Linear Tech 12-cell battery monitor/BMS

http://www.linear.com/ad/6802.jsp

FYI, This is close to release from linear. It can act as a BMS plus an interface to a laptop or the like for monitoring the pack. I believe they are developing SW for the latter as well. It seems to be geared more towards automotive applications with a a larger control system in place - but it might be of use for smaller e-bike apps too. I'm brainstorming a use for it in a future BMS for my a123 4p24s pack.

Mike
PS First post! (longtime lurker...)

_________________

"Linear Technology announces the LTC6802, a highly integrated multicell battery monitoring IC capable of measuring up to 12 individual battery cells. Using a novel stacking technique, multiple LTC6802s can be stacked in series without opto-couplers or isolators, allowing precision voltage monitoring of every cell in long strings of series connected batteries. Long battery strings enable high power, rechargeable battery applications, such as electric & hybrid electric vehicles, scooters, motorcycles, golf carts, wheelchairs, boats, submarines, forklifts, robotics and portable medical equipment.

The maximum total measurement error is guaranteed to be less than 0.25% from -40°C to 85°C and every cell voltage in a battery stack can be measured within 13ms. Each cell is monitored for under voltage and over voltage conditions and an associated MOSFET switch is available to discharge over-charged cells. Each LTC6802 communicates via a 1MHz serial interface. Also included are temperature sensor inputs, GPIO lines and a precision voltage reference. The LTC6802 is a small 8mm x 12mm surface mount device. The combined robustness, exceptional precision and tiny package directly addresses the critical requirements of emerging and advanced battery technologies."

.

Would that be suitable for NiMH series balancing?

I cannot work this out yet from their datasheets...

So it's almost a complete BMS on a chip. Nice.

Mr. Mik said:

Would that be suitable for NiMH series balancing?

I cannot work this out yet from their datasheets...

Click to expand...

Not this version, I believe. Judging by the pre-programmed HVC (4.2) and LVC (2.3) cutouts, this chip is made for LiFe.

But I'm betting they'll release versions for different chemistries, a programmeable one, or one with a set of legs used to vary the cutouts and balancing values.

Link said:

So it's almost a complete BMS on a chip. Nice.

Not this version, I believe. Judging by the pre-programmed HVC (4.2) and LVC (2.3) cutouts, this chip is made for LiFe.

But I'm betting they'll release versions for different chemistries, a programmeable one, or one with a set of legs used to vary the cutouts and balancing values.

Click to expand...

Link,

Where did you see that? I've just quickly scanned the datasheet and AFAIUI, the LTC6802-1 handles the cell level functions and communicates with a separate pack level controller. Undervoltage and overvoltage limits are programmable. Its says its for Li-Ion but it will do anything up to 5.175 V per cell. What it will do is to measure all the cell voltages and pass that info to the controller. It can apply a shunt to any cell or combination of cells, but only under command from the controller. It will also detect and respond to a variety of fault conditions.

Nice device.

Apologies if I've got it wrong - only had a quick scan so far.

Nick

.

Hi Wayne,

I think having little smarts in the chip isn't such a bad thing. It is programmable, but in the sense of loading registers rather than being able to take custom software. That means that a standard cell level hardware solution could be implemented, and all the software to customise it to a given pack or application goes in the master processor. I quite like that compared to the other ideas of a master processor and processors at the cells.

A lot of it is going to come down to the fact that some applications will be better with one board doing N cells and others will be better with a board per cell.

Nick

.

I'm awaiting samples for this chip through my LT contact - it's the ideal front end to the pack and it won't take but a small 8-bit micro to program the cell discharge setting regardless of cell chemistry. I will probably add a larger pack FET to the proto board to cut off pack discharge if any one cell hits it's discharge threshold (3V for A123). This can also throttle down the charging current as the cells gradually reach full charge.

I think the $1 PCB cost is a bit unrealistic (unless your building really high qty PCBs), but should still be a very affordable and scalable solution.

.

Nice looking chip. When it's available retail it'll go in my "useful unusual" chips. Actually, the demo-board shown on Linear Tech page looks like most of a good BMS for small cells just by its self. I hope the demo-board is available cheap.

Marty

Cool.
The actual datasheet is here: http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1003,C1037,C1134,P86662,D26880

They are pretty cagey about the specs for how much current the internal shunt FET can handle. It appears to be pretty wimpy, but they show a booster circuit that uses external FETs to switch the shunt resistor. No limit there.

