|Caged Frog ][|
I really like transconductance amps for driving headphones. The first one I built, the Caged Frog, has remained a favorite of mine. Even the Baby Frog is great for what it is. But, since I have a definite preference these days for differential amps, the next obvious thing to do was make a differential gm amp.
The easy, and perhaps obvious way to accomplish this would be to build a Firstwatt F1 clone, perhaps with some different parts. But, aside from the fact that that seems a little boring, I wanted to get back into some tubes, and perhaps use up a few parts out of the junk box.
I scrounged up a small stash of 7W7 tubes, a Pete Millett find, to try. These are loctal pentodes that can be thought of in spec terms as more or less a pentode version of the 7N7, which is the loctal 6SN7. That is, they are not actually particularly high gm tubes, but they are reasonably linear, and are reputed to work pretty well for audio. And, one does not really need that much gm to drive efficient headphones.
For the first version, I simply connected the cathodes, ran a resistor to ground, and hooked them up in pentode push-pull without any feedback. I also hooked up a bunch of big oil caps, but this had more to do with having a bunch of these, and not having anything else to do with them. This was actually a tolerably good amp. There was definitely a little noise creeping in, some from the unregulated power supply, and perhaps some from the unregulated screen grid supply. However, while the differential nature of the amp did eliminate much of the noise, better could be done. Better in this case meant a small regulator for the screens, which helped quite a lot.
For the experimental version, I used some Lundahl LL1663-PP 5K:8 output transformers. These can handle 40W and were really too big and clearly not the correct ratio for this project. The thing with output transformers is that as the signal level increases, the core size must also be increased. However, the bigger the core size, the lower the fidelity (something to keep in mind when you think about driving your high efficiency headphones with a 300B amp that can also power speakers -- it is a sucky way to build a headphone amp as you are just throwing away the low level signal). Anyway, back to the OPTs, the Lundahls had been sitting in a drawer for years. I kept meaning to do something with them, but I just could not find a project. But while they worked great for bread boarding, something better was definitely in order for this project. To that end, I picked up a pair of Lundahl LL1689-PP's. These are still pretty big, but they are much smaller and sound much better in this application.
These are better for two reasons. First, the 1663's have a 5K:8 Ohm impedance, which is a 25:1 voltage step down. However, they do not have the inductance to be loaded with 32 ohm headphones. That means that a large resistor needs to be placed in parallel. That means that lots of signal is being thrown away and that the tubes must then swing a larger signal. The tube in this amp are certainly capable of swinging this signal without clipping, but the more tubes swing, the higher the distortion. Thus, not only was low level detail being thrown away in the transformers, what was getting through was going to be higher in distortion (something else to think about with your 300B speaker/headphone amps as they have the same problem.) With the 1689's, not only are they smaller and generally higher fidelity (and cheaper), they can deal with the 32 Ohm load and their 18:1 voltage ratio requires less work from the tubes. Thus, a win-win-win.
With all that said, how the output transformer is loaded matters a lot. I did still find that some load in addition to the headphones seems to improve the sound. However, the resistor being used is now much larger, meaning that much less of the signal is being wasted across it. As with the other gm amps, this is a place worth experimenting as different resistor will change the sound quite a lot. One of these days I'll build a little switch system for changing this on the fly.
Anyhow, I also decided that a better tube might be in order. For the tube, the obvious best choice, because I had them, because they are good for audio, because they are quiet, and because they are actually a reasonably high gm tube, was the C3g. For the uninitiated, the C3g is one of those German super tubes designed to a commissioned spec, individually serial numbered, yada yada yada. Whatever, they are really good sounding, and they are pretty cheap. Not 7W7 cheap, but cheap when compared to sought after audio tubes. And they are loctals which meant I could reuse the same tube sockets without a lot of fuss. And they last forever.
The schematic shown is somewhat simplified. You'll need various stoppers, etc. to keep the tubes from oscillating. I did not specify any particular operating point (or tube for that matter) as this is worth playing with. For what it's worth, I believe that my B+ is around 200V and the screen grids are at ~150V. There is probably about 15mA through each tube, though I could be off on that.
I also built a super simple regulator for B+. It is basically the same design as the one for the screen grids.
|A simple yet effective regulator|
These are pretty simple things, basically a resistor and a zener diode to set the voltage, and a pass transistor, but they work to lower the power supply's impedance and greatly reduce ripple, so they work great. The amp has high enough PSRR that even a terrible regulator is overkill. I actually left the heaters as AC and they seem pretty quiet.
So, that's it. I built it without a volume pot as it is sort of a power amp for headphones. I figured that since I have more amps than I do sources, it makes more sense to put the volume control in the source. There is no tube glow, unfortunately, but there is plenty of tube sound. Seriously, if you have never played with pentodes because someone online convinced you that pentodes sound like solid state, and only triodes are capable of reproducing music, you are missing out (you might be missing out on some good solid state, too).