[2005-08-12]

While I am still enjoying my 2a3 monkey amp, I really wanted to design another amplifier, using concepts I had not previously tried. My introduction to really fantastic push-pull amps was at VSAC 2003 when I heard Allen Wrights fully differential 300B amps. I am definitely a single-ended believer, but the sound coming from these things was so amazing that I had to try a differential push-pull design. I messed around with a couple versions of a dialed-in balanced design aith a current source on the cathode and a negative voltage supply for the bottom of the current source and all that stuff, but never built anything.

Then I read through Lyn Olson's ETF presentation on differential topologies and his dissertation on ancient Western electric designs. I was inspired to try some of that old-is-new-again Western Electric stuff.

The third great idea that led to this design is Poinz' Tater Totter design. Suddenly my push-pull crush could be implemented without the hassle of multiple stages and interstage transformers!

Of course no minimalist push-puller can ignore the compact amp phenomenon, and especially the super cool work done by the enigamtic Keto.

So with this enormous body of work swimming around in my head I came up with this design. It is essentially a single stage Lynn Olson amplifier, with current source fed gas regulator power supply and all. But it is a spud (A spud amp is a one-tuber. Tuber, get it. Groan.). Well actually it requires 2 tubes per channel, plus a rectifier, so maybe it isn't purely a spud amp. Of course if you define a spud amp as an amp with a single gain stage instead of a single tube... Well, things get semantically confusing :)

I originally wanted to forgo a traditional phase inverter in favor of the compact amp style of "self-split", but that would have required a somewhat large resistor between one of the 6c45 grids and ground. I have been digging the grid-choke idea, and the 6c45 is supposed to like a low DCR grid-to-ground connection. My audionerd friend and local genius Matt Wiebe turned me on to a clever trick from Paul Joppa and Mike LeFevre (I am not worthy!) that suggests using a Peerless/Altec 15335 (which is a 15k:15k single ended input transformer) as a center tapped autoformer phase inverter. Minutes later I had cannibalized a pair of Ampex input transformers from a broken Ampex PR-10 that I have been wanting to get working again. These appear to be the same thing as an Altec 15335 and should work just as well. I think the 6c45 would have liked the radio frequency rejection of a conventional input transformer rather than an autoformer, but this was just too cool a trick not to use. I'll have to see if there is an octal plugin that would give me a 1:1+1 input transformer.

I have been collecting parts for a while, and have finally reached the critical mass required to build a parts-bin amp. I had the power transformer, power supply choke, output transformers (from an old Sansui I think), big honkin oil caps (Sprague Vitamin Q even), and gas reg tubes. All I needed to buy were the 6c45s and the current sources. I chose Kevin Carter's 2 terminal CCS assemblies for ease of use since they are at the outer reaches of the signal path, and should not affect the sound as much as if they were plate loads for example. I very much like Gary Pimm's tube based current sources, but they looked so complicated that I got scared. I probably should have stayed true to the Western Electric ideal and used a choke instead of a CCS, but I need to drop some voltage and add some power supply filtering, so a CCS just seems to fit the bill.

Construction of the amp was very straightforward. I love it when almost all of the parts mount to the chassis. I would have liked for the wiring to be a little bit neater, but this was the best I could do.

After all the worrying I did about the current sources and the heat they would be dissipating, it turned out to be a total non-issue. The power transformer gets warmer than the heatsinks I used. Chassis mounting the CCS would have been just fine without the heatsinks.

6c45 tubes can be a pain. I guess they are radar tubes or something like that, and so are very good at high frequency oscillation. To fight this oscillation, I used grid soppers on both grid pins on each tube, with the resistors forming a "tent" and connecting to the grid lead. This minor precaution was sufficient on the last amp I built, but not on this one. I think a combination of the little bit of RF generated by the gas reg tubes, and the reactive nature of the input autoformer is too much for these little radar tubes, and they just can't help but oscillate. The amp was stable when there was something connected to the input, assumably because the output impedance of the source would damp and resonances that try to get going in the input autoformer, but with nothing plugged in, it would go crazy. It would oscillate audibly, and the gas reg tubes would go out! That means that the output pair, which normally conduct 60mA, were drawing the full 80mA allowed by the current source! Nuts!

To fix this, I simply attached resistors across both windings of the input autoformer to mimic a source impedance. I started with 2.7k resistrs and that fixed it, but the low input impedance gave me noticeably less gain overall, and was likely taxing the source. So I started and 1M and worked down until the oscillation went away, ending up at 211k. Anything below 250k or so should work just fine.

I hooked this guy up to the (all tube tektronix 555) scope, and could definitely see some crazy noise in there. I think I will see what else can be done to calm these hyperactive little tubes down. Unfortunately I am not much of a scope jockey, and don;t know what a happy amp should look like, so I'm not sure how to distinguish between oscillation and plain old noise.

Suggestions I have had to help the oscillation problem are to add 0.1uF capacitors from both filament pins to ground and also between filament pins, ferrite beads on the cathodes, bigger gridstoppers, and voodoo rituals. I also want to try adding small caps across the gas regulators to hopefully shunt away some of thet RF they generate.