RoadTest: Maxim Integrated USB Type-C Autonomous Charger EVK
Author: Gough Lui
Creation date:
Evaluation Type: Semiconductors
Did you receive all parts the manufacturer stated would be included in the package?: True
What other parts do you consider comparable to this product?: Various ICs from Maxim Integrated, Monolithic Power Systems, Texas Instruments, Richtek Technology
What were the biggest problems encountered?: Unexpected current flow after charge termination, no jumper caps for header pins, LEDs impact quiescent current measurements, difficult to change current setting resistors.
Detailed Review:
Maxim Integrated MAX77751CEVKIT Evaluation Kit Review
By Gough Lui – December 2020 to February 2021
For most people, 2020 was a year they’d rather forget. For me, it’s been amongst the most productive. To cap it off and usher in the new year, here is my latest RoadTest review, with thanks to Maxim Integrated and element14.
It would seem that nowadays, lithium-ion and lithium-polymer batteries are everyone’s favourite chemistry. The cells are relatively inexpensive, offer better energy density than most other chemistries and are widely available in a range of different sizes. Likewise, USB has become a standard not only for data communications but power supply, especially with the USB Battery Charging Specification (BC) and the more recent advent of USB-C Power Delivery (PD). Putting two-and-two together, it’s clear that applications where USB recharging of lithium-ion and lithium-polymer cells are plenty – most portable devices you use nowadays will have some form of USB battery charging integrated circuit (IC) within it. The Maxim Integrated MAX77751 is a type of USB battery charging IC. More specifically, it is an autonomous, standalone, buck-topology switch-mode charger that supports USB BC, USB-C and power adapter inputs ranging from 4.7V to 13.7V while intelligently supplying power to the system and charging a single cell (or 1s battery) at up to 3.15A, adapting based on the charger input’s capability.
I’ve always been interested in batteries and power conversion. I spend quite a bit of my spare time characterising rechargeable batteries, building battery models and simulating batteries. It also happens that I have a passion for test and measurement, an accumulated cache of test equipment from prior RoadTests and directly-negotiated reviews, a growing proficiency in automation of test equipment and in data analysis. As a result, I applied to this RoadTest in the hopes I could put the gear I had to good use and deliver some insight into the MAX77751’s functional capabilities.
As always, if you found this review useful, insightful, interesting, helpful, entertaining or thought-provoking, I’d appreciate if you’d leave a like, rate the review, leave a comment or share it with your friends.
While lithium-ion chemistry batteries have a lot of conveniences, they also need some care in both charging and discharging to avoid potentially flammable results. In some inexpensive consumer electronic devices, these batteries may not be properly respected, thus leading to their early demise. For example, packs not being balanced, or being charged from modified power supplies with no termination capability, or with improper termination voltages. The following discussion overviews just the basics of lithium-ion charging of single-cells or 1s packs.
At a very basic level, charging of lithium-ion chemistry is usually achieved through a constant-current phase followed by a constant-voltage phase until the current falls below a threshold known as the termination current. The constant-current phase limits the current into the battery to a safe level that will not cause excessive heating and internal damage, while the constant-voltage phase avoids the cell being subject to overpotentials which would result in potentially explosive side-reactions occurring. Because these cells should not be subjected to the stresses of “floating” the cell voltage at this constant-voltage phase indefinitely, a termination current is set where if the current flowing into the cell falls below this level, charging is terminated and the process is deemed to be complete.
This process is complicated slightly by a few safety considerations. Cells which have been deeply discharged should not be subject to the full constant current and should instead be “trickled” up to a minimum level before applying the full current to avoid cell damage. Cells which aren’t within the right voltage range (i.e. fully charged, overcharged or overdischarged) should not be charged at all. Some packs may have a protection circuit attached which supervises the battery voltages and currents to protect against overcharge and overdischarge. When these packs are overdischarged, the protector “opens” which causes the pack to read zero (or very close to it). These packs need to be charged to “wake” up these circuits and reconnect the battery, but it’s hard to distinguish a shorted/faulty cell from one which has an open protector, so an even lower “trickle” is needed to check if the battery is alive or not. Finally, cells which are “leaky” and have developed internal shunts may cause a naïve charger to just charge them non-stop, never reaching the termination current – this should not be allowed to happen.
This is why a li-ion charger IC is often used to take care of all of these issues. The Maxim Integrated MAX77751 is no exception – it features cell-prequalification at 55mA with a 30-minute timeout, a trickle-charge at 300mA up to 3.1V, a fast-charge up to 3.15A (configurable), a top-off current of 100-350mA (configurable) with a top off timer of 30s and a total charge timeout of 6 hours. The charger automatically restarts if the cell voltage falls below 0.1V under the termination voltage. Termination voltages are pre-configured, available in a range of values from 4.10V to 4.50V in 0.05V steps. The MAX77751 also features USB-C CC and USB BC charger detection to configure charging currents and an automatic input current limit (AICL) feature that senses when adapters are being overloaded and limits current to prevent fold-back or shutdown. The input has a wide-voltage range of 4.6V to 13.7V operational, thus it is even possible to use the MAX77751 with non-USB sources of DC power such as a conventional barrel-jack input from a wall-wart.
The MAX77751 is an autonomous charger, configured by resistors and not otherwise needing any intervention from a host. This makes it convenient to operate as a standalone charger. It has two status lines for input status and charging status which can be used to drive LEDs or to provide basic charging status to a microcontroller if needed. There is a top off current control pin which can be used to shut-down the charger as well. While it does use the D+/D- and CC lines for charger detection, once detection is completed, an external switch can be used to take over the lines for data transaction, if desired. Additionally, it supports USB OTG reverse-boost, allowing it to boost the battery voltage up to 5.1V at 1.5A for providing power to USB accessories.
