3 Attachment(s)
True first order (phase perfect) vs. high order crossovers.
Quote:
Originally Posted by Ian Mackenzie
So are your drivers going to be in phase (polarity or 180 degrees out of phase)? The earlier Dynaudios worked like your proposed idea and possibly the Duntech's, but they were true 1st order designs.
Thanks again to Ian for raising the question on polarity, and 1st order slopes, It got me to re-check my system, and try out a few more ideas.
Following Ian's posting. I took the opportunity a couple of evenings ago this week, to re-visit the polarity, crossover slopes, and gain settings on my crossovers. In particular, I compared an implementation of true 1st order (6dB/octave) slopes at all four crossover points, against a high order (48dB/octave) Linkwitz-Riely crossover alignment, at the upper two crossover points, between midrange, and my two compression drivers. Both the 1st order, and the 8th order LR require all drivers to be wired in-phase. The first order xover has no phase distortion, but requires high power handling, low distortion, well behaved transducers, and relatively high crossover points, which, fortunately, my system employs. (but I still dare not turn it up anywhere near max volume). The second implementation uses 8th order LR crossovers between the midrange and the big compression driver, and also between the big, and small compression drivers. This has the best power handling performance (I can take the system up to near clipping if I so desire), and also helps to attenuate out of band problems, like cone break-up on the midrange, and diaphragm break-up on the 4" compression driver.
I have not had much chance to listen to the two systems (time for bed when I got done). But rapid switching (using the compare button on the DCX), indicated that the majority of sonic difference is related to the variation in frequency response, rather than phase variation. In other words, the differences were not subtle, and my preference went toward the 8th order slopes, because the 48dB/oct rolloff restricted the interaction/overlap between drivers better than the 1st order. Maybe I can tweak the xover point and response tailoring a little more to restrict the overlap zone in the 1st order xovers from producing a response bump around xover. (which I couldn't cure with simple inter-driver gain settings.)
The first image in this post shows a graphical comparison of the frequency response for the 1st order xover (top graph, both channels shown), and the 8th order slopes (bottom graph). Xover points are indicated with vertical arrows. The biggest effect/problem is at the 1khz xover point, where the 2123 and 2441 seem to sum constructively on the 1st order, and destructively on the 8th order.
Additionally, I could not (easily) implement the Sandrik method, as the Behringer only allows cascaded slopes using the rather cumbersome EQ modules, and the HP and LP shelving options only allow a maximum cut of -15dB. When I applied one of these on either side of the 1st order xover points, I started running into headroom problems in attempting to set output gains between drivers. When I build my next set of dedicated amps, for the compression drivers, I will include a 1st order pole in each amp, around 150-300Hz, so that the low frequency protection is "built in" to the amplifier, and any "fat finger" maladjustments on my part will not run the risk of driving the tweeters with wide band, or low frequency signal.
One other area that I revisited (while I had the laptop/mic test & measurement kit out) was the low frequency compensation for the Altec. After a simple change (2nd order boost instead of 1st order), I managed to get the low frequency -3 dB bandwidth down to an astonishingly low 5 Hz. This is also the approximate low end of the power bandwidth of the system, meaning that the speakers *could* produce around 110dB spl in the octave between 5-10Hz, with the amplifiers near clipping. Of course, there really is no music content down that low, but just in case musicians start producing/recording subsonic tones, I'm ready.
The 2nd image in this post shows the mandatory "eye candy" waterfall plot illustrating the usual room modes (under damped below 35 Hz), and the prodigious extent of amplitude down near the 2Hz end of the scale.
The final image shows a low frequency linear scale amplitude plot (generated using a 5s MLS signal, and 400ms analysis window over the impulse response), for both channels. I'm still amazed that the speakers, (and Panasonic mic), have response down that low.
Jack Bouska