It’s been a long time since I posted anything about power banks. Believe me, I would have liked to bring you some more reviews, or updates on cycle life testing, but due to my PhD commandeering my dual-measure multimeter I use for power-bank testing for a period of about 6 months, I’ve been stuck waiting. Luckily, since I built my USB current shunt, there is something that can be examined with a regular logging multimeter.
In light of the recent incidents involving self-balancing scooters (also incorrectly known as hoverboards) catching fire being blamed on overcharging, I thought it was a good time to quickly go over the fundamentals of Lithium-Ion charging and examine just how the various power banks in my collection perform in regards to this. Of course, we don’t expect to uncover anything dangerous as initial reviews would have weeded out a lack of charge termination, but we will come to see that there are many subtle variations from the ideal.
Both lithium-ion and lithium-polymer style batteries are rather picky about how they are treated, and will become dangerous if operated outside their specified conditions. As a general rule, most cells will not tolerate over-discharge (below about 2.8v per cell), or overcharge (above 4.2v per cell) with some rare exception for range-extended high capacity cells. Outside these ranges, permanent changes to the chemistry occur which can cause cell instability and “vent with flame”/explosion.
As a result, all commercially sold batteries generally have a protection PCB on board which contains at least one, preferably two lines of defense against overcharge and overdischarge, and will disconnect the cell from the electronics in case of anomalies. Rarely will cells be used without such protections, as they are inexpensive and can be bought as modules, although the level of protection varies.
Lithium-ion cells can provide high currents with low internal resistance, but if the current supplied is too high, can lead to damage and overheating of the cells. Better protection PCBs will have protection against this, and it is integrated into certain cells (e.g. 18650) where self-resetting PTCs are used to limit short circuit currents.
Charging of cells outside a 0 to 45 degrees Celsius window is also generally not permitted, and at the low end of the scale, requires limiting charge current to avoid plating metallic lithium which can cause cell instability and failure (e.g. explosion).
Finally, in regards to charging, high initial currents can also cause damage to cells and trickle charging is not permitted. An active form of charge termination and regulation is required, and charging is generally done in at least two-phases – an initial Constant Current (CC) phase, and a Constant Voltage (CV) phase followed by termination. The termination is usually done by current, when it falls below about 0.05 – 0.07C, or by time (e.g. 90 minutes). Better chargers will trickle charge a deeply discharged cell until it reaches 3.0v before switching to full charge current.
There is also an art to selecting the charge current, as higher CC phase currents can work against having a fast charge, as it results in a longer tapered CV phase. Generally, you expect to have the battery properly fully charged within 2-3 hours in most cases when meeting all the requirements.
An Aside: Self-Balancing Scooter Fires – Why?
Of course, despite all of these fickle tendencies, lithium-ion cells enjoy wide adoption thanks to the electrical engineering that goes into making them safe. Indeed, many are employed in high-drain applications, especially RC models and power tools, where their high energy to mass ratio makes them highly favourable.
So why did some self-balancing scooters catch fire? Was it overcharging as advised by the media?
Without having investigated one in person, it is hard to tell, but it seems to me that the fires are a result of several potential issues:
- The virality of the device itself seems to have spurred other Chinese manufacturers to clone it and manufacture their own devices under time pressure and cost pressure to make it in time for holiday season and to undercut their competitors.
- That may have led to poor quality designs, and specifically designs which eschewed some secondary protections as is common with aftermarket clone laptop batteries, etc. As a result, the battery protection may only be a single layer of protection – say only over-charge/over-discharge protection, potentially missing one-time fuses and thermal protection.
- Component substitutions may have been made where more expensive high-current capable branded 18650’s with proper PTC over-current protection were replaced with generic, Chinese domestic 18650’s intended for lower drain use which have higher internal resistance and possibly no PTC protection.
- Under use, lower quality batteries are likely to heat up more due to higher internal resistance, and the heat can accelerate failure of the batteries, or in sufficient quantities, cause the battery to fail and vent with flame.
- The design of the battery pack may not have considered the necessary protection for series-parallel strings which could have improved safety by preventing healthy strings from dumping current into a failed string which may have further exacerbated the condition.
- In the case of series-parallel strings, actual danger could arise where no pack balancing is provided and cells rapidly degraded or were poorly matched at manufacture time. This would cause cells within the pack to be both under and over charged at the same time, resulting in stresses accumulating on a few cells which would fail more rapidly especially under the heavy loading expected from wheel motors.
- Charging may be responsible if charge termination threshold was not properly set, or the timer protection did not exist, or had failed. If the maximum charge voltage threshold was not obeyed, this could cause rapid cell degradation and safety problems as well, although it could also be possible that poor quality power adapters and poor ESD design may have damaged the charging circuitry.
