Old BMS Hardware Thread

[retepsnikrep]

It's clear as a general rule we need to reduce the resistors for the optos to around 160-180R for the project.
I shall update schematics later if you can as well Greg.

[GregsGarage]

I've been thinking of updating the slave schematics for a while now. I'll draw up a 4 cell slave schematic to include the new resistor values. How about some other changes/improvements while we are at it like the 12f1822 pic with built in voltage reference. Another idea is to have some sort of active monitoring of the load resistor so that we can confirm it is operating correctly. Any other suggestions or things people would like to see?

EDIT: Of course it needs an ICSP header.

[retepsnikrep]

if you put up new proposed schematic i'll look at it and the pic and give my comments so we can then tweak it.

[isaac_alaska1]

here is the new scope plot, using my sub-par opto's with 160 ohm resistors for R1 and R3.



still only just barely over 4.5 volts, but it's working on the bench. hopefully it will still work when i install it in the vehicle...

[retepsnikrep]

That looks fine Issac can you take a pic on same scale 2/v per division as the others you posted.

It will never get to 5v as there is 0.6v drop through the opto transistor.
So that looks pretty good to me and might respond to an increase in baud rate later.

[isaac_alaska1]

yeah i can set it up and get a new pic. 2v/div on the y scale, what x scale do you prefer? should i stretch it out and just capture one with better res, or leave it compressed so you can see multiple slave signals?
or one of each

[GregsGarage]

Here's a quick sketch using the 12F1822 pic with internal voltage refeerence. I've moved things about a bit as the code will need changing anyway. The ICSP header shares connections with the master bus and slave out bus. Since these 2 connections are into diodes the pic should progam fine. Pin 5 on the ICSP can be toggled as a diagnostic output. I've also added a load monitor line so the pic can check the state of the load resistor. Just need to figure out how we want to do it in software. Please note, I have not checked the data sheet carefully for correct pin assignments, for example I am assuming the ICSP pin outs are the same as the 12F683.

Any thoughts on this or glaring errors? I've just

[retepsnikrep]

I think we should drive the load transistor harder as well and use one R value for that and the opto resistor. Say 180R or nearest preferred value. I like the way you have got rid of a resistor for the optos, sharing one is clever

We can prob bin one of the decoupling caps lets just have a simple 0.1uf poly cap.

The load transistor should be on the pic HPWM pin so we have an option to drive it proportionally. That means we could use a lower value for the load R and only drive if for a % of the time to acheive the same average current load.

Not sure how the internal reference works we will have to look at that, does it get tied to a pin internally.

I'll have a read up on the datasheet. We should be able to make it quite a bit more compact and shave off 3-4 components per slave.

We should bring out the cell connections to a D header pcb mounted socket as well rather than having wires flying from all over the board. We should also probably standardise on the slave board interconnects so they plug/stack together and have a standard number of cells on a board that fully utilises a D sub connector say 8 or 14 slaves per board for a 9/15way D sub or other suitable multiple pin connector.

We should also explore the idea of getting them built up this time in China as well as going SMD to reduce the size.
We can simply program the pics on recipt via 5 tiny ICSP pads and a pointy springy probe pushed against it for a few seconds each.


Issac a selection of pics would be good.

[GregsGarage]

I think we should drive the load transistor harder as well and use one R value for that and the opto resistor. Say 180R or nearest preferred value. I like the way you have got rid of a resistor for the optos, sharing one is clever

We should make both resistors the same value either 160R or 180R to reduce number of different parts on board

We can prob bin one of the decoupling caps lets just have a simple 0.1uf poly cap.

I thought that as well, but suggest leaving 2 pads on pcb, even if only one gets populated.

The load transistor should be on the pic HPWM pin so we have an option to drive it proportionally. That means we could use a lower value for the load R and only drive if for a % of the time to acheive the same average current load.

Good idea. As I said before I am not even sure if I got the ICSP header correct, just drew it the same as the 12f683. I read somewhere of driving the load at say 50% or less and sizing fuses at the cells so that the load at 100% blows the fuse. Shorted load blows fuse and protects cells.

Not sure how the internal reference works we will have to look at that, does it get tied to a pin internally.

Look at section 14 and 15 of the data sheet. The voltage reference is adjustable, I suggest 1.024 volt. This value should make the maths easier. Have a look at the first part of this video (about 7 minutes in) for an explaination of this. http://youtu.be/Lg6oYFerUlA It looks like the ADC can be configured to directly connect to the volt ref and Vss, so if I've understood it right, no external pins are used for this.

Code: Select all

I'll have a read up on the datasheet. We should be able to make it quite a bit more compact and shave off 3-4 components per slave.

In theory we could ditch the led, but I think it is useful. The pull down on slave 5 could be done internally in the pic I think, again we can leave the pad on the pcb and not populate it.

We should bring out the cell connections to a D header pcb mounted socket as well rather than having wires flying from all over the board. We should also probably standardise on the slave board interconnects so they plug/stack together and have a standard number of cells on a board that fully utilises a D sub connector say 8 or 14 slaves per board for a 9/15way D sub or other suitable multiple pin connector.

I am thinking of initially doing this as a 4 cell board 5cm x 5cm. It would have a male 9 pin d-sub on one end for cell connections. Pin 1 and 6 - cell 1neg : Pin 2 - cell 1+ : Pin 3 - cell 2+... Pin 5 and 9 - cell 4+. Pins 7 and 8 no connection. On the opposite side of the board a female db9 for master and slave bus. The small board would allow the slaves to be closer to the cells being monitored in a vehicle like mine where the cells are literally spread out from front to rear. A larger slave count board should be possible later once this is proven to work.

We should also explore the idea of getting them built up this time in China as well as going SMD to reduce the size.
We can simply program the pics on recipt via 5 tiny ICSP pads and a pointy springy probe pushed against it for a few seconds each.

I am also thinking smd, actually a double sided board with smd components on one side and the large through hole on the opposite, mainly the load resistor and DB9 connectors. Another advantage of a small board is that rework becomes less of an issue if a board develops a fault. A 25 cell board would have to be reworked, a 4 cell board is nearly disposable or at least easier to keep some spare boards and swap out leaving rework for a time that is conveinient.

[retepsnikrep]

Agreed same value resistors that's what i meant.

I like the load led as well and they need to be visible if we are stacking so perhaps near edge of the board.

I personally think we should aim for a db9 connector and 8 cells per board minimum, 4slaves per board is not enough for me. The less board interconnects we have the better.

The issue with DB9 family etc is isolation of the pins to the hood/metal D part we will need suitably DC rated parts for the full pack voltage from a pin to the hood. We might have to use a different connector. We should aim for an isolation/rating of at least 500V

RJ45 or whatever little modem loop back connectors on other side/end for master/slave bus.

Lots to look at. I'll study the datasheet later.