Low cost BMS

[Nevilleh]

Haven't been on here for a while, but I have been working on a BMS to replace my original 8-cell module one that proved to be awkward and clumsy to install, although it does work OK.
This one uses a single module per cell attached directly to the terminals. It has "daisy-chain" comms, opto-coupled from cell to cell and back to a master control unit. It was inspired by Weber and Coulomb's one, although built with PIC micros as they are the ones I have development tools for. It uses a thermistor for temperature sensing and has a shunt capable of bypassing 1/2 amp for cell balancing, if required.
The master has a 2x16 LCD that can show cell voltages and temperatures and also the max battery voltage. It has a low voltage alarm output and also an output to turn off the charger when a cell reaches its max voltage.
The cell modules cost about $5 and the master unit about $45 or so, depending on the LCD used.
It should be able to handle a couple of hundred cells OK, mostly limited by the RAM in the master.

Here's a picture of a cell module:


Note that it is all SMDs and this one is hand soldered. The next ones will be done in the oven.

The master control unit is not finished yet, waiting on some pcbs to test it with.

[evric]

Very nice.

[T1 Terry]

Haven't been on here for a while, but I have been working on a BMS to replace my original 8-cell module one that proved to be awkward and clumsy to install, although it does work OK.
This one uses a single module per cell attached directly to the terminals. It has "daisy-chain" comms, opto-coupled from cell to cell and back to a master control unit. It was inspired by Weber and Coulomb's one, although built with PIC micros as they are the ones I have development tools for. It uses a thermistor for temperature sensing and has a shunt capable of bypassing 1/2 amp for cell balancing, if required.
The master has a 2x16 LCD that can show cell voltages and temperatures and also the max battery voltage. It has a low voltage alarm output and also an output to turn off the charger when a cell reaches its max voltage.
The cell modules cost about $5 and the master unit about $45 or so, depending on the LCD used.
It should be able to handle a couple of hundred cells OK, mostly limited by the RAM in the master.

Here's a picture of a cell module:


Note that it is all SMDs and this one is hand soldered. The next ones will be done in the oven.

The master control unit is not finished yet, waiting on some pcbs to test it with.

"and has a shunt capable of bypassing 1/2 amp for cell balancing, if required."
Does this mean the shunt section can be disconnected or turned off? This is the bit that stops me from using a BMS, there have been so many reported problems with the shunt section latching on and dragging a cell dead.
Sounds very interesting and at that sort of price and simplicity I could have a market for you, or anyone else who has a similar simple low cost unit without any balancing gismo.

T1 Terry

[Nevilleh]

Yes, the shunt may be turned on or off by the master. It will only be turned on when you want to do cell balancing. The s/w will turn it off if there is a comms failure, regardless.
Note that even if it did fail "ON", provided the comms is working, the cell voltage is monitored continuously and an alarm will be generated if the cell volts falls to whatever the min is. I use 2.5V for my what-used-to-be Sky Energy cells.
If the comms fails, you better be checking out the whole system anyway!

[Johny]

Terry, what you saw in Neville's response was smart software engineering. Many BMS designs don't include common sense in that they won't fail "safely". My opinion based on the many forums I have read and folk who have had these problems is that the BMS systems that killed cells by enabling the shunt (or the device that switches it failing) were not that great to begin with. Just my opinion - I run 192 cells pairs with really cheap BMSs - that's my lookout!

[coulomb]

Neville,

good to see that you are progressing.

A few questions:

* How hot do those SMD 6R8 resistors get when bypassing half an amp at perhaps 3.65 V, or possibly more?

* Does PL1 have the cell voltage unfused on two pins just 2.54 mm apart?
* Is there a fuse on the board?
* How do you program the PIC microcontroller? I don't see anything that looks like JTAG, unless it's PL4. Can you program from the serial comms once the inital program is JTAG'd in?
* Have you tested this yet on a vehicle with a motor controller PWMing?
* Where do you intend to put the thermistor? On the plastic case of the cells?

It sure looks nice and simple.

