Tested: Two Naked Power Bank Internals (5600mAh, 4400mAh)

In some of my previous power bank postings, people have expressed interest in knowing which power banks are good and which power banks aren’t. Further to that, I did invite readers to donate power banks for testing.

Thanks to one anonymous donor (who I shall refer to as John Doe), I have been bequeathed a series of power banks to tear down and test. As a result, you can expect quite a few power-bank related postings in the near future. In this posting, we will be looking at two “naked” power banks – these had been disassembled before they were given to me, and all of the packaging and exterior are not available.

Samsung-based Power Bank (Pink, 5600mAh)

The first power bank is the Samsung-cell-based power bank, which I will call pink because of the cell shrink wrap colour. This unit is claimed to be a 5600mAh unit (as quoted by Mr. Doe).


The shrink wrapping seems to have been damaged and partially removed prior to receiving the unit. Unfortunately with it exposed, it can be quite vulnerable to short circuit and damage, so it was wrapped up in insulating tape after documentation.

The cells themselves are marked with ICR18650-26F, which indicates they are 2600mAh type cells. The two are connected in parallel, for a capacity of 5200mAh. I am not sure if the original capacity was misquoted by Mr Doe, or whether it was deliberately up-marked as a 5600mAh power bank but you can’t invent capacity out of nothing.

The conversion circuitry for this unit is made of a pair of PCBs interconnected. The top PCB visible has 5 blue LEDs arranged in an arc for indication of charge level, with a push button to invoke the battery indicator. The PCB is marked WST-A85 JHW. From this, I can only surmise from a quick Google search that the product likely originated from Shenzhen Wanshuntong Science & Technology Co., Ltd.


The rear of the rear PCB has very little to it, although, it is visible that the battery is connected by soldered insulated wire.


Unfurling the two halves of the PCB, it is noted that this power bank offers a single 1A output, and has a single LED as an emergency torch. The bottom PCB is marked WST-A86 JHH.

The top PCB seems to have some form of unmarked microcontroller U1, and a LM358 dual operational amplifier U2. The bottom half seems to have the bulk of the power conversion circuitry, including a few various MOSFETs, the inductor L1, SS54 5A Schottky free-wheeling diode D1, and a tantalum smoothing capacitor. It appears the compact size of the unit means that a larger electrolytic capacitor would not have fitted well, although the tantalum capacitor could be good enough. It’s also interesting to see that the inductor is not of the “enclosed” type, and would slightly less efficient.

Anonymous Cell Power Bank (Orange, 4400mAh)

The other power bank is a fairly similar form factor, and is built using orange-wrapped cells with no markings. There is some residue from adhesive that has been used to secure the cells inside the enclosure.


The top side of the blue PCB features the single 1A USB output and input for charging. The long black component seems to be a tilt sensor of some sort, which activates the indicator LEDs on being shaken. There is a 3R3 inductor, also of the open type, visible and a very common SK34 3A Schottky free-wheeling diode.


The underside also has an un-marked microcontroller, a charge controller and very few MOSFETs. Four surface mount blue LEDs are used to indicate charge level, and are activated whenever the power bank is shaken or moved. This can easily mean that power is wasted while transporting the unit around, although the LEDs do run at a low duty cycle to conserve some energy at the expense of introducing some flicker.


The cells are connected in a two-parallel configuration, utilizing the more common tab instead of soldered insulated wire.


A closer look at the top PCB silkscreening reveals the code XZD-M1896-V3. A quick Google search seems to suggest that this might be from XuZhengDa Group.

Performance Test: Pink Power Bank

Utilizing the new power bank test rig, I decided to run 5 runs at ~1A load, and 5 runs at ~500mA load to find what the effective capacity of the power bank is. Note that the effective capacity is the amount of energy we can extract from the power bank, and will be the capacity of the battery multiplied by the efficiency of the power conversion.

Load (mA) Run Capacity (mAh)
500 1 4361.916777
500 2 4241.938197
500 3 4150.036523
500 4 4154.647153
500 5 4201.861524
Mean 4222.080035
Range 211.8802534
StDev 86.73942327
Load (mA) Run Capacity (mAh)
1000 1 4034.506199
1000 2 4049.610516
1000 3 4020.476771
1000 4 3838.763246
1000 5 3922.779512
Mean 3973.227249
Range 210.8472694
StDev 90.06604885

From the test data, the resulting capacity is 4222mAh at 500mA load, and 3972mAh at 1000mA load. The range was a fair ~212mAh indicating relatively consistent charge termination. If we assume the cells are 5200mAh, then the conversion efficiency is 81% and 76.4% respectively, which is not particularly high but still relatively reasonable.


