Multi-chemistry charger: analog design practice?

[chrisc]

G'day folks,

I'm interested in building a micro-processor based multi-chemistry charger in order to experiment with firmware development for such devices. I have a client who re-sells OEM units from China into the radio-control market but is unhappy with the delivered firmware and is thinking of having their own custom unit designed. (Also, having seen inside one of the units myself, I'm a bit leery of the hardware - though they seem to work OK, my eyebrows were raised a few inches!)

However before he does that he wants to spend some time with a test setup to see what's involved in the software side (they have in-house software folks). So I want to put together a dev board that they can use.

This design needs to avoid using custom charger IC's; the desire is to put as much smarts into the firmware as possible (so as to make it flexible, and also lower ultimate production cost). It needs to support Pb/NiCd/NiMH and most lithium chemistry up to say 5 or 6 cells (with balancing). Input would be from a separate 15-20v supply.

Whatever I put together needs to do constant current / constant voltage and I suppose would use a buck converter. Things like the temperature monitoring, microprocessor etc are pretty clear to me, however having never worked on a charger before I'm not so comfortable with the rest of the analog side.

I'd be keen to hear opinions on what folks think would be the most appropriate approach to the analog design (in terms of concept, technology used), keeping in mind that while this board is not for production use, it needs to behave (insofar as the firmware is concerned) in a more or less similar way to something that they may get made for production down the track (i.e. it can't rely on off-loading any of the charge management to a custom charger chip if said chip is not economical to be used for production in consumer-level chargers).

[valentinc]

    What's the maximum charging current that all the 5 or 6 cells would draw ? 

[chrisc]

    What's the maximum charging current that all the 5 or 6 cells would draw ?

I'd say roughly 20A.

[reagle]

Speaking from past experience, closing the control loop in software requires a speedy micro or things get painful quickly. So you are best with that part done by something like a buck converter indeed, and then just using the micro to set current/voltage limits using DACs that drive current into feedback pin. Ti has an MSP430 based reference http://www.ti.com/lit/an/slaa476a/slaa476a.pdf illustrating the concept.

[IanB]

In my limited experience, once of the trickiest things in battery charger design is around the choice of switch mode designs vs analog designs. Switching designs are obviously efficient, flexible and cool running, but the switching noise plays havoc with the clean measurement of charging voltages at millivolt resolution as needed by charge control and termination algorithms.

So in the analog design part, your challenge is going to be to design a way of providing adjustable CC and CV charge control over a wide range while simultaneously being able to measure the battery voltage in CC mode and the charging current in CV mode with a sufficiently accurate and noise free signal.

[reagle]

Right,with Nickel chemistries you'd have to detect a very finicky dV spike. There is always a timer/temperature means to terminate though.
Lithium is easier as you just whack them in CC/CV mode and wait till current drops or battery (if smart) tells you to stop charge

[chrisc]

In my limited experience, once of the trickiest things in battery charger design is around the choice of switch mode designs vs analog designs. Switching designs are obviously efficient, flexible and cool running, but the switching noise plays havoc with the clean measurement of charging voltages at millivolt resolution as needed by charge control and termination algorithms.

So in the analog design part, your challenge is going to be to design a way of providing adjustable CC and CV charge control over a wide range while simultaneously being able to measure the battery voltage in CC mode and the charging current in CV mode with a sufficiently accurate and noise free signal.

This is a useful insight, thank you. I might have a play with deliberately injecting noisy charge into one of the battery pack protection/balance controllers I designed for a different customer to see how the hardware handles reading the values. It uses a TI BQ78PL116 (which I have gained a certain amount of respect for). It seems to have a quite decent measurement subsystem and if I can determine it handles that sort of noise satisfactorily it might be worth doing some investigation to see if I can gain more information about how they manage the filtering, and/or find another device that has similar measurement functionality without all the other overhead needed for a controller.

(It also has a charge pump system for cell balancing but I would suspect they take measurements when the pump is idle).

[chrisc]

Speaking from past experience, closing the control loop in software requires a speedy micro or things get painful quickly. So you are best with that part done by something like a buck converter indeed, and then just using the micro to set current/voltage limits using DACs that drive current into feedback pin. Ti has an MSP430 based reference http://www.ti.com/lit/an/slaa476a/slaa476a.pdf illustrating the concept.

Useful link, thanks! I had been trawling the TI stuff but for some reason had not yet come across that particular document.

[chrisc]

Right,with Nickel chemistries you'd have to detect a very finicky dV spike. There is always a timer/temperature means to terminate though.
Lithium is easier as you just whack them in CC/CV mode and wait till current drops or battery (if smart) tells you to stop charge

To be honest I'm half inclined to tell my client he ought to consider splitting the nickel charging off into a separate product (or just sticking with existing products). Handling nickel, I suspect, is more of a tick on the feature list than (at least in my opinion) something that is essential for his customers, most of whom use Lithium cells. The difficulty comes in selling a charger that doesn't handle nickel into a market full of ones that claim they do (regardless of how well/poorly).

I mentioned in my original post that my eyebrows were raised when I looked inside one of the existing Chinese-made chargers that are being sold into this market. I've attached a few photos to this post to illustrate why. Note that these are taken from a production charger from this particular manufacturer. It looked quite nice on the outside (typical consumer stuff with nicely molded case and custom LCD). Inside is a different issue.

