THOR performance decay new measurements

["Marco Denicolai" <Marco.Denicolai-at-xxxxxxxxxx> (by way of Terry Fritz <twftesla-at-xxxxxxxxxx>)]

Hello all.

I made yesterday new measurements in order to find the reason of the decreased
CCPS performance after the device has warmed up.

1. Resonance capacitors
They were all the time cold: not the cause.

2. The 560 VDC rectified from the 3-phase mains and used by the CCPS
No decrease of this voltage under any conditions: not the cause.

3. Drive frequency used to pilot the switcher bridge
I noticed a drift of about 1% after the SMPS power was switched on: turning
on/off the capacitor charging didn't affect the frequency while it does affect
the performance. Not the cause.

4. The diodes composing the HV bridges (4 strings of 5 diodes each, 7 kV 3A
250-500 ns)
As reported before, they were getting hot quickly. I scoped now (with a HV
probe) the voltage across one of these strings. See pictures #11 and #12 at

http://www.saunalahti.fi/dncmrc/work/tempmeas.htm

Now I see a CLEAR difference in the conduction region if the CCPS is cold or
hot. Please, correct me if I am wrong, but an increased direct voltage (over
-40V) tells me only that the turn-on time has grown, i.e. the diode has got
slower at turn-on. In any case, this is the only difference I was able to
measure.

I am now going to replace all diodes of one board bridge (20 pcs.) with 28 pcs.
of UF5408 (as Malcolm Watts suggested, thanks Malcolm): these devices have
got a
trr of only 75 ns but they are only 1 kV (thus I am going to use 7 of them in
series). Then I'll repeat the same measurement again.

Regards

P.S. The diode string I was scoping had the anode connected to ground and the
catode connected to the transformer. I was scoping the catode: then I should
have measured only positive voltages, negatives being clipped by the string
down
to -1.7 V * 5 = -8.5V (5 times the diode voltage drop).