Overclocking a UPS

I originally started this project in the summer of 2002, which was plagued with power outages. It reached a usable state in September of 2002, which was when most of these pictures were taken.

To the right you will see the poor subject of my latest experiment. This is a CyberPower Power2000 CP1500 UPS, or Uninterruptable Power Supply. If you want to get technical, it's really a Switching Power Supply, in that it runs household AC current through the outlets (not shown) in the back most of the time. Only when the AC power goes out (or you pull the plug) does it almost instantly switch over to batteries and run the outlets via an inverter. A true UPS, I am told, always runs the inverter, with household AC being used only to keep the batteries topped off.

This particular model is rated at 1500VA and 400W. The 1500VA says something about the total amount of power it can provide without needing a recharge, while the 400W says something about the amount of power it can provide at a given moment in time. Somebody who actually went on to get their Electrical Engineering degree can tell you to six decimal places what that means, but in practical terms it means that this model was one of the "biggest" sold for personal use at the time I bought it (2000?). In my case, I would typically get 20-30 minutes of run time before the batteries went dead. At least, I used to.

Seemingly overnight, but I realize now that it was after at least a couple of years, my run time decreased from 20-30 minutes to 20-30 milliseconds. That is, a minor power flicker that wouldn't even caues the VCR to forget the time was enough to shut down my UPS and, consequently, my computer system. I cracked the case and checked the batteries with my handy dandy voltmeter. One registered a healthy 14V. The other registered a sickly 10V. Obviously, the sickly one had a shorted cell or similar problem.

Meet our friend, the UPS battery. This is a sealed, gel-type 12V battery. Notice that it's over 5 inches wide and about that tall, but only 2 inches thick. Let's call that 50 cubic inches just for comparison purposes. Like most "pizza box" designs, this UPS uses two of them. In this case, they are wired together in series for a total of 24V. Unlike the average flooded lead-acid battery in your car, this one won't leak if mounted on its side (as it was in this design) or even upside down. However, like your average car battery, it is designed to supply a relatively large amount of current for a relatively short amount of time, followed by an immediate recharge and then being kept topped off.

How most home and small office UPS devices are "supposed" to be used is that the UPS keeps the computer system running through all the random outages and flickers that happen every day, especially in a rural area like the one I live in. These usually last only a few seconds. In the event that power goes out completely, the UPS kicks in, you save your work in progress, shut down the system, shut down the UPS. This UPS even came with software that does most of that for you. The UPS only runs off battery for a few minutes, then gets recharged as soon as the power comes back on.

But what if you're like me and see a UPS not as a graceful way to shut down, but as a way of thumbing my nose at power outages. I mean, if we can put a man on the moon, why can't I keep playing Quake III even after a tree falls over a power line a few miles down the road? I figured I simply needed (all together now) MORE POWER!

This is a deep cycle marine/RV battery. This battery is 10 inches wide, 8 inches tall, and 6 inches thick. For those keeping score at home, that's 480 cubic inches, or over 9 times the size of the UPS battery that it's replacing. Now, I've read that a flooded lead-acid battery like this doesn't have nearly the energy density of a gel-type, but even if it's only half as efficient, that's still 5 times the capacity. Also, straight replacement UPS batteries were going to cost me $30-40 each, plus shipping. Deep cycle marine batteries were available locally for $50 each.

(Regular automotive batteries are designed to put out a relatively large amount of power over a relatively short time. UPS batteries are designed to put out a relatively large amount of power over a short time. Deep cycle batteries are designed to put out a relatively small amount of power over a relatively long time. Notice the liberal use of the term "relatively." The power it takes to run a computer is nothing compared to the power it takes to crank a car engine. Though large for a UPS battery, it's small for something the size of an automotive or deep cycle battery. It's all relative.)