[Niklas]

NiMH and NiCd are not that difficult to charge. You will need a constant current, or at least somewhat constant, that you can turn off at certain intervals during the voltage measurement. The trick is to turn off the charge current and wait for a second or two before you measure the cell voltage. If the charge current is still on, then you will also measure the voltage drop in the battery cables.

For a flexible solution you would like to have a windowed voltage measurement of the battery pack voltage. DAC or a filtered PWM buffered with a voltage follower and then fed to a subtractor (Gain*(Ubatt-Uoffset)) before the ADC.

[Orpheus]

Am I seeing correctly? Did they manually grind out a trace that would otherwise directly short the + and - ?

If not, what is that?

[nukie]

It looks like one of the iMax B6 DUO variant
http://www.rcgroups.com/forums/showthread.php?t=1255637

Same as this
http://www.redrc.net/2011/03/venom-210-dual-output-multi-charger/

Same as this
http://www.bigsquidrc.com/hitec-x2-ultima-two-channel-charger/

Maybe this
http://www.amainhobbies.com/product_info.php/cPath/1574_116_3147/products_id/253492/n/Racers-Edge-TWIN400-Prime-DC-Balance-Charger-w-Aluminum-Case-NiCd-NiMH-LiPo-LiFe-Pb

There are a lot of fake iMax clones(google), are you sure you got the authentic stuff?

Anway SkyRC isn't the best out there their stuff isn't very efficient either, check out the quality in the RC forums. IMHO, you should make reference from www.revolectrix.com. Digesting their product lines will require at least a few hours. Powerlab6/8 will give you some idea.

I use their chargers, I think they are very simple to operate but awfully complicated software working in the background. That's what makes it so powerful and a beautiful design.

[chrisc]

Am I seeing correctly? Did they manually grind out a trace that would otherwise directly short the + and - ?

If not, what is that?

The large grindouts are where they disconnected the ground plane between the left and right halves of the board. On this unit the left and right are basically clones (it's a dual charger) and share almost no components. I suspect they just took a single charger design and plonked two of them down onto the same PCB, then discovered that they had issues with them sharing a ground.

As for the other alterations, frankly just looking at the board makes my brain hurt so I didn't try to figure out in detail what they have done 

[chrisc]

It looks like one of the iMax B6 DUO variant
http://www.rcgroups.com/forums/showthread.php?t=1255637

Bingo! The case is identical, the only difference being a different OEM label on it. Color me impressed ... I take it you are fairly familiar with these given you were able to finger it just by looking at the PCB? Was it the configuration of the connectors on the board edge that gave it away, or have you seen inside one before?

There are a lot of fake iMax clones(google), are you sure you got the authentic stuff?

Anway SkyRC isn't the best out there their stuff isn't very efficient either, check out the quality in the RC forums.

I will ask my client if they have considered this possibility. You mention SkyRC; I presume they are the originator of the genuine version of this design, and if so, that their hardware (while not the best) is generally not this bad?

IMHO, you should make reference from www.revolectrix.com. Digesting their product lines will require at least a few hours. Powerlab6/8 will give you some idea.

I use their chargers, I think they are very simple to operate but awfully complicated software working in the background. That's what makes it so powerful and a beautiful design.

This is very useful information, thanks. It's specifically the software complexity that we are trying to gauge.

[abyrvalg]

I've heard that all BC6/BC8 "originals" like SkyRC/Turnigy/Hobbyking etc are just copies of Korean Bantam chargers (bantamtek.com). Funny that Bantam themselves doesn't care about clones at all (they publish unprotected firmware updates on their site - just grab one, buy one charger and you are welcome at chinese "breeding" factory), but the "followers" have various protection stickers and online s/n validations

Edit: found some PCB views, not a full teardown, but quality level can be seen: http://forum.fonarevka.ru/showthread.php?t=6801

[NiHaoMike]

In my limited experience, once of the trickiest things in battery charger design is around the choice of switch mode designs vs analog designs. Switching designs are obviously efficient, flexible and cool running, but the switching noise plays havoc with the clean measurement of charging voltages at millivolt resolution as needed by charge control and termination algorithms.

So in the analog design part, your challenge is going to be to design a way of providing adjustable CC and CV charge control over a wide range while simultaneously being able to measure the battery voltage in CC mode and the charging current in CV mode with a sufficiently accurate and noise free signal.

You can have the ADC always sample at the same time in the switching cycle or temporarily disable the switching in order to obtain measurements. The latter can be done at a very low duty cycle and would not significantly slow down charging.

[IanB]

NiMH and NiCd are not that difficult to charge. You will need a constant current, or at least somewhat constant, that you can turn off at certain intervals during the voltage measurement. The trick is to turn off the charge current and wait for a second or two before you measure the cell voltage. If the charge current is still on, then you will also measure the voltage drop in the battery cables.

Turning off the current before measuring the cell voltage would be a bad idea with nickel chemistry and minus delta-V detection.

In order to reliably detect a drop in voltage you need to measure the voltage while the charging current is active (the decrease in voltage comes from a decrease in internal resistance of the cell due to internal heating that happens when the cell is approaching full charge). If you turn off the current before measuring you won't be able to measure the drop in internal resistance and you will be reduced to measuring a residual surface charge that does not give a strong indicator. This is a problem that Maha encountered in one of their chargers and they had to make a hasty fix after the first production run due to many reports of missed terminations and cooked batteries.