Diodes seem to be an important part in the sonic signature of a power supply. Some will argue that we can't live without soft recovery diodes, others will only swear by ultrafast ones, tube fanatics want tube recitifiers ; and there are also Schottky diodes, Silicon Carbide diodes, Plain old diodes, etc.
After reading this Article here (many thanks to the author), I had a crazy idea : I want a power supply without diodes. Of course this is impossible. Is it ?
I wondered if anybody had ever tried diode-wired MOSFETS. You can make a diode out of any transistor, just by connecting the Emitter with the Base (in a Bipolar), or the Gate dwith the Drain (in a Mosfet) :
Let us compare the currents flowing in both of these components, according to the voltage at their terminals :
The MOS (blue) is clearly a very bad diode, as it has a high threshold an a high dynamic resistance. The diode (red) lets through a lot more current.
But the switching characteristic of the MOSFET is also much nicer and softer than the Diode.
Example schematic :
This is a model of a transformer with its internal inductance (L1), capacitance (C1) and resistance (R1). We connect a diode bridge to it (right) or a diode bridge + mosfets (left).
We could have used 4 MOSFETS instead of the 4 diodes, but 2 Mos + 4 diodes is a cheaper solution, and it avoids reverse-biasing the Mosfets, which may not be appreciated.
What is important is that it is the Mosfets that control switching and not the diodes. Thus the diodes are, in fact "not there" as their recovery characteristic does not play any part.
The R/C in the Mosfet gates are here to soften switching a little more, but they are not mandatory. They could be removed without problems.
On to some simulation results... Here we have the voltage on the transformer, for the Diode case :
Here we have the voltage on the transformer, for the Diode + mosfet case :
It is clear that the MOSFET's soft switchoff does not excite the ringing in the transformer. These two curves were simulated without the snubber networks (R11/C12 and R16/C13) to make the difference more obvious.
With the snubbers in place, the curves look the same (not shown) but the Diodes still have an ugly glitch at switchoff whereas the Mos has no hint of a glitch.
This seems to be a nice thing to try for low current applications (be it low voltage, like in a preamp, or high voltage, like in a tube amp), if the voltage loss can be accomodated.
I did not try it myself (yet) but will if I get the occasion.
If you try it, I'd like to hear from it.