Saturday, June 16, 2012

More simulator FAIL

I decided to experiment a bit with LTspice and various MOSFET models. I've been getting results from manufacturer-supplied models that are radically different from what the datasheets would lead me to expect and what I see on my bench.

Here's one of the simple models I tried:

I figured, perhaps erroneously, that in the test circuit above, all the capacitances between the gate and other structures would be in parallel, giving me a nice, fat number. Of course, this this ignores charge transfers due to changes in the drain-source current (Miller charge), but it's a start.

The FDV301N datasheet gives these numbers:
Input Capacitance (Ciss)            = 9.5pF
Output Capacitance (Coss)           = 6.0pF
Reverse Transfer Capacitance (Crss) = 1.3pF
So I sorta figured I'd see something in the 10pF to 20pF range. I wasn't too far off: including the Surfboard adapter, the  FDV301N is a dead match for a 33pF NP0 capacitor. The Fairchild Spice model? It matches a 150pF capacitor pretty closely, with a few odd wiggles. In fact, if you pull apart the subcircuit definition you'll find these two lines:
Cgs 1 5x 78p
Cgdmax 7 4 75p
I was a math major for a while, so I know this one: 78 + 75 = 153.

Running the same test on the bench with the DMN26D0UT showed it's a good match with a 22pF capacitor. The BSS83 has such low capacitances that it looked the same as not having a C1 in place. Apparently the 9.5pF of the scope probe and the various stray capacitances from the solderless breadboard are so high by comparison that you just can't see a difference. In fact, the scope trace looks the same whether the Surfboard carrying the BSS83 is plugged in or not.

While I was munching on leftovers for dinner, it occurred to me that the same mechanism could be used to get a crude measure of the input capacitance of a real circuit. As I mentioned before, the test above didn't include the Miller charges. So I rearranged things into a two-stage inverter chain. There's clearly a brief pause in the RC charge curve at the FET's threshold voltage, just as there should be, but otherwise it looks just like the test above. I'd guess it'd be safe to call it a 50pF equivalent at most. Certainly not 153pF, at least not in the circuits I'm going to be using.

No comments:

Post a Comment