Ah, yes...hmm...
If you use IRF5851 as common source (i.e., just a regular unbuffered CMOS inverter), consider what happens for gate voltages between the threshold voltages. Suppose VDD = 10V. At Vg < 0.6V or Vg > 9.55V, no current flows (Id < 0.25mA). For gate voltages inbetween, one transistor will be fully saturated (~0.1 ohms) and the other stays in the linear range.
Referring to Fig.1 and Fig.16, as Vgs becomes biased on, the transistor turns on more and more, drawing
amperes fairly quickly. Curiously, the highest curve shown is 2.5V, drawing just under 10A in both cases; neither transistor is even rated for any higher peak current, for any duration -- yet they will quite clearly be capable of 20, maybe 30A peak current at 5V!
What's wrong with that? Besides the ludicrous demands on the supply bypass (you will need large ceramic capacitors and aluminum polymers to hold up the supply during such a massive transient!), exceeding peak rated current likely means spooky failures, like metal migration: the average temperature might not be exceeded, but with such extreme current density, the device will nonetheless fail after some number of cycles, maybe one, maybe a billion. But it is sure to fail.
Typical application as a gate driver replacing one of those "bipolar pair" devices might have ~20mA source drive, to expect 2A peak output. The total gate charge is around 10nC for 4.5V gate swing, but we're looking at a 10V application, so it'll be more, maybe 15nC total. A 20mA driver will then take 0.75us to commutate the gates, which means about 0.75us of >20A shoot-through. (Which will dissipate about 10V * 20A = 200W between the two transistors, or *0.75us = 150uJ per transition, for a maximum frequency of only 6.4kHz at rated power dissipation. Nevermind, maybe it will fail thermally first!) The output will swing a little bit faster than 0.75us, because most of the gate charge occurs during the Miller step, but it will still be much slower than with the bipolar device.
The fundamental problem with such a choice is that it's more than
ten times larger than necessary. Try a 1-2 ohm MOSFET instead; you'll still have to deal with 5A peak shoot-through, but it can be switched fast enough that it's not as big of a deal. This is comparable to the output drivers in proper monolithic CMOS driver chips. Even the largest (15A+) driver chips boast only of 0.2-0.5 ohm outputs; using as little as 0.09 ohms for a couple-amps application is ludicrous, as its performance shows.
Moral of the story: always try to pick parts with are suitably matched to the application. Excessive size leads to wasted money and poor dynamics, and rarely yields improved efficiency!
Tim