What Difference Does It Make?
A picture is worth a thousand words, so I fired up LTSpice and did a quick mock-up of a flyback circuit. In red, you have the output waveform using a capacitor with 60mΩ of ESR, which looks quite similar to what we saw across the Lelon capacitor. In green, we have an ESR of 25mΩ (typically in the form of two smaller devices in parallel) teamed with a 10µF MLCC to handle the high frequency content, as you will often find in higher-quality designs. If we could buy an ideal capacitor and put it in, we would get the cyan waveform. My mock circuit operates in open loop with a 10Ω load, and the differences in DC offsets are representative of reduced capacitor losses, meaning the 25mΩ and ideal capacitors make the circuit about 1.7% and 3% more efficient, respectively.
Every time you buy a device containing a power supply where manufacturers shaved pennies on capacitors, this is one of many compromises being made.
Analog And Digital Support
There is not much to get excited about on this half of the board: operational amplifier, linear regulators, tons of surface mount resistors, capacitors, diodes and transistors, a handful of Chang capacitors, and a 48-pin QFP micro-controller with a sticker identifying the firmware as “REV1 ©2011.” Just above the USB board header, between the bi-color (red/green) LED and buzzer, sits a lone OST 25V 22µF capacitor.
If you weren’t counting, that’s four capacitor brands so far, all on the lower half of the reputation ladder.
The Business End – Primary Side
The inverter involves a rather large chunk of the board, starting with the battery, the 30A battery fuse, a current measurement shunt (the thick piece of wire below the large capacitor), a CapXon 1000µF 25V capacitor to provide bypassing, and the DC-DC boost converter’s ST P55NF06 (60V/50A/18mΩ) FETs driving the boost transformer in a double-forward topology.
Just as it was on the AC input and output monitoring side, the boost converter’s DC output voltage gets sensed through a resistor network, this one tucked away along the bottom-left edge.
The Business End – Secondary Side
On the output side, we have a discrete diode bridge with snubber capacitors charging an OST RLG 450V 4.7µF capacitor, a handful of support components generating gate drive voltages for the output IRF610 (200V/3.3A/1.5Ω) FET bridge and their control opto-couplers, followed by an inductor and 100nF X-class capacitor providing some degree of output filtering. I do not like that OST capacitor’s survival chances: there is no inductor on the output of this double-forward converter, which means that the capacitor will get subjected to some harsh ripple current.
There you go, the internal tour is complete. This is another fine example of a generally great design brought down a few pegs to save $0.50 on parts.
How Electronic Inverters Work
In a nutshell, electronic inverters work by doing about the same thing a typical power supply does, but in reverse. Instead of taking line voltage, rectifying it, stepping it down through a high-frequency transformer and rectifying it again, DC voltage gets turned to AC through the high-frequency step-up transformer, rectified to HVDC, turned into line-voltage AC by the FET bridge operating in diagonally opposite pairs, and then lightly filtered across the load.
With fast enough FETs, it would be possible to use the exact same circuit to generate arbitrary waveforms by adding the appropriate modulation to either the top or bottom FETs, exactly as the CP1000AVRLCD did with its transformer.
Idle Efficiency
Why bother with the complexity? Because electronic inverters are far more efficient at light loads than iron-core inverters are. When there is no load, they top off their output capacitors and that’s it. Iron core transformers, on the other hand, have a significant magnetizing current that they require regardless of load. At light loads, this accounts for the bulk of the power draw.
I put my multimeter in series with my spare UPS’ batteries to have a look at their respective current draw with no external load attached. The BGE drew 200mA and could theoretically keep going for over 20 hours, while the CP drew 2A and would go dead doing nothing within four hours (despite having a 33%-larger battery). Am I being unfair by comparing a 125VA UPS to a 1000VA one? Not really: my ancient BX1000’s inverter still works correctly and it drew 325mA at 24V, less than a third of CP’s power.
Standby Power
With the unit off (charger only), the BGE90 draws 2.5W and 6VA versus 3.5W and 8.3VA in normal standby operation, which is 700mW worse than the BE550’s 2.8W and 5.6VA in standby. At first, you may wonder why APC could not be bothered to put a more efficient standby supply in there. But then again, this UPS does have a pair of USB ports with a combined maximum output of 2.5A powered by a DC-DC converter, which means that the standby supply also needs to power the 12.5W USB outputs.
Do Not Try This At Home
If you thought that using a UPS after a power strip was cringe-inducing, fasten your seat belts and hold on to your hats!
When I got around to looking at output waveforms, I ran into a little issue: the BGE90 is not intended for running devices that require grounding, and therefore only has two-pin polarized outlets. I used my simple and unsafe “universal lamp cord to alligator clip adapter” to get around this safety measure.
This is actually small-fry stuff compared to earlier in the piece when I had my hands millimeters from a live PCB.
Output Waveform
Instead of the 160-180V flats seen previously, the BGE90 only outputs 130V while delivering 64W, which seems a little low. Factor in the 72% duty cycle shown here, and the effective output voltage becomes 107VRMS. In the THD department, this waveform scores a surprisingly good 27% while under load. I added extra loads to see if I could cause an overload condition, but ran out of outlets at 88W, 13W over the specification and definitely more than sufficient to run the modem, router, and ATA it's primarily intended for.
I can think of at least one advantage to such a low peak voltage waveform: many small power adapters are still linear (use iron core transformers), where the lower voltage reduces the amount of power loss within them.
It Needs To Be About 20% Cooler
There aren’t many bad things to say about the BGE90M-CA. Its mechanical construction is sound except for the loose battery compartment, the wire gauges and breaker are in line with what the unit is intended for, and the board layout and assembly quality are as clean as they come. The single biggest issue is APC's choice of third- and fourth-tier capacitor companies, especially that lone Lelon eating the output of the UPS’ flyback power supply. I was thinking of buying type-A to barrel cables to power my modem and telephony adapter from the USB power ports, but if I did that, I doubt the Lelon capacitor would survive more than a year of 24/7 ripple exposure. Does that sound like a dare or a challenge?
In all seriousness, that 3.5W the unit is dissipating with nearly no ventilation does cause it to become noticeably warm, and I’m certain APC could have reduced its standby power consumption by more than 20% with a few design tweaks. Just look at the BE550G’s 2.8W.
Those two potential sore points aside, it should be one handy little UPS to have.
MORE: CyberPower CP1000PFCLCD Tear-Down
MORE: APC BE550G Tear-Down
