to the mike was...13 or 14 inches. Right in a place where dual slit diffraction would cause a big null. When I tested the 4648a-8 by itself, mike was at the same height as the upper woof, for a path length diff. around 9 inches, that and the woofs being 16.7 inches center to center both are implicated in the 817 dip...but I never looked even higher...measured on axis with the upper woof the system was measuring +/-1dB from 100 to 1000 except for the 817 Hz. cancellation.

I've now listened with fixed 12 dB pads on both systems, one set x'd at 600, the other at 1600 does sound a bit thinner, so the level dip from 500 to 1 k is obviously audible. Guess I'll go back down with them. Sam

Adam

I realize the rat meter suks for accuracy, all I'm hoping for is that the readings/calibration is not so crappy that it will change a lot over short freq. ranges. I just hope to see trends, such as in the xover region...great from 1.6 k upward, but depressed below that, usually confirmed by ear.

Listening to the 4648's run high, they still sound nice on axis, and to each side of the couch, at 12 feet or so. I think the dual woof array does exibit less vertical dispersion also...where I had the mike at 1 meter may have been too close, and at a larger angle off axis than was appropiate. You are right, in room rat meter is not much help, and "proves very little".

But these crude measurements DO seem to identify why the higher crossed pair sounds a bit thin, what I was previously subjectively attributing to level differences and/or room position was shown in the depression from 500 to 1 kHz. I have four different 902-8b's that exhibit the 19 kHz. peaking...none of the other hf systems I've ever checked had that, so IN SOME some cases I think valid conclusions can be made from rat data...

I think crossing as high as possible to get the HF farther above the 500 horn cutoff (less distortion), and staying below 800 (less "nasties") puts me pretty well back where I was going before being distracted...to use about 2.5 or 2.2 mH 14 ga. coils, and cross somewhere around 750... I had briefly considered a WWTWW stacked array of 4648a-8 on the bottom, 511 horn in the middle, and a second 4648a-8 above...then I could be the comb filter king of the west...

And all I really wanted to do was confirm those recently built 4 Pi Pro xovers work as advertised. They do. Sand, Sand, Sand...