A bit of a gray area are measurements in datasheets which were taken outside of the intended application range. E.g. switching time or Rdson of logic-level mosfets. The figure not given for when the mosfet is driven with a typical logic-level Vgs like 3.3V, not even 5V, but 10V. Bastards. Power ratings are also in such a gray area.
MOSFETs of all types over 20Vdss always work better at higher voltage, so you should try to drive at >= 10V if at all possible. Example: you have a 5V logic circuit and a 12V supply *somewhere*, but it's not nearby, so you figured, ah hell, just use logic level. And you discover the performance sucks. Later, you add a proper gate driver supplied from 12, and it works beautifully. If it's around, it's worth using.
If you must drive at 5V, do so, and use a very low impedance gate drive. The switching speed is dominated by Cgs equivalent and Miller effect, and during the switching transition (Miller step), you can't drive any more current into the internal gate node any faster than the total resistance allows. Helpfully, datasheets are providing R_g more often these days, so you can see what size gate driver is appropriate (if R_g is 5 ohms, a driver under 2 ohms or so won't help any), and better estimate switching speed.
Speaking of, calculate gate transition time based on Qg(tot). Max if provided.
Ignore Cgss.As for lower voltages, there is no such thing as a 3.3V logic level silicon MOSFET for high voltages (over say 40 or 60V). At best, you might find some that are specified for crappy performance there, but it's only very marginally in the saturation region, especially over Vgs(th) and T_J variation. If you absolutely must, well, whatever, but don't expect miracles. If you need actual switching performance, by all means, at the very least add a charge pump for ~6V, or a proper boost converter to get in the >= 10V range.
There are only two cases currently available where lower voltage is acceptable: low Vdss MOSFETs, which are made with smaller feature sizes and thus have higher gain, and GaN FETs ($$, small selection, small stock, often hard to use -- e.g. solder-bump dies), which have all-around way better performance than Si (especially for higher voltage ratings, currently up to 200V or thereabouts). The lowest Rds(on) I've seen is something like 700uohm (that's micro) for a 7V 30A Si part, rated at Vgs = 2.0V. This would be good for single or possibly two cell lithium battery management applications, and even some DC-DC converter action (beware of R_g, I don't recall it was particularly high for this example, if it was given, but with lower voltages comes greater sensitivity and all that).
FWIW, MOSFETs over about 40V all look the same, in that the gate voltage range is pretty much constant (Vgs(th) by design, as well as the amount required for saturation on top of that), and Rds(on) has a strong tempco (ranging from about +80% at 150C for the lower voltages, up to +180% for many ~1000V devices). Lower voltage devices (particularly 20V and below) exhibit higher gain (making logic level applications more practical) and lower Rds(on) tempco (more like +30 to +50% at 150C). Recognize that the only thing that really changes between "regular", "logic level" and (if you run across any,) "depletion mode", is the gate threshold voltage. The gain (change in drain current over change in gate voltage) isn't any different between any of them: the only difference is, logic level types are made with a lower Vgs(th), with a tighter tolerance. Let me put it this way: if you would never consider driving a regular FET with +2 to +7V (assuming the same Vgs(th) min-max spread), don't use a logic level FET at 0 to 5V.
Tim
Home
Help
Search
Login
Register
