MOSFET Steady-State Thermal Issues
The selection of a MOSFET to meet the maximum
continuous current is a fairly straightforward exercise.
First, the designer needs the following data:
The value of ILOAD(CONT, MAX.) for the output in question
(see Sense Resistor Selection).
The manufacturer’s datasheet for the candidate
MOSFET.
The maximum ambient temperature in which the
device will be required to operate.
Any knowledge one can get about the heat sinking
available to the device (e.g., can heat be dissipated
into the ground plane or power plane, if using a
surface-mount part? Is any airflow available?).
The datasheet will almost always give a value of on
resistance given for the MOSFET at a gate-source voltage
of 4.5V, and another value at a gate-source voltage of
10V. As a first approximation, add the two values together
and divide by two to get the on-resistance of the part with
8V of enhancement.
Call this value RON. Since a heavily enhanced MOSFET
acts as an ohmic (resistive) device, almost all that’s
required to determine steady-state power dissipation is to
calculate I2R.
The one addendum to this is that MOSFETs have a slight
increase in RON with increasing die temperature. A good
approximation for this value is 0.5% increase in RON per ºC
rise in junction temperature above the point at which RON
was initially specified by the manufacturer. For instance, if
the selected MOSFET has a calculated RON of 10mΩ at a
TJ = 25ºC, and the actual junction temperature ends up at
110ºC, a good first cut at the operating value for RON would
be:
mmRON 3.14005.025110110
Eq. 14
The final step is to make sure that the heat sinking
available to the MOSFET is capable of dissipating at least
as much power (rated in ºC/W) as that with which the
MOSFETs performance was specified by the
manufacturer. Here are a few practical tips:
The heat from a surface-mount device such as an
SOIC-8 MOSFET flows almost entirely out of the drain
leads. If the drain leads can be soldered down to one
square inch or more, the copper will act as the heat
sink for the part. This copper must be on the same
layer of the board as the MOSFET drain.
Airflow works. Even a few LFM (linear feet per minute)
of air will cool a MOSFET down substantially. If you
can, position the MOSFET(s) near the inlet of a power
supply’s fan, or the outlet of a processor’s cooling fan.
The best test of a surface-mount MOSFET for an
application (assuming the above tips show it to be a
likely fit) is an empirical one. Check the MOSFETs
temperature in the actual layout of the expected final
circuit, at full operating current. The use of a
thermocouple on the drain leads, or infrared pyrometer
on the package, will then give a reasonable idea of the
device’s junction temperature.
MOSFET Transient Thermal Issues
Having chosen a MOSFET that will withstand the imposed
voltage stresses, and the worse case continuous I2R
power dissipation which it will see, it remains only to verify
the MOSFETs ability to handle short-term overload power
dissipation without overheating. A MOSFET can handle a
much higher pulsed power without damage than its
continuous dissipation ratings would imply. The reason for
this is that, like everything else, thermal devices (silicon
die, lead frames, etc.) have thermal inertia.
In terms related directly to the specification and use of
power MOSFETs, this is known as “transient thermal
impedance,” or Z(JA). Almost all power MOSFET
datasheets give a Transient Thermal Impedance Curve.
For example, take the following case: VIN = 12V, tOCSLOW
has been set to 100ms, ILOAD(CONT. MAX) is 2.5A, the slow-trip
threshold is 50mV nominal, and the fast-trip threshold is
100mV. If the output is accidentally connected to a 3Ω
load, the output current from the MOSFET will be
regulated to 2.5A for 100ms (tOCSLOW) before the part trips.
During that time, the dissipation in the MOSFET is given
by:
P = E × I; EMOSFET = [12V-(2.5A)(3Ω)] = 4.5V
PMOSFET = (4.5V × 2.5A) = 11.25W for 100ms.
At first glance, it would appear that a really hefty MOSFET
is required to withstand this sort of fault condition. This is
where the transient thermal impedance curves become
very useful. Figure 9 shows the curve for the Vishay
(Siliconix) Si4410DY, a commonly used SOIC-8 power
MOSFET.
Taking the simplest case first, we’ll assume that once a
fault event such as the one in question occurs, it will be a
long time–ten minutes or more–before the fault is isolated
and the channel is reset. In such a case, we can
approximate this as a “single pulse” event, that is to say,
there’s no significant duty cycle. Then, reading up from the
X-axis at the point where “Square Wave Pulse Duration” is
equal to 0.1sec (=100ms), we see that the Z(JA) of this
MOSFET to a highly infrequent event of this duration is
only 8% of its continuous R(JA).
This particular part is specified as having an R(JA) of
50°C/W for intervals of 10 seconds or less.