The charger also supports Smart Power Selector™ which is a power path management system which allows for supplying power to the system while charging, only diverting excess power to the battery. It even allows the system to draw more power than supplied by the charger, taking the remainder from the battery. It provides an over-current protection function with automatic retry. It is also equipped with a true battery disconnect FET to isolate the battery in case of a fault. Best of all, all of the switching elements are integrated into the chip – all the support necessary is to provide a few passives and an inductor. It is also robust to overheating, automatically folding back charging current initially and shutting down upon reaching a maximum temperature.
Most of the time, a designers’ approach would be to identify their functional requirements and then go searching for parts that fit the bill. For this market survey, I’m trying to do the reverse – given the MAX77751 that I’m reviewing here, identify parts which might be comparable for a comparison. Because of the versatility of the MAX77751, this task is actually more difficult than it first seems.
The MAX77751 is amenable to use in several different scenarios, but users who might choose the MAX77751 may only need a subset of these capabilities. For example, while the MAX77751 can operate standalone, the INOKB and STAT lines can be used to provide basic charging status to a microcontroller. While the MAX77751 features USB BC and USB-C handling, it is also capable of operating with just a raw DC power input (e.g. from a power adapter) from a range of about 4.7-13.7V. Also, while the MAX77751 features USB-OTG reverse-boost mode for a 5.1V/1.5A output, it is equally possible to disable this mode entirely and just use it as a charger.
As a result, the MAX77751 actually straddles several battery-charger IC markets, including the “wider” input range chargers (not just 5V), USB controller + charger combinations, autonomous and “connected” chargers (to an extent), USB-OTG supplies (or power bank controllers to an extent) and higher current switch-mode chargers (>2.5A). Depending on what you’re looking for, the MAX77751 could be a choice amongst a plethora of devices – I collected about 80 datasheets on USB-compatible single-cell Li-Ion switching chargers alone.
To have a sensible market survey, I decided the essential features for this comparison would be:
Other capabilities were not considered essential but served as additional segmentation of the market, for example:
I began my search on component supplier websites, but found them relatively limited, so instead I visited the manufacturer’s websites directly. Manufacturers with products meeting the brief included Maxim Integrated, Texas Instruments, Monolithic Power Systems and Richtek Technology Corporation. The resulting list of ICs were charted based on their datasheet specifications and compared below with indicative pricing from electronics suppliers online at the time of the review.
The survey is broken up into two tables – the top-left containing standalone resistor-configurable ICs some with I2C capabilities. The bottom table contains chargers which rely on an I2C interface, although some are also capable of limited standalone operation (with the loss of some configurability). The MAX77751 is highlighted and on the left for easier comparison.
The MAX77751 stands out from the rest of the solutions for being the unit that has the lowest unit price, the smallest non-BGA package and support for USB 3.0 CC pins. It is also one of a few that have fixed termination voltage rather than configurable termination voltage, so ordering the correct part number is important. It has no I2C interface capability, being configured through resistors for autonomous usage, but still has status pins which can be read by GPIO or used for status LEDs. Charging current is reasonable compared to the peers, with some others reaching as high as 5.5A, while voltage window is enough to handle 5V, 9V and 12V (common fast-charge voltages) but not as wide as claimed by some of the others. Part of the reason may be that others specify a lower input voltage that is not sufficient to charge the battery but enough to enable their system rail output (SYS) only and some of the others may also specify a higher maximum input but may not be fully operational at that voltage. The things it lacks are the support for an NTC thermistor to control the charge voltage/current (requiring an external circuit to interface to the ITOPOFF input) and the internal resistance compensation feature (or remote sensing) to determine the actual cell voltage.
On the whole, the MAX77751 is a perfectly reasonable adaptable high-current autonomous charger for both USB and non-USB applications. If I were to have a wish-list, it maybe to integrate USB data line switching into the IC, along with overdischarge protection since the SYS output is already being regulated by the Smart Power Selector™ functionality.
The evaluation board comes in a small, plain brown cardboard box with a label on the underside.
Popping the lid reveals some pink anti-static foam, wrapped around a zip-lock static shielding bag containing the evaluation board. There are no other contents within the box – not even a disclaimer, warning about RoHS or California Proposition 65. There’s not even a link to the manual or documentation – although any resourceful engineer should be able to find that with ease. I think this is not a bad thing as it saves waste overall, but it does feel a little hollow given that the evaluation board goes for US$85.31 (Newark) / AU$130.24 (element14 AU).
The board is a four-layer PCB with an ENIG/ENEPIG finish and no major defects in terms of soldering, exhibiting only light scuffing of the solder resist. The board itself has a fairly large footprint, no doubt to improve thermal dissipation performance. Connection of inputs and outputs are all sensibly placed on the edges. Unfortunately, no jumper caps are included for the header pins.
Wire connections are provided as “loops” which eases the difficulty of soldering onto this multi-layer board. In my case, a preferable alternative would be the use of terminal blocks, however, those aren’t as “compatible” with alligator clips. Some of the loops do get a little close to one another, so some care is necessary in setting up the connections. Oscilloscope-probe test-points have been fitted which duplicate some of these connections for measurements. It is nice to see an excess of ground connection loops for probing convenience. The status lines have LEDs and a dropper resistor pre-fitted allowing for easy diagnosis of charger behaviour without using any test gear, but this will have an effect on the measurement of quiescent current and efficiency.
To make evaluating USB charger performance easy, the board has both a USB-Micro B socket and a USB-C socket preinstalled, with breakouts for the CC1/CC2 and D+/D- pins. These occupy one side of the board, which is a nice layout choice, although some of the traces could be straighter.
The board really makes a point to demonstrate how compact the footprint of the circuit is. The chip itself is 3x3mm, so this circuit is smaller than my thumbnail! That’s quite impressive. On the downside, changing resistances for the top-off and fast-charge current settings is made difficult (especially by hand), and realistic implementations are likely to need more footprint simply for thermal dissipation reasons. Thus, to assume that you could get away with only the enclosed area of the board is perhaps somewhat misleading if you expect to run the full current or higher voltage input.
