ARCHIVE OF REPAIR TRICKS, USEFUL INFO, ETC.

• Elba power supply in Bruker BLAX / BLARH
• Elba schematics found in databooks
• Fried 150ohm resistors in Bruker BLMX001 (B-LMX1, BLMX1 ?)
• Bad jacks in TI-83 calculators
• Spectra-Physics 2018-RM Laser
• Applied Kilovolts K9174 dual HV supply from mass spec
• Edwards Vacuum Valve PV25EK, IV25EK, blows fuse

Rather than paying the $1500, I've managed to repair a couple of these.  
Want some info?

First note that these "Elba" 220VAC supplies WILL RUN on 120VAC.  This 
makes testing easier.  But I wouldn't try drawing a high wattage load 
when the line voltage is so low.

Also note that you need to place a small load on the +15v, -15v, +5v 
supplies.  If you run them with no load at all, the switching-supply 
stops working and the output voltages will start falling, yet the main
+28V supply runs just fine.

On the first supply I repaired, the main bridge rectifier and the big 
APT5025 FETs for the 365VDC Power Factor regulation were fried.  These 
components sit right on the 220VAC line, so any major surge can kill 
them.  The diode bridge is a weird thin little thing under the main PCB, 
part number D20XB60, availible from Mouser Electronics, (600V 35amp bridge.)  
The MOSFETs were APT5025BN from Advanced Power Technology, advancedpower.com/ 
http://www.advancedpower.com//communities/apt/products/5025BN.PDF
I found an actual APT5025BN, but probably a similar transistor would 
work, if you can find a TO-247 package for 500V 23A 0.25ohm 300watt,
N-channel, gate threshold max 4v (such as IXFH24N50 or IRFP360
from DigiKey.)

The second dead unit also had a bad NPN transistor from ZETEX which, if 
I recall right, drove the gates of the main switching MOSFETS driving the 
220AC to 365VDC switcher.  The MOSFETS were dead, as was the UC3854N 
power-factor IC on the daughter board next to the 270uF 400V capacitors.  
(See below for a schematic that's similar to the daughter board PF-
correction circuit.) No doubt a surge on the 220VAC line shorted out one 
component, and the large current destroyed everything else.

A third unit had a stalled fan.  The heat had killed the two big 270uF
400V electrolytics.  Also, the values of tiny electrolytics were all wrong
on the power-factor daughter board (the board next to the 270uF 400v 
caps.)  I found I could replace all these tiny capacitors without having 
to unsolder the many pins of the daughter board.  Be sure to mark the 
polarity, since there's no silkscreen plus-signs on that PCB.  These
capacitors are part of a tiny high-freq switching power supply that runs 
components there, so high temperatures will bake out these capacitors and 
make the supply fail during a power-up.  (Apparently the Elba supply can
run almost forever when it's fan is dead, but the extra heat slowly ruins
the capacitors critical for power-up sequence.)  I replaced the two 47uF 50v 
(a volt doubler) that runs the power factor chip, the 100uF 50V cap on the 
corner next to the power transistor, and the two 22uF 35V caps next to 
the fan connector (they're essential for the fan's power supply.)


Now our fourth dead unit was very interesting.

There is an apparent design weakness in the 30V section of the "Elba" power 
supply used in some versions of the BLAX 300 and similar amps.  When the supply 
first comes up, the 30VDC switching regulator chip (UC3825) on the second
daughter board needs at least 9V to wake up and start making DC.  In normal 
operation it creates it's own supply voltage from its own 30VDC regulated 
output... it pulls itself into the sky by its own bootstraps!  But during 
startup, it temporarily needs another supply.  It gets this from a little 
"bootstrap power supply winding" on that big iron tapewound inductor in the 
center of the PCB which is part of the power-factor switching circuit.  
This winding is voltage-doubled with two diodes and three capacitors (two 10uF, 
one 220uF, partly hidden under a transformer,) making 20VDC, which supplies 
a LM7812 regulator, which supposedly puts out 12VDC for the UC3825 chip.  But 
the UC3825 draws 33mA, which can drag the 12v supply voltage of the 
bootstrap/LM7812 down to 8.9V...   and sometimes the UC3825 goes into continuous 
repeating reset and never starts.  It's waiting for its bootstrap-supply to rise 
up above 9.0V.  This all depends on temperature and on many component values.   
This is possibly bad design?  They should have given plenty of leeway (like 
designing it to actually put out 12VDC, even when it draws 33mA as it does.)

