Interesting transistor failure (long, semi-OT)

[WB3JOK]

kind of OT, also posted on Tekscopes2, but I figured it might be of interest to some of us soldering-iron jockeys, maybe the Class E types  

This is a PP-7482/G Rockwell Collins 1 kva 115v 400 Hz power supply aka inverter.I bought it in this "reparable" condition fromFair Radio about 10 years ago!



Even though the very ordinary design is from the 70's and you can buy identical inverters from Communist China for at least ten if not twenty years now, the USAF <sarcasm> in its infinite wisdom </sarcasm> won't release the schematic/repair manual because it has "military applications". A fellow Boatanchorite actually tried FOIA requests and appeals for years in the late 90's and got turned down at every step. Morons. But I digress...

I finally got around to tearing into it. It's a standard H-bridge PWM inverter. The non-isolated 60 Hz input power goes to a full-wave bridge and a hefty choke filter/capacitor bank (left side of the chassis). The center is taken up by a large heatsink that has the rectifiers, H-bridge (each arm has a 2N6306 driver, Darlington-connected to two parallel house-numbered TO-3 NPNs), and large snubbing diodes. The 400 Hz stuff is on the right (the output transformer, current transformer, filter components) as well as a simple low-voltage linear supply for the control boards (two Base Driver boards, one 400Hz Control board, and one PWM Control board). Most of that stuff would fit in three 14-pin DIP packages if I were inclined to redesign it! If it's still putting out the right signals I'll leave it alone - not broke, don't fix it, etc.

Now look closely at the steel lids of some of the output transistors (one each in three of the four H-bridge arms). Notice
the holes arc-melted right through them! That must have been quite a show when they blew. Lots of energy in that  capacitor bank augmented by the wall socket... I can only assume that a bonding wire vaporized, and the resulting plasma struck an arc to the lid  



A close inspection of the bridge wiring shows something I think is very inappropriate for bipolar transistors - the 8 output transistors are wired in 4 parallel pairs with no current sharing devices at all! No wonder they blew up... hadn't Rockwell Collins' engineers ever heard of emitter resistors?? Or perhaps it was for redundancy if one failed open?

I removed the eight output transistors (soldered directly to the pins, no sockets)   and found that the five without melted holes all had base-to-collector shorts. No point in trying to measure the exploded ones, although I may cut the lids off and inspect under magnification just for fun. I installed all new BUX48's from the same lot. That transistor has ample current capability (and higher Vceo) than a 1 Kva bridge operating at 150 vdc should ever call for.. I don't have a curve tracer, but as one per arm should be sufficient, then two even if not perfectly matched will be even better.

Surprisingly, the four 2N6306 transistors survived. They are pretty low Hfe so probably have a nice beefy junction. There are 1 amp picofuses from their bases to the driver circuitry, all of which are open! Hoping the smaller transistors on that driver board are ok... a little inspection and paper/pencil reveals two half-bridge circuits on each card, looks easy enough to diagnose/replace even without a schematic.

Anyway, just wanted to share. I considered replacing the bridge with MOSFETs but that would have meant redesigning the driver cards, too.

[KA0HCP]

Interesting and very well written!  Thanks for sharing.

Bill

[KB3DKS]

Nice !
Power device engineering has come a long way since those days. I'm surprised Solitron qualified for Mil Spec as I found very few of their 2n3055s survived in audio power amp use back then. There was a change in the RCA/Motorola 3055 family from epibase to power base or vice versa, die style that improved the situation but the good devices really started with the MJE series.

Bill,
KB3DKS

[WB3JOK]

Thanks for the compliment   For those of you still reading:

The H-bridge assembly is back together and seems to work, at least at 160 vdc and low currents - I can pull the outputs up and down into a resistive load with appropriate bias currents applied to each 2N6306 from a floating 12v battery. Leakage current isn't measurable on my Triplett 630PLK (100 ua range). So far so good.

Unfortunately there is more domino effect damage, not too surprising in a DC-coupled circuit... There are two driver cards, each containing two half-bridge drivers. The upper transistor is a 2N2905 PNP switching, and the lower is another house-numbered part, presumably NPN. I loaded each driver with a dummy load consisting of two diodes shunted by 100 ohms (basically what the H-bridge looks like to the driver), and connected my 453 scope in differential mode (ADD, Ch.2 Invert).

