When power bars with surge protection are not enough, and a little security from power loss is desired, the next step up is a UPS. APC sold more than a dozen different units in this form factor under various model numbers with similar feature sets. In my case, the original retail box says BX1000, the battery compartment cover says XS1000, and the USB device identifier string says RS1000. That's a nice little chimera.
The status-and-control panel has LEDs corresponding to On-Line, On-Battery, Overload, and Replace Battery. Below those LEDs lies the single On/Off/Cold-Start button. You get little more than the essential stuff.
The panels look exactly the same on both sides, with little more to see than APC's logo and an indentation for the floor or desk vertical stand.
On the rear panel, starting from the top, we have a USB interface, though APC decided to use an RJ45 jack instead of a type-B connector, a fan-less vent, a wiring fault indicator LED, the phone line protection jacks with a supplemental TVSS ground screw, a pair of surge-only outlets, six battery-backup outlets, a built-in breaker with an unspecified rating, and a 12 A input rating printed above the cable entry point itself. Again, there's nothing fancy to see; it's pretty much what you expect from a UPS' rear panel.
Instead of sticking model-specific labels on its units, APC printed basic ratings for all of the products using this housing (in both 115 and 230 V flavors). Verbiage on the right repeats the familiar "risk of electric shock" warning we see on practically all line-powered items and the "no user-serviceable parts inside" we see on just about everything.
The latter part of the print gives us some potentially useful information on the non-sinusoidal output: up to 67 percent THD with the strongest harmonic up to 40 percent. This is bound to add some depth and warmth to the 120 Hz hum on your favorite audio amplifier. On a more serious note, significant THD can confuse the active power factor correction (APFC) circuitry in some power supplies.
With the external details now covered, it's time to start stripping.
The battery cover hides this beefy-looking connector. At first, you may think there must be something fancy about it. But upon closer inspection, the housing is nothing more than a rig to help people safely plug and unplug the connector's three blade terminals, since the battery might not be isolated from mains or otherwise high voltages. APC reminds you of that by molding the warning on the shell itself.
The battery door also hides three of the 10 screws holding the UPS' enclosure together, two of which are fastening the status panel to the middle of the side panels and the other coming in from the side at the bottom-left corner. Popping the interface panel off reveals a fourth front screw coming in sideways in the top-left corner.
Right about now would be a good time to disconnect the battery and remove it (if possible) before diving in any deeper.
This is one of my BX1000's two long-dead original batteries; I simply haven't gotten around to dropping them off at the nearest Staples for recycling yet. If you are wondering why it appears somewhat oddly shaped, valve-regulated lead-acid accumulators (or VRLA), along with most other battery types, tend to bulge as they get old due to gas build-up. Pressure slowly stretches the casing, much like old and underrated electrolytic capacitors. The bulging is exaggerated in this shot; it does not look half as bad in person. But the effect was still bad enough that I had to partially disassemble the UPS to get the original batteries out when warnings started appearing. This is a somewhat common issue with UPS' that use snug-fitting slide-in batteries.
When you buy an UPS, keep in mind that typical lead-acid batteries have a shelf life of about five to six years under light use. Plan to replace them about that often if you want to minimize surprises, and even more frequently if your UPS kicks in on the regular.
Before you reject the original manufacturer's expensive replacements and substitute your own, compare specifications to get an idea of how well or poorly your improvised alternates will perform. I skipped this step when I bought the pictured batteries, since they were the only ones that fit at the local electronics store, and I was mainly interested in making the UPS stop complaining.
Since I only have about 200 W worth of equipment connected to the UPS, my ~$15 cheapies worked well enough for the last three or so years. However, I only get about four minutes of backup time instead of the 15+ minutes I should be getting with the OEM batteries, which tells me I am about due for a new pair. Without the plastic spacer frame between batteries and the large, tough stickers holding everything securely together, this hack job can be a chore to get in and out—one more reason to consider a proper replacement pack.
If you are curious to see how battery choice affects performance, proceed to the next slide.
I could not find specifications for my Solex SB1272, so I went hunting for the cheapest battery I could find as a plausible worst-case stand-in. The original CSB GP1272 can be found for $30 to $40, the BBB BP7-12 for $25 to $35, and the PowerSonic PS-1270 for as little as $12.
