Quad Cell Distributed Digital BMS design discussion

[Alan B]

I've looked at a number of BMS designs both here in ES and elsewhere, and it appears to me that a four cell module may be a fairly optimal size for a modular advanced BMS system. I'm plannning to configure my Headway cells into 4S blocks on my 48V ebike.

Why 4S?

1) voltages in range that can be read with simple dividers, 10 bits gives about 15 millivolt resolution
2) module is nominally 12 volts (in LiFePO4) which is easy to deal with
3) enough voltage for reliable microprocessor operation, but not too much for significant simple linear regulation losses
4) lots of available components can handle this voltage (FETs, regulators)
5) fewer overall parts than a per-cell solution without being too large and complex, fewer boards to build and manage, fewer connectors
6) modularity fits most popular pack sizes readily (divisible by 4)
7) better mechanical fit than 1S designs for cylindrical cells
8) more adaptable than analog designs - most upgrades in firmware, not requiring remaking or hacking boards
9) lots of applications for 12V battery power sources (aside from e-bikes)

Requirements for QC BMS:

1) low cost, low parts count, low power consumption, balanced per cell current consumption
2) programmable for flexibility of application
3) modular design - one or more quad cell BMS boards with isolated interconnects work together
4) voltage cutout function (LVC) isolated output bus (for ebrake)
5) communications for detailed cell info readout via data bus using standard network cables and optical isolated 5V serial signals
6) high noise rejection (compatible with motor and controller noise)
7) balancing and charge control from regulated current limited 15V power supply or a quality 12V automotive charger
8) charging direct from solar panel (current limited)
9) early balancing begins before high cells are full to compress full charging cycle time
10) cells maintained at full voltage for specific time to insure full charge
11) fuse protection against various faults
12) based on readily available free open source software tools
13) data readout from a PC with simple data bus adapter (get individual cell voltages, etc)
14) data readout using a uP terminal with LCD for onboard display of detailed info and higher level logging or management
15) no power reduction due to BMS (no shunt, etc)
16) adjustable thresholds for LVC, shunting, etc
17) programmable control of charging and balancing operation for improved control algorithms

QC BMS Design:

1) microprocessor based (my preference here is Atmel AVR, GCC-AVR, Python for host programs)
2) load resistors and FET switch/resistive load type balancing
3) optically isolated LVC bus for ebrake controller input
4) optically isolated 5V serial communications bus for data transfer
5) comm bus RJ45 based, two connectors per board for easy daisy-chain using off the shelf cables
6) boards individually addressable for readout and control
7) charge supply 15VDC regulated supply or current limited 12V automotive charger (or solar panel) up to approx 25A,
- charge control and balancing provided by the QC BMS board
8) 12V output with LVC FET switch for moderate loads and general 12 volt applications
9) motor output current to be taken direct from cell stack (LVC protection via ebrake output)
10) polyfuses for protection against various faults (no fuses to replace)
11) very low standby current regulator and low drain voltage dividers minimize overall standby current
12) uP in sleep mode to minimize power consumption when not active
13) can be disconnected from battery for longer term storage (2 plugs)

Here is a fairly complete version of the schematic:

http://picasaweb.google.com/lh/photo/Kbp2x_VflxtAC1oU44GxWuLi8wISrz31SJGZJ_3ODCg?feat=directlink

Is anyone else doing anything similar? Interested in something like this?

 

[Alan B]

First post updated and schematic added.

 

[rebelpilot]

Very interested.
4S - yes!

9) motor output current to be taken direct from cell stack (LVC protection via ebrake output)
9) fuse protection against various faults

I like this idea.

2) master readout using a uP terminal with LCD for onboard display of detailed info

What are the plans for the uP terminal? I have 10 pieces Optrex 16X4 character LCD - DMC-16433, just begging for a project. I have prior experience writing assembly language code for test equipment using Motorola 6805's, and would welcome the opportunity to jump back in, if you would like some rookie help. I am not familiar with python, but I usually learn fast.

 

[Gordo]

I am interested in this BMS. I have 20S LiFePO4 40Ah with no BMS. Currently a PIA to balance every 3-4 cycles.

 

[Alan B]

Sounds like a nice display. Would be great for this.

I don't have firm plans for a master terminal yet. That could be an off the shelf terminal such as this:

http://www.sparkfun.com/commerce/product_info.php?products_id=37

Hard to beat all those features for that price.

I'm planning for the slave board to do all the heavy lifting, so the master is completely optional and could be skipped. It is more for readout than anything else. The slave boards do the LVC and the charge control without help from a master.

However, a dose of reality. It takes quite a while to get a project like this going and done. So what I'm planning to do in the meantime with my almost built Headway pack is:

1) get a couple of CellLog's to do LVC and help me learn how things are working.

