Li-ion Battery Chargers: TI BQ24092 compared to Microchip MCP73833/4

[jpanhalt]

I want to include a battery charger in a device.  The battery will be a typical prismatic lithium ion/poly battery of about 1100 mAh (e.g, Samsung).  Fast charge rate will be limited to about 500 mA.  I have reduced my choices to the two chips in the title.  The chips are quite similar.  Links to the respective datasheets are:
http://www.ti.com/lit/ds/symlink/bq24092.pdf
http://ww1.microchip.com/downloads/en/DeviceDoc/22005b.pdf

My main concerns are thermal performance and ability to use the target circuit(load) while charging.

1) Thermal
The project will be hand assembled.   I am OK with "relatively" small stuff, like 0.5 mm pitch devices, but have not gotten into reflow and solder paste.  This is not a project I want to learn on at this stage.   The Microchip (MCC) device is an MSOP-10 package.  Its listed J-A resistance is 113 °C/W.  The TI device has an exposed thermal pad, and its J-A resistance is 71.2°C/W in the TI datasheet.  TI's E2E forum has discussions of errors in that datasheet that are not relevant to my question.  I have limited space and cannot include large copper pours for heat sinks.

Using the TI estimates, maximum dissipation will be about 800 mW.  The TI device should stay within thermal limits; the MCC device  will probably have some periods of thermal regulation.  I plan to include a copper area and add a small dab of thermally conductive epoxy under the MSOP chip to help.

I am looking for comments from anyone with experience with either chip at 500 mA charge rates.

2) Charging while under load
TI specifically states that its devices work while the battery is under load so long as current is sufficient.  My device will draw less than 10 mA, except when the GLCD backlight is on, which adds about 40 mA to 50 mA.  The default state of the backlight will be off, but the user can toggle it on for up to 2 minutes -- maybe less.  MCC does not comment on charging while under load.

Does anyone have experience using either charger while the battery is under load?  My concern is whether the load affects the charge cycle.  I can bypass the battery with a few components and power the load directly from the 5-V charger source, if need be.

I realize these are pretty specific questions about those two chips, but I seem to have exhausted what is available on the web.

Regards, John

[jeremy]

It’s a bad idea to power off a lipo while charging in my experience. If I remember correctly that MCP chip will not work in that way. The problem is that you can confuse the charger by having dynamic current spikes. Also if your battery is dead flat they can’t start up properly; the MCP uses a weak current to check if the battery is connected and/or how dead it is, and if your circuit sucks up all the energy it won’t work correctly.

For the price of a mosfet and a diode I don’t think it is worth risking. See http://ww1.microchip.com/downloads/en/Appnotes/01149c.pdf for a circuit.

I wouldn’t worry too much about the thermals beyond what you are doing.

[Siwastaja]

For the price of a mosfet and a diode I don’t think it is worth risking.

Talking about "risking", you make it sound easier than it is. But do this properly. Wrong dimensioning of the power path switch components against capacitive loads may cause a catastrophic overcharge situation. Been there, done that, followed appnote, FET blows short in certain conditions applying external power supply directly to the battery, had to recall. It's not nice.

It's one thing if your product fails by stopping working after an input voltage surge incident. It's another thing if it fails in a way of connecting the input directly to the li-ion cell, unprotected, from that moment on, while the product still appears working. With a 5V supply and one diode drop (assuming the diode didn't blow), it will be funny how it charges the cell to around 4.5V and everything kinda seems working....

There is no easy way out. Do a proper design with full analysis. Allocate a lot of time for this, especially if you are not experienced.

I hate many of these "looks nice&easy" li-ion products because you think they give you an easy way out of lithium ion safety issues, but when you start designing one in, suddenly you find that the IC does not integrate the functionality you needed (and what the product exists for), and you find yourself following an appnote written by an intern which requires putting in a large number of external discrete semiconductors, high-value capacitors, often large-value ceramic with no snubbing directly in the input, etc. These are recipes for disaster.