Different battery manufacturers report their products' performance in different formats. Here, I aggregate CSB's and BBB's 15-minutes discharge ratings and paste part of PowerSonic's graph, since the company does not provide tables. Since all three batteries have a nice common comparison point at or near a 14 A constant-current discharge rate, I extrapolated the other two other batteries' 14 A discharge curve on top of PowerSonic's graph to compare. Ideally, I should have used constant-power curves or tables, but PowerSonic does not provide either of those for this battery model.
The PowerSonic unit drops completely dead near the 12-minute mark, while the BBB unit runs out of steam at 15 minutes. But the CSB only starts to trail off, with maybe four or five minutes more to spare. In a constant-power discharge scenario, in which current rises as battery voltage drops, things would get much worse for the PowerSonic unit after the six-minute mark.
If you want the best performance and battery life out of your UPS, CSB is the mark to beat out of the three batteries I looked at. No wonder APC uses its batteries in so many small-footprint UPS models.
Having trouble unlocking your "slightly used" UPS to take a look inside? At the top of the BX1000, there is a snap about halfway in at the top holding the shell's two halves together. Its location is circled in red on this picture. You need a long, flat screwdriver to easily reach and disengage it.
Before you poke your hand in there, you may want to hit the Power/Mode button a few times after unplugging the unit. While I was working on the case, I accidentally hit the power switch. The unit beeped even though it had been disconnected for over half an hour. There was most likely harmless leftover charge in low-voltage circuitry, since I seriously doubt APC would forget bleeder resistors on the high-voltage side of things. In any event, it's always good to discharge everything you can before poking around, if only to prevent voltages from being applied to the wrong place in case of accidental shorts or other unplanned contact.
How do you like those spade connectors? APC appears to like them a lot. Everything except for the front interface module connects to the main board using crimped spade connector plugs with rubbery boots. Gotta love teardown-friendly hardware, as teardown-friendly as stiff spade connectors can be, anyway — nothing a decent pair of pliers and a little force cannot overcome.
If you have a penchant for tearing stuff apart without the diagrams to help you put them back together, it is a good idea to take pictures as you go for reference. In this case, APC was nice enough to use different wire colors to make connections and write them on the PCB's silkscreen near their respective spade headers. This will be really handy for reassembly.
All capacitors large enough to carry a logo or brand name are Nippon Chemi-Con, except for two HJ-branded 22 µF 200 V devices along the high-voltage output path.
Here's the front panel and its main board attachment mechanism. There are no detachable connectors with exposed conductors; APC uses clinch termination on the unstripped ribbon. Each terminal has eight nails punching through the cable, which are then flattened on the other side. That's not a common sight these days.
Once the spades are disconnected, the rear panel—complete with power cord, breaker, and outlets—comes off. While it is not easily readable in this picture, I can tell you that the breaker is a Rong Feng unit rated for 15 A at 250 VAC, which contradicts the 12 A input rating printed on the rear cover.
For those of you who were worried about the quality and reliability of outlets on APC's UPS after reading my article on the vintage SurgeArrest (where I complained that the middle outlets were nearly unusable), you’ll be relieved to know that the metal strips in the UPS appear to be slightly thicker. But even more important, instead of using the offset-stamped metal strips that I had issues with on the power bar, the UPS uses bent metal fingers for all connections on every outlet. The fingers have a slight angle to pinch plug prongs as they are inserted, providing a more even insertion and removal force. You have a clear shot at one of those just between the orange and ground wires, unhindered by the plastic framing around usable outlet positions.
Unlike the power bar, where electrical connections on outlet metal strips were spot-welded, these are soldered. All of the solder joints I looked at were clean and smooth, exactly as they should be.
This little critter resides in the middle of the rear cover, atop the rows of power outlets.
Two 600 150 FLEI (Or is that FLE1? Either way, I could not find specifications) devices in series with the phone input lines, one GNR 10D271K MOV directly across lines, and one more from each line to ground make up the BX1000's phone line protection circuitry. Two-hundred-seventy volts seems like a high clamping voltage for the phone line compared to the SurgeArrest's 175 V. There's no soldered ground wire; the ground wire connects on the tab partially hidden by the MOV labeled MV11.
The silkscreen implies that these two parts from Taiwan are positive thermal coefficient devices of some sort, and sure enough, their resistance does increase with temperature. It looks like these are intended to play the role of self-healing fuses.
The BX1000 comes equipped with APC's Automatic Voltage Regulation. AVR functionality works by inserting this transformer wired as an auto-transformer in the AC signal path to either boost line voltage by a few volts when wired in-phase or buck it by a similar amount when wired in anti-phase. Two relays control which configuration, if any, gets activated. It's a little crude, but the good old iron lump is a simple, reliable, and efficient way of doing coarse voltage regulation. Many dedicated AVR units are fundamentally the same design, but with multiple taps on their transformer for finer-grained adjustments across a wider input/output range.