To get started with the board, it is necessary to have access to test equipment such as a digital multimeter, power supply, DC electronic load, and various accessories such as a USB charger and suitable battery with the ability to handle the pre-configured 3.15A charging current and 4.20V termination voltage.
For the battery, I decided to use up to two 26650 Li-ion cells in series to be able to handle the charging current and discharge current. I built a carrier using Veroboard and two BLM battery holders which should be good for about 4A charge and about 10A discharge. This terminates to two screw terminals where wire would be terminated to crimp ring or fork terminals.
The other thing that is necessary is to make connections to the board. At a minimum, for testing charging, the battery must be attached between BATT and GND with charging provided via the onboard USB connections. As I desired testing of charging input from a power supply, I also used connections to CHGIN and GND. To compensate for voltage drop, I added additional wires for four-wire sensing mode for greater precision, resulting in more accurate figures.
Attaching the wires is best performed by soldering to the loops, although it may be possible to use alligator clips. For initial testing, I used terminal blocks that suited my test equipment, although some wiring was changed to be terminated to banana plugs or crimp lugs as necessary. Later on, for testing of the power path, attaching wires to the SYS connection is also necessary.
By the end of all this testing, my EVKIT PCB had a lot of wire hooked onto it … and a crimp snipped off as I went back and forth between battery and simulated power tests.
NOTE: In all tests, all measured voltages are relative to the voltage at the terminals of the EVKIT PCB measured by four wire sensing. Voltage drop due to leads to and from the EVKIT have been negated using four wire sensing for accuracy, however, any voltage drop within the EVKIT’s own PCB is not compensated for. In my opinion, this provides the most accurate evaluation of the kit’s performance, as voltage drops outside can be significant at high currents making two-wire connections less suitable.
Documentation consists of a datasheet for the MAX77751 (Rev 4, 11/20 at the time of posting) and a MAX77751CEVKIT User Manual (Rev 2, 9/20 at the time of posting). The datasheet is well-presented, fairly complete, easy to read and well laid-out over 40 pages. The User Manual is similarly worthy of praise, despite being much shorter at 9 pages. It details the various test configurations which can be used with the kit – with a real battery, battery simulator, electronic load and power supply. It provides test protocols for some of the other features, a full bill of materials and a list of component suppliers. Unfortunately, the design files for the evaluation kit board do not appear to be available for download, although the layouts are documented as images in the evaluation kit PDF and suggestions on layouts are provided in the datasheet.
This chapter will detail some of the performance parameters which were tested with my Rohde & Schwarz NGM202 two-quadrant power supply and battery simulator. The simulated cell in question is my Keeppower UH2655 IMR 5500mAh 3.7V 26650 Li-Ion cell using the extended model I created here. Because of the nature of battery simulation, noisy measurements are to be expected as the simulator “loop” adjusts the power supply channel’s voltage and resistance values based on the state of charge and present current but only with a limited bandwidth. Other tests were performed with the supply providing a stable voltage or current, or through the use of internal resistance mode to simulate an overloaded charger or poor-quality USB cable.
The first test I wanted to check was with regards to the MAX77751’s safety timeouts. To test the pre-qualification time-out, I set the simulated large cell to an over-discharged state of charge and let it “pre-qualify” at a trickle and measured the time until it stopped. To test the charge safety timeout, I set the sinking channel to maintain a fixed voltage, thus basically acting like a “stalled” battery and measured the time until charging stopped.
In both cases, the safety timeouts operated as intended, kicking in slightly earlier than expected. The pre-qualification current was measured at 58.5mA which is within the range in the datasheet (40-80mA, typical 55mA).
While I had intended to perform as much of, if not all, testing using a simulated battery for practicality, this was not possible as the charger is capable of 3.15A while the NGM202 is only capable of sinking up to 3.01A. It was not possible to get a proper test of the charging characteristic with the full input. Instead, charging “skipped” the CC phase entirely, as the voltage “flew” up to the point of being into the CV phase as the simulator could not sink the pre-configured 3.15A. Note that a positive value indicates current being drawn from the cell, while negative values indicate current being put into the cell (i.e. charging).
Because of this, it was not a good test of the MAX77751’s regular operation, however, it was seen that the charge terminated at 113mA with a voltage of 4.193V, within range of the datasheet claims (configured for 100mA ±5% ±20mA (=75-125mA), termination voltage of 4.2V with a tolerance of -0.9% (=4.1622V) to +0.3% (=4.2126V), typical -0.3% (=4.1874V)).
Reducing the charging current could be achieved by reconfiguring the charger’s fast charge resistance, but this was not something I had any intention or capability of doing due to the fine pitched components. Instead, I exploited the capability of the charger to “fold-back” current in case of a poor supply which I simulated by using the NGM202’s internal resistance function, simulating a 0.5Ω, 0.25Ω and 0.175Ω internal resistance. I computed these values to provide approximately 0.5V drop at the target current of 1A, 2A and 2.86A, knowing that the AICL function would likely be a little more conservative about allowing so much voltage drop.
The resulting curves have a lot of “noise” which seems to show the current falling to zero irregularly. This may be because of the way the simulated battery and simulated internal resistance response of the NGM202 interacts with the charger’s switching. However, the charging was successful and the charger did not seem to “glitch” at all. At 0.5Ω, charge current was about 700mA but peaked at 1A at the transition between CC and CV phases. At 0.25Ω, the current was about 1.4A, peaking at 1.85A and at 0.175Ω, it was about 2A, peaking at 2.5A. In all cases, the charger tries to regulate the input voltage to approximately 4.65V. There is some noise around the termination point which seems to “recur” periodically – perhaps this is indicative of interaction of the simulated battery with the charger.
