A couple of small low-voltage UPS units for 9/12 V and 5 V

Sometimes you need to keep a small device, such as a router, powered without interruption. Using a regular "UPS" for this is both expensive and impractical, so low-voltage UPS units are used instead. Today I'll take a look at a couple of them.



Quite a long time ago, about 5 years back, I reviewed a way to convert a regular switching power supply into a UPS, but that involved using a 12 V lead-acid battery. Here, not only do both units use lithium batteries, but those batteries are also rated at 3.7 V.

I ordered several boards, partly as spares and partly because shipping to the forwarding service cost extra. The seller packed them so thoroughly that the packaging probably weighed more than the boards themselves. You can see the size of the packaging compared with the boards in the photo.
The envelopes are labeled in a rather unusual way: 12 W and 5 V, although technically both are correct, just in different units :)


The boards really are tiny. On the left is a stack of three boards, and on the right, two.


The same boards, shown individually.


I'll start with the 12 V UPS.
It comes in several versions. Both support selectable 9/12 V output, but one delivers up to 12 W and the other up to 18 W.
The specified input voltage range is 5–12 V, with a maximum of 16 V. There's a catch, though: if the voltage drops below a certain level, charging continues, but the unit starts drawing power from the battery.
The battery voltage is 3.7 V, with a capacity of 3–15 Ah, although a higher-capacity battery will simply take longer to charge.

Unfortunately, the 18 W version wasn't available individually, and I didn't need a large quantity, so I had to settle for the 12 W version.
The price for a single unit was $1.88, link to the product page.


A few more specs for both board versions.


Everything looks quite neat. All connections are soldered; there are no terminal blocks. The overall build quality is impressive, especially for the price.


Here's a block diagram that gives a rough idea of how it works.
Charging is handled by the SY6952 chip, which is essentially a step-down converter with charge control.
The battery is connected through an XB8089 protection controller, which protects against overcurrent, overcharging, and over-discharging.
After that comes a step-up converter based on the XR2681.
There's also a small Ph/A4 chip, presumably a voltage monitor, since one of its pins is connected to the power input and another to the output (most likely) and the feedback input of the step-up converter.
One drawback I'll note right away is the rather inefficient step-up converter, which also uses an external diode.



The board is assembled on one side; the underside serves as a heatsink, and some of the high-current traces have an extra layer of solder.


Connections are very simple: input to the power supply, output to the load, and two wires to the battery.
There are also a couple of solder pads for setting the charging current and output voltage. By default, these are 600 mA and 9 V, but the charging current can be set to 1.2 A and the output voltage to 12 V.
My only complaint is the status LED. It's intended to be a two-color LED with a common anode, even though common-cathode versions are much more common.
There's also a 1.1 A fuse to protect the power supply from overload. The built-in charger is connected upstream of the fuse.


Here's an explanation in Chinese, but the image is useful for another reason: it shows the board dimensions.



To start with, I used some old battery from a tablet. I had to replace the two-color LED with two single-color LEDs.


The red LED lights up while charging, and the green one when charging is complete.
The important thing is that once charging is complete, the controller fully disconnects the charger instead of keeping the battery on "trickle charge". The thing is, UPS units designed for lead-acid batteries usually keep the battery topped up continuously, effectively using CV mode. But this mode isn't suitable for lithium batteries, both for safety reasons and because lithium batteries have a low self-discharge rate and simply don't need it.


In the load test at an output voltage of 9 V, the board delivered 1.6 A, then shut down when the protection kicked in.


At 12 V, the maximum current was 1.2 A. At both 9 V and 12 V, the voltage regulation was excellent and barely affected by the load current; shutdown occurred when the protection controller kicked in.


After giving it some thought, I decided the tablet battery simply couldn't handle those discharge conditions, so I got a more powerful battery.


Now I can run some additional tests. At this stage, the tests were performed with no load.
1. Input voltage: 6 V; input charging current: 383 mA
2. Input voltage: 12 V; input current dropped to 190 mA because the charger is a switching type, not a linear one.
3. I soldered the jumper that sets the charging current to 1.2 A. At an input voltage of 6 V, the charging current was 800 mA.
4. At 12 V, the current was almost 390 mA.
5, 6. As charging neared completion, the input current rose to 1 A at 6 V and almost 500 mA at 12 V.

All these operating conditions should be taken into account when choosing a power supply, since it will have to power the load as well as charge the battery. If you're using a 12 V power supply, add another 300–600 mA to the load current.