[Nevilleh]

Hi Coulomb,
They get up to about 100 deg C in a 25 deg ambient. Its still within their ratings though.
PL1 takes a jumper to connect power after you install the board. Probably not needed as the watchdog would make sure the micro started even if glitched up by fiddling with bolts.
The fuse consists of a bit of skinny track on the board and would be blown by the TVS if the volts got too high. And no, it doesn't have 150mm of clearance!
Yes, I have tested it on my car with just 4 cells - my original prototype modules - and it worked fine. Now I have proper pcbs, I am building 45 new ones, one for each cell. I did connect it to 4 cells that weren't connected to my old bms (which only covers 24 cells because I was too lazy to build any more modules when I discoverd how much work it was) and the first thing I discovered was that one of those hit 3.6V when charging quite a bit before the others. So I have been over-charging it all along! Doesn't seem any the worse for wear though.
The thermistor has leads of about 35 mm and I just stick it to the battery terminal with a blob of silicone.
I am trying out my "reflow" oven to build the rest. Actually a $75 benchtop oven with a proper temp sensor.

Edit: I forgot to add that the 5-pin connector is the ISP one. I've been looking at a bootloader, but haven't worked it out yet. I need to be able to trigger it by sending some sort of character stream and I haven't thought that quite through as yet.

Here's a photo of my first attempt at reflow - you will note that 3 resistors have moved. I don't know why that is, probably either too much solder paste or not enough! Any suggestions?

[Johny]

Does your benchtop oven blow air?

[Nevilleh]

It doesn't, just convection, so its not an airblast causing them to move.

[Tritium_James]

The pads look too close together to me. They're about the right distance outside edge to outside edge (ie, the space for the solder fillet) but they go under the component too far. I'd also say they look too narrow, they should be wider than the component.

The surface tension of whichever pad melts first will therefore pull the component to the position which minimises the surface tension, in this case diagonally. A wider pad would fix it I think.

If you're going to do PCB changes I'd put a LOT more copper around the power resistors, with a bunch of vias through to the back side, and more copper there. This will get your peak temps in the resistors down a long way. Don't forget because you're reflowing, not hand soldering, you don't have to worry about your iron managing to heat up all that copper to get a good solder joint.

[coulomb]

3 resistors have moved. I don't know why that is, probably either too much solder paste or not enough! Any suggestions?

It could be too much paste - it doesn't take much.

But it could also be a "torque" that results from not insisting that the tracks leave things like resistors at the ends only, inline with the resistor body. You seem to have gotten away with it on R9 and R10, possibly because you used less paste on those. R3 and R8 have the tracks running correctly, and these have had no problems.

You seem to have plenty of room on the boards, so maybe the next version can have the layout improved for this. But try less paste next board, as it's a lot less trouble.

[Nevilleh]

Thanks for the suggestions.
I'll have a look at the pad sizes, I just used the library ones in DesignSpark.
Also, vias and Cu on the back would obviously help with the power resistors temperature..

[Faz]

Why are you doing them in the oven? I would have thought it would take about the same time to hand solder them.

[coulomb]

Also, vias and Cu on the back would obviously help with the power resistors temperature..

Yes, though it will also carry the heat closer to the cells, where it is definitely not wanted. But there may be enough separation and/or air circulation for it not to be a problem.

[Nevilleh]

Well, it would if I did them one at a time, but I can get about 20 boards in the oven at once - a considerable saving in time and effort if it works properly.

[weber]

Well done on the new board. Are you gunna post the schematic? I agree with James re the reason why the resistors pull to 45 degrees. If these are standard DesignSpark library pads they must be pads for 0603 parts and you're using 0805s or something like that.

[Nevilleh]

Well done on the new board. Are you gunna post the schematic? I agree with James re the reason why the resistors pull to 45 degrees. If these are standard DesignSpark library pads they must be pads for 0603 parts and you're using 0805s or something like that.

