active crossover using 12AU7

[keto]

Though I should probably leave well-enough alone... I've been listening to a pair of JBL 4320's and 300B Paramounts, and I hear a slight, kind of "klangy" sound around the 800hz crossover (schematic of #3110 JBL network attached), well below the 1W output level, according to my multimeter. First idea was to replace all but the autoformer in the #3110 networks with really nice parts, but that would run over $1000. Then I thought, just the caps, with some nice Solens, which would run more like $100. But either option would still leave me with a 2nd/2nd order filter with the polarity of the tweeters reversed. Maybe part of the "klangyness" is from unfortunate overlap of the drivers? Maybe a 4th/4th order, in phase filter would sound nicer, to me? Maybe I could even... make it an active filter? Before I continue, I think the nicely re-coned 2215 woofers and 2420 tweeters are just fine, since the stereo image is really consistent and the "klangy" sound, I'm guessing, is more from aging caps than anything else. Of course I could be totally wrong about all this. And maybe its all just an excuse to get a second pair of Paramounts. Okay, so if all goes well, the generic active XO circuit should appear below. For a 750hz XO, and according to Pete Millet's site, I would use .001uF and .002uF caps and 150K and 300K resistors, and 300K grid leak resistors. Cathode resistors would be 50K and the 12AU7's would draw 2-2.5mA. The 5K PSU dropping resistors could be reduced so as to raise the V+ and V- and 12AU7 current draw. At first I was hoping to use a 3-triode compactron (6C10, 6D10, etc.), but it seems important to bias the first triode at at least -4Vg, and I couldn't find a compactron that was a triple 12AU7 or similar. Also, aside from the cool factor of getting a complete 4th order high- or low-pass out of a single tube, a row of 3 12AU7's for a high- and low-pass pair-o-filters or "parafilter" will probably offer better spacing for the R's and C's. Of course the circuit is planned so that I could add in C4S current draws, choke-input PSU, and maybe a little voltage regulation, but first I'd like to see if I can get away with AC heaters on the 12AU7's, since the PGP8.1 can accommodate 1.8A of AC but not 1.8A of DC @ 6.3V (heaters). My initial idea was to leave these 4320's stock while I worked on my other, multi-amping project, but now I'd really like to hear this 2-way speaker with a 4th/4th order filter and a Paramount direct coupled to each of these drivers. From the specs, I think the 2215's are 96dB efficient and the 2420's (WAG) 106dB. Also, I realize that for the HP, I could also use a 2nd order filter and re-size the interstage and parafeed caps in the Paramount, but I really don't want to touch the latter's circuit... yet. Any comments or suggestions on all this rambling would be very welcome. Thanks!

[Paul Joppa]

The 4320 was a seminal JBL speaker and a classic. However, there were some updates in later years. One of those was the horn.

The JBL literature is mostly written by their marketing people and hard to interpret sometimes. But I believe the 2307 horn/lens was originally intended for a 1200Hz crossover; its length is closely matched by the older "potato masher" horn/lens which crossed at 1200Hz. That length is about a half wavelength at 1200Hz, allowing the tweeter to be in phase with the woofer. It was also used at 500Hz for home audio, where the length is a small fraction of a wavelength. When they adopted this arrangement for professional studio monitors the crossover was raised to 800Hz, presumably to protect the tweeter from mechanical damage - the aluminum diaphragm surround is prone to fatigue failure in severe service. The 4330 and later models of this speaker used a re-designed horn, enough longer to be a half wavelength at 800Hz (some versions added the "slot" supertweeter as well).

Because of the above, it is possible that the crossover-region stuff you hear is due to the phase or delay mismatch. Many people have commented on the futility of "upgrading" JBL crossovers, which seem to have been designed by ear as much as by engineering.

Another speaker issue is the small horn mouth, only 4" diameter. This causes "organ-pipe" resonances in the lower range (below 2-3kHz) which can be audible as well. I call it a coloration or shouty character, but klangy is a reasonable term as well.

In case you wondered, I have studied this issue because I have a similar system - I use the larger horn driver and the shorter, 2" throat version of the horn. This combination has the same total acoustical length and the same mouth diameter as the 2307 unit that you have. I have become accustomed to the sound by now and I don't have a useful or tested suggestion - just wanted to share what I've learned or guessed.

[keto]

So, here's my latest attempt at the signal and PSU circuits. The original 150K and .001uF values, after morphing to 15K and .01uF, will now be 13K7 and .011uF, since those seem more "standard". In the PSU, I'm not sure if the two VR diodes should be grounded to earth or not. It seems like not grounding would be good because the node between the two diodes could then be made, say +25vdc by stacking an OC3 on top and an OD3 on the bottom, which would seem to help when it comes to biasing 6 C4S's as cathode followers.

Oh, and here's a link to an interesting-looking horn for the 2420 drivers... http://www.azurahorn.com/JBL2420_on_550.pdf

[Paul Joppa]

Your circuit shows +/-100v power, but the power supply shown will make +/-500v before the regulator tubes. The resistor loads on the gas regulators must be much larger, and will dissipate an awful lot of power. The power transformer center tap (not shown on your sketch) should connect to the filter center and to the gas tubes, both of which you have shown grounded. This is signal ground, and should be attached to chassis ground at one point only, preferable near the input RCA jacks.

You don't really need the first stage buffer, as long as the network impedance level (currently shown as 15K) is high enough. I'd go back to 100-200K as the design target if you want to cut out those first stages.

The component values for the filters show two cascaded Linkwitz-Riley second order filters, each of which is down 6dB at the crossover frequency. You will have -12dB at the crossover, and a big audible hole. If you cascade Butterworth filters (-3dB at crossover) you'll get a 4th order L-R (-6dB at crossover) which works pretty well if the drivers are in phase with each other at crossover.

Hope that helps - crossover are quite a challenge, we've used up an early prototype by re-soldering so many changes we can't do anything with it any more!

[Grainger49]

You might want a level control on the outputs to help match levels.  This will help tweak in that point that they focus best.

[keto]

Okay, I think this is an improvement... The 12AU7 CF operating points are about 140Vp-k, 2mA, -7Vg, and a 75K load. It looks like in the low-pass circuit, it would be best to direct-couple between stages. Any suggestions would be very welcome. Thanks! --keto

[keto]

Trimming either the high pass or low pass signals will be done with a stereo, 10K passive pre, and here is the power supply....