The soldering looks good overall, except for the “blobby” wire solder joints and the fuse solder joints, which show highly inconsistent solder coverage.
Were the solder mask tracks between fuse tabs intended to trap solder and form solid bridges (as they appear to have achieved between the bottom-left two tabs) or only neatly tin the exposed metal, which you see on the top row between the bottom-right tabs? I doubt the bubbles everywhere else were the intended result.
Su’scon All Around
Just like last time, CP used Su’scon capacitors all over the place. Of the electrolytic capacitors I found on this board, the only exception I noticed is a single Taicon HF on the standby/charging supply output. Three 40A fuses in parallel “protect” the #10 (40A) battery cables and feed the four pairs of Chino-Excel Technology CEP83A3 (rated for up to 100A of continuous current at 5.3mΩ RDSon) that make up the full bridge switching ~100A through the inverter’s transformer.
Take it as a reminder that trying to generate 1000VA from a 12V source yields somewhat crazy current figures.
Surge Protection And Other Nearby Bits
Along the left edge, we have the Viper22 controller and associated components for the charging supply, a bank of three yellow 20D271K MOVs next to a 220nF EMI filter capacitor, four relays handling power routing in the middle and a pair of 2.2µF X2 capacitors smoothing out the inverter’s output.
The UPS was plugged in and splayed out overnight. When I came to take this picture, I noticed that both white relays were a little warm. The same goes for the supply section’s EMI choke. If I had to hazard a guess, I’d say at least 3W are being lost in this component cluster. Let’s see what that translates to in actual measurements.
With no load attached, the UPS draws 26.1VA thanks to its output’s relatively large filter capacitor bank, while consuming 6.2W in actual power. That's still not as good as the BE550’s 2.8W, but much better than the SMART1000’s 12W or the LX1500’s 8.8W.
If you were wondering why the current waveform is so noisy, this is attributable to the X2 caps. They're shunting noise present on the AC feed, which translates into noisy current.
How does the 1000PFCLCD generate its “adaptive sine wave” output and what does it look like? There is only one way to find out: poke around. Here, I am tapped on both sides of the transformer’s low-voltage winding and the inverter output. The little blue capacitor provides AC coupling between the floating inverter circuitry and ground. While not ideal, it's better than blowing something up if there is an operating condition where the low-voltage side may end up connected to line voltage and good enough to find out how the modulation is achieved.
What Does The Scope Say?
There you go, simple as Π: the ‘B’ transformer leg gets switched alternatively between positive and negative battery voltage, while the duty cycle of the 26µs period on the ‘A’ leg gets modulated following a sinusoidal pattern. If you look at the purple trace’s intensity grading, you get a rough idea of how the high/low time varies with output amplitude.
Plugging the oscilloscope’s CSV data output into my analysis spreadsheet under a no-load condition, the output’s THD+N is a reasonable 7.7% while the output voltage is far out in the left field at only 46VRMS. Plug in almost anything, though, and voltage rises to a more reasonable 108V. On the third day of poking around, the low no-load output voltage became intermittent, occurring less than 5% of the time.
Cycling through the LCD’s functions while the 1000PFCLCD is on battery power, with the wacky output voltage showing up on the scope and my multimeter, the UPS still claims that its output is at 120V. It looks like the UPS may be hard-coded to report that figure on battery power instead of actually monitoring its own output voltage.
1000PFCLCD vs SL300
What happens to the output waveform when I plug in my SL300 PSU? The output voltage rises to 108VRMS, THD+N drops to an even better 4.4%, and the built-in display still reports 120V. If this was truly a 120V sinusoidal waveform, it would be peaking at 170V, not 150V. My spreadsheet tells me that the SL300 uses 64.5W and 89.6VA, while the 1000PFCLCD reports 66W and 90VA, which is within my measurement setup’s error margins. Oddly enough, this seems to indicate that the UPS’ power calculation may be based on real output voltage instead of the bogus 120V being displayed. If 120V was hard-coded in the calculation as it appears to be for the display, I would have expected my measurements and the UPS’ to disagree by about 10%.
What Have We Learned Today?
What goes into a $120 “adaptive sine wave” UPS? For the most part, exactly the same stuff that goes into a $50 regular “stepped approximation sine wave” UPS. CyberPower chose to use a full bridge driver with a single low-voltage transformer winding, but it could just as easily have gone with a center-tapped winding and only low-side drivers for each half. In other words: any UPS with fast enough FETs and FET drivers could potentially be converted to sinusoidal output with little more than a firmware upgrade and X2 capacitors on the output.
The overall build and design quality is on par with the LX1500, earning a point for including a #14 cord and then losing it for going cheap on the connectors. The biggest flaws I can find are falsely reporting the output voltage as 120V on battery power and failing to provide an output voltage within 5% of 120V under the same circumstance. Of course, there is also a question mark attached to those Su’scon capacitors, just like there was with the LX.
If you wanted to know the hardware differences between sinusoidal and non-sinusoidal UPSes, now you do: not much.