Thanks weber. I'll stick the schematics for the cell module and the master control unit below.
The DesignSpark libraries seem to be a bit odd here and there and I really should go through and make a library of the components I used with the correct pcb outlines. I found a couple of things that didn't fit, the optocoupler being another. That's why it hasn't been re-flowed as I have to bend its pins back under a bit and then hand solder it.
I used 1206 parts everywhere except the power resistors which are 2512.
Here's some pictures of my oven tray, first with 18 boards and 702 wee dots of solder paste, then with the components placed and lastly after "cooking". If you look carefully you will see 5 boards with resistors that moved. Not too bad, easy to fix with the iron.
The first hand-soldered prototype took me about 2 hours to do, by the time I opened all the packets, laid out the parts and then stuck on 1 at a time, soldering as I went.
Today, I took 2 1/2 hours to squirt the solder paste dots, place the parts and flow solder it in the oven. That's only a bit over 8 minutes per board, so that really answers the question above about why I didn't hand solder the lot. Mind you, I still have to add the optocoupler and the 4 headers - and the LM385. I couldn't get any LM385s in smd at the time, so I bought TO92 and I convert them to smd by bending the leads over and cutting them off short.
My oven is a Sunbeam "Mini Bake & Grill" with 2 elemts top and bottom and it is quite fierce - gets up to 220 in about 6 or 7 minutes.
My technique is to put 18 boards in the tray (as you can see) and then place the parts. Turn the oven on and wait for it to get up to temperature. I found the 210 setting on the thermostat gives me pretty well 220 inside. After it has stabilised (the elements go off and on a few times) I carefully slide the tray into the centre of the oven, shut the door and time 3 minutes. Then I turn it off and open the door to let it cool. That comes close to following the recommended temp/time curve. Anyway, it seems to work OK.

[Nevilleh]

Built the first iteration of the master control unit and getting the s/w sorted. Took a while as the first cell module I tested it with had a fault! One of my oven baked ones and I had connected the LM385 incorrectly. My own fault, moved it over one spot when hand soldering it after the baking.
But I now have two of my oven-baked modules connected to two Li cells on the bench and the control unit is talking to them and displaying their voltages (in millivolts)and temperatures, as below:



Note that the "h" and "l" mean high and low and the temps are in deg C.
I am wondering if displaying the voltage to the nearest 10 mV might be quite sufficient? Then I could just have 3.33 and 3.32. Just how fussy are these Li cells? The specs state a max charge voltage of 3.60V and doesn't that imply that +- 5 mV is good enough?

[Johny]

Nice - a lot better system for large format cells than your previous system. Speaking of that, do you still have PCBs for your 8 cells monitors around? I have started to realise my long term maintainance issues with the non-individual-cell-readable BMS that I have.

[weber]

Good work, Neville. Any chance of reposting the schematics at a resolution where the text is readable? You might crop them to leave off the frame, and/or save them as GIFs instead of JPEGs, and note that we're allowed to have images up to 175 kB now.

I'm particularly keen to know what resistor values you have chosen for your optocouplers, both LED current limiting and phototransistor pullup, and what the minimum CTR of your opto is. If yours is communicating reliably when your car is driven, while our string of 28 cell-top units corrupts a character every few seconds, I'm wondering if it's likely that we can solve the problem simply by decreasing both those resistor values significantly and running at 20 mA LED drive rather than 2 mA.

Of course it's also possible that our battery/controller/motor setup is generating more interference than yours.

[Nevilleh]

Yeah, I still have a bunch of un-assembled circuit boards. I never throw anything away! Do you want some?

[Johny]

Yeah, I still have a bunch of un-assembled circuit boards. I never throw anything away! Do you want some?

Can you tell me the physical size. W x L and populated height? From memory they do 8 cells each and I have 16 cells per box so I'd have to house 2 of them in the space Headway allowed.

Weber - did you see in my blog that the VFD output interfered with my BMS in the closest pack. The EMF that these things put out is wild!

[Nevilleh]

Good work, Neville. Any chance of reposting the schematics at a resolution where the text is readable? You might crop them to leave off the frame, and/or save them as GIFs instead of JPEGs, and note that we're allowed to have images up to 175 kB now.

I'm particularly keen to know what resistor values you have chosen for your optocouplers, both LED current limiting and phototransistor pullup, and what the minimum CTR of your opto is. If yours is communicating reliably when your car is driven, while our string of 28 cell-top units corrupts a character every few seconds, I'm wondering if it's likely that we can solve the problem simply by decreasing both those resistor values significantly and running at 20 mA LED drive rather than 2 mA.