To stress the AICL function, I decided to remove the internal resistance feature and instead set the NGM202’s output to have a strict 1.2A current limit. Because the supply reacts quickly to overcurrent, any attempt to go over the limit will result in the channel output collapsing rapidly.
To the MAX77751’s credit, this did not cause the charger to glitch or reset, despite rapidly bouncing in and out of constant current mode according to the LED indication on the front panel. The recorded data at 10Hz shows less oscillatory behaviour than in the internal-resistance case however, which is probably an artifact of the logging. With a current limit of 1.2A, the input current consumed was about 1A for most of the charge, delivering about 1.2A into the cell. The peak towards the end of the charge where it transitions between CC and CC is well limited to 1.2A to no ill effect. As a result, it’s apparent that the AICL functionality works well to accommodate simulated chargers of different capabilities.
Battery quiescent current was characterised by performing an I-V sweep from 4.2V down to 0V on the battery input with nothing connected to the SYS output and no input from the CHGIN input either. Battery quiescent current was measured with both the reverse boost disabled and the reverse boost enabled. Because of the connected LEDs to the status lines, the boost-enabled current readings are higher than the datasheet values.
With the reverse boost turned off, the quiescent current rests roughly about 27µA for voltages above about 1.6V. Below that, it seems the consumption drops off to below 2.5µA. This corresponds well with the datasheet claim of 30µA. With reverse boost turned on, above about 2.5V, the current consumption reaches between 6.4mA to 7.8mA depending on battery voltage under a no-load condition. This is higher than the 2mA rating in the datasheet, but that was achieved with the voltage forced above the regulation setpoint to force a no-switching condition while also not having LED indicators on the status lines. I would imagine that most of the measured current corresponds to the draw of the LEDs on the status lines.
Charger input quiescent current was characterised by performing an I-V sweep from 4.7V to 13.7V on the CHGIN input, while the BATT input was provided 4.2V (full charge voltage). Again, since the evaluation board has LEDs connected to the status lines and is likely the switching converter is still operational, the measured current is expected to be higher than the datasheet values.
The MAX77751 has some unusual tendencies in this regard. It was consuming current from both the charger and the fully-charged battery. Later on, we can see that the termination behaviour of the chip has a few quirks, so the results are a little difficult to interpret. However, it seems that the USB-consumed current ranges from about 3.5 to 6mA, while also consuming about 4 to 5mA from the battery within its operational voltage window. A spike of battery consumption was seen at the low voltage end, but it seems the quiescent battery consumption without USB supply (above) is fine.
Testing of the operational voltage range involved sweeping the input voltage through the range of 4V to 16V. The voltage and current for both input and output were recorded, to allow for a plot of current and efficiency.
The charger started up at 4.696V which is close to the 4.7V claimed. Unfortunately, the shutdown at 13.638V is a bit short of the datasheet claimed 13.7V typical but within the 13.4V minimum. This voltage is still sufficient for use with a regulated 12V source. As promised, the unit did not suffer any damage from being supplied with 16V. I wouldn’t recommend this for use with a raw automotive 12V supply as a result which could reach 14.45V on full charge and contain spikes which may exceed 16V causing damage to the converter. The “battery” was kept at a steady 3.01A sinking which resulted in a voltage of 4.181V and the unit being in the CV phase of the charge. The computed efficiency was greatest for the least voltage difference at the 5V end of the spectrum, reaching around 92% efficiency, falling to about 87.5% at 13.6V. As a result, the charger is expected to warm up more when charging at higher voltages as it is anticipated to dissipate about 1.1W at the low end and 1.8W at the high end of the voltage range. Testing with the IR laser thermometer showed the package levelling out at about 50°C while operating in a 25°C ambient, which is very comfortable as expected because of the large four-layer PCB design. The chip begins folding back the current once it hits 130°C junction temperature, which implies a case temperature of about 116.5°C.
Testing of the USB OTG Reverse Boost functionality was performed by setting the input voltage from the NGM202 to either 3V, 3.6V or 4.2V, simulating an empty, half or fully charged battery. The load was using the other channel of the NGM202, which was set to sweep and generate an I-V curve. From knowing both input and output powers, the efficiency of the converter can be calculated, noting that the LEDs on the board would have diminished the efficiency slightly.
The USB OTG output can be seen to have overcurrent protection, cutting off the output at a load of around 1.61 to 1.63A. The output voltage is about 5.094V unloaded, falling in almost a straight line as the load increases, down to a minimum observed in the 5.04-5.05V region. The voltage drop is slightly greater when the battery voltage is reduced.
Converter efficiency is excellent, eclipsing 80% efficiency above 40mA and reaching a peak around 94% in the half-full battery case. The efficiency is maintained above 90% for more than half of the load range, which is impressive.
The MAX77751 does not offer an over-discharge protection feature, however, it is still configured to provide power to the attached system through a SYS output. This output takes advantage of the MAX77751’s Smart Power Selector™ which routes current to the system first and takes the surplus to charge the battery, allowing the system to come-up without a valid battery and also supplementing inadequate current from the USB with current from the battery. Best of all, providing power in this way will avoid confusing the charger (i.e. battery charge never terminates because draw on the battery is greater than the termination current) or stress on the battery (e.g. from microcycling where the battery is discharged a tiny bit and recharged repetitively).
To test the SYS output dropout, I simulated a battery input by using the NGM202 as a voltage source and swept the voltage down from a high of 4.2V. I configured the other NGM202 channel to be a load, drawing 3A from the SYS output (which is capable of an absolute maximum current up to 4.5A). The point at which the SYS output is “lost” is deemed the drop-out voltage, and this also allows us to calculate the “loss” of voltage by routing power through the MAX77751.