In the top-right corner, we have the input noise and surge suppression filter that many enthusiasts are already familiar with from seeing similar device configurations in power supply reviews. Halfway across the top, there's a small component cluster consistent with a simple low-power switching power supply. Farther to the left, we see six thick, aluminum plate heatsinks and a substantial high-frequency transformer, hinting that something big must be going on over there.
Another pair of heatsink plates draws attention to the bottom-right corner, though their purpose is not immediately obvious from this angle (unless you know exactly how this type of UPS works). All heatsinks are crowned with a clear plastic clip to maintain spacing between plates; APC is not leaving anything to chance.
The bottom-left quadrant of the board contains the monitoring and control magic, with the USB interface micro-controller located along the bottom-right edge.
With the main board yanked out, the housing is stripped bare.
Both halves are extensively ribbed for stiffness, and the outer walls are thick enough that the recessed APC brand on the outside does not show through on the inside. Extra ribbing is present around the iron-core transformer and battery areas to help spread mechanical stress from their weight.
Near the center of the right half and above, where the main PCB used to be, there's a thin coat of soot-like residue. But I saw no immediately evident source on the board. My best guess is that a "hot potato" has been slow-roasting dust and possibly other materials, and the natural convection caused this fine particulate matter to slowly accumulate on the housing over the past 10 years.
Looking good for a decade-old board, isn't it? The PCB looks exactly as it should for a wave-soldering job, except for a slightly darkened area just above the hole near the center, indicating that whatever lies on the component side uses PCB copper pours as its heatsink and dissipates a fair amount of heat. This might explain the soot-like residue on the back cover, and the location is about just right, too. What could our mystery device possibly be?
Since the PCB blemish is the only detail that grabbed my attention, let's take a look at it first.
Our toasty chip here is a KA7812R, a Fairchild Semiconductor +12 V linear regulator. As anyone familiar with linear regulators can tell you, it is common for these little beasties to run hot. In this case, it likely takes the output voltage from the switching supply presumably used to charge the batteries and power relays, which means about 27 V, and drops it to 12 V for lower-voltage circuitry, such as operational amplifiers and possibly the low-voltage MOSFET drivers on the battery-powered side.
On both PCB surfaces, it is fascinating how the discoloration stops almost exactly at the copper pour boundaries, demonstrating how much more effective paper-thin copper pour and via-stitching (the row of tiny plated holes through the PCB around the IC) are at spreading heat than fiberglass alone. The technique becomes even more effective in board designs with one or more ground and power planes.
All devices mounted on heat sinks are riveted in place and won't go anywhere. Components with incompatible tab connections get the insulation strip and riveted retaining steel clip treatment, which hides most of the device information. In this picture, the slot on the leftmost device happens to line up with the device model, revealing it to be an IRF640.
Writing on the board reminds us that high voltage may still be present, even when the unit is unplugged from the wall, if the batteries are still connected. Another warning along the right edge of the board informs us that heat sinks may carry live voltage. Good thing the line power and battery were disconnected for a while before I even started removing housing screws.
Just like nearly every other line-powered piece of electronics, the BX1000 has one X-capacitor on each side of its large input choke, four Y-capacitors, three 20D361K, and one S18K150 MOVs, providing filtering and surge suppression along the top edge. Three hundred and sixty volts seems high to start clamping in a 120 V UPS. But then again, battery-powered loads get switched out when line voltage exceeds the configured limit.
Below those, we see the battery-power transfer relay, the two AVR relays, two surface-mount resistor ladders likely used by the micro-controller to sense line input voltage, and the wiring fault indicator LED. The small yellow transformer farther down provides current sensing for the battery-backup outlets, enabling the micro-controller and PC software to calculate how much power the attached devices are using.
Even power supplies need some power for themselves, and the BX1000 gets its juice from a TOP234Y highly integrated fly-back switching regulator rated for up to 45 W output on 85-265 VAC line input, which is more than enough to charge the UPS' batteries and run everything else. At a glance, the circuit configuration looks like one of the TOP23x reference designs, give or take a few application-specific tweaks.