The next step was to check the switching thresholds. In this case, a multimeter connected to the board's output was operating in logging mode, and the load was about 200 mA.

The voltage at the board's input was gradually reduced from 13.7 V to 5–6 V, then gradually raised back to its initial value.
Switching occurs at an input voltage of about 10.2 V, and the output voltage rises back to 12 V.
Note that if the input voltage is higher than the set 12 V, the simplified circuit design means the output will be the same voltage minus the voltage drop across the diode.


With a battery rated for a higher discharge current, the board managed to deliver about 1.8 A before the voltage started to drop gradually.


At 12 V, the current was 1.3 A; beyond that, the voltage also dropped gradually. Output power was about 16 W.


After that, I almost completely discharged the battery and ran the test again. The maximum output current at which the voltage stabilized was 1.1 A. I think that's fine, especially since the rated current was up to 1 A.


While discharging at 1 A, the converter temperature was 75–80 degrees, but toward the end it warmed up to 92 degrees.


Everything looked good until I got to the over-discharge protection. The thing is, it doesn't shut off in a latched mode; instead, it keeps trying to restart. The result looks something like this:


After a while, the converter shuts off, and all we're left with is the battery voltage through the converter's diode. This is actually the second drawback of this kind of converter circuit: the step-up converter can't completely disconnect the load.


The battery voltage graph shows that while the converter is running, the voltage drops. Once it shuts off, the load current drops (because the converter has shut down), and the voltage gradually starts to rise.


Ripple is fine: 45 mV at 0.5 A and about 75 mV at 1 A.
There are also no problems with the voltage dip when the converter starts up; the two lower oscilloscope traces show a small dip, and that's it.



Now for the second board.
This is a simpler and slightly more compact model, rated for a 5 V output (there seems to be a 6 V version too) and up to 2 A. It costs $1.17, product link.



Lots of specs, and once again they're all in Chinese :(


The circuit design here is noticeably different. The controller on the right handles everything: conversion, protection, and indication. Its markings have been erased, and the chip itself is covered with marker.
At the bottom left are a pair of battery protection transistors, and on the right are four LEDs indicating the battery charge level.


The underside of the board is empty. It only has the contact labels and specifications.


Board dimensions. It also says here that the only jumper sets the charging current: 0.6 A without the jumper, 1.6 A with it.


Connecting the board is extremely simple: it is essentially connected in parallel with the device's power line. As long as power is available, its job is to charge the battery; when power disappears or drops below a certain level, it takes over.


Like the previous one, the charger is a step-down converter, so the current drawn by the board depends on the voltage: the higher the voltage, the lower the current. Once charging is complete, consumption drops to 2–3 mA, i.e. LED power only.




The battery charge indicator works even without external power, but it only turns on when there is a load on the board's output.


The switch to battery power was disappointing, though: at the moment of switching, the voltage at the board's output drops to 4.36 V, so a sensitive load will almost certainly reboot, since the voltage should not drop below 4.75 V.
Here I gradually lowered the voltage from 5.1 V to about 2–3 V, then gradually raised it again.


With the battery fully charged, the board can deliver up to 2.8 A, despite being rated for 2.0 A, which is pretty good.


For the sake of the experiment, I fully discharged a fully charged battery. But the screenshot isn't meant to show the discharge process; it shows the periodic, small (20–30 mV) drop in output voltage. The intervals are almost identical, at one minute each, and get shorter as the battery discharges.


Near the end of the discharge, the voltage gradually drops to about 4.6 V, then the protection circuit disconnects the battery.


Unfortunately, the shutdown at the end of the discharge is also rather crude: the board keeps trying to restart, which could adversely affect the load.


At the end of the discharge, with a 2 A load, the hottest component on the board reached 78 degrees.


After fully discharging the battery, I charged it a little and repeated the load test. The maximum load current at which the output voltage remained within spec was 2.2 A.


The output ripple looks a little strange: the low-frequency ripple is modulated by high-frequency ripple. The total amplitude is about the same at load currents of 1 A and 2 A, around 90–100 mV.


I'll keep the conclusions brief today.
I liked the 12 V board, despite a few minor drawbacks. It reliably handles its rated current, and although the output voltage sags, it's within acceptable limits, so the board can be used for its intended purpose.
The 5 V version, on the other hand, didn't impress me at all. Sure, it's cheap, handles its rated load current without any trouble, and has an indicator, protection, and so on, but the significant voltage drop when switching to battery power was disappointing. Oh well...

That's about it. If you have any questions, I'll do my best to answer them.