Regarding the idea that heat can kill these supplies...  yep, if any of the
three electrolytic capacitors next to the LM7812 regulator in the voltage-doubler 
bootstrap supply should get baked out, so their capacitance value drops, or their 
if internal leakage gets large...   then these changes will push things over 
the edge.  The UC3825 on the daughter board never gets its 12VDC, and the 
supply cannot wake itself up anymore.

      The usual symptom: your system was running for many days or weeks, 
      but then after being switched off, or after a building power failure, 
      it won't wake back up again.

(But sometimes, if you switch it off and back on more than once, it will 
"catch," and start running.)  Our dead supply had a 220uF 50V electrolytic 
capacitor (next to the LM7812) which had changed itself to 20uF over the years.  
When replaced, the supply worked fine again.

So note well that these Bruker amps often die after a power failure, but very
often THIS ISN'T CAUSED BY A SURGE.   Instead, the power supply malfunctioned
weeks before, but the problem remained invisible as long as the amp remained
in operation.  If the AC power should drop for a second or two, then the power
supply goes to sleep and will never wake up until the capacitors are replaced.



Also about ELBA power supply schematic:

I discovered that the schematic for the front end, the 220VAC input section,
is very similar to the schematic shown in the following app. note for the 
UC3854 chip used in the supply's power-factor correction daughter-board.  
It uses a big MOSFET and an inductor to massage any line input voltage (50VAC - 
270VAC,) and it supplies 400 volts DC (365V) to the rest of the board.   The 
400Vdc is later switched as a 200KHz squarewave and applied to the ferrite 
stepdown transformer to make 28VDC.  Schematic:

  Advanced Power Factor Correction Control ICs (n.b. schematic on page 3)
  http://focus.ti.com/lit/an/slua177/slua177.pdf

Note that this power supply ***WILL RUN*** on 120vac, at least for testing.
Supposedly it up-converts any AC line voltage between approx. 60VAC and 
250VAC.   However, I wouldn't leave the whole NMR amp running on 120VAC, 
since the supply is probably out of spec for power factor, and might
overheat during a 600Watts load.

Other schematics for your reference:

  THE UC3823A,B AND UC3825A,B 
  ENHANCED GENERATION OF PWM CONTROLLERS
  http://focus.ti.com/lit/an/slua125/slua125.pdf

  UC3854 Controlled Power Factor Correction Circuit Design
  http://focus.ti.com/lit/an/slua144/slua144.pdf

  UC3854 provides power limiting with sinusoidal line current
  http://focus.ti.com/lit/an/slua196a/slua196a.pdf

  Optimizing UC3854 performance
  http://focus.ti.com/lit/an/slua172/slua172.pdf


Also see messages:


  question about Bruker amps and their power supply modules
  http://chemnmr.colorado.edu/ammrl/archives/February-2004/13.html

  question about Bruker amps and their power supply modules--Summary
  http://chemnmr.colorado.edu/ammrl/archives/February-2004/17.html

  Bruker BLAX/H Power Supply Cooling Fan
  http://chemnmr.colorado.edu/ammrl/archives/June-2001/7.html



Bill Beaty (beaty atsign chem washington e d u)
Mon, 10 May 2004 16:07:47 -0700

More news!

The design weakness in the BLAX Elba power supply is NOT in the designed
capacitor values as I thought.  But the problem is nearby:  the LM7812 
overheats the capacitors, causing them to slowly decrease in value over time, 
so the supply fails early.  There's a 10V Zener diode which seems to be the 
origin of the problem, and it might help things if we change it to a smaller 
value ( such as 4.3V 1N4731.)  This is not essential.  Instead just replace the 
three bad capacitors to get things up and running again.)

