Old BMS General Thread

[dillond666]

Peter,

I got the boards from pcb cart, just like you. Including VAT and postage they were £80 for 30 boards.
My proper name is Derek Dillon and I am a mechanical fitter in an oil refinery (boo hiss) .........but if anyone wants to offer me a job in renewables...........

D

[Jeremy]

Spooky! I thought another one of the regulars from the ES forum was here when I saw your avatar, Derek. "Doctor Bass" is a well-known and prolific poster over there on all things electric (plus he seems to collect vast quantities of lithium batteries). See here for one of his many posts there: http://endless-sphere.com/forums/viewtopic.php?f=14&t=2824

There are a few from this forum on the ES forum too; certainly Malcolm and Steve are regulars there that I know of.

Jeremy

[dillond666]

Jeremy,

Doh! somebody beat me to it. Back to the drawing board.

D

[GregsGarage]

Another quick update.

Didn't get a chance to finish the boards.

But did get a chance to test the new software.

V1_04 works as expected.

V1_05 was still sluggish at 31kHz so I bumped the speed to 125kHz and there is no noticeable difference in performance between the 2 versions. I will use V1_05 running at 125kHz and see how long it takes to balacne my 4 cells, one of them is quite badly out of balance due to some testing that I have been doing. Any way salvaged an old relay from a disused circuit board, not having any desoldering braid to hand, I just used a shot of compressed air, messy but the relay fell out

[nino500]

Jeremy,

Interesting example of Dr.Bass's posting.

Do you or any of the other guys here have an opinion on the viability of Dr. Bass's charging system?

It does seem magically simple.

[qdos]

Invite Dr Bass over here then

I do pop in to ES now and then myself but there's enough going on here to keep track of LOL

[Jeremy]

I did think about using that approach, rather than a "conventional" BMS, but encountered a potential problem with the selection of DC-DC converters. You can buy high power 3.3V (trimmable to about 3.65V to 3.7V) converters very cheaply on eBay, but they don't all behave well when operated at current limit (which is what is needed during the initial charge phase).

Most of these converters have foldback current limiting, which means they effectively turn off the output if asked to deliver more current than their limit. Some will use what's often called a "hiccup" mode to try and recover; the converter will pulse the output periodically to see if the over-current situation is still present. These will work OK as chargers, as the "hiccups" will eventually bring the cell up to a point where the converter isn't in current limiting, although it may take some time.

Unfortunately, some converters don't have a "hiccup" mode, but will shut down completely when faced with an overload. This type usually need to have the power or load disconnected to recover, so are not much use as chargers.

Buying DC-DC converters from a normal supplier, such that the spec can be assured, is prohibitively expensive. Buying from eBay carries the risk that you may get supplies that don't like a momentary over-current condition, but if they are cheap enough this may well be a worthwhile risk.

On a related point, I bought a couple of 37-72V input, 12V, 20A, output DC-DC converters from eBay and have found them to be superb value. They cost me around £45 including postage from the US for two.

Jeremy

[arnolde]

Peter,

This project is really awesome, thank you so much for sharing all this with us! I've started a new project which will require a 600V pack, so looking for a suitable BMS I just found this thread last night and now I've just finished reading it. I'm 90% sure I'll use it on my pack also, since I also love making PCB's (although I use Eagle) and programming PICs (in Assembler). Well... at least I did one project on a PIC18F 2 years ago, including a similar video chip like you're using (it actually is an OSD chip from a video recorder).

Anyway, here are some thoughts I had while reading, some of course are now irrelevant because meanwhile a month has passed, but I still want to mention them:

a) You mentioned taking out the slave interrupt routine because you had problems with data corruption when an interrupt occurred during a calculation. The way to cope with this is to save all the registers (or at least the ones you will use in the interrupt routine) in the stack (if you have one, otherwise somewhere else) at the start of the interrupt routine, and restore them back at the end. If you do this, you wont have any problems with interrupts. Some software compilers already do this automatically for you, on others you need to remember to do it yourself.

b) You mention adding 128 to a slave address to tell it to re-program instead of poll the measurement (or something like that). That would reduce the possible slave count to 128 (or 127), please consider that some people (like me) might need 200 or more slaves for a 660V pack.

