Maker Pro
Maker Pro

Cary 5000 spectrometer, repeated high voltage fail, Varian, Agilent

W

William Beaty

Jan 1, 1970
0
HIGH VOLTAGE SHORT IN CARY 5000 SPECTROMETER


TWICE! Our UV/VIS/NIR spectrometer lost its -1KV high volt source, and thecause both times turned out to be an internal short on the PMT-control PCB.. Both times this fault was easily found with a DVM: an unexpected leakagepath, well below 50Kohms, seen between the 1KV source and ground. The shorted HV trace and the groundplane are both internal layers in the 4-layer circuit board. Verify: when the 4-pin molex cable connecting the main PCB to the PMT-control PCB was removed, the -1KV at the main board reappeared again (actually measured as -980VDC when working correctly.)

The PCB with the actual problem is positioned vertically all the way to theright of the instrument. The -1KV switching supply is on the main board covering the left rear of the instrument. The HV cable is the cable with the 4-pin Molex connector, with one pin having a heavy red HV conductor.

The PCB design unwisely routes a 1000vdc power supply trace across an internal ground plane. Given enough years, the HV slowly eats through the paper-thin epoxy-glass layer and arcs to ground. Perhaps the same fault will occur with many other units?

The cure is to route the small 1KV supply trace across the PCB using a separate wire. But first you must use a Dremel tool to slice away the carbonized paths connected between the groundplane and the original 1KV trace that's found on an internal PCB layer. The resistance to ground for the HV signal should be around 3.4Meg, not 12K ohms! We had one carbonized short appear at the spot where that trace led away from the 4-pin Molex. Later another carbon path arose at the via hole near the end of the chain of white H11D1 optoisolators. While looking for the internal short I also drilled out the connection to the hot end of the orange disk capacitor, so used white RTV silicone so that cap would still be anchored mechanically by more than just its one remaining lead.

SEE: http://staff.washington.edu/wbeaty/cary1.jpg

The white arrow in the JPG above shows the added wiring. Another wire appears on the circuit side of the PCB between the orange capacitor and the 4-pin molex seen just above the white arrow.

The earlier failure also fried a couple of SMT power resistors in the switching supply on the instrument main PCB. I replaced these with 3watt through-hole versions.
 
C

Cydrome Leader

Jan 1, 1970
0
SPECTROMETER > > > TWICE! Our UV/VIS/NIR spectrometer lost its -1KV high
volt source, and the cause both times turned out to be an internal short
on the PMT-control PCB. Both times this fault was easily found with a
DVM: an unexpected leakage path, well below 50Kohms, seen between the 1KV
source and ground. The shorted HV trace and the groundplane are both
internal layers in the 4-layer circuit board. Verify: when the 4-pin
molex cable connecting the main PCB to the PMT-control PCB was removed,
the -1KV at the main board reappeared again (actually measured as -980VDC
when working correctly.) > > The PCB with the actual problem is
positioned vertically all the way to the right of the instrument. The
-1KV switching supply is on the main board covering the left rear of the
instrument. The HV cable is the cable with the 4-pin Molex connector,
with one pin having a heavy red HV conductor. > > The PCB design unwisely
routes a 1000vdc power supply trace across an internal ground plane.

Did they just forget to clear out the ground plane in that section? In
your photos it looks like they randomly took a paper punch to internal
planes by the optoisolator looking devices. weird.
 
P

Phil Allison

Jan 1, 1970
0
"William Beaty"

HIGH VOLTAGE SHORT IN CARY 5000 SPECTROMETER

( snip)
The PCB design unwisely routes a 1000vdc power supply trace across an
internal ground plane.


** That is quite an understatement.

Given enough years, the HV slowly eats through the paper-thin epoxy-glass
layer and arcs to ground. Perhaps the same fault will occur with many
other units?

** Guaranteed to.
The cure is to route the small 1KV supply trace across the PCB using a
separate wire.

** Yep.

Found myself doing the same once with a number of high power audio
amplifiers where the PCB had adjacent tracks, separated by only 0.4mm, that
carried +/- 120VDC. The fibreglass PCB was NOT coated with resist and dust,
grime and moisture settled on the PCB in normal use.

End result was "tracking" across the surface between the close spaced
conductors and eventual breakdown - leading to the output stage going into
full cross conduction, blowing fuses AND expensive Hitachi TO3 mosfets to
hell.

Part of the repair was to peel off the offending tracks and replace them
with ( light gauge) PVC coated wire.

If caught early enough, thorough cleaning of the PCB and spray coating with
clear "circuit board lacquer" prevented the disaster.



..... Phil
 
W

William Beaty

Jan 1, 1970
0
Did they just forget to clear out the ground plane in that section?

Looks like. I gave it some backlighting so the problem would be obvious.

They must have done some copper hold off clearance, but only for the vias and the various pins, but not for the trace itself.

MANY bad spacings here of course, but after a few years, only the adjacent planes have eaten through. Give it a few more years, see if the progressive HV rot will continue.

Usually I'm working on e-beam driver etc. boards with 5KV supplies. In that case the designers router out some actual 1mm slots between adjacent circuit sections, to interrupt any progressive growth of conductive material. "World's slowest lightning bolt," where a sharp carbon filament is converting epoxy at its tip into more charcoal.

((((((((((((((((((((((( ( ( (o) ) ) )))))))))))))))))))))))
William J. Beaty Research Engineer
beaty a chem washington edu UW Chem Dept, Bagley Hall RM74
billb a eskimo com Box 351700, Seattle, WA 98195-1700
campus x 3-6195 http//staff.washington.edu/wbeaty/
 
C

Cydrome Leader

Jan 1, 1970
0
William Beaty said:
Looks like. I gave it some backlighting so the problem would be obvious.

They must have done some copper hold off clearance, but only for the vias and the various pins, but not for the trace itself.

MANY bad spacings here of course, but after a few years, only the adjacent planes have eaten through. Give it a few more years, see if the progressive HV rot will continue.

Usually I'm working on e-beam driver etc. boards with 5KV supplies. In that case the designers router out some actual 1mm slots between adjacent circuit sections, to interrupt any progressive growth of conductive material. "World's slowest lightning bolt," where a sharp carbon filament is converting epoxy at its tip into more charcoal.

I had to send back a batch of HV (12KV I think) rectifier boards they had
to have the sharpest most snaggly solder and lead trim job possible. Just
holding the board in your palm would make it tear into your skin like
razor wire. I'd have just blob soldered over them, but no doubt with that
level of attention to detail during assembly there would have been
backwards or dud components on the board and messing with it would have
voided any exchange rights if it was bad.
 
Top