Mapping a voltage range?

[Things]

Hey guys,

This is probably a stupid simple question, but being a complete noob at analog electronics, I'm looking for a little assistance

Basically, I'm trying to monitor a 6S LiPo pack on a uC. However, the uC only has 10bit ADC's.

Using a 15:3 resistive divider, I get appropriate voltages for the ADC's, but I only get about 300mV range from the batteries being fully charged to fully dead.



So my question is, what would be the best way to monitor individual cell voltages on a microcontroller?

With the ADC too, as a lower cell's voltage drops, it'll also effect the cells connected higher than it also?

Cheers,
Dan

[BravoV]

So my question is, what would be the best way to monitor individual cell voltages on a microcontroller?

The best ? If I were you, I'd use the dedicated multi battery monitor IC, it will save you lots of time and energy. Also you didn't mention what kind of mcu you use and to monitor multiple cells, either you got many adc ports or it has internal multiplexer, if not you have to supply external one, and of course must get thru the trial and error too in building it yourself, hopefully the battery wont exploded or ruined in the progress. 

Example solution from Linear Tech, though other IC manufacturers may have similar solution.


Like LTC6801 only for  monitoring




while better one includes cells balancing LTC6803




There are many choices, these are just examples.

[Things]

Thanks for the reply

I have looked at those chips right at the beginning, but they're around $22 each in single qty's, not including shipping from LT themselves!. Element14 don't seem to stock the newer models.

The uC I was going to use is the ATMeg32u4 with 12 ADC's.

The main reason I want to do it with the uC is so that I can print the cell voltages on a LCD display. I can get a complete, 200W charger and balancer with a LCD from hobbyking for around $40, I'm tempted to open it up and have a look at how they monitor individual cell voltages onto the uC.

Hmm, looking at this charger I have here, and also basically ALL the multi-cell chips, they ALL seem to use a MUX. Is the purpose of this to allow the chips AREF to be changed to the next lowest cells voltage, so that your readings of the current cell are based on a "gnd" reference of the cell below?

EDIT: Thanks to Dave creepishly doing a teardown on a very similar charger to the one I literally just opened up, and providing a schematic, it appears I have my solution:





Just one thing though, will this still work properly if the ATMega is being powered from the battery it's reading? I don't think the grounds would be isolated in the charger, but I'm not sure.

EDIT: Looking at the schematic, I really don't see the point of putting the battery - through a divider then into the mux? What's the purpose of that? Possibly to isolate the batt - from the uC? Can anyone explain?

Cheers,
Dan

[Psi]

If you want something simple and aren't too worried about a slow discharge maybe try to do it this way using 0.1% resistors.



You must measure the cells in order from lowest to highest (since you need to know V1 to calculate V2 etc.)

Lets say you get 3.3V for R1 (cell1).
Next you measure the voltage on R2, in this example you will get 3.9+3.3=7.2V and through the divider = 3.6V
In software you double the value (so back to 7.2V) and then subtract what you have already measured for cell 1
7.2-3.3=3.9V

3rd cell you do the same but you multiple by 3 instead of 2 and subtract both the previous cells etc..

At least doing it this way you get to use most of the ADC resolution on each cell
There is an increasing error as you measure more cells but it shouldn't be too bad.

[Things]

Thanks for the suggestions guys. I managed to contact linear tech and get hold of a few LTC6803-4's, should work perfectly for what I'm doing .. if I can figure out the SPI section of the datasheet 

[KD7EEK]

Sorry for bringing up a really old thread from over a year ago but I was curious if any updates are available on the task of multi cell voltage monitoring?  I searched around and found the LTC680X line of chips to be the most ideal solution.  I also found demo boards from Linear that are available so you can get your hands dirty right away: http://www.linear.com/demo/?demo_board=&part_number=ltc6802&category_id=0&companion_board=&software_id=

The link above is for the older LTC6802 and they have boards for the 6803 and 6804 as well.

Just thought I would share and I would love to hear back from anyone who knows of another solution to monitoring individual cells in a battery system upwards of 24 or more cells.  Each LTC680X will monitor 12 cells and they can be chained together for very high cell count monitoring.

Did you get SPI going?  I am having a hard time determining which hyphen model number I need.  For example the, the LTC6803-2 or LTC6803-4?  Is it related to master / slave chaining?  All I could find on Linears website without reading every word in the datasheets is "The LTC6803-2 connects the bottom of the stack to V– internally. It is pin compatible with the LTC6802-2, providing a drop-in upgrade. The LTC6803-4 separates the bottom of the stack from V–, improving cell 1 measurement accuracy."

Thanks and I hope the reference to the demo boards helps, and I would love any feedback or other solutions to multiple battery cell monitoring for large cell count banks.

Alex.

[Psi]

I think some of the cheap lipo voltage testers use one of the mcu outputs as a pwm dac to adjust the mcus own ADC Vref to get around the problem.

[digsys]

...  I would love to hear back from anyone who knows of another solution to monitoring individual cells in a battery system upwards of 24 or more cells. .... and I would love any feedback or other solutions to multiple battery cell monitoring for large cell count banks. 