Most li-ion ICs are not providing any safety, have no safety approvals, and you as the designer need to understand every internal detail about them. It sucks, yeah.

In general, power path switching kind of sucks, more or less. Especially if the load current is small as is the case here, I would be very tempted to connect it "directly" to the cell, and find a charger IC that doesn't go haywire. If they do, they sound unreliable in any case.

The issue is, the controller FET (linear or switcher) is controlled with an actual current limit (and/or some kind of SOA protection); the power path switch FET which connects the input to the load directly should be the same, but for some strange reason, following the appnote examples, they always slap bare unprotected MOSFETs to switch the load, with no protections whatsoever. This is a recipe for a disaster once you understand what they are actually suggesting you to do! Outside of such use case, you would never do that, but use a protected IC load switch instead.

So, my bet is, use one specified to be operated with a directly connected load. You sidestep the whole power path switching issue, and only have one protected path (the regulator itself) from the supply to the battery.

Do note, they have even a paragraph of "Selecting The MOSFET" which fails to mention any of the required parameters, instead bullshitting about "gate threshold voltage", showing the poor guy who wrote it has never properly chosen a MOSFET for switching in their life.

[jpanhalt]

Thank you for that link.  That is a pretty common way to do it.  Although Q1  avoids a voltage drop from the battery to the load, a low Vf Schottky and LDO can probably replace it. Cost wise, very little difference. It's the claim by TI that a load can be powered during charging that attracted me to it.

[jeremy]

This is the chip you actually want to use: https://www.microchip.com/wwwproducts/en/en536670 but unfortunately it is QFN only.

I’ve honestly never had issues with LiPo batteries, but I don’t use cells which are unprotected. OVP should kick in here, even the cheapest of cheap protection ICs has that. https://lcsc.com/product-detail/MOSFET_DW01_C181096.html

[Siwastaja]

I’ve honestly never had issues with LiPo batteries, but I don’t use cells which are unprotected. OVP should kick in here, even the cheapest of cheap protection ICs has that. https://lcsc.com/product-detail/MOSFET_DW01_C181096.html

Yes, another layer of protection is very good to have. I don't like the trend of designing utterly catastrophic circuits in "let's put random components" fashion (like these appnotes invariably present), which are likely fail, then rely on the protection layer, though.

The good things in the cell-level protection circuitry are,
1) They won't have feature bloat, because they only protect that one single cell. No inputs, no outputs, no configuration.
2) They are designed by different people, using specific ICs for that job. Manufactured and assembled separately. So it's unlikely they fail at the same time.

They do fairly good job on catching problems caused by failed higher level.

The third, lowest layer of security is in the cell technology itself, so actual fire incidents are very rare. At least if you use good brand cells.

Do note that the posted appnote does not clearly specify using protected cells; what they show looks like a bare cell. They also fail to show a fuse, which is paramount, and a cheap component as well.

Instead of adding cheap MOSFETs that fail short, add cheap fuses that fail open

[jpanhalt]

@Siwastaja @jeremy

Those are great suggestions.  The warning about switching power paths with a single point failure mode is well taken (BTW, the MCP73871 uses a "synchronous switch PNP" that could have that same failure mode). 

Yes, I have considered adding a purpose-made battery protection circuit to the battery.  This device is not for high-volume production, and I had already planned on modifying the battery by taping an NTC bead to it and adding leads with a 3-pin JST plug.  I already have several of those inexpensive protection strips available.

Right now, I am leaning toward the TI chip with slightly better thermal performance and inhibiting the GLCD backlight (BL) during charging.  The power good (!PG) pin can be monitored so whenever an external source is applied, BL will be inhibited.  BL is already controlled by the MCU, and although there are not a lot of pins to spare, I do have a few.  I am using a 28-pin 16F1938.

John

[jeremy]

Where are you getting your batteries from? Almost all of the off-the-shelf prismatic ones on the market (which are not designed for RC vehicles) have the protection circuit built in. I think you’d have to try to actually find one without it if you didn’t want it for some reason. Cylindrical cells are the opposite, mostly unprotected.