Over-discharge protection was found to be functioning, with the power bank shutting down once the charge is depleted. This did not function as cleanly as could be expected when loaded at 1A, with the power bank cycling between off and on rapidly as it approached its end of discharge voltage, indicating a lack of sufficient hysteresis in the voltage setpoint.

The absolute voltage levels are very well regulated, with them sitting around 5.05v at light loading and 4.94v at the higher loading. Both are well within the expected ranges.


The ripple at 500mA loading was 57.27mV peak-to-peak, which is quite clean. The frequency of oscillation was about 655.6kHz.


Increasing the loading to 1A increased the ripple to 126.6mV peak-to-peak, which is still reasonable. The frequency of oscillation is about 644.3khz.

All things considered, the pink power bank is not bad. The efficiency of the converter could be improved, assuming the cells are really genuine Samsung cells of the expected 2600mAh capacity. The hysteresis at the end of discharge could be improved as well. However, it does produce relatively clean power, at the right voltage regardless of the level of charge.

Performance Test: Orange Power Bank

Following the same methodology, this power bank returned the following capacity:

Load (mA) Run Capacity (mAh)
500 1 4534.116486
500 2 4605.297565
500 3 4607.11821
500 4 4473.636598
500 5 4512.64021
Mean 4546.561814
Range 133.4816119
StDev 58.61018266
Load (mA) Run Capacity (mAh)
1000 1 4332.236016
1000 2 4313.675077
1000 3 4340.417733
1000 4 4276.400183
1000 5 4341.79924
Mean 4320.90565
Range 65.39905655
StDev 27.28854347

The capacity was determined to be 4547mAh at 500mA and 4321mAh at 1A, with a range of 133mAh and 65mAh. This is a fairly surprising result if the power bank is supposed to be 4400mAh, as the output energy exceeds the actual capacity of the cells. The results are fairly consistent as well. As a result, I’d have to guess that the cells are probably at least 2400mAh cells, but more probably 2600mAh cells, although their brand-less nature deserves caution as they may not last as well as the branded cells in terms of longevity.


The power bank works quite well for 500mA loads, however, increasing the load to 1A appears to show the conversion circuitry hitting a brick wall in terms of duty cycle, and thus the voltage slowly becomes “unregulated”. This is not a very good result, and while despite staying within the correct voltage ranges (4.75v to 5.25v) for the majority of the discharge, it is stressing the conversion circuitry to some extent. It isn’t something a well-designed converter should be doing.


At 500mA, the ripple voltage is 108.2mV peak-to-peak, which is reasonable. The oscillating frequency is quite high, at 959.8khz, likely to optimize for smaller inductors and capacitors but at the trade-off of efficiency.


At the higher load of 1A, the ripple voltage has increased to 148.1mV peak-to-peak, which is still rather acceptable but at the limits of what stock chargers will put out under load. The oscillation frequency is 962.2khz at this loading.

Again, a semi-reasonable effort in terms of capacity and ripple, although seeing the regulator run up towards its regulation limit at 1A implies that there is no headroom especially if heavier loads are inadvertently connected (e.g. tablets). Also, the lack of branding on the cells may mean questionable longevity.


These two power bank internals are actually fairly decent products with a similar form factor. Both suffer their own shortcomings, and unexpectedly, the 4400mAh orange tested to have a higher effective capacity than the 5600mAh pink. The pink was already over-rated initially, and has some hysteresis issues at the end of discharge, whereas the orange has regulation issues at high load currents. The shake to activate feature of the orange power bank may needlessly consume energy as well. Both power banks do a good job of complying with reasonable limits on ripple voltage, and offer a decent capacity.

About lui_gough

I'm a bit of a nut for electronics, computing, photography, radio, satellite and other technical hobbies. Click for more about me!
This entry was posted in Power Bank and tagged , . Bookmark the permalink.

2 Responses to Tested: Two Naked Power Bank Internals (5600mAh, 4400mAh)

  1. Nice work!
    Any chance you’ll be testing the batteries directly to determine how much of the capacity drop is due to the losses in the boost converters vs the batteries having lower than spec capacity?

    • lui_gough says:

      Dear Ralph,

      I have considered doing that, however, as it takes about a day to run a single run, it’s a substantial amount of work which I think is not justified. After all, most users will be concerned about how much “juice” they can get out of the power bank, not so much as to how much juice is going into the converter itself. A higher capacity battery with a ~72% converter would be equivalent to a lower capacity battery with a ~92% converter to an end user.

      I suppose that might be relevant if you’re looking to identify counterfeit 18650’s or to “reuse” the cells bare, but I’m willing to bet that most users aren’t going to do that.

      It’s also fairly frustrating as testing bare cells requires careful attention to “disconnect” the load before the cells are completely depleted and irreversably damaged, or to hack up a protection PCB to the cells directly. As a result, I might do it but only if I end up having an excess of time, which I doubt I will.

      – Gough

Error: Comment is Missing!