The overheating of the capacitors occurs because the little voltage doubler 
on the big series inductor winding (the two diodes and three capacitors) is 
only supposed to supply its 20Vdc to the LM7812 for a couple of seconds; 
during power-up until the main +30Vdc comes on line.  The output from this 
voltage doubler is passed through a diode, as is the main 30Vdc output.  Both 
are applied to the LM7812 regulator input pin, and whichever is higher, that 
one powers the regulator.  This lets the bootstrap supply send current to the
LM7812 during power up, then after a couple of seconds the main +28V takes 
over from the volt-doubler when it later wakes up.

But unfortunately Elba has put a 10Vdc Zener diode in series with the 30Vdc 
output to drop it down to 20V (no doubt because the LM7812 without 
heat sink gets quite hot when given 30V input, and the series zener shares 
some of the thermal wattage.)

So the little volt-doubler AND the main 30V are BOTH set to approx. 20V, and 
if you happen to be unlucky and have just the wrong circuit values, then 
BOTH ARE ALWAYS POWERING THE LM7812.  Or perhaps the voltage doubler "wins" 
and becomes the main supply for the UC3285 on that daughter board.   This is 
bad news for the capacitors in the voltage doubler, since they normally see 
two-ampere pulses at around 50KHz, and they will run fairly hot.  Over the 
months and years they get baked out, their values decrease, their 20Vdc 
output voltage decreases, and finally the voltage falls below the 9.0Vdc 
required by the UC3825 main 30Vdc regulator chip on the second daughter board.  

It also doesn't help that the capacitors are right up against the very hot 
LM7812 regulator; and that might even be the real trouble here after all.  But 
regardless, the temporary bootstrap power supply voltage gets too low, and the 
system gets flakey during power-up and can't wake up every time.  However, if 
it's ALREADY running, the bootstrap supply is not critical for operation, and 
the system will run fine...  as long as you never turn it off!

:)

Or, if your BLAX or BLARH apparently dies right after a power-off, be aware 
that in some cases the bootstrap supply is still VERY close to the correct 
voltage.  Try perhaps turning it off and on a couple of times (with luck 
it may "catch" and start working.)

The cure we used (your milage may vary!):  replace the two 10uF and the 
one 220uF capacitors (they're all glued together, positioned next to that 
LM7812 voltage regulator approximately in the middle of the main board.)   
Replace them with low-ESR, 105degC electrolytic capacitors.   But that 
doesn't fix the real problem.  So also look for a chain of resistors right 
at the edge of the main board (labeled DZ1, DZ2, DZ3.)  One is a 10V zener 
diode, the others are zero-ohm jumpers.  I replaced the 10V zener diode with 
a one-watt 4.3V zener.   This lets the poor little voltage-doubler circuit 
turn off when it's not needed.   But it makes the LM7812 regulator run even 
hotter than before.  If I see another one of these dead supplies, I think I'll 
also be putting a couple of little bitty TO-220 heatsinks on the LM7812 regulator 
(HS214-ND from digikey.)

• Message Archive: AMMRL Magnetic resonance lab managers' list
• My first "elba" msg on AMMRL
• AMMRL: more "elba" info

T. Pratum (pratum@u.washington.edu)
Tue, 21 Jul 1998 16:58:54 -0700 (PDT)


We have had 2 blax300rs failures in the past 6 months, and one of our
electronics engineers has found the same problem in both cases- the
MSA1023 amplifier in the blmx001 failed due to a failure of the bias
resistors R16,17,and R9 (150 ohm 2 watt composition). Their resistance
appeared to have decreased over time (in at least one case down to 10
ohms). I have checked 2 other blax300rs units that were delivered at the
same time (approx June 1995) and found the same problem in each of them
(although the MSA1023 hadn't failed yet). I also checked 2 other blax300rs
units which were delivered earlier (in 1994) and found no such problems.
So, my warning is that if you have a blax300rs that was delivered in 1995
it might have some bad resistors in it which will eventually stress the
MSA1023 into failure. You can check them in-situ with an ohm meter after
opening the unit up- they are quite obvious and the 3 in parallel should
have a resistance of 50 ohms (approx.). If anyone is interested in any
further information, please let me know.