You would still need some kind of 'host' processor to display the data and interface to the charger and controller or other LVC interface.

I wonder how much they will cost?

Nice. I've got my hands on a DC1331 demo board of the LTC6802 I also have the DC509 USB board that is used to eval various LTC circuits. I don't have the GUI SW yet, so I'm developing drivers for it on my own demo board using an ATmega168 8-bit uC (http://www.digilentinc.com/Products/Detail.cfm?Prod=NANOCON&Nav1=Products&Nav2=Embedded). Unfortunately the driver writing will be a slow process given I have to work for a living and my C is rusty.

The built in FETs don't look like the handle much at all - especially given the absolute max on any pin of +/- 10mA for the LTC6802. Fortunately the demo board has the external PFET and 15-ohm SMD 2512 1W shunt resistor to provide ~240mA discharge as well as an LED for any cell that starts conducting through the shunt. I found a 2W 10ohm replacement that could get it up to ~360mA if more is needed. See the attached schematic for the demo board. 2 of these and the nanocon board will be all I need for a preliminary (if not final) BMS and cell-monitoring solution

Initial plan is to PWM the charger current via the ATmega168 (accurately down to the shunt current) once any cell reaches the OV threshold (configurable to 24mV steps 0-6.1V) Once all cells reach the OV point (time dependent based on magnitude of shunt current) the PWM can shut the charger off altogether. Also while riding the ATmega168 could disconnect pack ground from the controller using a beefy FET if any cell hits the UV threshold (also user defined 0-6.1V in 24mV increments). More to follow...

Mike

Now you're talking...

I think I've pretty much proven that you can PWM the charger to control the average current. We have a circuit to cut off the pack also (thanks Randomly) if you don't want to just interface to the ebrake input.

oohh! parts in hand.

Mike_t, how much did the demo-board for the LTC6802 set you back?

Marty

.

Most of their demo boards are in the $125-$150 range.

The GUI was done by LT, I just couldn't get it to install for some reason. The drivers I'm writing are for the ATmega168 so I can use the demo board's on my e-bike.

Nice tip on the heatsink. I might also try and see if PWM'ing a 0-ohm shunt resistor is feasible as well. The onboard PFET (Vishay SI2351DS) can handle 2.2+ amps continuous and is designed for direct 3.3V uC driven PWM applications with it's low gate charge for a variable current between 0-2.2 amps. I figure the 1 MHz SPI port should be fast enough to PWM each cell individually, but I'd like to see real world results first. That would allow for some really quick cell balancing!

mike_t said:

Nice tip on the heatsink. I might also try and see if PWM'ing a 0-ohm shunt resistor is feasible as well. The onboard PFET (Vishay SI2351DS) can handle 2.2+ amps continuous and is designed for direct 3.3V uC driven PWM applications with it's low gate charge for a variable current between 0-2.2 amps. I figure the 1 MHz SPI port should be fast enough to PWM each cell individually, but I'd like to see real world results first. That would allow for some really quick cell balancing!

Click to expand...

Looking at the Rds(ON) (about 0.1 ohms) for that FET I wouldn't recommend using a 0-ohm shunt resistor. The 30 to 40 amp pulses would likely blow the chip right off the board.

Marty

Yes, I suppose. That little SOT-23 can handle 10 amps pulsed though, nice... A 0.25 ohm shunt should work (9-10A pulses @ 3.5V and using a 0-28% duty cycle, Iavgdc = 0 - 2.8A). Those are readily available in 2512, 2W package.

Using PWM instead of steady state yields 8x more average current for the same footprint - not too shabby. Or, more likely, a lot less heat for 1A balancing. Looks like this SW project just got more complex...

Mike

I'm not sure if you can get away with that? The heat has to go somewhere. Into the cell would spread out the heat, but the cell impedance might be pretty low. Cell heating could be an issue too. I opted for resistors that can withstand 100% duty cycle in case of a stuck switch.

Texas Instruments makes a real cool BMS chip, the BQ77PL900 which doesnt require any microcontroller to use, you use a computer with an I2C adapter and program its on-chip EEPROM as to all the voltage limits and how you want it to balance, etc and then it stands alone and runs with the parameters you put in it.
The only problem is it only does 10 cells and I dont think you can gang it with more chips to use more cells.
Intersil makes some too (ISL9217) but it can't be ganged either.
The MAX 11068 from maxim can be ganged together and is similar to the linear tech chip but its not available yet, you have to pester them to get a data sheet.

fechter said:

I'm not sure if you can get away with that? The heat has to go somewhere. Into the cell would spread out the heat, but the cell impedance might be pretty low. Cell heating could be an issue too. I opted for resistors that can withstand 100% duty cycle in case of a stuck switch.

