Load Bank Transistors Blowing out. 1

[mwebster]

I've been working on reverse engineering a resistive load bank control board that controls a 750 Amp load bank. My problem is that I don't see how the trim transistors keep popping.

Here are some of the fun details:
The board controls and trims a set of amplifiers that work to turn on the main load transistors. There are 30 discrete load steps of ~25A each and 6 analog trimmers of ~10A each when they're full on. All load transistors are Darlington PNP's rated at 50A continuous and 120V Vceo (MJ11033's). The control board will open the main load contactor if the voltage goes over ~35V, but it takes it 3+ seconds to do so. (It'll open faster with higher voltages, but I don't expect it to open much faster than a 1/2 second or so at 200+V. It would almost definitely miss fast transients) The nominal voltage source is 30VDC from 15-20kW airplane generator/starters.

The collectors (which is also their case) of the load transistors are all hooked together to the generator return (-) line. Their emitters are held off of the main rail by big honking power resistors that are tapped off of and shared by the various transistors. The steps get two 1/3 sections of resistor hooked up to the main rail and the trims get 1/2 a resistor each hooked up to the main rail. The resistors are 8.5 Ohm total. The bases and emitters are hooked up to the amplifier board which is hooked up to the control board. The base and emitter are held apart by a 100 Ohm resistor so when the amplifier tugs down on the base it turns on.

OK, all my simulations say that these transistors should be fine, ASSUMING that they're properly heatsunk and getting good ventilation. Big assumption, I know. I don't think that generator transients could cause this since the trims are isolated by 4.25 ohms from the main. It should require over 200V on the main bus to get more than 50A down through that transistor. Of course, if it's running on the hot side, that 50A handling capacity can be drastically cut.

The failures can often be dramatic as holes are burnt in the transistor can. They tend to happen in cycles, where one transistor goes out for whatever reason, then 2 or 3 more fail over the next several days. It's always the trim transistors (obviously they dissipate the most power as they often operate in their analog region).

So, my questions are these:
Which is more likely, overvoltage or heat (or a mix of both)?
How large of a transient should I expect to see on a running generator? (i.e. not at spin-up or spin-down, but after the load contactor has closed and the load bank is conducting at some level)
Am I missing a likely, or even an unlikely culprit?

Thanks for your time,
Mike

 

[waross]

Watch for load dumps. If you drop 50% of the load, expect the generator voltage to jump 100% until the regulation can reduce the field.
I would look closely at the power dissipation of the transistors running in their analog region. At 50% conduction they will be dissipating as much power as the load resistor.
itsmoked will be along soon with excellent advice.

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[itsmoked]

Ah yes barbarism is alive and well in the aircraft industry!

You realize you can buy one or two modern FETs stick them on a water cooled piece of metal and have this whole thing reduced to a hand carried device? I'm looking at one FET rated at 860A with a Vdss of 100V.


Anyway.. My guess would be over temp. It doesn't take long to toast one of those devices you're using. Once it starts to toast it's internal resistance rises which helps complete the barbecuing. Others are compromised, possibly by Bill's load dumps, when neighbors fail open, leaving them to toast soon after.

Have you waved a temp gun over your Army of Pass Elements while it was running?



Keith Cress
kcress -

http://www.flaminsystems.com

 

[mwebster]

Thanks for the replies.

No, I haven't checked the temperatures yet. This is for a client who's some 5 hours from my office, so it's pretty much an overnight visit when I go down there. I haven't dealt much with high power installations, so I wanted to get all my ducks in a row before heading down there.

Truth be told, I haven't even seen the load bank itself, it sits off in another room from the generators themselves. My company (a couple of years before I came on board) updated the generator test system from its early 80's incarnation (written in BASIC, no less) and the client figured the original load bank was "good enough" and wanted it left out of the quote. Of course, here we are, 3 years later with load bank problems.

From talking with the engineer that has been working on this system, the bank is fairly tightly enclosed and it's hard to get your head in there at an angle where you can see all the transistors when the fans are running. So it's hard to get a spot check from a temp gun. I'm hoping I can either get some thermocouples in there or convince the powers-that-be to spring for a FLIR rental.

As an aside, I'm very interested in your conception of a modern version with FETs and a water cooled heatsink. I'm sure I could do the research and run some numbers, but just to ask: Can a relatively small heatsink like that dissipate 20kW of power or are you talking in conjunction with the load resistors? (so that would be, say maybe 1/4 of peak with two FETs, or 5kW)
I guess the other big variable is whether you're talking about using a pump hooked up to a large external water reservoir for the cooling.

 

[itsmoked]

You can carry away just about any amount of heat with miraculous water. Think about an on-demand electric water heater! I recently was in here with one that uses 90A @ 240V = 21,600W and it's output was considered barely adequate for heating domestic water to 125F. You are talking about a few gallons a minute - very small pumps.

Very small cooling systems are used with large semiconductors in various power equipment systems.

As for your specific application a design would want to understand the daily test loading as well as local use of hot water or heat needs. This is because you can easily transport the heat, using water, elsewhere within reason.

If these tests only run for a few minutes a couple of times a day then a single tank of water with no extra heat dissipation needs, could be used. By the end of the day there's a tank with warm water in it that cools overnight ready for the next day.

If instead a unit runs longer than that say multiple hours a day you could pump to a small fan forced radiator somewhere where the waste heat is not an issue. Preferably outside where it could greatly reduce A/C loads otherwise incurred.

If even more heat is created by say 8hr test runs or even 24hr runs then environmentally you should really be doing heat recovery anyway. The hot water should be heating a stored source like a large insulated tank in series with the facilities water heater. This would directly reduce energy consumption used to just heat water.

Another example for the facility I am perceiving you to be describing might be a steam cleaner. Preheating the water fed to the steam cleaner again results in a daily reduction in the cost of running the steam cleaner.

Yes I'd still use resistors and not just use the MOSFETs as the only load as that would save a lot on cost. But you would want to use bar heaters that bolt to flat surfaces like your water cooled heatsink.



Keith Cress
kcress -

http://www.flaminsystems.com