The loss of voltage was quite limited – at a full charge, it was about 50mV which corresponds to a resistance of about 16.7mΩ. By the end of discharge, say 2.5V, the difference was about 70mV corresponding to a resistance of 23.3mΩ which is quite tolerable and close to the 20mΩ datasheet value. The MAX77751’s SYS output dropped out entirely when the voltage hit 1.572V, dropping the SYS current to zero. Thus, the drop-out voltage is low enough to use all the capacity of a cell, but is too low to use as an over-discharge protection measure.
Switching frequency is not noted within the datasheet but could be an important parameter for EMI/EMC reasons. This analysis was performed by providing power from the NGM202 at 5V, 9V and 12V and measuring the waveform across the LX and GND connections using a Rohde & Schwarz RTM3004.
To minimise probe grounding inductance, I very carefully laid the probe as such to allow the centre connector to touch the LX loop while the collar of the probe (ground) touched the GND loop. This avoided the need for a ground clip which may exaggerate switching waveforms due to inductance.
Starting at 5V, the switching waveform was seen to “dither” somewhat already which may help with EMI control (a sort of spread spectrum). The measure computed rise/fall times may not be accurate as the transitions are not monotonic, but the cursor computed frequency is about 727.8kHz. With a load as imposed (4.181V/3.01A), the converter is still working hard, with only short periods of off-time.
Increasing the input voltage through to 9V and 12V, the duty cycle of the converter can be seen to reduce and the frequency seems to increase. As a result, any potential EMI problems may be affected by the input voltage to the chip (and also potentially by the battery state).
To illustrate this, I tried simulating 0.5Ω internal resistance with an input of 5V, output being at 3.6V at 1A and the switching frequency is around 1.46MHz. Using persistence on the oscilloscope also allows the “dithering” nature of the waveform to be captured.
Using the FFT feature on the regular 5V input case spanning 0 to 30MHz, we can see the harmonic series of spikes with a “wavy” floor pattern possibly because of how the dithering affects the signal. A power-off noise floor is provided to illustrate that the waviness is not due to the instrument noise floor. Of course, how much this is a problem in practice depends on your circuit board design – measuring the switching node directly is not all that predictive of EMI issues as connections are usually kept short for the best performance and efficiency and less of this will be seen on the inputs/outputs which would be bypassed by capacitors.
As the battery simulator wasn’t capable of properly simulating the full current capability of the system, I had to revert to my “Plan B” of using a pair of Shockli 26650 IMR 3.7V 5500mAh Li-Ion cells. I had already been prepared for this possibility from the outset, as real cells may behave differently to simulations, but the choice of using two parallel cells was mainly to satisfy the need to charge at 3.15A (the cells are specified up to 2A charging only) and the need for testing the SYS overcurrent protection trip at about 6A. In return, the collective 11Ah capacity would make for a rather decent use of the charger’s 3.15A charging capability and corresponds to the capacity of some of the larger Li-Poly cells available today.
This set-up unfortunately means that I need the Rohde & Schwarz NGM202 to provide power and perform voltage measurements through DVM mode, but also my Keithley 2110 5.5-digit DMM to provide current readings (which introduces burden resistance) and my B&K Precision Model 8600 DC Electronic Load to drain the cells post-test. This also introduces some wear of the cells, thus slight variations in results are to be expected due to the non-idealities of real-world tests. Monitoring the USB connection and charger identification required the use of my Rohde & Schwarz RTM3004 oscilloscope as well, making for a rather unwieldy test set-up which should be able to suss out everything one would want to know about the charger IC’s behaviour.
Charging was undertaken at 5V, 9V and 12V inputs from the NGM202. The resulting data was plotted for voltage, current and efficiency profiles. The cell is connected such that a positive value indicates current being drawn from the cell, while negative values indicate current being put into the cell (i.e. charging).
Unfortunately, I did not anticipate the issue that the combined battery capacity was too big, thus the charge did not properly terminate in the top-up mode. On the upside, these graphs reconfirm the functioning of the 6-hour charge safety timeout.
A look at the current and voltage profiles shows a relatively typical Li-ion charge curve, albeit terminating at about 6-hours due to the charge safety timer. The curves are slightly offset because the battery was discharged to a slightly different level for the commencement of each test (a result of using a “real” battery as opposed to a simulated one and variations in dwell time between discharge and commencement of charging).
Looking at the CV and CC portions of the charge, the fast charge current was mostly between 3.14 to 3.15A which is quite good, although at 5V, the transition to CV did look a little messy. The continuous voltage portion measured between 4.181V and 4.192V, which is within the expected range and is likely varying due to temperature.
The initial trickle current was spot-on at 300mA, while termination was a little either side of 6-hours. The termination occurred when the cells were still soaking 350mA, so it was very close to fully charged but not quite. The termination in the case of 5V input did exhibit some strange non-zero current for a while post-termination, which is repeated in the following tests of charging restart and timeout. This behaviour was not expected but is not likely to result in any serious negative consequences.
The collected data allowed for the computation of the efficiency of the charger, which ranged from a low of about 80% in the trickle regime at 12V to a high of about 92% at near full charge with a 5V input. This is consistent with the single-point test result in the simulated battery chapter, but also consistent with the datasheet graph.
As this was not a complete test of the MAX77751’s charging capabilities, I decided to try again with a different cell – see the next chapter for full details.
Despite this little hiccup, it was a good chance given the cells are almost fully charged, to restart the charger, allow the cells to fill fully and examine what happens to the charger when the cells voltage starts to fall after it terminates after top-up has completed. According to the datasheet, once it falls to about 100mV below the termination voltage, charging should restart. To simulate this, I applied a load to the cells of about 30mA initially, stepping up to 80mA using the B&K Precision Model 8600 connected directly to the pack (bypassing the DMM). This was chosen so as to be below the termination current to ensure that the charger can restart and the top-up can complete successfully, but also so the DMM could indicate how much current was flowing from the charger at any given moment.
