Review, Teardown: Xiaomi (Mi) NDY-02-AL 16000mAh Power Bank

Xiaomi, or Mi for short, is a Chinese mobile phone manufacturer that have been known for delivering affordable, high quality products, with an aim to become the next “household brand”. Their product lines have diversified away from only phones and tablets to now include accessories and household appliances. To date, I have tested a few of their products, including power banks, a fitness tracker and a Bluetooth speaker and so far, they have produced very commendable results.

Many readers, looking for larger power banks, have suggested that I try the Xiaomi NDY-02-AL 16000mAh power bank. Unfortunately, as a hobbyist running a personal website, I really had no need for power banks above a certain number of units and nobody was willing to donate a unit for testing. Luckily for you all, Gearbest have stepped up to the plate, and were willing to provide me the Xiaomi NDY-02-AL 16000mAh unit for testing and review.

This review has been a long time in the making, requiring over one and a half months of testing time, and producing over 120Mb of data which needed to be processed to create the performance test results. Will it carry on the Xiaomi heritage of being a quality product offering excellent value for money? I finally have an answer. Read on.

Unboxing

The unit arrived in a matte finish thin cardboard box, with a grey print Mi-logo on the front and website on the rear. Stocking labels seem to have been placed on the front, indicating relatively fresh stock dated 3rd April 2015.

On one side of the box, there is a basic list of specifications for the product.The unit claims to have 16000mAh capacity at a nominal voltage of 3.75V for LG cells, or 3.6V for other cells. It accepts a 5V 2A charging input, and can put out two outputs at 5.1V at up to 2.1A with a maximum combined output of 3.6A.

On the other side, there are barcodes and the Mi authenticity verification label. The scratch-off panel on the authenticity verification label was removed and the resulting numbers were entered into the authenticity checker which proved that the unit was authentic.

Inside, the unit sits on a moulded cardboard tray, along with a short flat cable, and a Chinese only instruction manual. The front of the power bank is emblazoned with the Mi logo. The power bank weighed 346 grams on my cheap set of scales.

The rear of the power bank is emblazoned with the website mi.com. For protection, the ports of the power bank are taped over.

The power bank ports are beige, and the Mi logo can be seen in the plastic moulding on the top port. The two USB ports are configured to be in an opposing direction.

The other end of the power bank has all of the specs and details.

The supplied flat cable has a black tongue on the USB A port side, with no moulding marks or logos on the USB A port side. The micro B port side has a black tongue as well, with a lightning bolt logo on the moulding on one side.

Teardown

The assembly of this Xiaomi power bank is very similar to the 10,400mAh models previously reviewed. Removing the self-adhered plastic surrounds on the front panel allows access to screws which secure the body of the unit to the aluminium outer shell.

We can see the use of white diffusing tape over the LED holes, as with previous Xiaomi power banks. Removing the four Philips screws allows you to pry out the center unit using a flat-head screwdriver in the little plastic gap in the top left.

The unit consists of five LGABE11865 3200mAh cells. The construction is of high quality, with liberal use of kapton tape for insulation, quality tabbing between cells with clean spot welds, and foam tape to ensure a snug fit. Carrying on from the previous power bank, this unit maintains its thermistor monitoring of the pack temperature as well, for safety.

The main PCB is marked PB160_MB_44 dated 28th September 2014. The PCB uses a BQ24195 2.5A/4.5A single cell charger with 5.1V 1.3A/2.1A synchronous boost operation by Texas Instruments on the top side, with large MLCC capacitors sited close to the inductor for good ripple performance. It also appears that an 8A fuse is present on the top end, fusing a pair of parallel cells at the negative.

The PCB and connectors are supported by a plastic frame, which can be unclipped to reveal the underside of the PCB.

The underside seems to reveal a stack of paralleled MOSFET switches to disconnect the cells, likely to ensure sufficient capacity under high current loading and less waste heat. A second inductor and un-identified surface mount power converter is seen on the underside, which results in this unit having a unique design where the two USB ports are being provided by two separate independent converters. The whole unit is controlled by an ABOV semiconductor MC97F1204S microcontroller.

