AnsweredAssumed Answered

LTC4368 weak gate drive

Question asked by aydc on May 27, 2018
Latest reply on Jun 8, 2018 by Manpreet

I have implemented the LTC4368 with the following configuration:

VIN(MIN) = 40V

VIN(MAX) = 52V

OCP = 5A

 

Nominal input voltage is 48V. This is part of a larger system, however all that is currently populated on the PCBA is what is shown in the attached schematic.

UVLO/OVP works as intended. However the gate drive voltage in the valid band is extremely weak; only around 2.4V. When Q13 (N-FET pair) is unfitted, GATE drives to about 12V in the valid band, which agrees with the 10-13.1V range indicated in the datasheet for 12V - 60V VIN. Removing C71 has no apparent effect on drive strength.

 

Interestingly, when this circuit is run in LTspice as drawn, the UVLO threshold doesn't comply. In fact, the FETs don't turn on anywhere across an applied 0-55V ramp input. I don't understand how, as the UV pin threshold exceeds 0.5V at ~40V in. However, when R85 and R87 are swapped, the FETs turn on (albeit at a slightly lower 37V than the programmed 40V) and correctly turn off at the OVP threshold of 52V. Even more intriguingly, implementing the typical application circuit on the cover of the datasheet in LTspice also doesn't work. I don't know which to believe at this point. On one hand, the gate drive potentials are as expected in the model (when the FETs are operating); on the other, the real world circuit behaves correctly with UV/OV thresholds but has a weak gate drive. The FETs used in the model (FDS5690) aren't the same as what's used on the actual design (SQJ974EP), but their max input capacitances and gate charges are very similar (1107pF/1050pF, 32nC/30nC) so I doubt this is the cause of the discrepancy. I do find it a little concerning that the model doesn't seem to work correctly.

 

Furthermore, it seems to have killed a few MOSFETs already, with their gate source resistances ranging from 500R-200k after being used in the circuit. I have already replaced the 4368 with another in case the first was a dud only to have the same operating behaviour on the replacement.There must be something simple that's wrong here, but it's about as close to the typical app note as you can get as far as I can see. Has anyone used this part with success?

Attachments

Outcomes