The stubby line input capacitor is about half the height of a typical 680 µF 200 V device, but actually is only a 47 µF part rated for 450 V, which is surprising large for such a low capacity. The 450 V rating was also unexpected since there is no input voltage doubler; a 200-250 V device would have been sufficient. Continuing the overrated component theme, the surface-mount diodes are S1K types rated for 800 V reverse voltage and 1 A forward current. It looks like the circuit is clearly designed for reuse across APC's 115 and 230 V products, and the company decided to forgo shaving pennies by tweaking the BoM for individual line voltage ranges. That's not surprising, considering how the housing, and presumably the PCB, span at least two rating models in both 115 and 230 V flavors.
This board area may not look particularly exciting, but the micro-controller hidden under that sticker controls battery charging, AVR relay configuration, the high-voltage DC-DC converter, the line/battery transfer relay, the high-voltage "modified sine wave" output MOSFET switching sequence, and a piezoelectric buzzer to wake you up at 2 A.M. when there is a power blip if you forget to either turn that feature off or at least configure a delay on it.
Also present in the area are a few operational amplifiers to help with monitoring various signals, such as battery voltage, output voltage, and output current; a PWM controller for the battery side of things (more on that later); and a ton of (de-)coupling capacitors.
Here are those six aluminum plates and transformer again. Each of the plates on the left carries a lone IRF3205 HexFET device rated for 55 V and 110 A driven by IR4427S dual MOSFETs controlled by a MicroSemi 3525ADW pulse-width modulator located half-way across the board — you can see it in the bottom-left corner of slide 17. Devices on the plates to the right have two pins each, giving them away as diodes, and the board's copper artwork says they are in a full-bridge configuration. Rectifiers on the front plate are hidden under a riveted steel clamp, but the second row showcases OnSemi U820 markings, which is short-hand for MUR820 rated as ultra-fast rectifiers for 600 V and 8 A. Near the bottom-right corner, we have the rectifier bridge's 22 µF 200 V output capacitor and a pair of 470 kΩ resistors in series providing 940 kΩ of bleeder resistance.
Nestled between the four left heat sinks are the battery spade terminals, a pair of 35 V 2200 uF capacitors (to decouple high-frequency switching ripple from the batteries), and a pair of 30 A fuses in parallel configuration for 60 A combined rating (to accommodate the 50+ amperes the UPS may draw from its 24 V battery pack under full 1000 VA load).
If you have not guessed already, this circuit configuration is a 1000 W peak battery-voltage to 160+ VDC step-up converter. I could fire it up to make actual measurements, but guessing is much safer than poking around with the board upside down — that is, assuming the UPS would even work with everything else unplugged.
This is what a "modified sine wave" output looks like from a PCB layout point of view: ~160 VDC comes in from the step-up converter through the fat traces at the top-left, pass under a large Falco-branded inductor, gets decoupled by another capacitor before reaching the UPS' six output MOSFETs—one IRF640 rated for 9 A and 200 V with 350 mΩ RDSON, along two PHP20NQ20T rated for 20 A and 200 V with 120 mΩ RDSON on each heat sink. All of these MOSFETs are driven by a trio of IR2101S dual independent high-side-low-side drivers, which are, in turn, controlled directly by the micro-controller. The MOSFETs' output passes through the largish inductor near the upper-right corner, and then the yellow current-sensing transformer, before reaching the output.
Since driving a modified square wave should only require four switching devices, you may wonder where the two weaker IRF devices fit in. After following the traces, it appears the two IRFs are actually there to switch that Falco inductor in and "snub" inductive load current from the high-side devices as they turn on or off. In-circuit observation with an oscilloscope would be necessary to find out exactly what the drive sequence is. At the very least, the IRFs are definitely related to the Falco inductor, since they form a half-bridge driving one of its two terminals and nothing else.
Managing to get part numbers off those second-row TO-220s was more challenging than it sounds. I could not do it by eye due to extremely shallow angles. But after some trial and error with my camera and lighting, I managed to get readable shots without seriously straining my vision. The U820's markings were particularly difficult to get, with the laser etching disappearing under the tiniest amount of dust until I found the right lighting angle to make it somewhat readable.
For a 10-year-old UPS, this BX1000 has held up quite well. Aside from some discoloration around the KA7812 and the soot-like residue that accumulated behind it, the rest of the board and its components still look good-as-new after a little bit of cleaning. The unit would likely also work good-as-new if I got a pair of replacement batteries that met or exceeded OEM specifications instead of grabbing the most convenient things that fit.
To end this story, here's a nice picture of shiny, candy-colored components I ended up not using elsewhere. You can see how APC labeled spades with a header identifier, wire color, and function, along with the TOP234Y riveted to its heat sink.