Cool tutorial: Converting an APC UPS into a powerful sinewave inverter

[cte7ds]

After researching my latest purchase, an APC Smart-UPS SUA1500i I ran into this cool little tutorial video that made me feel even happier with my choice



This guy had an old SUA1500RM2U laying around, designed to deliver 120 VAC pure sinewave with 1500 VA at 980 watts for ~7.4 minutes. And with a surge capability of around 1760 watts. And efficiency of 84% with a 872W resistive load.

After he was done modding the unit was able to deliver 1500 watts continuously and provide a surge power of at least 3200 watts, which is enough for a lot of household appliances. Also as a side effect the battery charging current increased from ~3A to ~7A which is a nice bonus when using bigger batteries (the unit however is known to be capable to accept an external charger in parallel with its own). Overall efficiency of the inverter stayed the same.

Here's a little overview of what he did:

He modified the H-bridge by soldering extra TO220 mosfets (and their gate resistors) to the unpopulated spots on the PCB. Stock configuration had 8 (IRF4104 is my best guess) mosfets populated out of 16 possible. He added 8 IRF3803's and used thermal grease (stock model didn't have any). Interesting thing about the H-bridge is that when not being used as an inverter they are actually being used on their linear region to charge the batteries.

He added an unpopulated low-esr capacitor near the battery terminals (probably in parallel with the batteries)

Added cooling directly over the mosfet heatsinks because they were  the "high thermal mass"-kind, instead of having maximized surface area. This is because the UPS wasn't designed to run for extended periods of time.

Because the transformer was bolted directly against the chassis he added thermal grease under the transformer and increased airflow over the transformer portion.

He used a secret programming mode to change the calibration of the output load sensor (current transformer) to offset the overload detection. Also the buzzer could be turned off permanently using software (phew..)

Overall the biggest issue was the thermal characteristics of the transformer. It's the most expensive part of the UPS so that's where the manufacturer tries to save as much as possible :-p

My use case for the UPS is gonna be like 1000VA maximum computer load (mostly SMPS with active PFC and some with passive PFC) extended run-time, so I'll probably be doing at least the thermal grease mod for the transformer and possibly adding more forced airflow. I'm getting the tower model so I'll be doing load testing of my own to see what is necessary. From what I know it should be using all the same components, just more tightly packed.

I did already install a special Ground Fault Circuit Interrupter Protected 230VAC wall socket and 10 meters of triple insulated 3x1.5mm^2 copper conductor from my lab to the living room so I can have all my computers from two different rooms behind the same UPS. So I'm pretty dedicated in having a decent UPS system and might just do all the mods as what this guy on youtube did to get a bit more performance

Bought my unit from the United States (someone was selling 230VAC models cheaply) so I'm never gonna be mailing it back for warranty anyways because of the cost, so I'm not that concerned about immediately voiding the warranty, and I could never keep my nose away from the internals anyways 

Aaanyways I just thought I'd share this tutorial with you because I thought it was really cool what he did and he clearly seemed like a talented person. And also a little heads up that I'm gonna be working on this so if someone is interested in the results or in helping out that would be cool

[poptones]

Here's a bit more on that from one of my favorite youtubers

Cheap Backup Power

[krish2487]

He used a secret programming mode to change the calibration of the output load sensor (current transformer) to offset the overload detection. Also the buzzer could be turned off permanently using software (phew..)

You dont have to do any black magic to mod the output of the ct.

Note down the burden resistor present in your UPS and note at which load level it trips.

scale the resistor to follow the projected increase in power.

:-)

Of course the UPS will still read the and compare the old set point of the CT but will trip at a higher load level.

Think of it as changing a voltage divider  ratio but keeping the multimeter mode constant.

[cte7ds]

He used a secret programming mode to change the calibration of the output load sensor (current transformer) to offset the overload detection. Also the buzzer could be turned off permanently using software (phew..)

You dont have to do any black magic to mod the output of the ct.

Note down the burden resistor present in your UPS and note at which load level it trips.

scale the resistor to follow the projected increase in power.

:-)

Of course the UPS will still read the and compare the old set point of the CT but will trip at a higher load level.

Think of it as changing a voltage divider  ratio but keeping the multimeter mode constant.

Hmm, I'm not quite sure what you mean by the burden resistor tripping? I think a schematic would be helpful.

