I have a system that is powered by a 7.4V Lithium-Ion battery. The system, which is a combination of processor module and sensors, draws a varying amount of current at varying voltages (5V, 12V, -12V, 6V...) but pulls about 10 watts. The battery I chose is quite a hefty 10 amp-hour one.

This started off with a simple off-line charger which is no problem. Then feature creep set in (sigh) and now I'm tasked with a dual battery + charger with gas gauges. That means there are up to 3 DC sources to be managed - two batteries and a charging port. And I've two months.

One way to do this is use a fully featured chip like the LTC1760. This has support for dual smart batteries and a power path architecture, all managed from SMBUS. It's in TSSOP-48. That creates another issue - you have to come up with the smart batteries but they can be bought off the shelf (example). I don't believe it's reasonable to make smart batteries from scratch in this case!

The question is:

  • Is this adding complexity where it isn't warranted? If I have to put in a gas gauge type system, the smart battery has that but really requires a smart battery charger chip.
  • I can't see how to (simply) support two batteries, multiplexed with on-line charging, except by using the LTC1760. Is this really a common scenario?
  • The LTC will switch-mode large currents. This easily leads to RFI issues and supply noise. Is it a requirement for a TSSOP-48 chip like this that I should go to a 4-layer board? Or could I get by with 2-layer? I haven't done 4 layer before but there's some good information on stackexchange!
  • 1
    \$\begingroup\$ Go for 4 layer. It sounds like it's needed. It may seem daunting but it isn't. \$\endgroup\$ – Andy aka Dec 2 '14 at 22:32
  • \$\begingroup\$ It does indeed look daunting. On the other hand it also looks easier to route a complex board in less space. If I read some of the tips on this site, I think I'll have a go and send the board out to a pcb pooler. A 4 layer board won't get any arguments on that front! \$\endgroup\$ – carveone Dec 3 '14 at 9:36

The LTC1760 is an all-in-one device. This means that you will want to look at board size, parts cost and engineering costs of both options to determine if you want to use it or not. If you have very low volumes, the engineering costs will dominate. In any case, the chip has three main functions:

(1) Combine power sources.

Since you only need to switch about 1A, you may use diodes for that. Use shottky diodes if you can tolerate inefficiencies, or use 'ideal diodes' such as LTC4358. If your DC input is higher that battery charging voltage, you will not need any sort of smart switching/control -- the diodes will close when external power is applied.

(2) Charge the batteries

For the simplest solution, give each battery its own charger circuit. If you want to be more fancy ('I want to fully change one battery before switching to the next one'), add dual MOSFETs or use 'enable' input on chargers.

(3) Control everything from Smbus/I2C

This may be the most important part. If you have a big CPU and few GPIOs available, the LTC1760 is better, as it will only require a single I2C bus and contains both mosfets and drivers for them. If you have a small microcontroller, it may be easier for you to avoid un-needed I2C programming by controlling power switching directly.

This brings us to the last point -- how much controllability do you need? If you want to control every aspect of the system (which battery do I discharge first? This DC adapter can only provide 20W, so do not charge all batteries at once. etc...) then LTC1760 is better. If you just want all functions to be automatic ('We discharge the battery with higher voltage. We change when the device is plugged into the wall'), stick with more discrete solution and save yourself work of programming firmware.

Finally, the smart batteries are great if your project can afford them. I have been using inspired energy ones just as regular packs -- discharge directly, charge with CC/CV charger. They are great because they provide safety cut-off, and because they provide easy-to-use metrics on smbus interface such as 'capacity left' and 'time to full discharge'

  • \$\begingroup\$ I do need more control - discharge a then b etc - but there are prioritising ideal diodes too. Of course if I'm going for that kind of control I could control the switching directly. I'd prefer more control over automatic functioning which does lead me back to the LTC1760. The smart batteries are pricey but they do look the part, provide the safety circuits and, as you say, I can read all the metrics. All for a bit of extra $ - I think it's worth it! Thanks. \$\endgroup\$ – carveone Dec 3 '14 at 9:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.