Wolverine Front End Redesign
The main problem with this sumbitch is the need to swing the big volts (some 310VP-P) that the 845 requires since it's such a low-u type, while keeping the distortion down. These really are conflicting requirements. It's also nice to have some open loop gain for dropping in a bit of gNFB, which will sound like heresy to all the DHT fanatics out there. That, and the inclusion of SS circuitry really ought to have the audiophoolz howling. Who cares, so long as it gets the job done right.
After considering a bunch of possibilities, it looks like the way to go is small signal pentode front end. The main problem here is that the vast majority of these small signal pentodes like to run very thin (low IP (1.0mA and even less) with eeeeeeenormous plate resistors (>= 200K)) which you see all the time in designs like this. None of that's gonna cut it in this design.
Or you can go for the vid pents: designed to swing the Bigvolts to drive CRTs and do so with decent linearity. The only other problem is that these types like to pull big currents as well. This is, of course, required since a video signal has an upper end at around 4.0MHz. Even a few pF's are gonna require a healthy current to charge up at rates like that. After looking into a bunch of these types, the 12BY7A looks about the best. It doesn't require the high VPK's that some do to hit a linear part of the characteristic. This type being designed to kick colour TV CRTs in the ass, it needs to pull quite a bit more current than what you commonly find with small signal types. Unforch, these guys aren't exactly cheap since boatloads of them found their way into ham radio gear as oscillators and drivers. Forch, I have a bunch of 'em to play around with. Being that this is a vid amp type, I wouldn't expect microphonics and/or heater leakage to be big problems (it can be with small signal RF types like those 6AU6s).
12BY7A Loadline
Quickpost this image to Myspace, Digg, Facebook, and others!
Looking at the plate characteristics, the lines centered on the VGK= -3.5V look to be about the most linear, and the resulting loadline gives a THD= 0% estimate (based on h2 only) which is a good sign. Of course, this loadline doesn't leave the saturation region at the desireable VGK= 0 point, but rather at around -2.0V. Bad news since it will be out of clipping, so that there won't be any aural warning that something's about to poof. That region to the extreme left of the characteristic is red screen territory. We can handle this problem handily by dropping the screen voltage. Since this characteristic is developed on the basis of a VSGK= 180V, we have a ways to go. Digging out the forward characteristics charts, it's pretty obvious that dropping VSGK to 95 -- 100V will get the job done, and push that loadline up to the VGK= 0 point before leaving the saturation region. It is also desireable to get that screen voltage as low as possible for improved sonic performance.
Still, almost no h2 doesn't guarantee that it won't be making h3 or higher order harmonics that'll sound like:
. So what to do about that? The answer here is to go with a semi-active load. Let the plate resistor handle about a third of the total plate current. That way, you don't have a CCS fighting a CCS, and making no sound. You also get the benefit of increased linearity. By using an N-Channel, power MOSFET based CCS design, you also get one additional bonus: you can use the thing almost like a Mu-follower, and get a Lo-Z output. This also helps with your bandwidth when connecting to your load, in this case, the triode section of the 6BX7 -- a vertical deflection oscillator/amplifier type that can handle the high voltages required here for swing and headroom. Even though this type has low gain, what gain it does have leads to a nasty Ci + CMiller + Cstray problem. Making the CCS do double duty here fixes that.
To eliminate a cathode bypass capacitor, and to provide a gNFB summing node, use the other triode in the 6BX7 bottle as an unbalanced differential. That way, what would be your cathode bias resistor becomes a passive tail load. (This particular topology usually isn't as useful as some designers seem to think it is.) Since the cathode current of the 12BY7A is so high, and the grid bias (-1.0V) so low, its cathode resistor doesn't need to be bypassed since it's already quite small (40R7) and that little bit of cathode degeneration will help the distortion situation. About the only possible problem here is sky high gain since the 12BY7A has a huge gm. The solution is also quite simple: make an anode follower to reduce that gain to a more useful level. That would require including a cathode follower up front since any varying series resistance from the volume control will also change the gain. This might be the place to put those 6AU6s to work as trioded cathode followers.
The only thing that sorta sux here is the need for two RC couplings. These will have to be staggered so as to not cause phase shifts that can lead to instability under gNFB. It shouldn't be a problem since power MOSFETs like to see gate resistors much bigger than the usual grid DC return resistors that VTs can tolerate due to the inevitable partial vacuum fill that leads to ionization grid current problems. So frequency staggering isn't gonna be a problem, nor lead to having to resort to huge coupling capacitors.
Main Schemo (Revised)
Quickpost this image to Myspace, Digg, Facebook, and others!