[chickenHeadKnob]

I started viewing this guy's video(s) but sweet jesus its over an hour(s) long!  so I went straight to part six where he changes the fets. OK I am not a hardware guy but even I know you do not tug on superman's cape, and you do not just randomly parallel unmatched fets. What is going to happen to the current sharing(balancing). Power fet cognoscente please enlighten me, shouldn't this thing turn into  a smoke generator?

[SeanB]

The current transformer is loaded by a resistor in the unit. If you make the resistor lower value the current required to reach the trip point is correspondingly higher. Current transformers do not worry much about voltage ( within limits) as they provide a constant current into the load in ratio, much like the common voltage transformer. thus a lower value load resistor means the primary current has to be so much higher to reach the required voltage on the resistor.

[ptricks]

Good project if you want to do it for the learning experience but not really the best way to get a powerful inverter.
Look on ebay, auctions locally , military auctions.
A few years ago I attended a local school surplus auction , they had a pallet of  25 APC UPS for auction, got the entire lot for $20, most people didn't know what they were and the lot was big and most people didn't want to load it on a truck.
They were all rack mount, 3000 watt units , rated for full time operation complete with internal fans and jacks on the back for extra batteries. Every single one of them had the same fault, batteries were bad and wouldn't charge. Easy fix.

[cte7ds]

The current transformer is loaded by a resistor in the unit. If you make the resistor lower value the current required to reach the trip point is correspondingly higher. Current transformers do not worry much about voltage ( within limits) as they provide a constant current into the load in ratio, much like the common voltage transformer. thus a lower value load resistor means the primary current has to be so much higher to reach the required voltage on the resistor.

Like in this slightly different model the R133 next to CT1?

Full schematic: http://www.upsclub.org/schem/APC/735/E735Gschem.pdf

[cte7ds]

I started viewing this guy's video(s) but sweet jesus its over an hour(s) long!  so I went straight to part six where he changes the fets. OK I am not a hardware guy but even I know you do not tug on superman's cape, and you do not just randomly parallel unmatched fets. What is going to happen to the current sharing(balancing). Power fet cognoscente please enlighten me, shouldn't this thing turn into  a smoke generator?

Yeah I'm also amazed how easy it was, usually when repairing a device with fets the end result is like this:
I've had that happen to me a few times with Chinese inverters :p

Good project if you want to do it for the learning experience but not really the best way to get a powerful inverter.
Look on ebay, auctions locally , military auctions.

As long as you stay away from the cheap Chinese no-name brands and go with something that has a known good design. Here's a great example of what eBay can get you: http://www.ludens.cl/Electron/chinverter/chinverter.html :-p

Products like this: http://www.ebay.com/itm/1500W-3000W-peak-24v-220v-Power-Inverter-Charger-UPS-Quiet-Fast-Charge-C7A-/360464406988?pt=UK_Home_Garden_PowerTools_SM&hash=item53ed5a59cc sure seem like a great deal but I doubt that they are well designed.

A few years ago I attended a local school surplus auction , they had a pallet of  25 APC UPS for auction, got the entire lot for $20, most people didn't know what they were and the lot was big and most people didn't want to load it on a truck.
They were all rack mount, 3000 watt units , rated for full time operation complete with internal fans and jacks on the back for extra batteries. Every single one of them had the same fault, batteries were bad and wouldn't charge. Easy fix.

I wish I had luck like that. You must have made quite a nice amount of extra cash with those units?

[SeanB]

Like in this slightly different model the R133 next to CT1?

Yes. like R69 is the load resistor for CT2. If you make it 5R6 it will roughly double the current required to trip the UPS.

[cte7ds]

Just received the package from USA. I'm amazed how relatively well it survived the trip with such inadequate amount of packing material, only one dent in the product and the styrofoam in the package was completely shattered.

Also the Smart Card and USB-daughterboard connectors had become lose. Nothing majorly wrong at this point. That is until I checked the battery pack voltage... a whopping 1.443 volts for two lead acid batteries in series.... great, never seen a battery that discharged ever before. Well I didn't even need the batteries in the first place but I would have assumed that a product listed as "Brand new" would have brand new batteries. Funny thing with batteries as dead as these, the UPS refuses to start up at all. So basically the unit is dead out of the box if you didn't know any better :p What do you guys think should I ask for a partial refund or something? Had the batteries been brand new I would have sold 'em and now I can't do that

<dave> Don't turn it on, take it apart!</dave> I took the cover off and inspected for shipping damage and attached all the connectors that had gotten lose and then did a quick test with my bench power supply and a 40W light bulb as a load. The unit seems to be working at least, that's really all I wanted anyways hehe.