Click to expand...

It's certainly feasible, but not worth SW complexity and risk in the end. The cells would be fine with 10A continuous (4C discharge rates) at 16 mOhms cell resistance - so 1A average DC current would certainly not be a problem. The heat is lost in the reduced resistance of the shunt resistor from 15 ohms to 0.25ohms w/ P=I*I*R. At 1A average DC current (10% on time) it's average power dissipation is only 0.25W - 1/8 the rating of the resistor. The key like you said is ensuring the 10% duty cycle. I don't think it's worth the risk though - code hangs and the watchdog takes a sec to reboot the micro while a switch is stuck ON for 1 full sec at 10A.

And really who needs ultra fast balancing. If they are balanced every time it is charged it shouldn't take that long even with 240mA charge rate for the lagging cells. I was just exploring the possibilities with this chip and fine tuned digital control...

mike_t said:

It's certainly feasible, but not worth SW complexity and risk in the end. The cells would be fine with 10A continuous (4C discharge rates) at 16 mOhms cell resistance - so 1A average DC current would certainly not be a problem. The heat is lost in the reduced resistance of the shunt resistor from 15 ohms to 0.25ohms w/ P=I*I*R. At 1A average DC current (10% on time) it's average power dissipation is only 0.25W - 1/8 the rating of the resistor. ...

Click to expand...

There are large cell applications where something more like 1A would be good. If you have a 3.7v cell with 16mohm internal resistance and a FET with, say, 5mohms, then you would have a peak current (ignoring wire resistance) of something like 175 amps. That would be too high.
You also need to dissipate 3.7 watts of heat to somewhere. If it went into the batteries, they might have enough surface area to effectively get rid of quite a bit of heat.

I've looked at a number of possible non-dissipative shunting schemes including switched capacitors and various inductor arrangements. None of these would be cost effective on smaller (<40ah) packs I suspect.

The best one I've come up with is to have a small high frequency transformer on each cell. The cell voltage is PWM'd into the primary of the transformer to. All of the secondaries are rectified into a bus line that goes to the main battery terminals. When the PWM switch opens, the collapsing magnetic field will generate a high voltage pulse that gets circulated back to the pack. The transformers would be similar to what you find on a small dc-dc converter.

Not only does this approach eliminate most of the heating, but allows the current to be sent to the low cells. The downside is cost of the tranformer is relatively high.

.

fechter said:

mike_t said:

It's certainly feasible, but not worth SW complexity and risk in the end. The cells would be fine with 10A continuous (4C discharge rates) at 16 mOhms cell resistance - so 1A average DC current would certainly not be a problem. The heat is lost in the reduced resistance of the shunt resistor from 15 ohms to 0.25ohms w/ P=I*I*R. At 1A average DC current (10% on time) it's average power dissipation is only 0.25W - 1/8 the rating of the resistor. ...

Click to expand...

There are large cell applications where something more like 1A would be good. If you have a 3.7v cell with 16mohm internal resistance and a FET with, say, 5mohms, then you would have a peak current (ignoring wire resistance) of something like 175 amps. That would be too high.
You also need to dissipate 3.7 watts of heat to somewhere. If it went into the batteries, they might have enough surface area to effectively get rid of quite a bit of heat.

Click to expand...

Hmm...that would be too much, but you left out the 0.25ohm shunt. The FET on this board is about 0.1ohms @ 4.5V and the cells are 0.016ohms. Total resistance is 0.366ohms (0.354ohms in my case, with 4 cells in parallel). This gives a current pulse of 10A peak worst case. It will be less on avg due to miller gate charge effect (ramps up to 10A) and slightly less due to 3.3V switching instead of 4.5V. If PWM'd to 10% on time, this would give 90% cool down time. That is why I use avg current of 1A with 0.25W avg power on the shunt, 0.1W avg on the FET, and 0.004-0.016W on the cells (4P or single, respectively). The frequency of the PWM is critical of course. Ideally it would be 10ms on, 90ms off or faster. 10sec on, 90sec off would clearly not be good!

As long as care is taken to have a short timeout on the watchdog to restart the micro if code ever hangs with a FET on, then you should be golden.

I ordered 25 of the 0.25ohm, 2W, 2512 resistors. This is a prototype after all, and now my curiosity has the best of me.

Mike