Earlier, in my article about modifying a USB charger doctor as a USB current shunt, I elaborated on the possibility of microcycling occurring in the battery of my mobile phone due to the irregular current profile after the phone had reported full charge.
In order to investigate this thoroughly, one should measure the current into and out of the battery of the phone to be certain as to what is happening to the battery itself. This is necessary as there may be intervening circuitry and capacitors on the charging side that obscure the behaviour if you’re measuring the source current like I was.
Buoyed by the comment left to me by an anonymous commenter which also expressed some concern about rapidly reducing battery capacity on their LG handset, I decided that an investigation was worth my time.
The problem is that most phone battery compartments are designed quite tight without any space to maneuver, and it’s definitely not an easy job to modify a phone just to test out the hypothesis. Any low-current current measurement requires severing the connection between the battery and load at either pole. Higher current/less accurate measurements can be obtained using magnetic clamp style devices, but that also requires one nice conductor to clamp around. In a phone battery compartment, I came up with a method.
Gough’s Hacky Method
To do this, I resorted to a kitchen friend, namely aluminium (or aluminum, depending on where you’re from) foil. Using a sharp knife, I cut it into small strips, which should be wide enough for some strength, and thin enough to avoid bridging across terminals (depending on the pitch of the terminals).
These would be slotted in-between the battery terminals to provide a “interception point” where all the current can be metered. To prevent the back-to-back foil from shorting out, I decided to insulate the pieces where they will touch with paper, and superglue, in essence, forming a laminate.
After the glue had set, the excess paper was cut off to give a nice thin profile “tap”.
That can then be slotted in-between the terminal spring and the battery, with the spring force pushing on the sandwich causing the other side to contact with the battery fairly well. You don’t want to make this too thick as you may damage the spring contact on the phone by causing metal fatigue which results in inadequate contact pressure and reboots when the phone is vibrated or dropped.
The next part of the equation is to make the shunt resistance. In theory, you would just clip a multimeter in current mode across the two pieces of foil, but the burden resistance of the meter is likely to be a problem.
The burden resistance is literally “a resistance” the meter presents to the circuit under test, in order to convert the current into a voltage which can be measured. Most meters present burden resistances which vary with which current range is in use, and can be quite high for small current ranges (ohms to tens of ohms). For example, my Keysight U1461A meter has burden resistances ranging from 1 ohm for the 60 to 600mA ranges, and 100 ohms for 600uA to 6mA (and even higher for uA range). The effect of this burden resistor needs to be taken into account.
Assuming I use the 600mA range, with a burden resistor of 1 ohm, the voltage drop would be 0.6V. This, compared to the battery nominal voltage of 3.7V is quite big and will affect how the phone behaves with the battery. It will cause the phone to think the battery is almost depleted at high-draws even though it may have half its charge. This would affect the test results and make it completely invalid. Another complication is that the range is likely not sufficient in case the phone consumes more than 600mA, which is quite possible on an instantaneous basis (think peak load current draw).
As a result, we would like an even smaller burden resistor. Ultimately, the game is to make it small, but not too small as to have noise become the major impact on your desired signal. I decided to use 0.1 ohms as the burden resistance, planning for a 60mV voltage drop at 600mA. In reality, 60mV is pretty minor, and a 60mV deviation near the end of discharge for a Li-Ion cell represents only a small capacity loss as the cell voltage drops steeply near the end (~2.8-3V only represents <2% of cell capacity). It should also not be a big impact to the charging algorithm because as the charge current tapers off, we expect to see the charge terminate at 0.02-0.07C which for a 3140mA battery is 62.8 – 219.8mA or a voltage difference across the burden resistance of 6.28 to 21.98mV. For charging, this result can be considered borderline as charger IC accuracy to <100mV is considered required if maximising battery capacity, and 1% is common (i.e. 42mV at end of charge), so we have increased the voltage error by a significant fraction of the IC’s capability. Regardless, if I went any smaller, the risk would be that the signal would be buried under random noise, so this is a good compromise.
Unfortunately, 0.1 ohm resistors aren’t units I have on hand, so the solution was to use 11 x 1.1 ohm resistors in parallel – namely this very ugly twisted-together hulk of resistors.
After adding solder to the ends to bind them all together, the resistor was measured on a Agilent U1733C as 0.1006 ohms, after null offsetting the lead resistance. This was pretty much on the dot.
Then comes the other issue – you can’t solder to aluminium. Believe me, I’ve tried before, and I know full well that the oxide coating on aluminium prevents any good contact – so the solution was just to use the test lead clips to clip the foil to the resistor like below, while making sure that the ends don’t touch metal components they shouldn’t (otherwise they would ground the battery).
There is, of course, some resistance in the foil itself, but for all intents, this was considered negligible. I tried to measure it with the Agilent U1733C, which reported a resistance of 0.02 ohms for the combined length of foil. With this knowledge, it’s likely completely possible to eliminate the burden resistor and fashion one out of “cutting” foil to the right dimensions, although, as foil is fragile, you might have trouble with preventing it from tearing.
