Tuesday, December 6, 2011
The Baby Frog -- A MOSFET Based Single Ended Transconductance Amplifier for Headphones
As I mentioned in the article about the Caged Frog pentode based transconductance amplifier, in a lot of ways, a pentode is pretty similar to a depletion mode MOSFET.
Obviously, one is a tube and one isn't, but if you look at the plate curves of a pentode, they look more or less like a FET's curves.
Thus, there is no reason that a simple transconductance amplifier cannot be built from a mosfet. For this project, I used a DN2540, which is a depletion mode mosfet. One can also use a more common enhancement mode fet, but because, unlike with a depletion mode device, the gate must be positive of the source, the enhancement mode device's input needs to be capacitor coupled, or there needs to be a negative rail. These both being extra work, I opted for the easier path. And I had a stack of DN2540's around.
Obviously, there are a lot of ways to accomplish this. But, as a bit of an experiment, and because I had the parts, I opted to basically build the same amp. Other than switching to a mosfet from a tube, the only other topological changes are that one can get rid of the screen grid supply as well as the heater supply. Additionally, while the DN2540 is a high voltage device, it does not need to be run at a high voltage. To that end, I found an old unused 24V LM317 based power supply and a wall wart. I also removed the source resistor's bypass capacitor.
Now, this brings us to a short discussion of output transformers. If you want to build this, you'll need some. And, since we are running a bunch of current through them, they'll need to be fairly large, which means transformers built for a power amp. Generally, with output transformers, you get what you pay for, and good ones can cost a small fortune. This being a cheap-o project, I didn't want to spend a lot of money. The only real option for decent, yet not too pricey, OPTs, is Edcor. You are never going to confuse them with Electra-Print/Lundahl/Magnequest/etc., but they are fine for this project. I used the XSE series which cost about $17 each.
My transformers are 5K:8. Since I am using 32 ohm headphones, this would reflect as a 20K impedance. Unfortunately, the transformers are not inductive enough to act as 20K:32 transformers, so a resistor load in parallel with the phones is necessary. I played around with a bunch of values and ended up with some 10R resistors in parallel with the output. Exactly what value to use is really a matter of experiment. And in fact, switching these around can change the sound quite a lot, so using something like a 20 ohm linear pot can be a good option.
Transformers are limited by how much current they can tolerate. The Edcors seem to max out at around 60mA, beyond which point they will saturate. If one were simply running the phones off the drain**, this would be perhaps a little low. But, transformers convert voltage to current while conserving power, so a big voltage swing on the primary will translate to a small voltage, high current on the output. Additionally, the gain of the stage is the ratio of the load (the transformer) to the biasing resistor. A smaller resistor will not only increase current, but also gain. To increase it even more, the resistor can be bypassed with a large capacitor. Finally, if you need more swing room, increase the power supply. The DN2540 can tolerate up to 450V which means that, so long as there is sufficient heatsinking, the power supply can be quite high.
* What are those tubes? They are 6H6's and are not connected. They are there to fill some sharp holes that were in the metal top ... and to look like frog's eyes.
** And why aren't we just running the phones off the drain? Because you would need to cap couple the amp, and I hate capacitors.