The refitting was almost trivial. I removed the old batteries. I drilled a couple of holes in the top shell over where the batteries used to sit. I connected the battery wires to bolts which I fitted through said holes. Meanwhile, outside the case, I connected two deep cycle marine batteries together in series and then bolted the cables to the same bolts. I double-checked a few things with the voltmeter, crossed my fingers, and fired it up. The UPS appeared to accept the batteries without the "Check Battery" or any other warning lights coming on. It also supplied power in battery mode, as tested by a small television. I didn't do any heavy testing because the batteries were fresh from the store and probably had very little charge after sitting on the shelf for who knows how long.

A word about wires. The internal battery wires were 12 ga. I had to lengthen them so I could get them connected to the bolts. I used some 8-10 ga wires I scavanged from some junked cars. Remember, smaller numbers equal bigger wires. Outside the box, I splurged and bought some new 4 ga cables, designed to run from car battery to starter. In the South, we call this "overdesigning out the yin-yang." After all, instead of only a few inches of wire between the battery and the inverter, there were now a few feet of wire. I didn't want wires getting hot or losing a lot of voltage, so I figured that bigger was better. Besides, the starter cables already had connectors on the ends, and that simplified things immensely.

Here's a pic of the UPS all opened up. The four outlets on the back turned out to be two regular household electrical recepticles like you'd find at any home improvement store. For some reason that surprised me.

Notice the big empty area where the two batteries used to reside. Over half the UPS is battery space. Even so, the batteries are small and light compared to what could have been used. This is a classic example of form over function. For a few dollars more, you could have 5 times the battery capacity. However, is the average consumer going to purchase a UPS that requires a dolly to get it out to the car? Is the average consumer willing to take three trips to move the UPS from one room to the other? Is the average consumer willing to check battery electrolyte levels monthly and re-fill quarterly?

In addition to capacity, I had another reason for wanting to move the batteries outside of the case, that being heat. As I've stated before, heat plus electronics equals expensive. I had only recently re-installed the software that came with my UPS. It allows my main computer to speak with the UPS via a serial cable, and reports on incoming AC line voltage, battery charge, backup time remaining, and temperature of the unit. It was the temperature part that worried me. It was telling me that the unit was running about 114F just sitting keeping the batteries topped off. After 5 minutes in battery mode, it was 140F and climbing. Admittedly, this was late Summer in the deep South without air conditioning, but I still thought that was more than a bit excessive.

When I first took my UPS apart, I must have spent 20 minutes locating all nine screws and finally prying the two halves of the clamshell apart. Even though the UPS had not been plugged in for that long, the batteries and various other components were still quite warm to the touch. In addition, there was some discoloration to the inside of the top clamshell piece. It looked like soot. I figure that the dead short in the one of the batteries caused the charging circuit to go full out for a long period of time, which caused everything to get way too hot in an enclosed space. It was even worse running the inverter. The only ventilation was a set of louvres in the back. I decided right then and there to add some cooling power, though it would take me a few weeks to get around to it.

The UPS has some really massive heatsinks in there, but they don't have a lot of nice fins like the heatsink on your computer's CPU. These are essentially solid blocks of aluminum which no doubt suck heat from the components very effectively. They are not so good, however, at releasing that heat to the air, particularly air trapped inside a small box with only a few louvres in the top. Using the UPS like it's "supposed" to be used, for only a few seconds or minutes at a time, this suffices. The heatsinks have plenty of time to cool back down. Using the UPS with two deep cycle marine batteries for a couple of hours, on the other hand, and we're talking meltdown.

I briefly considered removing the heatsinks and using a drill to honeycomb them or perhaps use a cutting wheel to add some more slots. It didn't take me long to realize that I'd probably damage some vital component during the removal process.

Though there was plenty of space internally since the batteries had been removed, I realized that just sticking a fan in there wouldn't do much good because it would only scatter the heat around inside the case. The whole case was already getting heated up anyway. Plus, there wasn't space right above the heatsinks which is where the fan needed to go. The answer? BLOWHOLE!

As you might have read in some of my other articles, a recent trend in computer cooling is one or more holes cut in the case itself so that a fan can blow outside air directly onto a component that requires cooling. This is called a blowhole because, well, suckhole just doesn't sound nearly as cool. After a bit of measurement, I found the perfect location for a pair of 80mm fans and figured it would be no problem to cut holes in the plastic case.

