48V Linear supply series pass element replacement (BUK436-100A)

[nostril]

Hi all, I've diagnosed a linear 48V@6A power supply as having shorted one of the three N-MOSFETs forming its series pass element. Being from the early-to-mid 1990s, it uses a now-obsolete transistor, BUK436-100A from Philips.

I've removed the shorted MOSFET and the supply is correctly regulating with its two remaining transistors driving a light load. To replace the blown transistor, I am considering finding a modern replacement from Digikey/Mouser etc, but I imagine I need to match transfer characteristics, gate capacitance etc of the replacement to the remaining two transistors as well as the obvious SOA stuff. Is this correct? How closely would I have to match the transfer characteristics when there are load sharing resistors in place?

Alternatively, if I replace all three transistors with a modern part, how closely would I have to match new with old? Is it likely that the control circuit will have been tuned such that instability/overshoot/undershoot would be meaningfully increased without careful attention? Is there anything else I should be considering?

Schematic of the series pass board attached, still working on schematic for the control board.

Thanks

[Weston]

Three characteristics would be important in finding a replacement transistor, the SOA, threshold voltage, and the transconductance. The loop bandwidth for a linear power supply is most likely going to be low enough where the capacitance's of the MOSFET will not impact any of the major poles/zeros that are important for control loop stability, so unless there is an order of magnitude difference you can probably ignore matching that.

At a maximum current of 6A the sharing resistor is only going to drop 0.44V per part, and less that that at lower currents. The threshold voltage is not typically specified this tightly in datasheets. Ideally the three transistors would be the same part number from the same manufacturing lot. It would be hard to get a replacement part matching the threshold voltage within the ~0.2V or so required to effectively share power given the value of the current sharing resistor.

Loop stability is going to be impacted by transconductance, which determines the gain (ratio of drain current to gate voltage). In order to switch faster, modern MOSFETs often have a higher transconductance. Even if the threshold voltage of a replacement part was the same it may not share current equally due to a different transconductance. Transconductance is often given at a single current which can be different between datasheets, but scales with the square root of drain current, so parts can be compared. The current sharing resistors provide negative feedback and reduce the impact of the transconductance.  As transconductance increases with drain current any stability issues will happen at the maximum load current.


Due to variations in threshold voltage I would recommend replacing all three transistors together, ideally with parts from the same lot. Control loops are generally tuned with enough phase margin where even a multiplicative increase in the loop gain should still be stable (notice that the BUK436-100A does not even specify a maximum transconductance!).

I recently fixed an Agilent linear power supply that used a IRFP150N so I am fimilar with the part. Looking at the datasheet it has almost the same transconductance as the BUK436-100A. https://www.digikey.com/en/products/detail/infineon-technologies/IRFP150NPBF/811534 

It seems like the part should work in your application. However, it technically does not have a DC SOA curve.

[nostril]

Thanks @Weston for the detailed advice, and especially the example part number that fits the role fairly well. It has slightly narrower Vgs, but ±20V should probably still be fine for this application given the zener clamp at around 16V. I'll do a search to check I'm not missing any part that is a slightly better fit for the price, otherwise I'll go with the IRFP150N.

It looks like the transistors would be run so far within their SOA that even if there is no official DC SOA from the datasheet it'll probably work out fine. The unloaded unregulated voltage on this supply is approximately 60V and there is foldback current limiting in case the regulated output is yanked below 48V.

[nostril]

A final follow-up for anyone finding this thread in future. The IRFP150N are excellent replacements in practice for the BUK436-100A in this power supply. I replaced all three and load tested the power supply afterwards, and it performs as well as can be expected.

Thanks again