The truth will only be known when someone tears down and examines the units and all of their variants. Of course, it is very possible to build safe units, as Li-Ion cordless tools have been around for quite a while without as many incidents, although component selection can be critical and choosing the right battery, charging algorithm and equipping a full set of primary and secondary protections is what it takes to mitigate the risks. Most Li-Ion batteries supplied with branded products don’t take the risk of missing these protections, for fear of litigation.
Power Bank Charging Profiles
Using the USB current shunt provided me the opportunity to measure the charge current consumption for power banks. To do this, I first depleted all the power banks that I currently still own (modified and unmodified), and then charged them under the supervision of my Keysight Technologies U1461A. This measures the current from a 5v 2A USB charger via the shunt. I noticed contact resistance had been a slight issue in higher current runs, so I deformed the USB connectors to add higher pressure to the contacts to reduce the resistance as much as practicable.
Due to there being two main variations of charger topology, a note about that is needed. The termination current measured is at 5v on the 5v input. The actual current into the cells depends on the charger topology. The cheapest and nastiest (efficiency wise) charger ICs are simple linear-style regulators which drop excess voltage, thus the current into the cell is equivalent to the current on the 5v line, although during charging, for the most part it is about 74% efficient (3.7v / 5v). The other sort is a buck converter, which means that the excess voltage is traded off for current. During the termination phase, this would mean the current into the cell would be 1.19 (5v / 4.2v) times the current on the 5v line assuming 100% converter efficiency, and during the initial CC stage, the current would be 1.67 (5v / 3v) times the current on the 5v line. Many buck converter designs can be >85% efficient, making for faster charge times and better efficiency.
Because of the time taken to charge, and the time taken to analyze and produce graphs, this experiment took the better part of a week and a half to do, but the results seem rather interesting (at least, to me).
Charge Profiles: The Conventional
If we’re talking about ideal charging profiles for Li-Ion, then the Powertraveller PowerMonkey Discovery power bank has pretty much the conventional ideal shape. During the CC phase, it has a very flat current profile, with a slight wiggle in transition to CV phase and a smooth taper, followed by termination.
Sadly, as it’s an older power bank, it is limited to 500mA both ways, but it does adhere to the current limit quite well, and the taper region is relatively short, meaning a more optimal charge strategy. One criticism would be that the charge termination current seems to be set a bit on the low side, with the current most suitable for cells about 743-3094mAh in size, rather than 3500mAh.
Another one with a pretty ideal curve was an anonymous 6000mAh power bank based on the latter design in this page. Despite being rated for a 1A charge input, it was conservatively set to 700mA which means slower than expected charging. It seems probable this is a linear-charger. However, despite this, it has a very flat curve and decently short tapered region. The termination current is, again, set a bit low and seems suited for 971-4046mAh cells.
Charge Profiles: The Opportunistic
As it turns out, the Xiaomi power banks are special in the sense that they were the only power banks in my collection to exhibit opportunistic tendencies. They were also both buck converter type meaning that they had good charging efficiency as well. This seems to periodically (5-10 minutes) increase the charging current and see what happens. It quickly backs-off and finds an equilibrium point, probably measuring the impedance/voltage drop of the charger vs current to sit at a current which is safe for the charger and optimal for the charging speed of the power bank.
Being rated for a 2A input, it was clearly trying its best to draw as much as possible. Some of the deviation from 2A is explained by the additional connectors and current shunt causing voltage drop, resulting in a slightly compromised charge speed. The dips, however, are not likely to be material in terms of affecting the battery life but are a characteristic of the charging algorithm. However, due to the flat-ish current on 5v, with a buck type charger, the actual current into the cell is likely to be high initially, and slowly decreasing over time rather than following a strict CC regime. This isn’t likely to be damaging due to the large cell capacity versus current into the cell, as it would adhere to the maximum initial current specifications. The charge termination point is a little low again, suited for cells ranging between 2286-9520mAh.
This was the larger power bank and it had a very similar profile, although I accidentally bumped the connector resulting in a dip in rate, until the next “test” which shot the rate back up, further than it started with. This one drew even closer to the 2A claim, and this algorithm optimizes the charging rate for the battery with a conventional 5v, 2A adapter. Charge termination appears suited for 2857-11900mAh cells, so is again, slightly low.
Charge Profiles: The Close-To Optimal
The Apotop DW21 was pretty close to optimal, and very similar to conventional design except it had a strange discontinuous transition to CV mode. This isn’t necessarily a bad thing, but may mean a slightly longer charge time. It had good charge rate regulation to keep close to the rated 1A. The charge termination current is set correctly in the range of 0.02-0.07C.