2) get some connectors set up so I can use my Cellpro 10S charger to balance charge them. Maybe get a second one so I can do all 48V (four 12V sections) in parallel.

What I'm looking for at this point is feedback on the slave requirements, design and schematic. I need to flesh out the Bill of Materials, and detail the design.

 

[Alan B]

One element of this design is the requirement for a 15 volt supply to charge from. Here's one that might be good. Lambda LS-75, $29 each. 15 volts at 5 amps. Voltage adjustable.

Not sure it can handle the load of depleted batteries though, they caution about drawing too much current.

 

[Alan B]

Here is a power supply that can handle continuous shorts. But it does this by hiccup mode which could be a problem...

http://products.cui.com/getpdf.aspx?filename=VGS-75-XX.pdf

And still only $33.

 

[rebelpilot]

Both supplies appear to require an additional current limiter circuit. Is it possible to add a simple shunt and opamp circuit to provide feedback into a switchers' voltage sense circuit to add foldback current limiting? A quick search on the web didn't turn up any references so maybe I'm out in left field here. Is there something I'm missing?

 

[Alan B]

If we can't find supplies that have current limiting I may have to add it to the BMS board. I was hoping to keep it simple.

 

[Tiberius]

... Is it possible to add a simple shunt and opamp circuit to provide feedback into a switchers' voltage sense circuit to add foldback current limiting? A quick search on the web didn't turn up any references so maybe I'm out in left field here. Is there something I'm missing?

Hi Rebelpilot,

This has been discussed quite a few times here on ES and sample circuits posted. I don't have links to hand, but a detailed search should turn them up, or look in the threads about Meanwell PSUs.

Nick

 

[Jeremy Harris]

Just what I was going to say, Nick. This is very much re-inventing the wheel! The forum search function is your friend............

Peter Perkins has already designed, developed, built and tested a very nice PIC based modular BMS, that uses individual PICs on slave boards (you can have as many or as few as you wish, I believe) and a Master board that drives a very nice little graphical display. Peter has already posted on here about it a few times, but perhaps the best source of info on his BMS is on the BVS forum, here: http://www.batteryvehiclesociety.org.uk/forums/viewforum.php?f=53 It's not specifically a quad cell system, but could be any number of cells you want and is modular. The slaves all communicate with the master via a serial link, which makes the design extremely flexible. In essence, a change of code and you can alter pretty much anything you like. What's more, I think Peter may even have boards available, I know that he's developed this pretty much as an open source project over the years.

As for power supplies, a few years ago I pointed out on here that the cheap switching supplies from Meanwell et al were near ideal (the thread is still current and is just a few posts down from this thread, on the first page: http://endless-sphere.com/forums/viewtopic.php?f=14&t=4125). Since then Fechter has made a very neat little current limiter board that allows them to work in continuous current limit mode, rather than revert to hiccup mode. I've been using a Meanwell without the current limit mod as a charger for a 40Ah, 51V nominal pack and despite the current limit being about 7A or so I've never had it actually stay in hiccup mode for more than about 30 seconds. After that time the pack voltage has risen enough to ensure that the current drawn is less than the hiccup mode limit and it just charges smoothly. It's well worth noting that even at just 7A you can get a significant voltage drop on the charging leads - enough to allow these supplies to normally work OK as they are (but Fechters mod makes them really fail-safe).

Jeremy

 

[Alan B]

Thanks for the comments and links Nick and Jeremy. Those are very helpful.

I had not seen Peter's PIC setup. While it is interesting I don't think it meets the requirements I'm trying to meet (in post 1).

The Meanwells appear to be a good deal, and very popular.

I'm planning to use off the shelf standard power supplies available from many sources, and not require something different depending on the system voltage, or depending on a particular supply from one manufacturer. So whether I'm running from 36V or 48V I want to use the same charging supply. It must be less than 50 volts to meet safety rules at work. A 48V battery is too high in voltage (and cannot be worked on de-energized) so I break my packs into lower voltage. The only time a voltage over 50V is developed is inside the enclosed wiring on the bike, which is acceptable.

I picked 4S/12V because there are a lot of other applications for 12V outside the e-bike arena, and the multi-use potential is important.

This morning while driving to work I decided that it makes the most sense to standardize on the common 12VDC supply for charging as well, avoiding the less available 15V supplies. That would also allow charging from a 12V car or RV battery, and 12V supplies are often the most cost effective supplies available. I adjusted the first posting accordingly. This will require some design changes.