So time to tear it apart and mod the sucker!

[SeanB]

Charge up the batteries from a power supply, they possibly might still work. If you get to 15V per battery and it is still not taking more than 500ma then they are toast though.

[cte7ds]

Charge up the batteries from a power supply, they possibly might still work. If you get to 15V per battery and it is still not taking more than 500ma then they are toast though.

0mA @ 15V :-p

[SeanB]

Time to sell them to the scrapyard then..........

[cte7ds]

Here are some pictures of the SUA1500i internals and the mods I've done so far.

First off I installed a connector to the 8 pin ribbon cable connecting the main board and the front panel because it was annoying having it dangling from from the main board. I don't understand why this wasn't done by the manufacturer, after all, everything else on the PCB has connectors u_u

I took off the transformer and attached an aluminum plate under it with thermal grease to get much better thermal conductivity between the transformer and the chassis.

Replaced the battery disconnect plug at the back with a 20mm IP68 rated strain relief socket meant for 6-12mm cables. It's just big enough that it can fit the disassembled battery cables through it. Also had to make a little aluminum square with a round hole for the socket because the existing hole was so weird.

The chassis wiring for the batteries in this unit is made out of AWG10 wire, there's about 3 feet of it because the negative wire goes through the battery disconnect plug, but since I removed it I was able to reduce the wiring length to 2 feet. The connector used is a Anderson SB-50 forklift battery connector with a contact resistance of 0.2 milliohms and the 2 feet of AWG10 add about 2 milliohms to that.

The mosfets used in this unit turned out to be 8 x HRF3205, a 100A 55V 0.008 Ohm N-Channel power mosfet with a power dissipation of 175W and a pulsed drain current of 390A. The total gate charge on this thing is 170nC max. with turn-on delay of 14ns and turn-off delay of 43ns. My local electronics store has IRF3205 which is basically the same part and IRF1405ZPBF which is almost twice as good. The 3205's cost 2.70EUR a pop and 1405's cost 3.50EUR, so I'm thinking ill just go with 8 x IRF3205 for 22EUR. It's really expensive compared to eBay prices, but at least they are guaranteed to be genuine and ill have them in my hands really fast

For the battery filter capacitor I was thinking of using a Panasonic FR-series (low-esr) 2200uF 35V. The one that's already in place is a Jianghai CD298 2700uF 40V.

Looking at the design of this things and comparing it to the schematics of the older models it's quite clear it's still pretty much the same design. There's a good explanation of how these units work on youtube here: So all the power stuff is quite nice and simple, I don't know why they have so much control circuitry for it though, maybe because the design is almost 2 decades old :p

One thought I had is if these units could be turned in to grid-tie inverters, because from what I understand the inverter on these things is always running in sync with the AC mains input, so it might be as easy as modifying the output of the inverter in parallel with mains input and connecting your solar or wind power to the battery input. Not that I wanted to do that but just a thought

[cte7ds]

I desoldered all heatsinks and installed 8 x IRF3205s in the empty mosfet spots and used 18-ohm 1206 SMD resistors for the gates. The HRF3205s had 20-ohm gate resistors, but 18-ohm was the closest value I had so I used those. I added thermal grease to the heat sinks and bolted all the fets with M3 screws, washers and nuts. Kind of a cool trick how I ended up tightening them I did have some doubts about putting the thermal grease in because I thought perhaps it could increase the resistance between mosfet tabs and the heatsinks and that way have a negative effect, but I'm not sure how plausible that is?

Added the 2200uF 35V 105C 0.01ohm Panasonic FR capacitor in parallel with the 2700uF 40V 105C 0.02ohm capacitor that was already on board. I can't figure out why the stock capacitor is physically so huge. My extra capacitor looks ridiculous next to it.