At the beginning of the graph, the charger completes a proper charge termination. The voltage of the cell falls somewhat at that point, because it may have been a little elevated by the resistance of the wiring from the charger to the cell. To help it along, I then apply a small load to the cells, simulating a “leaky” cell.
Unfortunately, some unusual behaviour was exhibited by the circuit. Instead of actually “hard” terminating, as the voltage of the cell started to fall, the charger started to push a tiny bit of current into the cell, thus countering my efforts to slowly deplete the cell. This current reached almost 15mA and at a point in time, it was even drawing a tiny bit of current from the cells around 4.17V. This was not an error of the test equipment – the current consumed by the charger follows the same trend in reverse! This seems to be somewhat “analog” behaviour and is not entirely desirable, but also is not catastrophic either.
After many hours of equilibrating, I decided to push the leak rate higher, up to 250mA, which rapidly dropped the cell voltage down to 4.11V where the charger restarted. This is close to the 4.10V expected, but because the leak rate is now greater than the charge termination, the charging does not terminate normally and instead “equilibrates” with the charging current being about 250mA which is all being sunk by the load until the 6-hour charge safety timeout kicked in and “latched”. This stopped the charger and prevented the charger from restarting, as expected.
As it turns out, I was a bit too optimistic with the 2p arrangement in the previous section, resulting in a battery pack that was too big to observe charge termination. As a result, I swapped over to “Plan C”, using a single Keeppower UH2655 26650 IMR 3.7V 5500mAh lithium cell. The advantage of this cell is a manufacturer rating allowing up to 4A of charge current, thus a single cell would be able to safely tolerate the charger output. The test setup remains the same, just that the tests had to be repeated.
This section repeats the tests in the previous chapter, except for using one 26650 cell that can handle the rated current instead of two in parallel.
Charging begun at approximately the same cell voltage, but it seems that in the 5V supply scenario, the shift from trickle charge mode into fast charge was somewhat delayed compared to 9V and 12V. It is listed as 3.1V typical with a range from 3.0 to 3.2V, so it is close to being within specification but the cause is unknown. Charge time ranged from about 3h38m to 3h47m. The charge current profile shows a traditional clear trickle-fast-taper-terminate mode of operation, as promised.
Zooming into the CV and CC sections, the CV voltage is seen to be around 4.181V to 4.191V. The CC current was mostly within 3.14A to 3.16A although the transition from CC to CV in the case of a 5V input featured a region of “noisy” sloping through 3.12A to 3.14A. This is inconsequential to the battery charging process, but is likely due to the internal architecture of the charger.
The trickle charge current is perfect, being pretty much spot-on with the 300mA claimed in the datasheet. Termination, in this case, did occur being approximately one minute after it passed through the 100mA termination current. This is longer than the top-off timer specification of 30 seconds, but is likely because of a current measurement difference – i.e. when does it detect passing through 100mA. If we consider the portion when the “noise” in the current trace stops, then the top-off timer is about 30 seconds, however, the termination current is closer to 98mA.
The unusual behaviour with termination at 5V input is again seen, with the termination not being a “hard” termination. Instead, the current drops but about 6mA is still being put into the cell at this time. Eventually the “leakage” characteristic slows depending on voltage of the battery, but this was not expected behaviour especially when compared to 9V and 12V curves. I suspect it is unlikely to be consequential to the battery’s lifetime as the current is rather small and will only persist while the item remains connected to USB power.
The real test of the MAX77751 is from using it with commercial USB chargers. For this, I tested the evaluation board with the following mixture of chargers of different types and current ratings:
In order to observe the signalling detection routine of the MAX77751, I used my Rohde & Schwarz RTM3004’s four analog channels to monitor CC1, CC2, D+ and D- and used two digital channels to monitor INOKB and STAT to determine when the charger had successfully completed its detection routine. This required a rather extensive set of probes to be clipped onto the board.
Testing with the USB-C ZMI HA932’s 65W USB-C port showed the following:
The charger detection routine completes in about 500ms. The CC1 and CC2 pins are reversed if the cable is reversed at the charger end (right screenshot). The CC pins measure about 1.75V with periodic pulses towards zero. The unit has a 5.1kΩ pulldown and it is expected that the charger would have a 10kΩ pullup to 5V to signal 3A capability, thus the expected voltage would be (about) 1.69V. If the charger had 1.5A capability, it would use a 22kΩ resistance and the voltage would be 0.94V. Finally, if the cable is a USB-A to USB-C cable or it can only offer “default” level (500mA USB 2.0 or 900mA USB 3.0), then it would be marked using a 56kΩ resistance and the voltage would be 0.42V.
The D+ and D- lines show equal signals (indicating they are connected together) which signals the presence of a USB DCP. The charging starts almost immediately, and the STAT pin begins its blinking at around 1s. The charger itself seems to generate quite a bit of noise once it starts up! Charging begins and runs at about 5V/3A input.
Using the 18W port showed a virtually identical result. This is not unexpected since the charger is about 15W (5V/3A) and the USB-PD specification states that such loads should use 5V and only larger loads are permitted to use 9V, 12V and 20V levels. Because of this, the wide voltage range capabilities of the MAX77751 seem to be a little moot assuming the use of USB-C inputs only.
Testing using the USB-A port on the HA932 using a USB-A to USB-C cable shows that the CC line is pulled up to about 0.48V signalling “default” USB power, and the D+/D- detection routine appears to find a DCP which results in charging at 1.5A. Using a USB-A to USB micro-B cable to this evaluation board makes little difference, as now the CC lines are “floating” zero and the D+/D- detection routine confirms that a DCP was detected and charging occurs at 1.5A.