But the signals aren't there except for some 120Hz ripple originating from the 15 vdc supply, and one looks very different from the other three Examination of the board shows that the 2N2905 PNP transistor provides base current via a 16.9 ohm resistor (showing signs of heat), so the transistor's almost certainly toast. The 1 ohm resistor to the pull-down (base discharge) transistor is intact but most likely that "mystery" transistor is fried too. I think a 2N2270 will be a good replacement here and even the same package, too.

However - the boards are coated with a thin, clear and very tough insulating coating. Definitely *not* MFP. I'm guessing it's Parylene. So I can't measure anything even with firm pressure on test prods! Does anyone have a recommended potion for removing it?

EDIT: A little Googling found an interesting link on coating removal... looks like I have my work cut out for me!

thanks
-Charles
WB3JOK/0

[KA0HCP]

All pcb's in navy equipmentwere conformal coated to protect from the humid salt environment.    This is something that hams have little experience with.  I've asked for comments on other boards and got no replies.  My impression has been that the navy used urethane coatings, though I never actually saw them being applied at the intermediate repair level.

For DIY trials I would suggest acetone and then MEK, methyl ethyl ketone; with a swab in a clear location.

Jensen/Stanley has an excellent catalog of tools and process supplies:  http://www.stanleysupplyservices.com/

Here are two two small pen applicators that remove the major types of conformal coatings.  I have no experience with them.   Page 403

Techspray 2510-N  $11,    pn  413-714

Circuit Works  CW3500 $16,  pn 408-374


Thanks for the white paper on coatings; outstanding.

Bill

[KA3EKH]

I have worked on some smaller inverters that I have around here for running mill equipment that were built in the seventies and eighties and have not seen anything in the emitter leg of them, just grounded emitter and collector to transformer with any current limiting in the feed to the center tap of the inverter transformer. All failures I have had were shorted switching transistors or shorted pass transistors in the feed to the center tap of the transformers that result in excessive output voltage, those little servos and fans run real fast at 160 volt 400 cycles! The control circuits are always mysteries with lots of little parts and transformers. Hard to see how something apparently so simple can be built so complicated. Would not have thought of them being PWM being waveform of output is important. Just curious on what your running that requires that big of a inverter?

Ray F

[WA1GFZ]

We removed Parylene with a walnut shell blaster. Walnut shells crushed into powder does a great job taking it off. Very tough coating that handles high voltage. We used it on all our boards that carried high voltage. Humiseal holds moisture and pretty useless but cheap. Consider corona dope when the repair is done. Parylene is an expensive process so it isn't used unless necessary. I think it is rated 1000V/Mil
Conap was proven bad for you so avoid it since causes cancer. We went to parylene when conap was banned.

[W3SLK]

Bill said:

All pcb's in navy equipmentwere conformal coated to protect from the humid salt environment.

Indeed! The best thing I found to remove conformal coating was the use of dental tools. These were standard issue for our 2M station. I can't remember its formal name, but the tool that looked like a minature spade was best at scraping it off. Sometimes you could lift it with heat but most of the time you lifted the pad-eye with it and had to refab a run.

[Jim KF2SY]

Back in the day some of the guys in our non-Mil world used this spray for conformal coating removal. 
The standard thin goopy coatings washes away easily from PC boards with this stuff.....IIRC

http://www.chemtronics.com/products/product.asp?r=1&m=2&id=4

 

[WB3JOK]

Would not have thought of them being PWM being waveform of output is important. Just curious on what your running that requires that big of a inverter?

Actually PWM is a very good (high-efficiency) method to generate a low-distortion sine wave output. I hear it works on ham transmitters too 
The switching frequency is typically 20-50 KHz, easily filtered from the 400 Hz, being 100 times higher in frequency so a simple low-pass filter will work...

Square-wave or modified-square-wave inverters are very simple and inexpensive, but also have large amounts of THD especially higher harmonics. Some loads don't like that...

I don't actually NEED 1 kva of 400 Hz power right now, but who knows, I might come across a deal on a mil-surplus transmitter or other piece of equipment that is going for peanuts because it requires 400 Hz! 