After I put everything back together I did some load testing. The efficiency of the unit stayed at the same 80% as it was with just 8 mosfets. Ran a 230V 1000W electric heater fan for about 30 minutes and the mosfet heatsinks stayed at around 43 degrees celcius and the transformer barely had time to warm up. I did notice my extra capacitor got quite hot, around 65C. So I'm thinking maybe I need to find some better caps? Maybe I should go overkill and do something like 2 x 4700uF 63V 105C low-esr caps? :p

Interestingly the 1000W heater (measured 1005W) showed up as a 94.2% load with stock current transformer calibration value of 0xDB. The unit is rated at 980W so the calibration was quite a bit off. I changed the calibration value to 0xBC which made the 1005W load show up as 79.3%, there's still room to go higher but I'll do so if I ever need it.

The battery charger is now putting out 6.4 amps, which I think is decent enough for 120Ah battery bank alone so that I don't need an external charger with my setup. The transformer makes a really nasty audible buzz during charging. The more amps it's putting out the more loud the sound is, I believe it's normal because I can hear the exact same buzzing noise in knurlgnars videos. The buzzing noise however made me wonder what kind of a horrible waveform is the charger gonna have. I figured there's probably a lot of AC ripple in the charger output and did some AC ripple voltage and current measurements. There indeed seems to be quite massive AC ripple especially when charging with a lot of current. I included some annotated scope screen captures of my measurements. I think the charger is still "good enough" but just barely. I haven't yet done testing with the big batteries, just a 100ohm load resistor and much smaller VRLA batteries. There's a good paper about this whole subject here: http://www.emersonnetworkpower.com/en-US/Brands/Liebert/Documents/White%20Papers/Effects%20of%20AC%20Ripple%20Current%20on%20VRLA%20Battery%20Life.pdf It basically says 5 amps of AC ripple current per 100Ah of battery capacity and ripple voltage should be < 0.5% of normal float voltage. I'm guessing that even if the floating AC ripple voltage is higher (~3%) yet if it's not causing a big ripple current and there is no heating it should be ok. In any case I can always hook up the external charger that has a slightly higher voltage and that seems to really make this whole problem go away. I set the APC charger to output 27.10VDC so I can easily hook up an external charger set to 27.30VDC so it will dominate the integrated charger.

My cabling currently consists of 2 meters of 10mm^2 and 0.6 meters of AWG10 joined with a 0.2 milliohm SB50 connector. This gives me a voltage drop of 0.41 volts at 50.8ADC load. That is the draw from the battery with a 964W AC load at the output. I don't have plans to draw more than that continuously, so that is why I'm not gonna upgrade the cabling any further, even though there's certainly a need to do so. Currently the system is fused with a 100A ANL-fuse which is ok for the 10mm^2 cable but probably too big for the AWG10. The rule of thumb is 10A per 1mm^2. I'm guessing I should at least replace the AWG10 with 10mm^2.

[peter.mitchell]

I'm curious as to why the FET heatsinks have little "ears" on them, also, if it is going to be running full time, it maybe worth while to get some quiet fans to replace the ones that are in there (and add a few), so you don't go crazy and there is good airflow for continued use. Definitely swap the caps if you're going to be using it full time. Is the 10AWG connected to a different battery bank or is it an extension? 2ft at 50A is pushing your luck for 10AWG, I'd switch it, but really, it depend son how hot i gets under continuous usage.

[SeanB]

Ears are for the drain connections to the transformer, the heatsinks are used as the one connection as there is no need to try to route a second high current plane in the board then, so you an have lower losses with more copper in the board as a common plane.

[cte7ds]

I'm curious as to why the FET heatsinks have little "ears" on them, also, if it is going to be running full time, it maybe worth while to get some quiet fans to replace the ones that are in there (and add a few), so you don't go crazy and there is good airflow for continued use. Definitely swap the caps if you're going to be using it full time. Is the 10AWG connected to a different battery bank or is it an extension? 2ft at 50A is pushing your luck for 10AWG, I'd switch it, but really, it depend son how hot i gets under continuous usage.

The fan on the unit is a 24VDC 2200RPM 29.4CFM 80x80x25mm brushless fan made by Sunon (KDE2408PTS3). It's mounted so that all the airflow passes straight between the mosfet heatsinks, so the cooling is quite good already. Also I'm only gonna be using the unit during power outages and maintenance, that for some reason requires interruption in mains input. But obviously it's nice to have a device that can do more if required. The fan only spins when the mosfets have any sort of significant load on them.