The measured current over time is annotated above. Initially, the phone draws from the battery as it is unplugged and runs the process of boot-up. As soon as I sleep the screen, the power draw reduces, but only by half, while the background processes continue to execute and keep the SoC “busy”. Once they have finished, the current draw settles to a much smaller draw, close to 0A with spikes where the SoC is woken up. This is the expected behaviour.
Once the charger is connected, the consumption goes up as the SoC is awoken along with the screen, then charging commences soonafter. As the battery is almost full at the stage I started, the charge current tapers off, but an observation is made that discharge spikes occur even during the charging phase, and occur at an interval of every few seconds.
Even after about an hour and a half, the charge current is plateauing but the spikes continue, representing discharge spikes during charging, or microcycling as I had suspected. These spikes were not an aberration of the measurement apparatus being in error, as the polarity of the current is as expected (i.e. during the no-charger connected phase, currents were positive indicating draw, and at deep sleep, currents were still positive but only slightly as expected). There is a potential that the burden resistor may have impacted the charge termination strategy, by introducing a current dependent voltage swing, although the effect this should have is not to introduce current draw but instead to potentially cause oscillation in charge current (i.e. charger goes on sensing voltage drop after the charger stopped supply current, sees the voltage rise to the threshold and then turns off, and repeats). This should result in an oscillation between negative values and zero, rather than positive values as seen here.
Towards the end, in the last few minutes, I took the data and averaged the current in and out of the battery. The result was a net charge rate of 13mA, which is very small and below that of the expected 0.02-0.07C charge termination threshold, namely 62.8 to 219.8mA. As a result, the phone was not treating the Li-Ion cell correctly from a conventional standpoint, namely to stop charging current at the 0.02C threshold, and is instead operating it on a sort of trickle balanced with discharge pulses. This is potentially stressful to the battery and was not my expectation.
I would have to theorize that most handsets on the market, LG or not, have this behaviour based on the fact that they use the battery as an integral part of the power system and refuse to operate when no battery is inserted. In essence, they rely on the battery as a form of “infinite capacitor”.
Whether small microcycles on this magnitude have any effect on the battery longevity is a question that needs more information, however, the constant flow of current into and out of the battery is likely to cause small heating effects with cell internal resistance that would likely increase the rate of degradation, and chemical processes which may not be reversible. Constant maintenance at high states of charge with such trickle effects may lead to long term separator voltage-related stress and lithium plating in the case of incorrect termination voltage or failure to terminate resulting in unsafe explosions of cells.
The advice often given by aftermarket sellers of clone batteries is to never leave them on charge overnight. The cited reason is often to avoid overcharging. The truth is that lithium ion cells should be designed as to be impossible to be overcharged due to onboard protection boards and charger algorithms strictly designed to prevent this. Instead, there is a distinct possibility that their advice actually stems from observed reduced battery life from maintenance at high states of charge due to separator stress combined with the effects of microcycling? Perhaps aftermarket cells with poorer quality construction and higher internal resistances are more vulnerable to early failure due to microcycling and high average states of charge? I’m not entirely sure, but this sort of advice is stuff I often get from eBay and ‘flea market’ sellers, with most OEMs never offering such advice. Food for thought, I suppose.
As mentioned earlier, the foil itself was quite fragile, and after the first test, on repositioning the phone, it tore under the load of the test clips. A second “tap” was fashioned, this time out of electrical tape being stuck to the back of one piece of foil, with another piece loosely placed behind and then taped where it emerged out of the battery for mechanical support. Due to the slight stretchiness and better adhesion of the tape to the foil, this provided better mechanical strength allowing the phone to run vertically and be controlled by the user.
A battery run-down test was performed to measure the capacity of the battery. Voltage was not measured, only current (by measuring the burden resistor voltage), integrated over time to provide the battery capacity. The minor loss in the burden resistor was taken as negligible (and in practice, it was), and the phone was run with a Flashlight app running permanently. This consumes about 600mA from the cell, which is roughly C/5 rate for a 3140mAh cell – thus is a good load to confirm the cell capacity. Some spikes in current usage due to background apps was experienced, but were not really significant to the findings.
After 20 months of moderate usage (two to three days partial run-down a week, estimated 80-200 full cycles at the most), the cell only recorded a capacity of 2067.486mAh or about 65.84% of its original capacity. This explains the disappointingly short run-times, and is considered failed for a lithium-ion battery (below 80%). Whether this diminished capacity is due to microcycling or not is hard to tell, but at this stage, it certainly appears to be a contributor.
Conclusions
In 20 months of ownership, a rapid marked decrease in the operating time was witnessed. Some investigation surrounding the charging behaviour uncovered unusual microcycling-like behaviour and trickle-charging below recommended Li-Ion charging termination threshold currents. This may be related to their use of the battery as an “infinite” capacitor, without any need to incorporate larger bulk capacitance which may make a phone physically larger.
Ultimately, better battery management is probably desirable, and can be had by incorporating enough bulk capacitance (possibly at the cost of slimness of the phone) to allow the phone to run off external power solely when the battery is charged, provide enough charge for graceful switch-over of power source, and to have proper charge termination rather than a constant back-and-forth balancing game that is currently seen. A larger hysteresis value for the charger may help, as it will avoid re-activating the charger on a small consumption from the battery, although complete elimination of consumption from the battery while plugged in is probably the first step.