However, I did pull the low-voltage power supply, disassemble it and trace out the circuits. Two 1000uf axial electrolytics were completely open... and both were in the floating supplies that provide 18 volts to the driver stages. Hmmm. After replacing them the drivers now have good DC supplies, and two of the four are putting out a narrow pulse at a 400 Hz rate. BUT they are on the same ("bottom") legs of the H-bridge so that's still not normal behavior - I'm confused. 

The smaller transistors on the driver cards appear to be intact. I'm wondering if the 140 vdc has to be hooked up in order for the driver card to turn on, or something... they're down in a little card-cage and can't get to all the circuit points without disassembling more of the chassis.

At this point I'm just going to fix it because I'm stubborn 

[Opcom]

Bad filters play havoc with any kind of switching power supplies. The paralleling of transistors has been done if they were closely matched. We used to buy batches of Motorola TO-3 transistors and match up sets of them when repairing high powered amplifiers.

[DMOD]

Does this circuit have a Bi-polar supply voltage, or single supply with bottom grounded and top going to 150VDC?


I guess I am a  bit confused:
Are there two PNP's in parallel at the top for each node and two NPN's in parallel at each of the bottom nodes?

For H-bridge inverters, proper operation of the snubbing diodes and the timing (phasing) of the base waveforms is critical.

The control circuitry usually performs a soft turn-on and samples the current via one of the components or through a sampling transformer. The control circuitry is supposed to shut down for any fault condition.

I know it seems silly, and anyone with an EE background can reverse engineer this circuitry in half a day or less, but this inverter is still used in some Military apps.

The real complex part is the transformer with it's special windings, controled leakage inductance, saturation characteristics, etc.

[WB3JOK]

Does this circuit have a Bi-polar supply voltage, or single supply with bottom grounded and top going to 150VDC?

Neither. Single supply, floating. The incoming 60 Hz line goes through a bridge rectifier, no isolation transformer. The entire innards float. Only the frame itself is grounded to the power line ground. Makes it a PITA to do measurements when there is no ground reference for the scope... time to rig up an isolation transformer!

I guess I am a  bit confused:
Are there two PNP's in parallel at the top for each node and two NPN's in parallel at each of the bottom nodes?

No, it uses eight identical transistors, two NPN's for each bridge arm.

Incidentally, a little thought shows that three floating bias supplies are required for the four drivers: a common one for the bottom two bridge legs, whose emitters are connected together and to the neg side of the 150 vdc; and one each for the top left and top right bridge legs. And indeed that's what the power supply has 

The control circuitry usually performs a soft turn-on and samples the current via one of the components or through a sampling transformer. The control circuitry is supposed to shut down for any fault condition.

that's what I figure is going on here. Both a current and a potential transformers are on the 400 Hz side, providing feedback to the PWM controller, as well as a tap on the capacitor bank's bleeder resistor (presumably for sampling the rectified input voltage). There is an idiot light (two actually, green for ok, red for fault). I plan to hook everything up and see what happens... no erroneous drive signals that can blow up my new H-bridge, anyway. Although a fast-blow fuse for a couple of amps in line with the 150 vdc might not be a bad idea!

I know it seems silly, and anyone with an EE background can reverse engineer this circuitry in half a day or less, but this inverter is still used in some Military apps.  The real complex part is the transformer with it's special windings, controled leakage inductance, saturation characteristics, etc.

I have such a background  The pain is in the mechanically compact design, house-numbered IC's, multilayer PC board for the PWM card, etc. Should I get tired of messing with it, it'd be relatively simple to just make a new control card from one of National Semi's PWM-on-a-chip IC's 

Precisely because anyone can reverse engineer, or just buy a brand new one, the fact that it's still used in the military is irrelevant! Except to the bureaucratic "orifice" who blindly cites that factoid as justification for denying the release of schematic and repair info.

[DMOD]

Precisely because anyone can reverse engineer, or just buy a brand new one, the fact that it's still used in the military is irrelevant!

I feel your pain, but it is ITAR controlled and the manufacturer had no control over its classification.

Don't ask me how I know.   

[WB3JOK]

Oh, I'm not blaming Rockwell Collins... the problem is with the USAF Brain Trust 
I served almost 10 years and am well acquainted with the rule-book mentality therein.