Been making BM systems for a while now, and the LT1990 is the perfect IC for the job. High accuracy, +/- 250V Input range etc, and
you end up with 0 - 4.25V for each cell, NO varying impedance due to resistor dividers, just a fixed low impedance output.
My current PCB is at - www.pbase.com/digsys/image/154176002
Each BM can do up to 22 cells (or ANY other Input - as they are Isolated), and is bussed (isolated) to as many more BMs as required.
That way, each BM ONLY has to keep 120V max Isolation to the BUS. The simple inverter is required for -5V

[Clear as mud]

There's a different concept here: http://www.edn.com/design/analog/4327228/Novel-measurement-circuit-eases-battery-stack-cell-design
It uses a small transformer on each cell to provide isolation from the high common-mode voltage.  One side of each transformer coil is in parallel with each cell, through a reverse-biased diode to prevent discharge.  Pulses can be fed sequentially into the other side of each transformer, and the characteristics of the return pulses measured to get an idea of cell voltage and condition.  I don't know that it is really very practical, but at least it's an interesting idea.

[KD7EEK]

Been making BM systems for a while now, and the LT1990 is the perfect IC for the job. High accuracy, +/- 250V Input range etc, and
you end up with 0 - 4.25V for each cell, NO varying impedance due to resistor dividers, just a fixed low impedance output.
My current PCB is at - www.pbase.com/digsys/image/154176002
Each BM can do up to 22 cells (or ANY other Input - as they are Isolated), and is bussed (isolated) to as many more BMs as required.
That way, each BM ONLY has to keep 120V max Isolation to the BUS. The simple inverter is required for -5V

Thank you for sharing.  It is very helpful.

Alex.

[digsys]

  Thank you for sharing.  It is very helpful

NP. Forgot to add. I've tried those LT "do everything" chipsets from beta. In MY opinion ... they're crap. LT do some very nice stuff,
but I'm afraid, these were a FAIL. If you had the exact niche set-up they may made for, and you were lazy or didn't feel competent,
then fine. Too restrictive, too many parts to add, too inflexible etc etc. The LT1990 solves 90% of the the problems of HVBM in one go.

[KD7EEK]

NP. Forgot to add. I've tried those LT "do everything" chipsets from beta. In MY opinion ... they're crap. LT do some very nice stuff,
but I'm afraid, these were a FAIL. If you had the exact niche set-up they may made for, and you were lazy or didn't feel competent,
then fine. Too restrictive, too many parts to add, too inflexible etc etc. The LT1990 solves 90% of the the problems of HVBM in one go.

Thanks.  One requirement I have is small form factor.  I don't have a lot of space to work with.  I am reading more on the LT1990.  It is much more simple and could be integrated into a solution that fits my exact need with careful planning.

Not gonna lie, I am "lazy" and a ready made, fits all solution is attractive.  I am going to be trying a few different options.

Do you offer an assembled version of your battery monitor for sale?

Thanks again,

Alex.

[digsys]

  One requirement I have is small form factor.  I don't have a lot of space to work with. 

The pic I posted is my "deluxe" version :-) .. with all the bells and whistles. Others have stripped my cct right down into
2x2" squ, double sided, with only the BM (and a 2Wire data port). It's only 120V per module, Max, if you use 22 cells.

Not gonna lie, I am "lazy" and a ready made, fits all solution is attractive. I am going to be trying a few different options. 

Of course :-), just offering a solution and that IC may fit your design. I didn't mean to rubbish it outright, It's just a
sore point with me ... a long story :-)

   ... Do you offer an assembled version of your battery monitor for sale? 

I just haven't had time to get around to that yet, and the first version would be the "deluxe" one, mainly for Solar race cars.
Most my stuff is kind of open source, so I usually send clubs the circuit and a reference layout. No code but a flowchart of
what I* deem important "specs". It's a quiet time right now, so not sure if any of them have their version available.

[KD7EEK]

Others have stripped my cct right down into 2x2" squ, double sided, with only the BM (and a 2Wire data port).

........

It's a quiet time right now, so not sure if any of them have their version available.

This sounds like exactly what I need to start with; small and simple with only voltage monitoring and comm to microcontroller.

Are you aware of anyone that has one of your stripped down, smaller, simple voltage monitoring solutions available?  Or a project name / developer I can search for?

Don't want to make you winded, but I am curious of your long story - or the highlights - as I am going to be evaluating the LTC680x chips.

Thanks,

Alex.

[dannyf]

Does it really need to be this complex?

What about a pnp + 2 resistors for each cell you want to monitor? Base to the cell negative, emitter to the cell's positive through resistor #1, and collector to the common ground through resistor #2. The voltage drop on resistor #2 is roughly (Vcell - 0.7) / R1 * R2.

It can be done with npn as well but slightly more complex.

[digsys]

   Does it really need to be this complex?  ... 

Complex to some - clean, simple and accuracy for others. Yes, there's 100 ways to skin a cat :-)

[digsys]

Are you aware of anyone that has one of your stripped down, smaller, simple voltage monitoring solutions available?  Or a project name / developer I can search for? 

Bad news ... I had a box full of various versions of my BMs and others, so I went to find it. Looks like it ended up in one of the
batches of old-stuff I gave away recently. No one is around right now, so I'll have no idea for a few weeks at least.
Club and University members often change every year, so it's VERY time consuming trying to find previous links. .. Sorry

[dannyf]

Forgot to mention that there are current-mirror chips that do precisely this, in a manner similar to the pnp solution I proposed earlier, but in a chip format. Their break-down voltage usually is more limited, however.

4-20ma transceivers can be used this way as well - they are more expensive and harder to find.

High-side current sensors would be another solution, but with its own issues too.