PRELUDE - and being the devil's advocate
Copyright 2010 Troels Gravesen

Audio Technology 10C77-25-10-KAP + AT 15H52-06-13-SDKAM + ScanSpeak Revelator D2905/9900

Some people may find the PRELUDE crossover a bit complicated and let's play the devils advocate and go through the crossover, bit by bit. To some people a low order filter may imply a coil and a cap, but as discussed on the "Siri's Killer Note" page, low order filters may take quite a few components to reach our goal.

Bass section:
Q: Do we really need the RC circuit across the 3.9 mH coil?
A: Good question! We have a bump in the bass driver's response at around 2.5 kHz and although at a very low level, it's clearly audible when listening to the MLS signal, but hardly visible on overall frequency response (fig.2) and it cannot be stated for sure this bump will impact the overall sound, but the cost of these two components is very low compared to the overall cost of the speaker, so leave it in.

Fig. 1, left: Fine-tuning of bass coil RC circuit. Blue = no RC circuit. Green = 0.56 uF + 10R. Yellow = 1 uF + 10R, purple = 1.5 uF + 10R.
Fig. 2, right: Frequency response w/wo RC circuit. Take your pick!

Q: Can't we use electrolytic cabs instead of expensive polyprops ? (47 uF in tweeter section, 270 uF in mid section and 100 uF in bass section ).
A: Yes, but considering the overall cost of the speaker I wouldn't compromise. But I agree this would save you some 100 EUR in total.

Tweeter section:
Considering the fairly high point of crossover between tweeter and mid, do we really need an impedance flattening LCR circuit? I mean, another expensive PP cap and 2.2 mH coil don't come for nothing!
A: Another good question... Based on listening tests and distortion measurements, it cannot be concluded this LCR circuit pays off, but the choice is yours.

Fig. 3. Distortion test w/wo tweeter LCR circuit (47 uF + 2.2 mH + 5R6). Only 2nd harmonics shown. Green and yellow is without LCR circuit if you really want to know. Measurement done at 1 meter distance and at ~95 dB level.

Left: Fig. 4. CSD at 40 dB scaling with tweeter notch filter in place. Right: Fig. 5. CSD at 40 dB scaling without tweeter notch filter. No difference at all.

Q: Do we really have to use Superior Z-caps for the tweeter?

Mid section:
Q: My goodness, 3 notch filters! Do I have a cousin producing coils and caps?
A: Nope! The mid section could be made much simpler as seen on fig. 6, but it impacts the ever so important midrange negatively. The high-pass section LCR circuit is the one that could be left out although it does compromise bass-mid phase integration and increases cone excursion due to the middriver's 1st order high-pass section, thus higher distortion at high sound levels.
Next the two low-pass section notch-filters could be reduced to one by increasing value of L1041. This does however create a minor depression in the 1-2 kHz range and a minor notch at 2.5 kHz. Looks irrelevant on paper, but the full monty just sounds better, smooth and easy on the ear and vocals simply sounds better. View Fig. 7 below.

The minimal crossover seen below, fig. 6, surely looks tempting, but serious listening tests left no doubt, the V2 crossover simply sounds better.
The drivers alone for this construction will set up back some 2,000 EUR and the best option crossover (V2) some 850 EUR. By using standard polyprops (6 x 22 uF for mids) you may save some 340 EUR or ~10% of total cost. By using electrolytics for the three large value caps may save you another 100 EUR.

Fig. 6. PRELUDE cut-to-the-bone crossover.

Fig. 7.