30
14 parts in stock at the time of this writing - but switches
made from it command an even higher premium than
does SiC for the same voltage and current rating.
Whether a premium in price is worth it depends on the
advantages a particular technology delivers, and the value
proposition for SiC, at least, is very compelling, especially
for devices rated 650 V or higher. Sure, there are the “Su-
perJunction” Si-MOSFETs that can achieve very respect-
able on-resistance values in the 600 V to 900 V range, but
at 1,200 V there is simply no comparison anymore: SiC
blows away Si, and manufacturers are even now roll-
ing out 1,700 V rated parts, which Si doesn’t even try to
compete with. GaN should be capable of a similar voltage
rating vs conduction loss figure of merit, but the GaN
switches available so far seem to be rated for 650 V or
lower. GaN can switch much faster than SiC or Si - 10 ns
rise/fall time, with minimal turn-on/turn-off delays, too- but this may be a problem masquerading as a solution,
as switching 300 V or more this rapidly results in massive
amounts of ringing, voltage overshoots and RFI/EMI. The
RFI/EMI issue may very well be insurmountable when
practical considerations such as, say, passing electromag-
netic compatibility (EMC) testing is are factored in. The
wise (or the battle-scarred) know that nothing ruins a
development schedule and budget quite like failing EMC
testing so badly that you have to do a total mechanical
and electrical redesign.
There are other peculiarities of each of these new ma-
terials that can also trap the inexperienced (or unwary).
The first is the difference in behavior of the intrinsic
(aka “body”) diode that is found in both Si and SiC
MOSFETs (but not in GaN switches). This diode is a pn
junction type formed as a consequence of the way these
modern vertical-structure power/switching MOSFETs
are constructed. The body diode is infamous for be-
ing slow in the Si-MOSFET - taking 300 ns or longer to
turn off - so engineers learn to set up circuit conditions
so that it either does not ever conduct, or at least, does
not have to endure high dV/dt when reverse voltage is
reapplied. The body diodes in SiC MOSFETs are con-
siderably faster, however, with reverse recovery times
that would give a hyperfast discrete pn diode a run for
its money (60 ns or less). One might naively assume that
there’s no need to worry about reverse recovery losses
(or outright failure) from body diode conduction in
the SiC MOSFET, but take a closer look at the forward
voltage spec in the datasheet and you’ll likely rethink
that assumption: 2.5 V to 3.5 V (or more)! As a result,
overall losses might be higher than expected when using
SiC MOSFETs in circuits in which body diode conduc-
tion is inevitable (such as bridge and two-switch forward
or flyback converters). Other than that one caveat, SiC
MOSFETs (and Schottky diodes) are compelling choices,
especially when operating at high bus voltages (>600 V)
and/or temperatures (>100° C).
In contrast, GaN definitely has some gotchas to watch
out for, and this technology is just too new for most en-
gineers to have had any experience with it. Complicat-
ing matters is the fact that different structures have been
employed to make GaN switches with equally different
functional (and pathological) behaviors. The first thing
to note, though, is that GaN switches are not MOSFETs,GaN Transconductance Curves
Whether a premium in price
is worth it depends on the
advantages a particular
technology delivers, and
the value proposition
for SiC, at least, is very
compelling.
THE TECH