c) Have you considered putting everything on one single big board (all the slaves also), electrically seperated (except for the opto-connections) but mechanically connected on one board? Then you would have the same circuit electrically, but the advantages:
1) just one wire per cell goes to each + pole (and one to the pack - pole)
2) cheaper board manufacturing costs, just one board layout/drill/soldermask/etc., less total board area because you save the terminal screwon areas
3) board can be easily protected mechanically, no danger of bashing any slaves when handling the battery pack
4) you could move the BMS to/from different packs if you wanted, because the intercell-wiring can stay on the packs with a 50-pole plug for plugging in the BMS
5) alternatively make one masterboard and one slaveboard, then you can make a slaveboard for i.e. 64 slaves which is the same size as the masterboard, and if someone needs more slaves they can just stack as many as they need, like i would need 4 of them = 256 slave "slots", i wouldnt put the parts onto all of them (leave 40 or so blank) but it would still be cheaper than 200 single "boardlets".

d) for the floorplate of your spare wheel well, may i suggest putting your board as you want it and then using spray-foam (used for fixing windowframes and doorframes) to fill out the space underneath, it's light and strong and expands to fill every small cavity, will distribute the weight very evenly over the most uneven surface (cover the inside with plastic foil first (like buttering a cake pan), so then you can lift out the whole lump for maintenance if you need to) - just make sure you only spray about 50% of the area you want to fill, as it expands by 200% during hardening. If you leave a hole (several cm diameter) where the foam can escape, it wont push your mold out of place though (it expands softly). I recommend using 2-component-foam (with the twist-wheel on the bottom to mix the 2 components) as it hardens through in a few hours, as opposed to the one-component-foam which can take days to harden in the center of a big block.

regards,
Ethan

[retepsnikrep]

Arnolde

Thanks for your comments.

I got over the interrupt problem by preserving the value obtained from the previous adc conversion loop and using that when an interrupt occured. So it's actually using the previous adc result after an interrupt. As it loops so quickly this doesnt matter.

I though about a big board but it sounded a nightmare and would be massive for 50 or 100 slaves. It may be the way to go though, feel free to knock one up based on the overall design if you wish. In your setup there would be 200 wires going from the giant slave board to the pack. In the current setup with individual slave boards there are only four wires from the pack to the Master. I like the individual boards as the visual indications they also provide is there on top of the cell in question, so you can see what's happening.

One of the reasons the Master is limited to 128 cells is due to the amount of available Pic scratchpad ram which is 128bytes, this is used to store the incomming serial voltage data from the Slaves and keep execution speed up to a reasonable level.

The total pack voltage is then calculated by adding all the values stored in the scrathpad ram together. This could be removed but calculations would have to then be done on the fly, inbetween serial transmissions from each slave, slowing the data stream probably, but meaning you could have an unlimited number of cells in the pack. The Pic I'm using the Picaxe 28X1 has an upgrade comming soon (28X2)with 256 bytes of ram which would allow 256 cells using my current software design. Again feel free to try the on the fly approach by modifying the software and see how fast it works, I will probably do this later on my 50 cell pack, so we can get an idea of the speed using the two approaches. Remember the Master software is still untested and very much in alpha form, so a comlete change of direction as to the way it calculates/operates could happen at any time. The speed taken to poll the pack would simply increase as more cells are added, my design criteria were that it should poll the pack once per second ish

I've already bought some expanding foam for the battery tray issue

Regards Peter

[arnolde]

Peter,

Well, okay maybe 50 slaves on one PCB is a bit many... One could still make batches of i.e. 14 or 28 and then you'd only have handfulls of wire, and 4 wires between each board. Well, I'll try different variations and see. I'll load up my Eagle and see what I can do

Another advantage of non-single slaves would be, that they can then also be used with cells that dont even have screw terminals (like lipo pouches).

Are you still happy with the hardware layout as on page 11, post #2 of this thread? Or have you made any updates/bugfixes since then?

I wonder if I could import your layout files into my layout editor? Have you uploaded them anywhere?

Ethan