Testing with other chargers showed that chargers were categorised correctly into types, and the current rating applicable to the type was applied with automatic input current limiting for the weaker adapters. Unfortunately, this means many of the 2A/3A capable adapters that signal as a DCP are capped to 1.5A charging with this solution, unless you deliberately force the MAX77751 to “unknown” detection by tampering with the USB connections. On the whole, this may not be such a bad thing especially for USB-A and micro-B based solutions as the plugs are rarely rated for more than 1.8A. In this respect, Apple 2A/2.4A and Samsung 2A chargers are at an advantage for this particular chip as they are proprietary types that are detected and configured for the full current limit automatically (aside from USB-C chargers). Unfortunately the solution doesn’t take advantage of modes such as Qualcomm Quick Charge which offers 9V/12V or quasi-continuous voltage adaptation which would be able to provide quicker charging while respecting a 1.5A current limit because of legacy cable or connector limitations.
Examples of DCP detections –
The ZTE signals as a DCP but can only handle 700mA of sourcing. After starting up, AICL reduces the current loading to about 691mA. The HTC signals as a DCP as well, but is capable of 1A of sourcing. In this case, AICL reduced current loading to 1.1A.
The Chicony is a 2A capable DCP, but because of the standards-compliant DCP implementation, 1.5A is drawn from the adapter. Similarly, the Motorola TurboPower adaptor is a Qualcomm Quick Charge compliant supply which is capable of 3A at 5V, yet because it is detected as a DCP, it too only charges at 1.5A.
Examples of SDP/CDP detections –
A standard USB-A port is detected and 500mA charging is enabled, respecting the port’s capabilities.
A CDP is detected in both USB-A and USB-C varieties and charging at 1.5A was observed. It seems that the detection waveforms are inconsistent with the datasheet examples, however, functionally the MAX77751 did correctly identify the port and set an appropriate charging current in both cases.
Examples of Proprietary detections –
An Apple 10W USB Charger is correctly detected and charges at 2A. An Anker PowerCore+ 10050mAh power bank uses a controller that claims to be “intelligent” and adaptive (PowerIQ). The detection result is atypical, but in the end, it seems the MAX77751 concludes it is a DCP and charges at 1.5A.
Similarly, a Xiaomi Power Bank Pro is detected as a DCP and charges at 1.5A. However, note how the intervening steps do not reach 0V - so there probably is some kind of non-standard signal that is being put on the output to "improve" compatibility with other devices.
Charging was run using the ZMI HA932 USB-C Port, Apple 10W charger, Motorola TurboPower, HTC and ZTE chargers to spread the full gamut of dedicated charger charging currents. In all cases, the charger operated as intended, although the time taken is dependent on the available current.
The slowest charger in the lot is the ZTE which could only source 700mA. It was so slow that the charger quit on the charge safety time-out, as designed, at the six-hour mark after entering fast charge. The constant current section was a little bumpy at the beginning and tapered slightly possibly because of changes in efficiency with changes in cell voltage and charger capability changes as it warmed up. The input voltage was dragged down to about 4.5V for the CC section.
The slightly-more-capable HTC charger was sourcing about 1A and completed the charge in just over five-and-a-half hours. Again, a bump in the CC section is seen along with a slope. Proper termination was seen, with the adapter voltage being just about 4.5V for the CC phase.
The Motorola charger was hamstrung to DCP limits of 1.5A but it was able to complete charging in just over four-and-a-half hours. In this case, the bump and slope in the current can be seen at the beginning of the CC phase, but because the charger has a negative internal resistance (designed to compensate for USB cable voltage drops), the voltage actually increased to about 5.15V as the current draw increased.
The Apple charger was quicker-still, being allowed to charge at 2A, it achieved a full charge in under four-and-a-half hours. The voltage was pulled to about 4.9V in the CC phase, as it was within its specifications, and the bump at the beginning of the CC phase is less pronounced.
Finally, charging with a proper USB-C source and USB-C cable resulted in a charge current of 3A and delivered the quickest charge of all, completing in just under four hours. The voltage on the adapter fell to about 4.75V, but the bump and slope of the CC section are notably reduced. The CV section now dominates the charge time, as is usual for lithium-ion batteries when the charge rate is increased.
The MAX7751’s SPS allows for intelligent routing of USB power to the host system while taking surplus power for charging, even under fully-depleted or no-battery scenarios. This set-up also allows current draw beyond the USB input’s capability to be supplemented from the battery. This section will investigate what the SYS output voltage looks like under various loads and how the current routing occurs under a no-battery, empty-battery and full-battery scenario.
Note that this test is deliberately left to the end of my evaluation as the test will exceed the absolute maximum continuous current ratings for SYS and BATT which is 4.5A. It is noted that the OCP for BATT and CHGIN are both set for 6A to accommodate current spikes on start-up, however, these tests show that the OCP on its own will not guarantee respecting the maximum continuous currents thus device lifetime impacts could still occur with currents below the OCP threshold.
Under the SPS regime, the SYS output voltage is dependent on the installed battery.
Where there is no battery or a full battery, the output will sit at the full voltage (4.2V or thereabouts). If the battery installed is being charged, the SYS output will sit limited at the voltage of the charging battery.
With no battery, as the load increases to 3.15A, the voltage starts to fall and by about 3.6A, the SYS output collapses entirely as the USB input (the sole source of power) drops out. This is under ideal circumstances where the unit is connected to a 3A capable source, a lower current source with no installed battery would (obviously) drop-out at a lower current. If a battery is available, the voltage will collapse to the battery voltage in which case the battery then augments the USB current.
As the current load increases, a discharged battery may have its voltage collapse too far as it becomes discharged causing the solution to drop-out. Otherwise, the current could reach and exceed 9A before a shutdown occurs – in this case, as the battery OCP was reached (6A + approximately 3.15A from USB). Note that these values are well beyond the absolute maximum continuous 4.5A current rating, however, the charger IC did not fail in this test.