It’s usually wrapped in that brownish kapton tape at the end of the battery.

[jpanhalt]

I had planned to use the Samsung AB553446GZ (https://www.amazon.com/Samsung-AB553446GZ-AB553446GZB-AB553446GZBSTD-Original/dp/B00DP0RO5G).  I got some at a reasonable price for my cell phone and they happened to fit the battery compartment of the off-the-shelf case I found.

The battery has 3 contacts as usual, but rather than try to make a spring loaded fixture for it, I figured I would just pop the plastic off and see what's there.  Then add a 3-wire cable w/plug for my device.  Center pin to negative shows considerable resistance ( a 100 k NTC?).  I am not absolutely committed to that battery, but it is a convenient size.  If it is a 100 k NTC, the TI chip offers that as an option. 

As far as I know, protection may be built in.  I just don't have the details on it.  I wanted to get the first prototype PCB sent off relatively soon so the project can be tested before Spring.  It is basically just a data logger with graphical display for roasting meat.

[jeremy]

You might find that battery is completely potted, so it could be hard to look around inside. Also if the contacts are spot welded or crimped directly to the cell outputs it might be a good idea to avoid soldering to them to add your own cable as you will be transferring heat directly into the middle of the cell.

The most common cell is something like this; it has protection built in and has the JST PH connector already crimped:

https://www.adafruit.com/product/258

Obviously they can be had for much cheaper elsewhere but with the Chinese supply chain issues at the moment I think we’ve all got to take what we can get 

You can attach these batteries directly to a PCB (double sided tape, cable ties through slots, screw down bracket, etc), and PCB mount the thermistor right next to it if you want to go down that path. But don’t let me tell you what to do, just sharing what has worked for me in the past.

[jpanhalt]

I saw that battery the other day at Adafruit.   I do have a battery tab CD welder I built years ago for NiCd's and have used it for Li-ion tabs, but the Li-poly batteries I have used were already set up with leads.

I am not too worried about the specifics of the battery right now.  With all that's going on right now, I suspect PCB delays will increase , so I wanted to get into line.

[mariush]

You should consider NOT using OEM batteries like that Samsung one.

There are companies selling such batteries in volume, and they're reasonably priced.

For example:

8.3$ at 50pcs 900mAh, 850pcs in stock : https://www.digikey.com/product-detail/en/jauch-quartz/LP603443JU-PCM-WIRES-70MM/1908-1365-ND/9560987

8.5$ at 50pcs 1350mAh, 870 pcs in stock : https://www.digikey.com/product-detail/en/jauch-quartz/LP503759JU-PCM-2-WIRES-70MM/1908-1369-ND/9560991

same 1450 mAh 840pcs : https://www.digikey.com/product-detail/en/jauch-quartz/LP103048JU-PCM-WIRES-70MM/1908-1370-ND/9560992

They also have an Adafruit battery at 9.5$ at 50pcs, 1200mAh : https://www.digikey.com/product-detail/en/adafruit-industries-llc/258/1528-1838-ND/5054544

tme.eu  has them for less than 4.5$ if you order 50-100 at a time, but shipping them to US would probably cost or take a lot of time as they would not be shipped by air

450mAh 3.86$/100 https://www.tme.eu/ro/en/details/accu-lp502248_cl/rechargeable-batteries/cellevia-batteries/l502248/
800mAh 3.86$/100 https://www.tme.eu/ro/en/details/accu-lp453350_cl/rechargeable-batteries/cellevia-batteries/l453350/
1100mAh 3.85$/100 https://www.tme.eu/ro/en/details/accu-lp623255_cl/rechargeable-batteries/cellevia-batteries/lp623255/
900mAh 4.2$/100 https://www.tme.eu/ro/en/details/accu-lp523450_cl/rechargeable-batteries/cellevia-batteries/lp523450/