My goal here is to refine the requirements and the design, then decide if it is worth building. It is good to get links here to other solutions, it would make them easier to find for everyone and it is important to compare the choices. I have done hundreds of searches and read thousands of posts and still not found everything related.

 

[Jeremy Harris]

The Meanwell type supplies come in many different voltages and are made by many different manufacturers. They are cheap and do the job very well (I'm using a modded 12V one to charge my 4S, 80Ah pack, with my own homebrew 4 cell BMS). If you read the switched mode power supply thread you'll see that there are many variants of these supplies, but all share a common topology and all come in a range of standard size cases, making them interchangeable. Add in the fact that they are easily available anywhere in the world, and they seem ideal.

Jeremy

 

[rebelpilot]

Nick, Jeremy, thanks for the input. "Fechter's mini Meanwell limiter board" is exactly what I was looking for. I was specifically searching the web for "foldback current limit" and "switching supply" and nothing relevant turned up. This seemed like an obvious solution, but my search only found references to hiccup mode limiting. I've reinvented the wheel many times :lol:
for others whose search leads them here:
http://endless-sphere.com/forums/viewtopic.php?f=31&t=21768&hilit=meanwell+psu

 

[Alan B]

Even in the Meanwell supplies the 12V is lower cost than other voltages for the same power level, at least in a quick study on on ebay.

 

[Alan B]

Perhaps the best way to handle the 12V charging supply is to use 12V automotive chargers. They are current limited and have about the right output voltage. The QC BMS board can control the charge current and implement the balancing which works best when done together.

So perhaps the best power supply is the common non-microprocessor type 12V automotive charger?

Or the standard RV converter supplies are also set up to handle the current limiting and charging loads of large lead acid batteries. One that I have will deliver 60 Amps of charge current.

I also see over in the Meanwell thread they have come up with some nice newer boards to turn the Meanwell into a charger. So if you want to hack a power supply that is always an option. But balancing must be provided from somewhere else either with an automotive charger or with a modified power supply.

The Role the QC BMS board can have in this is:

1) starting balancing early when the high cells cross say 3.5 volts, and signal "charging/balancing"
2) turning the charge current off when the highest cell reaches full voltage, and signalling "charging/finishing"
3) bleeding the hi cell(s) down a bit
4) turning the charge current back on to complete charging the lower cells
5) alternating the above steps 2-4 until all are full by holding them at full voltage for a specified time, and then
6) turning charging current off and signalling charge complete

 

[liveforphysics]

Perhaps the best way to handle the 12V charging supply is to use 12V automotive chargers. They are current limited and have about the right output voltage. The QC BMS board can control the charge current and implement the balancing which works best when done together.

The 12v automotive chargers I've seen are always just a multi-tap transformer with a rectifier and a cap. You combine taps when you select the desired current level. IMHO, it's a much much worse choice than a meanwell in all respects.

 

[Jeremy Harris]

The 12v automotive chargers I've seen are always just a multi-tap transformer with a rectifier and a cap. You combine taps when you select the desired current level. IMHO, it's a much much worse choice than a meanwell in all respects.

Seconded. I have taken apart three or four over the years and they were all, universally, pretty poorly built, had little or no smoothing and no form of voltage regulation or current limiting. They rely on the fact that lead acid batteries are tolerant of high ripple current during charging and have a natural characteristic that limits charge current when they are near full charge. Many automotive chargers just use the winding resistance of the transformer as the current limit, some use ballast resistors. Either way they're not fit for purpose in this application.

It seems to me that, instead of heading in a worthwhile direction, this thread is trying to ignore a lot of hard-won knowledge that has been discovered on here by many forum members working together over the years. For example, several people here have designed, built, extensively tested and refined BMS systems, ranging from simple shunt systems to systems with multiple microcontrollers. Those members have taken a lot of time and effort to detail their work and discuss it openly with others, with the result that there is now a very good overall body of knowledge on what works and what doesn't. Similarly, collectively we've scoured the world (literally, as this forum is global) looking for suitable power supplies to use as chargers. Forum members have tried just about every power source commonly available to drive chargers, from directly rectified household supplies (scary, but true....) through multiple modified wall wart type supplies, ex-PC and laptop supplies to the freely available switched mode blocks that many have now settled on as being the best compromise between cost and effectiveness.

I'm not saying that you're not going to come up with something new, but I have to ask, why invent a new solution when there are a plethora of good, open source, working and proven solutions that we've all collectively worked on and refined for years?

Jeremy

 

[rebelpilot]

I used to work for a contract pcb assembly house. The two most demanding customers may suprise you. Black and Decker's group that managed the 24 volt DC mower, and Exide battery chargers.
These are the same chargers available at Walmart.
View attachment IMAG0181.jpg
These products demanded very close tolerances on the charge voltage levels requiring 0.1% resistors and TL431 precision voltage references. A word of caution: all these supplies are temperature compensated to match the temperature dependent charging characteristics of SLA's. They are easily modified to charge lithium, but they are not efficient, or reliable.

