There is a current limit, implemented by sensing the current passing through the MOSFET bridge. In normal circumstances this is equal to the current drawn by the motor. Yet, the MOSFETs do fail, by going short-circuit (which in turn destroys the current sensing resistor). Somehow, the FETs are getting blown up. This can be a) excessive voltage b) excessive current c) Safe Operating Area (SOA) limits are exceeded Regarding a), there could be transients on the MOSFET bridge supply rail. The IRF9540/IRFI540 FETs have a maximum voltage of 100V. This doesn't even meet the ISO design specification for 24V vehicles (remember the 24V aircraft electrics are basically 24V commercial vehicle electrics) which calls for 150V transients lasting 150ms. I don't have the spec to hand though I have designed to it in the past; it is well known to car electronics designers. For 12V vehicles the max transients are 60V. However, on the TB20 the Master AP switch disconnects the servo power and this switch is always OFF during starting and shutdown. Also, the servo contains rudimentary overvoltage protection in the form of a 50V clamp (on input only, not on the bridge itself) and some fairly big capacitors, which should suffice. Regarding b) I think this is unlikely since the FETs will withstand 11A with just a thermal/SOA limit, and 100A pulsed. Regarding c), there are several possibilities: c1) Under some circumstances (perhaps a specific brush condition) the motor presents an unusual load and this exceeds the SOA limits of the FETs. c2) A problem further back the circuit, which turns on both FETs (the two FETs making up each vertical leg of the bridge) together thus shorting the supply rail. This type of drive failure rapidly destroys any MOSFET bridge unless it's designed for it. The circuit which generates the complementary pulses uses quite complex components and could be misbehaving under unusual circumstances; for example during EMC conditions (ground based radar inducing transients into wiring). This eventuality should still be covered by the current limiting circuit IF it acts fast enough, but the FETs would be exposed to a very high current, with a much greater impact on their SOA limits than anything coming from the motor. There is however a defect in the design there: the 2N2222 base resistors (R23,R24) are too low in value (they are 28R, should be at least 1k) and if both FETs were turned on there would be about 1A flowing into the B-E junction of the 2N2222 which would blow it up (or degrade it) pretty fast, rendering the current limit circuit permanently ineffective. I would have used much bigger FETs (a very easy and cheap fix), and fixed the current limit so it works properly and within component limits even when the FETs get incorrectly turned on, in any combination. *********** I've just thought of another way to blow up the KS271C roll servo. The current limit circuit is no good. It isn't the pulse-by-pulse circuit used in normal switching-mode power supplies (but it should be). It just holds the MOSFET Vgs at the point where the volt drop across the current sensing resistor is about 0.6V; at this point the MOSFET could be dissipating power of the order of 120 watts (depending on the voltage across it) which will rapidly destroy it. There are several schematics in the KS271C manual 002-09835-0000 002-09656-02 002-09087-0000 002-09366-0000 which vary in detail but all use the same current limit principle. Let's say the duty cycle to the motor is 40% then if there is a motor fault the current limit will be active 40% of the time. This will be enough to destroy the FET. If this was done like in any normal switching power supply, the current limit circuit would cut the drive to the FET off *completely*, and not restore it until the start of the next PWM cycle. The question of course is what could cause the current limit condition in the first place. *********** The final instalment is this: The current limit circuit is a negative feedback arrangement and given the absence of any frequency compensation it may be unstable. In the event of instability, the oscillation will be in the MHz region and this together with the motor inductance will create havoc. Such a condition, even if it lasted microseconds, would cause a massive amount of RFI to be emitted internally to the servo, upsetting the operation of the op-amps and other low current components, and very possibly flip the circuit into a condition from which it doesn't recover. Has anyone characterised the closed loop behaviour of the current limit circuit, with various types of motor load? I bet not.