Tom Pratum
Dept of Chemistry
Box 351700
Univ of Washington
pratum@u.washington.edu  (old addr, No longer at UW NMR)
List archive: Bruker Users Mail (BUM), for 1998 Archive

The UV Laser in our mass-spectrometer department had a strange failure: blown supply fuses, and a huge splotch of golden mirror coating on the PCB connected to all the large power components. Somewhere there was a direct short across the main power line on the DC side. Two metal terminals had touched together between PCBs in the supply, and the point of contact was explosively vaporised, coating everything around it with metallic copper!

The failure apparently came from a MOV device which had been crushed between two circuit boards, where a big screw with DC main power on it had pushed through the epoxy coating on the MOV. This placed a short on the output of the main 3-phase bridge rectifier, which blew out one of its diodes before the 50-amp cartridge fuses blew.

I found that Spectra Physics apparently was expecting this failure, since they sell a repair kit: the bridge module plus cables, along with screws which are shorter than the original ones. The 50-amp cartridge fuses can be had from Mouser. (and if Spectra Physics should discontinue their kit, the three-phase bridge is similar to a known part available elsewhere):

Bridge Rectifier Upgrade Kit, 4801-1108UPG, Spectra Physics
Bridge Rectifier, 3-phase, EH80, EH100, Microsemi.com
NON-50 cartridge fuse, 504-NON-50 mouser.com

After replacing the bridge, the laser still didn't work: it ran at full power (50 amps,) and the current-adjust pot on the remote console box had no effect. Occasionally it went into overcurrent shutdown. I traced this to a shorted zener. Besides killing the bridge module, the transient had shorted out CR-23 on the logic/driver board, a 1N4735 6.2V zener associated with the main series regulator (it's on a transistor driving the base pin on one of the big darlingtons in series with the main 50amp supply.)

One last note. I had one of the big darlingtons removed from the water-cooling heat sink and noticed that the heat sink has large holes allowing direct water contact with the bottom plates of the transistor bricks. If you should try removing one of the bricks while the cooling water pump is running, you might get a big wet suprise!

I've now seen two of these supply modules with the same problem. One side of the dual HV supply has an incorrect voltage with large noise signal. Is the noise from 60Hz, or from switching-freq leakage? Nope, it's a varying-frequency spike signal ...caused by arcing!

Unfortunately the pinout is unlike any of the others on the http://www.appliedkilovolts.com website. To operate the supply I traced through the circuitry to find the +24v and ground pins on the connector. The four linked pins at one end of the 32-pin connector are ground, while the two linked pins near the other end (one position in from that end) are the +24v input. There is also an Enable pin which normally has +5 while floating, and needs to be grounded to run the supply. If pins 16 and 15 are the supply ground, while pin 2 is +24, then Enable is pin 7. Apply a +24DC supply at about 1/3 amp, ground the enable pin, and HV will be present on the two output cables.

The failure I observed was from arcing caused by a weak point in their design. It's located right where the ground braid of the HV cable is twisted around the center conductor in the cable (right near where the cables dive into the silicone-embedded HV section.) Apparently the voltage rating of their HV cable is too low, so the strong e-field caused by the edge of the tightly-twisted ground braid will create corona discharge which eventually eats its way through the thick polyethelene insulation of the center conductor. Eventually a continuous arc punches through and burns a slot in the insulation.

The cure is easy: Unscrew the single screw which holds the HV cables' ground braid lugs, carefully cut off about an inch of the black outer jacket of the HV cable, then untwist the ground braid and open it out so it's not bound tightly around the center conductor. You'll find a black-edged hole burned through the white polyethelene insulation there. I ran a drop of cyanoacrylate crazy glue to fill the hole, let it harden, then ladled in lots of RTV silicone caulk around the failure point and between the ground braid and the center conductor. (Avoid making bubbles which trigger arcing!) This keeps air away, and keeps the ground braid smoothly flared out so no high-field spots exist. I didn't even have to cut the ground lugs off, just screwed them back into position.