 

[Jeremy Harris]

Nice to see at least one automotive charger manufacturer has got their act together, those chargers look a bit better made than the one sitting in my garage at the moment.

Interesting to to see that they also use the TL431, that's the preferred shunt regulator used on many of the BMS designs on this forum (and elsewhere). My own two home brew BMS' use TL431s, driving darlington power transistors to get better high current handling and power dissipation capability.

Jeremy

 

[Alan B]

I want to thank people for constructive criticism and discussion here.

I have always avoided those low quality unregulated automotive chargers. The ones I have are regulated and current limited. They can be connected backwards without damage and left on the battery indefinitely without overcharging. The suggestion of using lead acid chargers on 4S configurations came from this board.

Some of my applications are 12 volt. What off the shelf efficient and effective solutions do we have for that? Single cell PICs are not appealing, the parts count is pretty high with a master and four slaves for 4S and a very high opto count. Single cell designs are not mechanically well suited for Headways. There's nowhere to mount them. Single cell boards are better for prismatics but still not well suited for 4S. The other problem with single cell solutions is that they rely on cell voltage to operate, and they need to operate below low voltage cutoff to continue to protect the cell. But if the cell drops below their terminal voltage requirements they cannot operate. So there is a built in failure mode. I've had TL54 type designs fail because at low voltage they don't work correctly. The opto won't work at low voltages either. So if a cell gets low enough these BMS' are going to fail to assert LVC. So 1S designs have some built in problems and this includes the popular analog designs. I think we can do better.

I'm familiar with applications that use 14V 5A off the shelf current limited supplies driving several farads of capacitors. These supplies are not quite as cheap as the Meanwells but require no hacking to be used in this design. A stock Meanwell is apparently hit or miss, according to experience here on ES and according to its specifications. Adding a regulator board to a Meanwell makes it suitable. I'm not ignoring the Meanwell, it is just another supply choice.

I'm looking for a good, available, effective 4S modular BMS solution. I have not seen it yet. The best I've found so far is a CellLog but that is incomplete and needs additional circuitry to control the load, and an RC balancing charger plus power supply for charging. That's pretty expensive and not easy to carry around, and I would need several of them. I have about eight of these 4S setups to manage, and I need to charge in several locations. See requirements in post 1.

 

[Jeremy Harris]

My boat pack of Headways is 4S, 8P (12V nominal, 80Ah) and the simple BMS first devised by Gary, Richard and Bob works flawlessly. I have either solar charging (at up to around 10 to 12 amps absolute maximum) or mains charging from a Meanwell type switched mode 150 watt supply (without any mods at all). The 4 channel BMS circuit board is fairly small, is fitted on the end of the battery pack and does exactly what's needed; keeping charge voltage to 3.65V per cell and shutting the controller down (via the ebrake line) if any cell gets below 2.5V. The parts count is about as low as you can get, as is the cost. Why won't something like this do the job for your pack?

Jeremy

 

[Alan B]

Where are these available now?

 

[Jeremy Harris]

If you read the BMS threads you'll find full details for every variant of this basic design that's so far been developed. I've stuck with the original, which uses a TIP105 as a shunt power transistor, but the newer versions that Richard and Gary have developed is a fair bit smaller, as they don't dump all the heat in the shunt transistor any more. From time to time Gary sells units on his TPPacks web shop, although I don't know what he has on offer at the moment, as I've not kept up with that thread for a fair while.

When we were all first discussing all this, back in around 2007/2008, I went and bought a stock of parts for the very first design. It's these I've been using ever since, as it works perfectly well. Only last week I made up an 8 cell version for another Headway pack - it took me maybe a couple of hours to build and works just as flawlessly as the 4 cell one on the boat pack.

FWIW, here is the circuit for my 4 cell boat charge management board (I use a Cellog 8 for LVC, with the alarm output driving the EB line on the controller), although I can't take credit for the individual shunt design, as that was a joint effort between Richard, Gary and Bob some time ago. The crude, simple and very effective charge current limiting circuit is mine, I just wanted something as simple as possible to reduce the shunt disspiation during the balancing phase. In my experience, charge management is the only really critical thing with a 4 cell pack, you can pretty much get away with just monitoring overall pack voltage for LVC if you wanted to, rather than use a cell level system.



Jeremy

 

[Alan B]

Looks like a clever design Jeremy. A good starting point for a simple setup.

It meets very few of the requirements I have outlined in post #1.