Finished: Proac 2.5 active xo

Well, it took a fair amount of work, but I completed my active xo project.

For some background on the design, check out this thread:

http://www.diyaudio.com/forums/showthread.php?threadid=8541

I was initially targeting crossover points at 1500Hz and 3800Hz. My simulations indicated that a good fit electrically was about 2000Hz and 4000Hz, and I was told that 1500/3800 were the xover points in the commercial design.

However, despite the dreaded "hump" at 2kHz, there was not enough response in the 2000-2500Hz range resulting in a suckout. I then tried crossing at 2000 and 2500Hz on the low end, which did fill in the gap in the 2-2.5kHz region. I still had a suckout at 3.2kHz. To fix this, I tried moving the high pass crossover point down a bit. Unfortunately this was not successfull; as predicted in simulation, there was a significant hump in the response at 1kHz, and the suckout at 3.2kHz was only slightly ameliorated!

At this point I was grumbling a bit; however, I started to look at Jacq's published measurements on the Proac clones and noticed that I had achieved nearly identical acoustic response in one of my trials. (Jacq: Thank you for publishing tweeter and woofer as separate measurements!)

I then referred to the Stereophile review of the commerical speakers (http://www.proac-loudspeakers.com/stphres2.htm) and the in-room response was nearly identical to my measurements: flat to about 38Hz, suckout at 3.2kHz, rise in response from 5kHz-10kHz, fall in response from 16kHz-20kHz. (Note that the anechoic response of this speaker was very flat despite these in-room measurements)

I then concluded that this xover arrangement was very, very close to matching the passive version.

Final xover points:
LP 6dB @ 1500Hz, 12dB @ 2500Hz
HP 18dB @ 3800Hz

You can see below the actual in-room measurements of 1500/3800, 2500/3800, and 2500/3000. I think the remaining "error" in the response needs to be solved by fixing the room and/or speaker placement, which is beyond the scope of this xover project.

Final note: ^_^ Although I built a passive xover for subjective comparisions, I didn't trust it very much for measurements because I used junkbox parts and high DCR coils. The active and passive xovers do match fairly closely.

Crossover components:

2n5457 as a buffer, biased at 1.3mA (CCS)
SLA battery supplies
Dale resistors
Wima FKP in the HP sections
Also Wima MKP and ERO 1813 in the LP

Distortion is about -85dB 2nd harmonic SE, which of course disappears entirely when the differential output is used. 3rd harmonic appears to be way down in the -100dB arena. Noise is certainly under the noise floor of my test equipment.

I am using Rod Elliots SLA charger design:
http://sound.westhost.com/project98.htm

Initial listening impressions: the bass is even tighter in the active version. Complex passages seem cleaner. Imaging is very, very good.

Click here for a (large) pic of the complete unit:
http://www.anidian.com/imgs/xover.jpg

Or one of the boards. Each board is a single differential channel. I didn't want to do any Diff - > SE -> Diff type conversions, nor did I want to use something like a diff. pair here, so I simply duplicated the circuitry. I matched each differential "pair" of circuits to 0.5%. Over the entire range they track within 0.25dB, which is about as far as I trust my measurement setup.

P.S. That dreaded 2kHz hump is gone. I think this is validation that at least a good bit of the hump is due to electrical interactions between the driver and the passive xover. The active version is immune to these problems.