I cut the holes with a Dremel tool and couple of cutting discs. I want to make it perfectly clear that this is not the type of tool you want to use for a job like this. The cutting disc does not so much cut as burn its way through plastic. I had to make several passes, then go back and clean up the cut with a knife, scraping away all the melted, gummed up plastic. My hands still smell like burnt plastic. In the end, the holes looked a little rough, as you can tell from the picture. The holes could actually stand to be a bit larger, but I was afraid that I'd make them too big and there'd be big gaps visible everywhere. Enlarging them is a project for a later date.

Wiring in the fans, I made a decision not to let the wires enter the case right where the fans were located. This is also where the heatsinks are located and where all the 120V AC wiring is located. I didn't want to risk shorting something in there. Instead, I ran the wires outside a few inches and fed them through a hole I drilled in what used to be the battery compartment.

Here's the finished product. There wasn't enough room to mount the fans internally where I wanted, so they're outside. Unfortunately, I had to cut away most of the product name. I added some chrome fan grills I had left over from another project. They both look good and keep me from losing any more fingers, though they do add a little to the noise.

The fans themselves are a pair of identical 80mm no-name 12V "hobby" fans, rated at something like 0.13A so they shouldn't have a big effect on charging or battery life. Like the batteries, the fans are wired in series. They each "see" half of 24V, so they run fine. The charging circuit actually puts out about 28V, so they each get 14V and they're running a little bit faster than they should. That's actually a good thing, because they weren't moving a whole lot of air at 12V, which is why I hadn't put them in any of my computer systems yet. I don't think the minor overvoltage will be a problem and, if it is, the fans are right out where I can see them so I'll know right off if they start to break down.

Additions in March of 2003

I took out the "fussing and griping" section because it turned out that I was actually right about most of it. The UPS's charging circuits accepted the load of those huge batteries without complaint. Now that it was actively cooled, the UPS could supply emergency power for a few hours. Life was (mostly) good.

A problem was the dual UPS fans were making more noise than I liked. I tried using the "Using Battery" LED and a photocell to turn the fans on, but it didn't put out enough light to trigger the photocell. I tried using a small relay hooked to the LED wires, but I could never get it to trigger consistently. I wasn't sure enough of my electrical skills to poke around with a voltmeter looking for some place inside the UPS circuitry where the 24V from the battery pack was switched (entirely too close to where the 120V AC was located). I eventually "solved" the problem by untwisting a pair of fan wires. I figured I'd just twist them back together if there was a lengthy outage and the UPS started heating up.

Then, one day, while I was out, the power company decided to swap some equipment just down the road. Power had been out for about an hour when I got home, and the computer was still humming along nicely. There was also a faint odor of burning electronics. I checked the UPS and hot air was rising from the blowholes. I turned on the fans (twisted two wires together) and loaded up the UPS monitoring software. It told me that the UPS temperature was 158F, that the overheat level had been exceeded, and emergency shutdown was proceeding in 5, 4, 3... Well, you get the picture.

Next day, I headed to Radio Shack and found a 12V SPDT relay. DT in that it had connections for both normally closed and normally open, meaning that I could wire something to the normally closed side so that it would be turned on only when the relay wasn't getting power. I wired my UPS fans to this normally closed side.

To power the relay, I used the molex connectors and wiring that I'd chopped off an old fan. I added about 10 feet of wire which I ran from the UPS to one of my "secondary" computers which was not serviced by the UPS. In other words, when the power goes out, this computer goes dead. This computer goes dead, the relay goes back to "normally closed" and the fans turn on. That's the plan, anyway, and it does seem to work when I cut power manually.

Update May 3, 2006: After many years of faithful service, I have retired this UPS and sent the batteries in for recycling. They were pretty much dead anyway since I failed to keep the electrolyte topped off most of the time. I plan to hack the new UPS to within an inch of its electronic life as soon as I finish with some other projects.