With the DC input capacitors I'm not quite sure yet what I'm gonna do. I'm guessing I should go with the biggest capacitors the PCB can fit (or that I can afford). The more capacitance, voltage and lower esr, the better? right? I ordered a few capacitors in my recent tme.eu order that should be arriving in the next few days. But tme didn't really have anything that I would consider "perfect", that would be >= 4700uF, at least 50V/105C rated and low-esr. But I'm gonna experiment with the ones I got and see if they make a difference. The most suitable ones I was able to find from tme were:
SAMWHA HE1J478M25040HA (4700uF 63V 105C Snap-in) which are likely not very "low-esr" and SAMWHA WL1V338M1631 (3300uF 35V 105C low-esr) which are fairly low impedance but I would have to bodge two of them in series to increase the voltage capability.

The 2 feet of AWG10 is what is being used to get the battery connector outside of the chassis. One end is soldered to the main PCBs high current traces and the other terminated with an Anderson SB50 connector. From there it continues to the battery cable I made using 10mm^2 cable. I could simply solder the cable I made straight to the PCB and get rid of the stock Anderson connectors and AWG10 entirely. Then I would have 2 meters of 10mm^2 copper conductor going straight from the battery terminals soldered to the main PCB. I'll do some more load testing to see if this is necessary. During my initial testing the cables just got a bit warm to touch, nothing severe. The AWG10 wire is 105C rated and my 10mm^2 wire is 90C rated.

Ears are for the drain connections to the transformer, the heatsinks are used as the one connection as there is no need to try to route a second high current plane in the board then, so you an have lower losses with more copper in the board as a common plane.

Exactly, the black and white wires from the transformers low voltage primary are screwed to the ears with ring connectors and locking washers. I wonder by how much I increased the resistance between the heatsinks and drains by adding thermal paste?

[SeanB]

You probably reduced it by excluding corrosion products that had built up, and the grease is keeping the metal contacts from further oxidation as well.

[T4P]

Mate, samwhas are part of the badcaps 

[cte7ds]

Mate, samwhas are part of the badcaps 

I don't feel like paying the $30 shipping fee for a digikey order that is under $100 in value, especially when I'm not sure what kind of capacitors would be best and what effect will different esr and capacitance have. Local stores don't carry big capacitors like this and the ones on eBay although cheap are often fake and the delivery takes ages (especially now that its xmas time) at which point the unit is gonna be already hooked up and I'm not gonna be bothered to take it down to change caps :p

If someone knows what the role of these DC input capacitors is and could recommend what esr and capacitance to use, that would be great help. I'd imagine something like this would be great: http://www.digikey.com/product-detail/en/ECO-S1HA153EA/P6693-ND/131595 hehe

[cte7ds]

Decided to make the cabling better after all. I desoldered the stock battery wiring, and re-drilled the ~3.5mm hole with a 5mm drill bit. It was a double sided plated trough hole so by drilling it I removed the plating of course. So I had to scratch some of the solder mask off around the holes and solder my 10mm^2 cable to the PCB from both sides. Had to use two soldering irons at once and apply solder with my third hand :p So now the battery wiring is 2 meters of 10mm^2 soldered to the PCB and quick release battery terminals at the other end. I was able to reduce the voltage drop (from battery posts to PCB) at 49.5ADC to 0.20V (half what it used to be). The link between batteries consist of quick release battery terminals soldered to 16mm^2 cable and in between there is a ANL fuseholder with a 100A fuse. I decided to also solder the 16mm^2 cable to the fuseholder because the screw terminals on it were horrible, I had to use a blowtorch for all these connections :p

As a little bonus feature I added a JST RCY connector with a 1 amp in-line fuse connected to the battery input traces on the main pcb so I can get accurate voltage readings when the unit is back together. I also though about adding an external charger port by having heavy gauge wire going to the battery traces trough a beefy diode, but my cable strain relief thingy couldn't fit any more wires trough it.

I decided not to do anything about the input capacitor, I left the stock one in and didn't bother to add a crappy capacitor in parallel with it, figured it might just cause problems. I'll order some quality replacement caps next time I need something from Digikey so that I have them in case I ever need to/want to replace them.

