ltc4364 - Inrush and FET Selection

Wayne E. said:

I suspect that the 4V threshold of my FETs are too low and residual voltage on HGATE is allowing Q4 to be partially turned on.  If I probe HGATE directly with an o-scope probe, the issue goes away.  I suspect the 1M load of the probe drains off the residual voltage. 

If you hotplug the first time, inrush control work, since Vgs ramps from 0V to ~12V.
If you unplug Vin and hotplug again, before Vgs has enough time to drop from leakage, inrush control fails.

As you found with the scope probe, the solution is a bleeder resistor from GATE to ground. 1MΩ works.
Don't use too small of a resistance, since the LTC4364 only has 10uA of gate drive (min).

Can you comment on the minimum Vgs rating I should use for FET selection?

Looking at the LTC4364 datasheet, I see M1 FETs with similarly low Vgs thresholds as I chose.  I expected to see FETs with higher thresholds therefore confirming my FET selection was in error but that but that does not seem to be the case.  Am I on the right track that a FET with a higher Vgs will resolve my issue?  Will the designs in the the datasheet also fail repeated hot plug events?

In terms of adding a bleed resistor... I found that I can replace C137 with a resistor that corrects the hot plug inrush I am failing.  But I am now getting a 3A inrush spike when the LTC4364-2 turns on the FETs.  3A is below our failing threshold but not by much. 

I did find a 1Meg was too small of a resistor for my circuit.  It does not appear to let the Q3 fully turn on.  A 5.6M seems to allow the Q3 to fully turn on.  I also tested 11.2M (2 x 5.6M) which also worked.  If I move forward with this solution, I'll use a 10M. 

I am hesitant to replace C173 since it removes my inrush control.  I am also looking for a solution that wont require a board spin. 

Thanks for the help. 

Wayne E. said:

I found that I can replace C137 with a resistor that corrects the hot plug inrush I am failing. 

Don't replace C137. The ratio of Cgate to Cout determines inrush current.

Wayne E. said:

I am hesitant to replace C173 since it removes my inrush control.  I am also looking for a solution that wont require a board spin. 

You can use a through-hole resistor to bleed off the gate charge between hotplugs.

Unfortunately, a through hole resistor would not be acceptable.

Please reply to my Vgs threshold question.  Even a "I do not know" answer would be helpful. 

Wayne E. said:

Please reply to my Vgs threshold question.

Higher threshold for NFETs is preferable in hotswap applications.

I tested a FET with a higher threshold but it did not seem to help.  The NTTFS022N15MC I attempted to use has a threshold between 2.5V and 4.5V.  This was only modestly higher than the original FDMC86260 which had a range of 2.0V to 4.0V.  (I did not test the thresholds directly so it it possible their tolerances were working against me.)  This was the only FET I could find that would work with my existing footprint and had a higher threshold.

I ended up solving my issue with your recommended solution of putting a resistor in parallel with C173.  A 10M resistor and 10nF cap gave me good results.  We plan to stack the two 0603 sized components on top of each other to avoid a board spin. 

If available, please point me to an app note that discusses how to address multiple hotplug events.  Multiple hotplug events is part of DO-160 testing.   

Thank you for your help.