FET maximum temperature for short bursts of pulses

need

uses

duty

Greetings.

Okay, you've obviously got the right general idea, but it appears you are using the wrong thermal resistance figure for the above calculation. There are three different thermal resistances that need to be considered and handled differently, junction to case, junction to ambient, and case to ambient. The junction to ambient is the sum of the junction to case and case to ambient thermal resistances.

If we look at figure 3 of the datasheet:

Notice that the units on the graph are normalized Z(theta)JC. Or in other words, thermal impedance from junction to case. In other words, figure three is useful for determining how much higher temperature the junction will rise above the case temperature, but directly what the peak junction temperature will be. Then notice in the "notes" of that figure 3, that is has a formula: Peak Tj= Pdm * Z(theta)JC * R(theta)JC + Tc.

This is the formula you want to be using to find the peak junction temperature. So, in your case the formula should work something like this:

Peak Junction Temperature (deg. C) = 12W * 0.3 * 3.125 C/W + Tcase average

Note that the difference here between this and your calculation is the R(theta)JC figure is actually 3.125 C/W instead of the 100 C/W figure you used. Note that both the R(theta)JC and R(theta)JA are both specified in the datasheet.

So in other words your peak junction temperature should be about 11.25 degrees Celsius above the maximum case temperature expected. If you ran your MOSFET continuously in this pulsed 12W 10% duty cycle application, then the average power dissipation would be 1.2W. This is when you would use the R(theta)JA (or perhaps case to ambient thermal resistance, which isn't directly specified in the datasheet, but the R(theta)JA is dominated by R(theta)case to ambient in this case, so the numbers are nearly the same) to estimate case temperature. In this case with a 100 C/W thermal resistance the case temperature should be roughly 120 C higher than ambient (although the 100 C/W figure in the datasheet is a maximum figure and may be somewhat conservative depending upon how you mounted the device). So, if you ran this in an ambient of 25 deg. C, your peak junction temperature should be roughly 25C + 120C + 11.25C = 156 C. This is below the datasheet maximum of

175C for this device, although you might not want to run it continuously like this so as to have extra safety margin and so as to preserve the quality of your PCB. The PCB will probably turn brown and degrade somewhat if kept continuously at substantially high temperature (noticeably above 100 C for instance). Solder joints will also be more stressed (particularly during thermal cycling) and may fail over time.

There

Hmmm... After reading what I just wrote, it appears i wrote a whole bunch of highly confusing gobble-de-gook and quite possibly obfuscated a relatively simple concept.

Anyway, the important conclusion was to study figure three of that datasheet more closely and use the proper thermal resistance values for the formula. I'd refer you to a application note on these issues and how to figure peak Tj under pulsed conditions, but none perfectly suited immediately come to mind. MOSFET manufacturer's such as Fairchild and International Rectifier do have some application notes on thermal issues, although they may not all be exactly relevant to your situation at hand.

Can anyone please tell me how to calculate the peak temperature of a FET when it is diven with short bursts of pulses for a short period?

I have an existing circuit that is driven from a microcontroller and I need to know how hard I can drive it without blowing up the FET. The circuit uses an RFD14N05L and switches the FET to draw a 1A 1ms pulse every 10ms for

100ms from a 12V supply (i.e. 10 pulses). The FET is a TO-252 package with no heatsink (just soldered to quite a small area of copper) and does not even get warm. I want to try variations on that by increasing the pulse width and/or increasing the number of pulses, but I haven't been able to come up with any sensible calculations for the maximum temperature.

Following the normal calculations I have seen gives the following-

1A at 12V is 12W and the duty cycle factor is 0.1

Thermal resistance junction to ambient = 100 °C/W

Normalised thermal impedance (from the datasheet) is 0.3 (for 1ms pulse duty factor = 0.1)

So peak junction temperature = (12W x 100°C/W x 0.3) + 25°C = 385°C

That doesn't seem right and I know that it doesn't even warm up with those pulses.

I am assuming that the datasheet "transient thermal impedance" gives values to calculate what happens if you continuously pulse the FET (as in a typical PWM application like motor control) and that this indicates that if I kept my 1ms pulses going for long enough it would heat up the FET over the safe junction temperature.

I found a Fairchild app note (AN-7516) that indicated that a TO-220 package might have a thermal capacity of about 0.54 J/°C. Using this figure, each pulse is dissipating 12W x 1ms = 0.012J, so each pulse would heat the silicon up by 0.012 x 0.54 = 0.00648°C. If this calculation is correct I could use 100's of pulses. Is this calculation correct?

Any help for a software engineer, who's reaching the limit of his electronics knowledge, would be greatly appreciated!

- Charles

For pulsed work, C/W is not the steady state figure.

If it's a one-off, 100ms pulse train, there is no need of a heatsink, as the heat won't get outside the package in 100ms.

There are several thermal impedences to consider. For very short pulses, in the single millisecond range, no heat gets outside the silicon.

If you plot temperature rise at the junction for a given energy pulse against pulse duration, it's nowhere near linear.

This is because at very short pulses, all the energy stays inside the silicon, and only gets out after the pulse is over. As time goes on, the pulse has enough length to begin heating the plastic and metal of teh case, and as more time goes on, it can heat the heatsink.

However. As a "always safe" figure, look under safe operating area on the datasheet.

with

,

No, I should have mentioned that the FET is used in linear mode - there is a control circuit with feedback and two op-amps to get the required current and when switched on there is 12V across the FET and a 0.01 Ohm current sense resistor, so it really is 12W.

have

of

are

other

junction

is

datasheet

all

Well, it may have been gobble-de-gook, but I got something useful out of it, so thanks!