Converting an APC Smart-UPS 2200i to 24V

[marcone]

Hello there,

Just got one on these things which almost broke my back while taking it from the car to the house.


Want to use it as a back-up for power failures but since I only have 2x 12V large capacity batteries and don't want to fork out the extra cash for another 2, I want to convert this to run out of 24V. Plus there is plenty of space inside for mods, not cramped as the SUA1500.

Checklist of things to do so far:
*Change secondary transformer windings from series to parallel. This thing has 2 huge power transforms with the 230V side in parallel and the low voltage side in series. - easy to do
*double up the Mosfets count (currently only 3x4 MOSFETS) and filtration capacity (currently 2x1500uF) - it requires to disassemble the heat sinks in order to solder gate drive resistors for the additional Mosfets - moderate effort
*Beef up board current traces and battery wiring (will use 10mm2 wire). Also double the battery pack fuse (don't know the current AMP rating) - easy
*Add additional dissipation area to Mosfet heat sinks and install a large 120mm fan on the side of the case pulling out air of the enclosure (the back fan sucks air in) - easy

What I am missing is a way to trick the unit's brain to think it's still running on 48V. I know that there is a resistive divider somewhere on the board that is used for monitoring battery voltage and changing one of those resistors would do the trick but I don't know exactly where that is located and what kind of resistor value modification it needs.

Any of you have an idea in this last point?

[BradC]

Yep. The resistive divider generates 5v at the charger asic when the desired float voltage is reached. Shame it's an SUA as the schematic is near impossible to come by. The earlier SU are available.

The SUA also use a dedicated mosfet driver, so you'll want to look at that closely to see if it needs mods to work correctly on 24V. There are some other funky bits that will trip you up. The ups has a switch mode converter to drop the 48v to 24v for the logic and drivers. That converter is not going to work with a 24V input. The charger ASIC has a shunt there somewhere to measure fet current. That will need modification as you are planning on doubling that. You are going to be looking down the barrel of 100A, so you'll seriously want to look at the dc switching. Once you get that done you'll find that battery constants 4,5 & 6 in the firmware are waaaaay out and you'll probably find the ups goes into almost instant shutdown on battery or grossly miscalculates your run time (in either direction). Those constants compensate for the dc resistance, battery impedance and load/volt characteristics of the designed battery pack and are very specific to each model/battery configuration.

Have you even looked inside it yet? The SUA models are all quite small SMD and a lot harder to trace out than the older generations. I'd be flogging it and picking up a 1500 ( which is the largest 24v unit they make).

If you do decide to go ahead with it, I might be able to help with a bit more technical detail on the firmware and asics, but I reckon you're in for a much harder battle than you've planned for.

[MagicSmoker]

Terrible idea, converting a 48V UPS to 24V... If the weight is a concern (and really, why would it be after you've already moved it, unless you plan on taking the thing with you everywhere, that is?), then I'd figure out how to replace the crappy lead-acid batteries with lithium. Lithium iron phosphate (LiFePO4) is a pretty forgiving chemistry that can be charged with a typical lead-acid profile if the cells are stacked up appropriately (IIRC, a max of 3.4V per cell is recommended, though 3.3V will give you longer life at slightly reduced capacity). In this case, float voltage from the existing charger will likely be 55.2V so 17 LiFePO4 cells in series should be safe. Other lithium chemistries are less well suited to performing as direct replacements for lead-acids so be warned.

[marcone]

Yep.

Thank you very much for the reply and offer to help, indeed it's not as straightforward as I though, lots of thickening required and all this without a schematic (and proper knowledge on my side might add) so it looks like investing in 2 more batteries is not such a bad idea after all.

My problem is that I need to provide back-up for a load drawing up to 1300W in peak moments.
I played around with a 1500VA unit but could not get the H bridge cooled properly and ended up blowing Mosfets on 3 of the 4 arms after about 1 week of operation. The 1500VA unit can't be made for operation at such high loads as there is no space around the Mosfet bridge for beefing up heat sinks or installing additional fans.

Terrible idea.

Well I had to ask Weight is not the problem, cost is. I got this UPS for less than one battery's cost and the modifications I had in mind do not really add much on top.

[BradC]

Go and buy 4 CSB 12170 batteries from wherever you can get them cheapest. Generic replacements just won't last as long or perform as well. Those are fantastic UPS.

The 1500 are pretty rugged units and are generally limited by transformer cooling rather than mosfet, so you must have beat the snot out of it to have killed the mosfets. You do realise however they are only rated to just over 1000W. The 1500 rating is VA at about a 70% power factor, so it's only just rated to provide 1000 real watts.

[marcone]

The 1500 are pretty rugged units and are generally limited by transformer cooling rather than mosfet, so you must have beat the snot out of it to have killed the mosfets. You do realise however they are only rated to just over 1000W. The 1500 rating is VA at about a 70% power factor, so it's only just rated to provide 1000 real watts.

I kind of did that by putting a constant load of 300w with peaks of 800-1300w. Transformer was able to handle this load however I guess the heat kept building up in the Mosfets until they let the magic smoke escape.