What appears to be an 8A fuse is also seen on the underside of the PCB which connects to the negative of three parallel cells. This is then bussed together with the fuse controlling two parallel cells on the other side of the PCB.

This is another unique arrangement that I haven’t seen before, and it overcomes a normal restriction by manufacturers on paralleling Li-Ion cells. Normally, most manufacturers will not warrant or recommend the connection of more than three 18650 cells together, as a cell failure will lead to significant energy dissipation with the two other cells dumping into the failed cell. Xiaomi’s answer to this is to segment the system into two strings – one having two cells in parallel, the other having three, each protected by their own fuse. In case of a cell failure, a fuse from the non-failing string will blow limiting the fault current to just the ones in parallel (at most, two other cells), thus maintaining the safety recommendations by the manufacturer. A smart solution indeed – especially in a market where some batteries unashamedly try to parallel eight cells together with no safeguards at all!

It is nice to see that the tradition of using potted inductors continues, although a shielded unit might offer even better efficiency. This reduces potential for acoustic problems.

In all, the construction resembles a high quality laptop battery pack with notable attention to safety and performance aspects.

Performance Testing

Performance testing was performed using the same configuration that has been used for most of the power bank tests, based on the “new” test rig. As this unit has a particular design, the testing was performed twice – with results for Rail A (USB port on the Mi-logo side) reported separately to the results for Rail B (USB port on the website-printing side) as they are based around two separate switching converters internally. This resulted in 30 cycles being put onto the power bank, and over a month and a half of continuous testing.

Due to the large number of data points, plotting was performed using Matlab instead of Excel, as it kept choking on the data which had over 112,000 points for a single test run at 500mA.

Charging the power bank from a 2A power adapter took approximately 14 hours, but sometimes longer if the power bank did not detect or use the full capacity of the charger. High quality cables and a high speed charger are recommended to avoid lengthy recharge times.

The Xiaomi power bank uses cells with a nominal voltage of 3.75v. Readers should remember that the number of mAh that a power bank has is not a measure of energy without knowing the voltage – i.e. 1000mAh at 3.75v is MORE than 1000mAh at 3.6v. Because of this discrepancy, this creates a reporting difficulty as all the previous tests were reported using a nominal voltage of 3.7v. As a result, the tabulated capacities are calculated on a 3.7v nominal voltage basis so as to be comparable with previous testing, with the 3.75v converted figures used to calculate efficiency and provided in text.

Capacity Results – Rail A

The performance of Rail A’s converter is excellent, with an average of 14,824mAh at ~500mA, 14,725mAh at ~1A and 13,986mAh at ~2A loadings (all adjusted for 3.75v nominal voltage). This results in a calculated efficiency of 92.7%, 92.0% and 87.4% respectively. The discharge termination worked well, and produced very consistent figures with all ranges below 65mAh.

The underside of the unit claims the usable output to be 10,800mAh at 5.1V at a 1A loading as a typical figure with efficiency above 90%. Through my calculations, it was determined that the Rail A performance came out to 10,827mAh at 5.1V at 1A, and the efficiency assuming 16000mAh of cells of 92%, easily exceeding these claims. They also claim a 93% maximum efficiency, and the measured result of 92.7% is close enough especially considering the error contribution of the test rig.

Capacity Results – Rail B

The performance of Rail B’s converter appears to be measurably slightly less efficient than Rail A, although not significantly so. The capacity is 14,793mAh at ~500mA, 14,551mAh at ~1A and 13706mAh at ~2A, all adjusted for 3.75v nominal voltage. This results in a calculated efficiency of 92.5%, 90.9% and 85.7% respectively, which is lower than that of Rail A. Discharge termination seems slightly less consistent with loading on Rail B, with the range of measured capacity up to 144mAh.

