APC Smart UPS SU3000RM Overloaded when started from batteries

You could put an ohmmeter from drain to source and from gate to source, and see if any of them are shorted or have blown gates.

If your power supplies overshoot badly after a large load transient, you could have had a transient overvoltage. Doesn't take much energy to blow a gate.

Cheers

Phil Hobbs

In the APC SmartUPS, some of these run the switching power supply section full time, whether on AC or battery. On AC, the voltage from the switching section is slightly below the input AC voltage, so the load draws most of its power from the AC line. Of course, they are phase locked. When the AC line voltage disappears, the switching power supply section takes over running the load with zero switching time. I'm not sure if the SU3000 SmartUPS works this way but it's worth checking.

A schematic should be useful: (175K)

Offhand, I would guess(tm) that you blew something up with your 48V power supply. The SU2200/3000RM uses a pile of 4ea 12V 18AH AGM batteries. I believe that they are wired 2 in series and 2 in parallel for a total of 24VDC, not 48VDC. You probably blew up the electronics with the overvoltage.

If it draws lots of current from the battery section, the fets are the likely culprit. However, as I vaguely recall, the SU2200/3000 series has 4 strings of 8 FET's in parallel. It's not going to be much fun finding the one FET out of 32 that is blown. I built a crude current probe using a Hall effect device at the end that is useful for finding which device is sucking the most current.

Most of the value and cost of a UPS are in the batteries. In this case, a set of batteries is about $130 plus shipping. At the local recycler, such large UPS's without batteries are plentiful because the price of a new or refurbished UPS is only a little more than the cost of decent batteries. Incidentally, beware of junk batteries and look at the weight of the battery for a clue as to how much lead is inside. I have my own collection of UPS's with dead batteries.

Good luck.

On 7/3/2018 9:43 AM, Jeff Liebermann wrote: I built a crude current

I've been considering doing that. Can you share details about which parts you used? What kind of sensitivity can you get?

Thanks, mike

Everyone thanks for the advice. It is for sure a 48V unit,(batteries were wired as such and it puts out 48v charge current) but the SMPS power supply may have overshot that a bit under load.

My Hakko 936 doesn't quite have enough thermal mass to desolder the FETs (even with the giant chisel tip I put on it) because of the huge double sided traces.

Getting a new hot air rework station this week and that should be able to do the job without cooking the components. I had one before but didnt make it with me on the last move.

Strictly speaking the unit works as a UPS as it is, but I can't use it as an off grid inverter because it won't start off batteries only.

I will check back in after I get the FETs out and check them.

Look on the top of the first page of the schematic. It says +24v and is connected to the diodes after the AC power transformer. The external battery connectors J5 and J9 traces to a collection of power relays (RY3, RY4) that are labelled as 24V coils. The 4th drawing "Main Circuit BD" also shows +24V. While most of the electronics runs on +12V, I found no evidence of anything labeled +48V. Either the schematic is the wrong unit, there might be a different between "3000" and "SU3000RM", or your batteries were wired incorrectly.

Digging deeper, the SU3000RM uses two 24V RBC11 battery cartridges: that are wired like this: The connectors go to J5 and J9 which connect the two connectors in parallel for 24V, not 48V.

Anything attached to a heat sink is going to be a problem removing with a hot air desoldering station. I suggest a big chisel tip and a solder sucker.

Do you still have the old batteries? If so, can you test the SU3000RM with the batteries installed and see if it survives the initial self test? During this test, the UPS runs on battery for a few seconds. If the batteries are dead, it will so indicate with the front panel lights. It might be a good quick check of the inverter section.

I tried to find the one I built but couldn't. So, I do it from memory.

The "probe" is a wooden coffee stirring rod that I stole from the local coffee shop. Stuck onto the end was a linear Hall effect device in a TO-92 package. As I recall, it was a TI DRV5053: It comes in several sensitivity ranges. Methinks it was EA or +45 volts/Tesla. That's not very sensitive, but more on that later.

The schematic was very simple. See 8.2.2. on the data sheet or: The output went to crude x1000 LM358 DC amplifier and then to my TEK

The reason I threw this together was that I was trying to fix a no-name Chinese solar inverter that had about 32 IGBT devices in parallel in 4 strings. One device was probably shorted, but I couldn't tell by probing. I also couldn't run it long enough to check which device got warm. Checking the gates showed nothing useful. So, my last resort was to try and measure the collector current without breaking any wires. Looking for a switching waveform on the collector would have been so easy, except that some protection circuit turned off the drive oscillator because of the high current. I tried an inductive pickup, but it just picked up junk from the adjacent working devices.

I had just enough exposed collector lead accessible to sense a magnetic field. I had a sample DRV5053 device, so I quickly threw it together on the wooden stick and looked for some indication of a magnetic field. After some tinkering, I could just barely see the current on the scope. Walking it down the line, I found that 3 devices had shorted. I removed the shorted device but did not initially replace them. It worked. I then made a guess as to some potential replacements, actually found some in my pile, and proclaimed the inverter to be repaired (until some better parts arrived in the mail). Note that this repair was done from start to finish in about an hour.

The key here was that it was only going to work with fairly high currents. Judging by my automotive magnetic field ammeter, the inverter was sucking about 30A with the blown devices which was probably distributed as 10A per device. So, how much sensitivity will it take to get a 1 division (2mv) change on the scope? Plugging into a handy Ampere's Law calculator at: for 10A at 2mm radius (my guess), I get a 4x10^-8 Tesla field. The Hall effect device has a sensitivity of -90V/T for an output change of a fabulous: 4x10^-8T * 45V/T = 1.8 uV with a gain of x1000, that becomes 1.2mv, which I could just barely see under the noise and garbage on the scope.

So, you're welcome to build such a device, but I don't think it will be a troubleshooters dream current probe, that can measure DC currents without breaking the line. You might want to look into IMC (integrated magnetic concentrator) type Hall effect structures. By grabbing more of the magnetic field and concentrating it into the tiny Hall effect chip area, it produces a higher reading and therefore a higher sensitivity. The problem is that the pickup toroid or core needs to go around the wire, which defeats the basic purpose of the no-contact current probe. However, I suspect that some kind of probe can be contrived with will sense more of the magnetic field than the tiny chip, which will help with the sensitivity problem. "Measuring Current with IMC Hall Effect Technology"

Thanks, Looks like that's not gonna be worth the trouble for me. Too little sensitivity. I had some success with AC fields using the tiniest torroid core I could find. I cut a slot in the core and wound it full of tiny wire. Took about a dozen tries to cut a slot without breaking the core...good times... In your case, proximity to other strong fields would be an issue.

For stuff that's actually shorted, I've had great success using the HP current tracer and pulser. I can stick the pulse on a trace and tell which direction the current is going. Also useful finding bad caps.