Sticking to the SUA1500i, i went ahead and replaced all of the Mosfets in the unit with new ones and tried to power the thing up:
* If I just connect the mains cable, unit does nothing (as it's supposed to do without batteries) however
* If I feed power to the battery cables through a power supply (providing 27.5V at 5A) with mains plugged in, unit blinks all the LEDs on the front panel, pulses buzzer and relays at a frequency that varies with the voltage applied at battery terminals. I know that unit does a battery test at startup drawing about 10A which my power supply is not able to provide but it should nonetheless power up and only blink the "replace battery" LED - tested that when I first got the unit.
*If i unplug mains cable and connect power source to the battery cables it only acts as above when i press the ON button.

I am guessing it does that because it senses an internal fault, Mosfet driving circuit maybe?
When reassembling the unit I checked:
*transformer (outputs 14V - I guess it does not need to output the full 24V because of the switched voltage provided by the H bridge)
*Mosfet gate resistors, tested fine at 20 ohms
*filtering caps, also tested fine.

My question is, what should I be looking for next?
Does anybody have a schematic of the driving circuit for the Mosfet bridge?

[BradC]

I might. PM me your email address. I'm not being cryptic. I have a schematic that might have the info you want, but as I've not had an SUA1500 in bits I can't be sure.

[marcone]

Thanks for your offer to help.
I took another look at the board and Mosfets are driven directly by this guy: http://www.intersil.com/content/dam/Intersil/documents/hip4/hip4082.pdf
I guess the next thing would be to get that replaced and see if it makes any difference.

[marcone]

Update on the repair of the SUA1500i: transformer is bad.
Although there' are no outside signs of heating/melting, there is definitely a short between 2 winding layers on the low side , that's why the low 14V output when high side is connected to 230V and the winding  impedance of just .14 ohms.
Bad part is that it took me another 8 MOSFETS and a driver IC to reach that conclusion. At least I got to desolder and solder my first SOIC!

[wraper]

Sticking to the SUA1500i, i went ahead and replaced all of the Mosfets in the unit with new ones and tried to power the thing up:

Bad idea. Never ever do this when reparing any sort of SMPS.

that's why the low 14V output when high side is connected to 230V and the winding  impedance of just .14 ohms

0.14 ohm is not low at all, about what you would normally expect. IMO you came to wrong conclusion. I would expect something faulty in what is driving the MOSFETS.

Although there' are no outside signs of heating/melting, there is definitely a short between 2 winding layers on the low side

And how did you definitely check this? To check this without disassembling the transformer, you would need to check inductance as minimum. All you need to blow the mosfets is DC on the MOSFET gates instead of AC. HIP4082 won't protect anything as it does not generate any PWM by itself and will happily output DC. It's just a dumb gate driver.

[marcone]

I measured the low side of one transformer from a SUA2200i I have which measured 0.21 ohms on my LCR meter, so 0.14 ohms looks like 1/3 of the low side winding is shorted, especially since the transformer outputs 14v when high side is connected to 230V.
All Mosfets and driver circuitry were changed and the unit actually seemed to work for half a sec before the Mosfets in charge of one transformer side of the H Bridge went boom.

[MagicSmoker]

Sticking to the SUA1500i, i went ahead and replaced all of the Mosfets in the unit with new ones and tried to power the thing up:

Bad idea. Never ever do this when reparing any sort of SMPS.

Yes, what the OP did was dangerous, but merely telling him that without offering an alternative approach isn't exactly helpful. And it is most definitely possible to test for a shorted turn in a transformer without disassembling it or measuring its inductance... See below.

So, to that end @marcone, you should always use a current-limited power source when testing a SMPS. In this case, a current-limited DC power supply instead of the batteries would be ideal, but inserting a 1-10 ohm wirewound resistor in series with the positive input to the UPS should work.

I measured the low side of one transformer from a SUA2200i I have which measured 0.21 ohms on my LCR meter, so 0.14 ohms looks like 1/3 of the low side winding is shorted, especially since the transformer outputs 14v when high side is connected to 230V.
All Mosfets and driver circuitry were changed and the unit actually seemed to work for half a sec before the Mosfets in charge of one transformer side of the H Bridge went boom.

If you have an LCR meter why didn't you measure the inductance of the windings?!? 

Note that a turn-to-turn short in a transformer winding will cause the inductance of all windings to drop dramatically, nearly down to their leakage value. And not to point out the obvious here, but a shorted turn is equivalent to shorting a single turn secondary, which causes primary current to increase tremendously, of course, possibly killing the switches driving it.

So applying a known AC voltage/frequency to each winding and measuring the voltage on the other windings is a good test, just use a much lower voltage than the winding is supposed to operate at under normal conditions (like, 1/10th) to avoid having to supply a large current. For example, apply 23VAC to a 230VAC winding and look for ~4.8VAC on a 48VAC winding. The ideal tool for this sort of testing is a Variac, but you can also use another small mains transformer, say, 230VAC to 12.6VAC or the like.

And not to point out the obvious yet again here, but you only use 50/60Hz AC to test mains transformers, not high frequency switchmode transformers! For the latter the only practical test is to connect a LCR meter to one of the windings and then see how much the inductance drops when another winding is shorted externally. All true transformers and multi-winding inductors will show a dramatic drop in inductance when any one winding is shorted; if you don't see much of a drop then that means there is likely an internal short already present. This test is valid for mains transformers, too.