Note: Potential Issues of Using a USB Powerbank as a UPS

On my recent review of the Blitzwolf BW-P12 10000mAh Power Bank, a reader commented asking whether that particular power bank could be used as a makeshift uninterruptible power supply (UPS) for USB-powered devices. This is an excellent question and one I’ve always wanted to convey some of my thoughts about, so I decided that it deserved its very own post.

Does it? Does it not?

If you want a simple answer, unfortunately, you’re usually out of luck. When buying a power-bank, rarely do they say very much about whether they work as a UPS or not. Increasingly, products might occasionally state that they support pass-through charging, but that is not a universal practice and what that means in practice can affect certain applications.

For example, the Blitzwolf BW-P12 is a great example – the manufacturer’s website does not mention pass-through charging, but the iSmartware SW6208 datasheet states the following:

The SW6208 supports charging the battery and supplying extern device simultaneously, and this function can be disabled if not needing. When only one port turns on, it supports fast charge input and output. When two or three ports turn on, it only supports 5V input and output. In this situation, charger hold voltage is set to 4.8V. if input voltage drops due to the weak power supply of adapter, charge current will automatically fall down so the input voltage holds to 4.8V and supplies extern device in priority.

It would seem that from the datasheet, unless explicitly disabled by the manufacturer, the controller would support pass-through charging capability.

But what does this mean in practice? The easiest way to know is to try it out. I enlisted my “on-loan” Rohde & Schwarz NGU401 SMU to act as a load on the output of the BW-P12, sinking a constant 500mA to simulate a moderately light load. I used my ZMI zPower Trio USB charger to provide power to the BW-P12 while monitoring the output voltage using a Rohde & Schwarz RTM3004 oscilloscope (although FastLog would have been sufficient for this purpose as well).

The results are interesting to say the least. The above is the trace from a USB-C load attached to the power bank while a USB Micro-B charge cable is attached. The output voltage collapses to about 4.2V for 80ms, then down to absolutely nothing for 960ms. This means that whatever load is connected will see its power out of spec for about a second during a charger plug-in event. The output power does return after this, but sits at around 4.8V as the system tries to “balance” the power going into the battery and out to the load without overloading the power adapter. The output voltage does have a loss because of the resistance of the MOSFETs in the power path as well.

Doing the same but instead having the load on the USB-A port and the charger be on the USB-C port results in a slightly more favourable result of no output for about 125ms. This is a much shorter time, but when combined with the 80ms of 4.2V, means that the attached device will be exposed to out of specification power for about a fifth of a second.

Herein lies the problem – just because you have pass-through charging does not mean that your output is uninterrupted. If you’re just powering some system that needs to be running and could handle short power interruptions, this might be acceptable (subject to other caveats which I will explain in a moment). But if your device absolutely requires uninterrupted power, such as a Raspberry Pi that would incorrectly shut-down with possible data-corruption results, then this is not going to work on its own. You could design something to work around this to some degree – perhaps a nice Schottky diode feeding a large supercapacitor bank with some charge current limiting to avoid over-current during charging, but now you’ve literally designed yourself a small UPS!

Design Matters

Perhaps it is a good question to ask ourselves why this behaves the way that it does. The typical application schematic in the datasheet is of great help.

In this particular design, all the inputs and outputs are essentially commoned together into one net, but each of the inputs and outputs has a MOSFET switch allowing for some control and is directly connected to the IC so it can sense if the port is attached to a charger or load. The IC is also the only switching controller in the whole show, so it’s going to be busy thinking about what to do.

From what I can gather, while supplying a load, the unit is boosting from the Li-Poly cell to generate a steady output at the requested voltage. Once a charger is attached, the IC can detect the presence of voltage and signalling lines – it has to stop boosting the voltage (hence the drop to ~4.2V), turn off the output (well, not necessarily but it does), start the charging into the cell (perhaps acting like a buck converter) and then turn the output back on. This power path management takes time, in part to ensure the right currents/voltages are monitored and that inrush is controlled, thus the output drops-out for a bit.

But it does not have to be like this – some of my favourite Xiaomi power banks are not designed this way. Take for example, the Xiaomi Power Bank Pro – this one does not use an “all-in-one” chip and is actually built on a more “discrete” principle.

By that, I mean that the unit has a separate charger and a separate boost converter, all orchestrated by a microcontroller. This means that they can operate independently – the two inductors also means that it’s capable of different voltage fast-charge in and out simultaneously. The price for this is additional complexity.

