The soldering looks good overall, apart from the Tripp-Lite’s trademark solder flux residue all over the board. The only two noteworthy components, aside from the passives and discrete ones, are an LM324 and a USB micro-controller.
It looks like the wave soldering process at Tripp-Lite’s factory might require a little tweaking. It missed a strip of bare copper between the two negative battery connections. Solder failing to stick in the wake of large component pins is a common manufacturing problem. From the amount of flux around the bottom wire, I’m guessing the board has been reworked due to a cold solder joint or joint under-fill on that wire. The top wire’s connection looks great.
When dealing with electronics, you often see chip markings defaced to make reverse-engineering more difficult. But when you are a large company and order chips factory-programmed by the tens of thousands, you can spare yourself that trouble by adding custom markings to your order. Here, Tripp-Lite had the chips marked with its name and firmware version. At a glance, the pinout appears to match the Microchip PIC18.
Surge suppression and filtering are provided a trio of 20D271K MOVs and a 100nF X2-class capacitor in the background to the left. Behind the MOVs, you can see the bottom of a ½W 680Ω resistor next to an axial film capacitor, the yellow part slightly visible behind the right MOV. At first, I thought it may have been a capacitive dropper power supply for the control circuitry. But after following the traces, it turned out to be a simple snubber between the AC input and UPS output to reduce arcing when relays break contact.
Cheesy Capacitor Brands Of The Day
Peppered throughout the board are Gemcon- and Jamicon-brand capacitors. There is not much information about Gemcon in consumer electronics, but Jamicon is considered slightly better than Fuhjyyu or Su’scon.
Can you hear that noise? That’s the sound of my confidence crashing.
Switching 80A is too much for a single FET to handle reliably, so each heat sink gets two P80NF55s. With 6.5mΩ of on-resistance, that’s up to 20W getting converted to heat by these devices. Because each of the two heat sinks is responsible for driving one side of the “PWM sine wave,” this reduces average dissipation to 10W per sink.
In the LX tear-down, some readers wondered if a UPS could run continuously if given extra batteries, and I answered that components in most consumer UPSes are not sized to support such a taxing workload. This is a prime example of heat sinks just large enough to handle the stock battery under significant load.
The Weakest Link
So, what's the missing function? How about a power supply for the monitoring circuitry and charging the battery? Instead of a dedicated switching power supply, which costs less than a dollar in parts to implement, Tripp-Lite powers the SMART1000LCD through the inverter transformer’s battery winding. This means the UPS’ 1000VA transformer is always in-circuit in every operating mode other than off.
Do you know what else it means? That the UPS cannot power up with a dead or missing battery since it needs battery power to activate its AC input relay and energize the transformer.
Let’s see what this translates to in measurements.
How much power does this monstrosity use while turned on with no load attached? I measured 26.8VA, 21.9 of which at the fundamental frequency, an integral power of 12.4W and a power factor of 48%. Basically, half of the power going in serves no purpose other than magnetizing the transformer core. A large chunk of the other half is copper and iron core losses incurred while doing so. At least 12.4W is better than the 18W specification on the labels.
Why is the waveform so odd? The first current peak between 0V and 90V is caused by capacitors and the battery getting topped off. The UPS had been plugged in for over 10 hours though, so the battery should have been as charged as it was going to get. Once that initial peak is over, current dips to whatever the core’s magnetizing current is at. The second peak occurs when line voltage reaches 0V as expected since AC current through an inductor lags 90° behind voltage.
With my SL300 connected and powering the same 80W LED strip as last time, I turned off power to the UPS using one of my spare power strips. Unless Tripp-Lite has different versions of the SMART1000LCD for the USA and Canada/Mexico, its “PWM sine wave” translates to the usual bipolar rectangular waveform, and is practically identical to the LX1500’s output.
When I paid $160 CAN for this unit, I was expecting exceptional design and component quality to justify the price tag for its modest specifications compared to slightly more expensive units. Aside from the great battery and neat movable display though, there are too many important things to dislike about Tripp-Lite's SMART1000LCD:
- high standby power draw due to using the inverter transformer for operating power instead of a power-efficient standby power supply
- high standby temperature due to said high standby power dissipation
- third-tier capacitors
- no breaker
I was originally planning to give this unit to my mother to power her VoIP installation, but with its horrible 12W of standby power, this UPS would waste more power than the modem, router and ATA combined. I would not normally return a tear-down guinea pig to the store out of respect for whoever might end up buying it, but in this case I simply do not want to encourage the manufacturer by keeping it.
By the way, and for comparison’s sake, I wrote in the LX1500 story that it drew 9W. That was actually incorrect: the unit was drawing 9VA while its real power was 4.5W. That includes its built-in USB power supply.
My plan for the next installment? Get the cheapest UPS I can find and see how it fares compared to the SMART. Want to bet on whether the sub-$100 UPS will win?