I tried to do a runtime calibration but it turns out my battery bank has so much capacity that the unit simply cannot be calibrated for so much capacity. Even manually setting the runtime estimation calibration value to 0xFF in the eeprom gave me way way low estimates. The calibration also gave me an opportunity to do some stress testing of the unit. I ran my 1000W electric heater fan for over 2 hours from the inverter until my batteries were completely discharged. The fet area on the backside of the main PCB never got above 58C but at the end the Class H rated (180C) transformer was at around 100C with hot spots of 118C. It was like a frying pan and would burn fingers very quickly if touched. There was also the smell of nicely cooked transformer in the air, it wasn't smoke or anything so I didn't think it was burning. I certainly would never dare to run this unit continuously with any more load than 1000W, and even with 1000W it might not last for very long in 24/7 use... Or does anyone know what kind of abuse can a Class H rated transformer take? Is it ok as long as there is no visible smoke? :p The low voltage cutoff engaged at around 19.5VDC which was a bit surprising because I had previously specced it at <20.70VDC from my bench PSU. Anyways, when the cut off engaged, the voltage was dropping like a stone so we certainly got a 100% DOD.

I hooked up my Brymen BM869 100kHz AC+DC TRMS current meter in series with the empty batteries and the UPS to get a better idea about the charger. I got readings of 4.0AAC and 6.3ADC and the UPS was pulling ~260W from the mains with a power factor of .98

This concludes my modifications and testing of the unit so I was able to finally put the cover on and put the unit in use. For now it's feeding to the input of my old UPS system but I will be retiring the old system once I dare to power everything down for upgrades.

Here's a list of caps this this thing eats:
Electrolytic:
4.7uF 25V, 1pcs
22uF 25V, 5pcs
47uF 63V, 1pcs
100uF 35V, 1pcs
330uF 25V, 2pcs
2700uF 40V, 1pcs (room for 2)

Filter capacitors:
224k 400V red film capacitor
105k 400V red film capacitor
.33uF X2 plastic film capacitor
1uF X2 plastic film capacitor
2.2uF X2 plastic film capacitor
4n7 Y2 blue ceramic capacitor, 9pcs

Overvoltage protection:
TVR 14681, 1pcs
ZNR V14471U, 1pcs (room for 3)

[dark_hawk]

Thank you so much for this "Really" Cool tutorial.

If you had a penny for each time I viewed those pictures you'd be a very rich man.

Quick question: How did your project go? Are you still using it? Have you ran into any problems that you didn't report?

Thank you.

[cte7ds]

Thank you so much for this "Really" Cool tutorial.

If you had a penny for each time I viewed those pictures you'd be a very rich man.

Quick question: How did your project go? Are you still using it? Have you ran into any problems that you didn't report?

Thank you.

Yay, I'm glad someone found it useful

So it has now been two years since I set up the system and I just finished load testing the 2x120Ah batteries (with a normal ~100 amp resistive battery load tester). Both of the batteries showed a rating of around 1000CCA (Cold Cranking Amps). Back when I bought the batteries the reading was the same, so looks like there hasn't been any huge immediate deterioration from the float charging applied by the UPS.

In the past few years while this system has been operational I've experienced many many short power drops/brownouts and a handful of longer outages and this unit has gotten me trough all of them without any sign of troubles *knock on wood*, it simply seems to be a rock solid UPS with a nicely regulated output (big-ass iron core transformers ftw?) :-)

I've been loading the UPS with 500-600 watts and it even got me trough a 6 hours power outage last winter! I was really impressed at how long it was able to last. My only regret is that I didn't go with the 48V 2200W rackmount model I once saw selling for cheap, could've had even a longer lasting system :-P

One other modification idea that has crossed my mind is that if I ever find another Smart-UPS 1500 for really cheap, it would be interesting to take the transformer out of it and hook it up in parallel with the existing transformer. This would essentially convert the UPS into the XL (Extended Runtime) model, I think... (I remember seeing the internals of the XL model and it was just two transformers in parallel (but maybe those transformers need to be carefully matched?)). There would actually be room inside the UPS for another transformer because I don't have the internal batteries.

Anyways, I can highly recommend this UPS and the battery mod if you want something that can keep you working trough a long power outage. Also the linux support for this UPS (apcupsd) has been great to work with. You can get a small web-page that displays all the relevant details about the status of the UPS.

P.S. Someone asked how to access/modify the calibration values of the UPS so here are a few links about that:
http://www.jjoseph.org/notes/apc_smartups_battery_float_voltage
http://www.crufty.net/sjg/apc.html