The underside of the unit claims the usable output to be 10,800mAh at 5.1V at a 1A loading as a typical figure with efficiency above 90%. Through my calculations, it was determined that the Rail B performance came out to 10,699mAh at 5.1V at 1A, and the efficiency assuming 16000mAh of cells of 91%, which falls slightly below the claim on usable capacity but well within the margin of error. As a result, Xiaomi should be praised for providing a figure of actual usable energy and actually meeting their claims.

Voltage Results – Rail A

It is important to remember that these voltages were measured inclusive of voltage drop in a short segment (20cm) of USB 3.0 lead and the connectors, which means that some of the voltage drop under loading is due to resistive losses in the rig.

On the whole, the output voltages showed some signs of regulation “fuzziness”, but the voltages were very well regulated on the whole. Even under the ~2A loading, the voltage did not fall below 4.85v, which is ample margin from the 4.75v USB minimum voltage requirement.

Voltage Results – Rail B

Rail B showed slightly lower voltages compared to Rail A, with more stringent regulation resulting in mostly nice straight lines with one exception at the 2A loading where it was somewhat more “frantic”. It is important to keep the scale in mind, as the jagged variances being recorded here wouldn’t really even show on a cheap 3.5 digit multimeter.

On the whole, while the voltages are slightly lower, they still don’t violate the 4.75v USB minimum voltage requirement, and are perfectly adequate.

Ripple Results – Rail A

At 500mA, the ripple was 53.48mV peak to peak, which is just above the ATX 50mV “standard”, but still quite significantly below the 150mV value that most stock chargers put out. The rate of oscillation was 343kHz.

At 1A, the ripple measured an average of 22mV peak to peak, which is a record low for my testing. Not believing my eyes, I zoomed in …

… and close up at a small time base, the average was 18.34mV peak to peak with a frequency of 475khz. An exemplary result.

At 2A loading, the ripple averaged 55.57mV peak-to-peak, again slightly more than the ATX 50mV standard but much less than the 150mV from standard OEM phone chargers. The frequency of oscillation is 946.5khz. This is a excellent result overall on Rail A.

Ripple Results – Rail B

The ripple characteristics of Rail B are completely different from Rail A. For the most part, there is negligible to no ripple on the output, except for bursts of noise which occur periodically. At 500mA load, these bursts came at a 25.2Hz repetition frequency and resulted in an average peak to peak reading of 92.15mV, mainly due to the bursts. The ripple voltage would be better than Rail A if the bursts were not present, but even with the bursts, the maximum ripple is below 150mV that stock chargers can put out and is safe for usage.

The characteristics continue with 1A loading, with 91.72mV peak to peak average ripple, and 25.2Hz repetition rate.

At 2A loading, the ripple averages 91.15mV with 25.2Hz repetition rate. As a result, Rail B has a higher average ripple due to the spikes, with mostly quite low-ripple voltage in-between bursts. The ripple characteristics are largely independent of loading.

Conclusion

Xiaomi have delivered once again, by producing a product that performs well, looks good and is available at a very competitive price. The unit shares similarities with the previous Xiaomi power banks by employing LG cells and TI power converters.

Quality components have been used throughout, and the performance very closely meets the stated figures. Ripple figures are quite commendable as well. The capacity of the power bank is very adequate for even power-users, and its design seems to be intelligent with a safety focus as well, which makes this unit hard to fault.

Interestingly, the two-rail design means that the performance of the two ports are slightly different, with different ripple signatures and slightly different efficiencies. On the whole, the differences are small and won’t worry most users, but it also means that it’s a bad idea to connect the output of the two ports together using a Y-cable.

Of course, it is important to note that such large capacity power banks can really stretch your patience, as the large capacity requires a long time to recharge (approximately 14 hours, subject to the capacity of your charger and the quality of your connection cable).

Those looking to purchase Xiaomi products should be aware of fake counterfeit products which don’t achieve the same quality or performance despite commanding similar prices. Look for the authenticity check scratch-off sticker and check the code to make sure you’re receiving a genuine product.

Gearbest are definitely worth considering, as they’re currently selling this unit on special, for AU$34.13 which I can vouch as being genuine based on the item I have received.