To more clearly illustrate where current is coming from, I have broken out the above two graphs of the charged battery and discharged battery scenario. It is clear that in the charged battery scenario, the current is provided by the USB supply up to about 3.15A of output when it remains constant (at its limit). The battery soon provides the shortfall, with the curve becoming noisy above 5A (perhaps the chip is being stressed due to heating and the cell contacts are not as adept at high currents). Reaching above 8.5A of SYS output current, the USB current falls for the first time, which is picked up from the battery, pushing it towards the OCP where it shuts down entirely. This test pushed the chip well past its rated maximums.
In the discharged battery scenario, the current from the USB remained high throughput, pegged at the maximum as it was initially charging the battery until all the current had to be delivered to the SYS output. Then, the shortfall was provided by the battery discharging until about 6.5A when the battery voltage may have fallen too much causing it to drop out. These tests prove that the SPS system was operating as intended, actively routing the power as necessary to meet the demands of the system and battery.
Lithium-ion batteries and USB are both ubiquitous and frequently used together. Properly charging a lithium-ion battery from a USB supply safely requires careful respect of the cell’s characteristics and is most easily achieved using a charger IC of which the Maxim Integrated MAX77751 is an example.
Compared with the competition, the MAX77751 is a low-cost, capable and versatile standalone charger IC. A key distinguishing feature is the onboard autonomous handling of USB 3.0 CC pins and D+/D- detection of SDP, CDP and DCP sources in addition to some proprietary chargers. It has a wide input voltage range of 4.7-13.7V, charging capability up to 3.15A and the ability to provide USB-OTG reverse-boost output of 5.1V/1.5A. It is capable of Smart Power Selector power path management and automatic input current limiting to operate with weaker adapters or poor cables. It can even be configured for use with non-USB sources of power, such as barrel plug inputs from wall-wart adapters. Another benefit is that the package is not a BGA package, which may reduce assembly complexity and cost.
Among the downsides is that some competitors offer higher charge currents and a mixed autonomous (with limited configuration)/hosted option with full configurability through an I2C interface whereas the MAX77751 has only two status outputs, resistor configuration and fixed termination voltage. However, in the intended autonomous standalone usage scenario, these are not major issues. Reuse of the USB lines is possible, if an external switch is used after the INOKB signal indicates the input detection routine is completed. Another downside is the lack of inbuilt thermistor input handling, thus issues could arise when charging or discharging under extreme temperatures.
The evaluation board itself makes the advantages of the chip clear by demonstrating a very small footprint layout with both USB-C and USB micro-B connections in addition to test loops and LED indicators. The board is a high-quality four-layer PCB, which seems to be thermally advantageous to the solution, with a convenient arrangement of connections for evaluation with quality documentation available online. On the downside, the documentation does not include PCB design files, only images of the layout, schematic and layout recommendations. The inclusion of indicator LEDs also affects certain measurements while reconfiguration of charge current resistors is made difficult by the small size of the resistors and layout.
On the whole, the charger proved to operate as described from both commercial USB charger supplies and bench supply inputs across its voltage range. Charge safety features of pre-qualification timeout and charge safety timeout were demonstrated as functioning. Charging current and termination voltages were within specifications and USB detection proved reliable and robust across a range of SDP, CDP, DCP and proprietary chargers. The operation of the automatic input current limiting feature was capable of handling both gentle and “hard” current limits without causing sustained overload of the input or interruption of the charging process. The capabilities of the charger to automatically restart, route power (SPS) and shut down on over-current (OCP) and overvoltage (OVP) / undervoltage (UVLO) were also demonstrated. Testing of the USB OTG reverse boost mode also provided a good result with high efficiency and good voltage maintenance with current capability exceeding specifications.
The only qualm seems to be a bit of unexpected behaviour on charge termination especially when operating at 5V. Instead of having no current into the cell, a very small charge current or discharge current seemed to be present (a few mA either way) and is voltage dependent. Quiescent current measurements were complicated by this and the draw of the onboard LEDs. This would not seem (on the face of it) to be a major issue, however, most chargers will “hard” terminate. Whether this characteristic was specific to my sample is not something I can determine. Another note is that the OCP settings while designed to allow for current spikes, will not protect against sustained overcurrent which could exceed the absolute maximum SYS and BATT current ratings as demonstrated in the test. While no ill effects were recorded on this once-off test, a degradation of lifetime or reliability could occur.
It is unfortunate, however, that even though the MAX77751 is capable of charging from 9V or 12V inputs, that it is not capable of interfacing with Qualcomm Quick Charge capable supplies which could benefit from this (especially when respecting 1.5A current limits on legacy cables and connectors). The use of higher voltages with USB-C is not permitted by USB PD specifications due to the total power requirement being below 15W, meaning that it seems that the wider voltage range is really only useful for non-USB applications. Furthermore, it is unfortunate that detection of DCP limits the current to 1.5A. While this is definitely compliant to USB BCS standard and is the safest option, many DCP supplies are capable of 2A or even 3A. The MAX77751 has to be forced to detect an unknown adapter (e.g. by severing the D+/D- lines) to fall back to AICL and charge at the maximum available source rate. Given its position between the system and battery, it also seems to be an ideal position to also integrate overdischarge protection, however, this does not seem to be a feature of any charger ICs.
When I started this RoadTest, I thought it would be a relatively simple, short and straightforward one. However, it proved to be an interesting RoadTest with a few twists and turns, which I have included to show that things don’t always go to plan.
If you’re still reading – thanks for sticking around and I hope you’ve found this review to be comprehensive, detailed and useful. Feel free to leave a rating, like and share with anyone who may be interested. As always, I’d appreciate it if you would leave any questions in the comments and I’ll try and answer them for you as best as I can. Thanks as well to Maxim Integrated and element14 for this RoadTest.
Top Comments
Gough Lui Awesome, detailed write-up and Roadtest. Incredibly thorough. Nice work.
I agree, excellent road test report.
DAB
Thanks Gough,
Another comprehensive and useful Road Test.
MK