The Hillo Power is where some strangeness starts to occur. It very much tries to be an optimal charge algorithm, but it has noise all over the place likely due to the use of a buck converter design. The duty cycle of the converter seems to have limited adjustment steps, hence the visible stair-stepping in the current consumption. In the bulk charge, it does adhere relatively well to the 1A limit, and it charges extremely quickly in the bulk region with only a short taper region, but I can’t help but wonder whether the noisy charging could cause overvoltage during the tapered region. High accuracy in limiting the maximum voltage is required to make sure the batteries have optimal cycle life, but so far, units based on this design seem to perform quite well for cycle life in my testing, and the charge time is very aggressive. The termination current is set correctly.
Charge Profiles: The “Needs Help” and Downright Odd
This was a modified power bank, originally 5000mAh but replaced with 2x3000mAh cells for 6000mAh capacity. Name has been censored as the product is modified and the company had objections to me naming their brand, but this shows several problems.
First of all, the CC phase is extremely short, and falls below the claimed 1A current input. The tapered section makes up the majority of the charge time, resulting in a slow charge. I wonder if the tapered section is truly CV, or whether the charge IC has a crude algorithm that reduces the charge current depending on the difference between cell voltage and target voltage. Otherwise, maybe it’s a limitation of the switching IC? Termination current is, however, set in the correct range.
It seems that WST could have done a better job of optimizing the charge time on this model, as the taper CV section is more than half the charge time. There is also an unsteady start, and it seems the 1A current limit isn’t quite that accurately set. Charge termination current is set a little low, best for 1714-7140mAh cells.
We go back to the nasty unbranded stuff, the 30000mAh unit seems to show very unexpected behaviour. The charge profile seems to be made of an initial spike, followed by linear segments which form a hump shape. The available current from the charger is poorly utilized, with the charge current varying but very rarely staying near the maximum of 700mA. Any doubts as to the capacity can be allayed, as the linear charge shows a very similar energy as to the discharge test results, but at least, over the years of keeping it in a drawer, it hasn’t changed much.
The 5000mAh unit has a more acceptable charge profile, but again, has its own take on how it’s done. There is an initial deep-discharge trickle phase, which is good practice. Then it switches over to a CC mode, with a spike then ramp-up. There is a sharp transition before it changes into CV mode. On the whole, this may be because it simplifies the IC design and it’s not particularly harmful for the battery, but it represents poor utilization of the available charge current meaning longer charge times as it could have been pulling 1A most of the time. The charger type also appears to be a linear-charger meaning poorer charging efficiency, and the charge termination current is a hair on the low side again.
The unbranded lipstick 3000mAh unit seems to be all taper. This is capped at about 700mA, but it means slower charging than otherwise possible. It also seems to be a linear charger as well. The charge termination is set correctly though.
Finally, the generic 12000mAh is another all-taper algorithm, which seems to be based on voltage-difference from target voltage with half-current blips. The current seems to follow the voltage-difference between the cell voltage and input voltage. The termination current is suited for 7786-32428mAh cells, so maybe it’s a bit on the high side (as the cells are less than the advertised capacity), resulting in somewhat incomplete charging. Again, it’s obvious this power bank doesn’t have the claimed capacity.
Lithium-ion chemistry rechargeable batteries are quite fickle and need care to ensure their safe usage. From testing a bunch of products, we can see that the charging strategies employed by these units are rather diverse, with quality branded products adopting more conventional strategies, and unbranded units pretty much going with crude charging strategies. Other products are somewhere in-between.
On the whole, while all of the units showed charging termination meaning overcharging is not a real possibility, few of the units made good of their input specifications. The Xiaomi units were the most aggressive in using the available input current, resulting in the best charge times for the capacity. Most other units in my collection are limited to 1A, and only the Apotop and Hillo Power managed to use the claimed input current. Others were conservatively limited to lower currents in the 600-850mA range which slows down charging and can cause inconvenience to the users. Only the last tested unbranded power bank showed a charging strategy that did not appear to have a proper current limit, especially at initial charge, which could be somewhat risky and damaging to the cells. Others appeared to limit the current satisfactorily.
As a whole, when choosing (or properly reviewing) a power bank, it’s a complex task as you need to evaluate the design, available capacity claims, the output power quality and the safety and efficiency of the charging strategy and compatibility with your devices in terms of fast-charging signalling. However, I hope this has been an interesting and informative post that inspires others to test and report on the charging behaviour of their power banks.