More than one of these high-vacuum solenoid valves has now had the same problem. These valve contain two DC coils: one connects to points numbered 1 and 2, and is pulsed at high current to trigger the valve coil. The other is connected to points 1 and 3, and is continuous at low current to hold the iron core in position.

Rather than pulsing on for 50mS as desighend, the driver turns the pulse coil on full blast, which instantly blows the internal fuse. In both dead units the problem was the same: the small NPN transistor in TO-92 case was shorted, and the 220K resistor in series with the 22V zener diode was open. I replaced the transistor with a 2N2222A, and replaced the 220K with an old half-watt carbon resistor. Since these valves spend their whole lives in the turned-on state, probably the failure was caused by line spikes on the 220VAC supply. But the 220K was a film type resistance element, and the high voltage might have slowly degraded the film and lowered the resistance until it fried the transistor and itself.

Notes: the coil driver is a fullwave bridge with two SCRs as the diodes on the positive side of the bridge. A second pair of diodes supplies the +310VDC for the holding coil and for the rest of the driver circuit. A diac in series with a 120K resistor with .047uF capacitor is used to trigger the SCR gates, and the NPN transistor pulls the diac signal down in order to turn off the coil after the pulse time has completed. When power is first applied, the NPN transistor is off, so the SCRs turn the coil on. Then the tantalum capacitor charges up through the 220K towards 31VDC, until the 22V zener diode turns on and raises the NPN base voltage, turning on the transistor and pulling down the SCR gates to zero. The coil in the smaller unit sees 400 watts, and 1000 watts in the larger unit. So obviously the driver needs to turn these coils off after just a few AC cycles have passed.

Fairly old TI-83 calculators give a communication error because the internal jack in the TI-83 dies from old age (especially it wears out with heavy classroom use.) The tiny gold leaf-springs get crushed out of the way so they no longer make contact. Sometimes you can open the case and use a needle to bend them back again, but this doesn't always work. And in an education lab environment, they will quickly go bad again.

For an electronics person, the connector is not that difficult to solder in on their own. However, only a somewhat-similar part is available. I've never found a source for the original gold-plated surface-mount 2.5mm stereo mini jack. Digi-key sells a useable replacement, their number CP-2523SJ-ND, but one pad is in the wrong place, and the tiny posts on the bottom will not match the holes in the calculator circuit board.

WARNING: TEXAS INSTRUMENTS WANTS YOU TO SHIP THESE BACK TO THE FACTORY FOR FREE REPAIR!!! If you want to break into these yourself, you take responsibility for possibly messing up the electronics. WARNING: OPENING THE CASE WILL DISCONNECT THE MEMORY BACKUP BATTERY AND WIPE OUT ALL STORED SOFTWARE.

To get inside the thing, the screws require a #6 spline screwdriver such as Xcelite 99-62. Fortunately a 1.5mm hex (allen) screwdriver works great, and probably you can even use a 0.050" hex wrench if you tilt it so the points grip the inside of the screw.

Before soldering in the new jack, slice the tiny pins off the bottom. You might wish to affix the jack to the board with a bit of cyanoacrylate glue before soldering. The rear terminal on the new jack does not match the PCB pad, and it also touches against the PCB ground. I bend these terminals upwards out of contact with the PCB, then solder a piece of resistor-lead between the PCB pad and the terminal. When reassembling the calculator, make sure all the keys line up with the holes in the case before replacing all the screws.

Note that these calculators will "forget" their LCD brightness setting. Don't panic if the display seems blank when you reassemble the unit. Hit the ON switch, and look for the phrase "Mem Cleared" on the display. Then type [2ND][UPARROW][2ND][UPARROW] over and over to find the correct brightness setting.

The same Digi-Key part can be used to repair bad jacks in Vernier LabPro units, but the larger size of the new jack makes the fit even more of a problem. If you really want to repair your own LabPro using the Digi-key part, you'll have to take a chance and do some slight case-modifications.

• TI website: tech forums: TI-83
• TI calculator tech support
• TI Graphing Calculator FAQ
• TI support phone number: 1-800-TI-CARES