LM5116: Question on powering up/down with Vccx and low UVLO

We have a supply using the LM5116 (final target is the LM5116-HT).

Basic specs:

Vac:                 70Vrms

Vrectified:        100Vdc

Vin (chip):        70Vdc

Vout:                60Vdc

Vccx:               11Vdc

Vuvlo:              50Vdc

Iout:                 Imin = 30mA through 2.15K resistance on board,
                        Imax = 8 amps

Fsw:                   120KHz

Capacitance:

Vrectified:        3553.7uF/200V

Vin:                  8398.9uF/100V

Vout:                2199.6uF/100V or more

Input to the board is 70Vac. The rectified AC (Vrec) is run through a series pass voltage regulator/limiter adjusted for 70Vdc feed to the switcher (Vin). Below 70V, Vin tracks the Vrec. This regulator/limiter feedback control is powered from Vccx.

Vccx is derived from Vrec through a small series pass regulator set for 11V. Full Vccx level occurs when Vrec reaches about 20V. Output of the Vccx regulator is directly connected to the Vcxx pin of the LM5116.

Power on sequence:

Vrec = 20V          Vccx=11V, fully up

Vin = 50V            Vuvlo is reached and LM5116 comes alive. Running max duty cycle. Output is present but less the Vout

Vin = 63V            At about 63V, Vout reaches the spec’d level and the LM5116 starts to regulate the output

Vin = 70V            With only the on-board load of about 30mA (no additional external load) the duty cycle is about 33%. LM5116 at about 85% D.C. (1 amp load).

Power off sequence:

Vout = 70V         LM5116 at about 85% D.C. (1 amp load)

Vin = 63V            At about 63V, the LM5116 starts to run at max D.C. and drops out of regulation

Vin = 50V            Vuvlo is reached and LM5116 switches off.

Vrec = 20V          Vccx starts to drop down.

Due to the large input capacitance, it takes about 45secs or more to bleed Vin and Vrec fully down.

Issue:

We can power cycle the board with no load (except for the on board 30mA)

Under light load of 1amp the chip will blow. Have not tested at higher loads but assume it will be the same.

Pin checking the LM5116 will show either the HO pin (18) shorted to SW (20) or the LO pin (15) shorted to PGND (14). Output FETs are OK. Replacing the LM5116 brings the supply back up.

I need to get an updated schematic or make a mark-up to post.

Questions:

1. Are there any issues with the Vccx being fully present as Vin comes up? Are there advantages to having Vccx track Vin or Vo during power up/down?
2. Are there any issues with the Vuvlo being set so much lower than the Vo level?
3. Are there any advantages/disadvantages to a significantly longer hiccup time (larger cap on UVLO) and/or a longer soft start time? (I am thinking 1-2sec verses the 300’ish msec we currently have.)