Indeed, I tried measuring this on the oscilloscope when the input was being connected and disconnected – absolutely no interruption whatsoever. A straight line is not so exciting, so I didn’t bother with a screen-grab.

Oh My! Look at the Time!

Having just discrete elements is not sufficient to keep things running smoothly. The simple matter of “balance” means that you need to ensure the charging rate is at least as fast as the depletion rate, otherwise the pack may deplete itself under load. This is especially relevant to older generic power-banks which may have used cheaper TP4056 linear regulator charge ICs which often had currents of 800-1000mA into the cell.

But you can also be bitten by another “feature” of charger ICs – the charge safety timeout. Most charger ICs on their own expect to be charging a cell without any additional load on the cell. However, if you’re operating like a UPS, you will be taking current from the power bank. In an ideal design, the power path may be able to juggle demands assuming a single voltage and have the load served from the input power, but this is not the way the Xiaomi power bank above works.

Instead, in that power bank, it behaves more like an “online” UPS – everything coming in has to go through the charger and into the cell, and everything that does out has to come from the cell and through the boost converter. While this ensures a clean and steady output, it comes with efficiency losses.

Many charger ICs have a safety timeout which is usually set to some time between eight hours to under 40 hours to ensure they do not keep putting energy into a potentially partially-shorted cell. This is to reduce the risk of catastrophic danger from lithium-ion cells. I’ve not used my Xiaomi power bank as a UPS for long enough to know if there is a time-out, but I do know many standalone charger ICs do have a timeout, in which case the charger will simply stop running leaving the cells to deplete for no good reason.

Treating Li-Ion/Li-Poly Batteries Well

Another issue with the whole concept is simply that it will often abuse the lithium-ion and lithium-polymer cells. This is because they are usually designed to be charged in a CC/CV regime, terminating at a set current level to avoid overcharging. When you run them in a UPS role in the way the Xiaomi power bank does, then it is exposed to the heat from the circuitry and is going to experience “micro-cycling” at a high-state of charge.

In fact, it is well known that keeping Li-Ion/Li-Poly batteries at high states of charge for long periods contributes to rapid loss of capacity and problems with swelling. My personal experience seems to show this is more severe for extended-voltage Li cells such as those used in the Xiaomi power banks and modern mobile phones which try to maximise energy storage density. The result is often swollen cells which present a safety risk.

To treat the cells well requires sacrificing some capacity to reduce the voltage stress on the cells. I managed to dig up this table from a Samsung SDI datasheet for their INR18650-30Q cell, but the guidance is applicable in general for battery pack development:

It may be slightly difficult to read, but I want you to notice the charging voltage for 4.20V and 4.35V termination voltage cells is recommended to be 4.00V or 4.05V for Energy Storage Systems (ESS) or Uninterruptible Power Supply (UPS) applications. The same threshold applies for extended voltage cells suggesting that high voltage stress from being maintained at high state-of-charge conditions are indeed likely the cause for a loss of capacity and swelling of Li-Poly packs especially.

Unfortunately, most power banks prioritise extracting the maximum capacity from the cells and thus mostly have charger chips or controllers that are either fixed to a nominal 4.20V or 4.35V termination, or can be adjusted from 4.20V and up rather than allowing for lower capacity, greater safety and longer storage life. This is a big shame as most power banks are stored fully charged in anticipation of a need, and thus they are essentially sitting under high stress that causes them to degrade more rapidly in terms of cell capacity.

Conclusion

Understanding if a given power bank can perform the role of an uninterruptible power supply for USB devices is not an easy task, especially relying on manufacturer information alone. An advertised “pass-through charging” capability does not necessarily imply that the output is free of interruptions.

Even if you did find a power bank that was seemingly capable of uninterruptible operation, this may not be under all conditions as the input charger IC could time-out on over-time protection. If it doesn’t, what it may be doing is mistreating the lithium-ion/polymer cell enough that it will shorten its lifetime by holding it at a maximum state of charge, producing voltage stress on the cell which may contribute (along with temperature) to premature capacity loss and swelling of lithium-polymer cells.

While there may be some power bank products that will seemingly operate in the UPS role, unless they have been carefully engineered to best avoid these pitfalls, they are unlikely to be reliable, safe or kind to the lithium-ion/polymer cells in the long-run and I would advise against long-term unmonitored use as an unhappy lithium battery can be a pretty catastrophic event.