Of course it's also possible that our battery/controller/motor setup is generating more interference than yours.

The optocouplers are Vishay VO615A - 6 which have a CTR in the range 100 to 300 %.
The resistor driving the emitter is 330R and the load resistor on the output is 2k2. You will note that the wires between boards are on the driver side (the 330R side) which means that the impedance is pretty low and hence there should be very little interference pickup - if any.
On the cell modules, the Vs is as low as 2.5V which means the diode current could be under 4 mA. Works OK, but.
And on the control unit, the Vs is 5.0V so lots more current available.
In the car, I have a pair of wires running from a cell in the boot to the control unit inside the car and there is no problem with that. All my comms wires are twisted pairs too.

[T1 Terry]

I am wondering if displaying the voltage to the nearest 10 mV might be quite sufficient? Then I could just have 3.33 and 3.32. Just how fussy are these Li cells? The specs state a max charge voltage of 3.60V and doesn't that imply that +- 5 mV is good enough?

I charge my cells to 3.45v this is to give me a big safety margin should they go out of balance. I don't start to worry till they get up around the 3.85v, then I stop charging and let the load of the inverter that powers my 2 250ltr fridges pull the rogue cell back down.
As a result of a few tests I have determined that any cell voltage above 3.4v doesn't carry much meat with it, very little ah capacity but all that is below 3.35v per cell has virtually the whole cell ah capacity behind it. (Don't ya just love techo talk ) The result of this high cell voltage having little capacity is each time it's drained off the recharge brings the other cells up much closer together, the variation between the cells diminishes each recharge.
I guess I only get away with this because I only have a 12v nom pack so the scope for a cell to go much out of balance to the other 3 but still stay under the 13.8v float limit.
As an acid test I built up a 360ah 12v pack (16 X 90ah cells- 4 X 90ah in parallel X 4 sets in series,)) that had been in the packing case for 2 1/2 yrs. Within 2 weeks, with absolutely no external balancing the cells have come from 350mv separation to 25mv. The spread of actual ah capacity was probably much higher because the cells with the lower voltage (below 3.45v) could have been many ah lower without a big voltage difference showing at the terminals.
About the only take away from that bit of rambling is that if you let the voltage climb to 3.8v and then drain it back to 3.45v while hlding the pack on a float charge the cells will balance much faster.

T1 Terry

[woody]

Hi Terry,

not sure if I'm misunderstanding what you are saying, sorry if I didn't...

As a result of a few tests I have determined that any cell voltage above 3.4v doesn't carry much meat with it, very little ah capacity but all that is below 3.35v per cell has virtually the whole cell ah capacity behind it.

What you are saying is that the charge curve kicks up at the end - I agree with that.

The cell is like a fat bottle with a long thin neck - as you pour water in steadily, the level doesn't change much until just near the end.

The result of this high cell voltage having little capacity is each time it's drained off the recharge brings the other cells up much closer together, the variation between the cells diminishes each recharge.

I disagree with this.

If the cells are in series, the same number of amp-hours is taken out of them when discharging, and put back in when charging - they are locked together. Any variation comes from variations in the self-discharge rates.
Top-balancing is the equivalent of increasing the self-discharge rate above a certain voltage.

I guess I only get away with this because I only have a 12v nom pack so the scope for a cell to go much out of balance to the other 3 but still stay under the 13.8v float limit.
As an acid test I built up a 360ah 12v pack (16 X 90ah cells- 4 X 90ah in parallel X 4 sets in series,)) that had been in the packing case for 2 1/2 yrs. Within 2 weeks, with absolutely no external balancing the cells have come from 350mv separation to 25mv. The spread of actual ah capacity was probably much higher because the cells with the lower voltage (below 3.45v) could have been many ah lower without a big voltage difference showing at the terminals.
About the only take away from that bit of rambling is that if you let the voltage climb to 3.8v and then drain it back to 3.45v while hlding the pack on a float charge the cells will balance much faster.

If you are talking about draining individual cells down to 3.45V while the pack is on a holding charge, then you are top-balancing and it should work.