Meanwhile I decided to trace out the entire schematic of the driver card, and enter it into Eagle 4.16. With that in front of me it was easy to see how it's supposed to be working. I'm using my Tek 114 pulse generator to simulate the PWM input, and a 100 ohm resistor in parallel with two series diodes for an equivalent dummy load (recall that the H-bridge has a Darlington-connected pair with 100 ohm shunt resistance). For power supplies, the inverter's own internal unit!

With my P6019 current probe on the line to the dummy load, I see about 300 ma positive drive, a huge but very short negative spike around 2.5 amps (reverse recovery of the diode load) rapidly settling to less than 100 ma negative (the resistor load). That ought to be enough base drive in both directions and may also explain the house-numbered special transistor in the totem pole used for the pull-down (it has a 1 ohm series R vs. 16.9 ohm for the pull-up 2N2905). I was planning to use a 2N2270 for a replacement if necessary, but I'd better check and make sure it has enough pulse current capability! Of course it may not need to sink quite that much current when hooked up to the actual H-bridge (faster junctions), but then again why would they use such a small resistor if not anticipating large discharge currents...

The input stage to each driver uses a 2N3439 (BVceo 350v) and the return/common lead (from the emitter) does not connect anywhere else on the board. There has to be a level shifting circuit somewhere on the PWM board since both the base and emitter connections come off the driver card and they used a high-voltage transistor. This also implies that I have to have the 150 vdc hooked up to make it work. Hopefully will know more later this week... watch for "smoke report"

It feels good to flex the mental muscles of circuit design, reverse-engineering and debugging 
-Charles

[WB3JOK]

I installed the new picofuses and reassembled the bridge into the chassis. To bring it up "slow" I put a 100W bulb in series with the H-bridge rail, put a 40W bulb on the output as it may not be designed to run unloaded, crossed my fingers and hit the breaker. There was a 2 sec. pause and then the 100W damage-control bulb lit up and the red "Fault" light came on. Crap!! Looks like I toasted all my new parts! 

But then I also noticed the 40w bulb load was half-lit So I put a 20 ohm power resistor in line and tried again... the  load bulb was brighter that time. Finally a 4 ohm, and the 40w bulb lit fully AND the green "Normal" lamp illuminated!


The startup delay of a second or two is to allow everything to stabilize before turning on the drive, I assume. Lastly I put the wiring directly back to the cap bank and used the 100w bulb as the load.




The output transformer really "sings" at 400 Hz with plenty of audible harmonics buzzing too! I expected the output waveform to be a bit cleaner given the 20 KHz PWM carrier frequency. A quick test with a 20 ohm load resistor (700w at 118.5 Vrms) showed the identical waveforms, just higher current.  This is with the 100 watt bulb load; top trace is load voltage 100v/cm, bottom trace is load current 1A/cm (P6019 current probe), timebase 0.5 ms/cm.



This has been bugging me off and on for ten years and I'm glad I finally got around to de-bugging it! So now I need some equipment that requires 400 Hz power Maybe I can get a deal on an aircraft radio setup...

-Charles

[Ralph W3GL]

   Good job, Charles, this project was very well documented as well as
   well done...    It's worth a whole handful of "attaboys"...

[KA0HCP]

Congratulations!

[DMOD]

Way to go Charles.

Phil - AC0OB

[WB3JOK]

I appreciate the compliments, even if all I really did in the end was change eight transistors and four fuses 
At least I know how most of the circuit works now!

Meanwhile, a 2N3725 would be a much better totem-pole driver transistor (than the 2N2270 I originally recommended).
Here's the schematic of the driver card, which may be of use to someone.

There are several people on the Boatanchors list who have these inverters, in various stages of repair... no one has the tech manual either. One ham actually got permission to visit Tinker AFB, OK and lay hands on the holy grail manual and even remove it from the library - but he was only allowed to copy it on copiers designated for classified documents, and they wouldn't give him access to one! Kafka must be snickering bitterly somewhere... 

[Opcom]

I'm glad you got it worked out. "house-numbered IC's" - nasty. As nasty is the old practice of scraping off the numbers on ICs to thwart anyone but the factory from fixing whatever it is.

400Hz gear - -old Collins aircraft rigs!