using voltage dividers to manage a 24s battery pack

[BM11]

I'm going to be using a PIC microcontroller to manage a 24s battery pack. Since it can only take a maximum input of 5V, I need to scale down the voltage of each cell. The most apparent solution is using 24 voltage dividers. Is there an easier way to tackle the problem or some way to stagger the voltage divers in order to use fewer components?

 

[Jeremy Harris]

The problem with this approach is resolution of the PIC A/D converter.

For the bottom few cells things are reasonably OK, but by the time you get to the top of the stack you'll find that 1 bit on the 10bit PIC A/D will be too big a voltage jump to keep the cell in limits.

Here are a couple of worked examples to illustrate the problem:

Cell 2, nominal charge cut off voltage 2 x 3.65V = 7.3V Divide this by 2 (using your potential divider approach), say, to get the voltage into the working range of the PIC A/D, gives 3.65V at the PIC A/D input. The PIC A/D will resolve to 5v / 1024 = 4.9mV, which is effectively 9.8mV per bit at the cell, so this is fine, as a couple of bits of A/D error won't cause the cell voltage to be out by much.

Now, let's take a look at cell 24, where the charge cut off voltage is 24 x 3.65V = 87.6V. To get this down to the working range of the PIC A/D we need to divide it by around 24. This brings the A/D voltage back down to 3.65V again. However, a single bit error on the PIC A/D will now result in a voltage error at the measurement point of nearly 118mV. Realistically, you have to allow for at least a 2 bit error, plus whatever reference drift you may have, so the likely error at the top of the stack will be around 0.2V, too big for accurate balancing.

There is another option using PICs, and that is to use a single small PIC per cell. This works OK, as a PIC will run on the voltage from one cell and each PIC can communicate via an isolated serial bus to a master controller. Peter Perkins has done a good job of designing a system like this, take a look at this very long thread on the BVS forum: http://www.batteryvehiclesociety.org.uk/forums/viewtopic.php?f=3&t=1245

Jeremy

 

[7circle]

What about a sample and hold circuit that can cope with 100V.

You would also need a dual channel MUX to 24 cells.

They discuss similar issue here
http://www.endless-sphere.com/forums/viewtopic.php?f=14&t=8941&p=138557&hilit=sample+hold#p137917

I'm sure there is IC device set to do it for 8 channels but parts get costly for 100V devices.

A descrete circuit could use 24 pairs of optos via 100K res's to charge 1uF with 6V zener limit.
Then another opto pair to link cap to analogue input.

Trying to find example on web....

 

[heathyoung]

Another possibility is to use multiple voltage reference diodes - Like an LM4040 - in series to limit the voltages on the ADC inputs. Then use these known values to determine your resistive dividers.

Using purely resistive dividers is a nightmare at higher voltages, due to the input impedence of the PIC becoming an issue as well - needing low value resistors (or opamp buffers) or you have issues with heat dissipation in the resistors.

 

[7circle]

I'm sure there is IC device set to do it for 8 channels but parts get costly for 100V devices.

A descrete circuit could use 24 pairs of optos via 100K res's to charge 1uF with 6V zener limit.
Then another opto pair to link cap to analogue input.

Trying to find example on web....

16 way analogue switch at 100V - Circuit would need 3 of these.
http://www.supertex.com/pdf/datasheets/HV2601.pdf
Their online shopping has them at US$18.88 1 up
OR
http://www.clare.com/home/pdfs.nsf/www/CPC7220.pdf/$file/CPC7220.pdf

The thread "Re: A PIC based Battery Management System"
http://www.endless-sphere.com/forums/viewtopic.php?f=14&t=3486&hilit=hold&start=195#p53289
has the sample and hold approach has come up now and again.

Finding a low cost analogue switch solution is tricky.

I also like the CPU/ADC per Cell with opto linked com's.

But a paired MUX approach to measure the cells is worth taking to a complete schematic.