This article is turning into a bit of a long story. It was originally published on 31st March 2015, but there are lots of updates since then. Check the bottom for new content. Latest update is 7th September 2015
Part of the fun here is playing with new technologies, which includes Lithium Batteries. As any fule know those are the ones that explode, are horrendously expensive and whose performance doesn't live up to their advertising.
There are those better qualified to cover the theory: here's an excellent article which I found after I'd already ordered the cells, which pleased me a lot as he has more knowledge than me and came to the same conclusions.
So no theory from me, just installation details. My goal is a combined starter/house battery for a small-but-electrically-intense boat with an outboard motor equipped with an alternator. The outboard changes the game a bit because a) I can use one battery - if it goes flat, I'll use the pull cord, and b) the alternator will always be running when the motor is running. This has implications for LiFePo4 batteries, which don't like being overcharged.
First some conclusions:
- It's much, much cheaper to build your own battery from cells than buy one in. At the 2014 Southampton Boat Show I wandered the stands and found only one supplier (Mastervolt) that dealt with Lithium batteries, who were offering an esoteric, outsized and wildly expensive option. I gather availability is better in the US, but building your own battery is still comparatively easy.
- If you're going to go with Lithium, plan this in at the start. Initially I was looking at it as a drop-in replacement down the line, but it rapidly became clear that I had to build the system around Lithium - specifically, the chargers (wind, solar, alternator, mains) and the means of measuring how much power in the battery must be specified with Lithium in mind. There is no such thing as a drop-in replacement here.
- LiFePo4 is dangerous when it's overcharged. Yes there are other ways to damage the battery, but for me the big one is the spectre of an unbridled lithium fire. With LiFePo4 this risk is minimal unless you pump too much power into the battery (not the case with other Lithium chemistries, but they're not applicable here). So, you need a way to disconnect the charger and/or dump power, which implies a Battery Management System and some careful thought.
And if you still need convincing, take a look at these numbers:
|Lifeline GPL-4DL||CALB Cam 72 x 8|
|Available Capacity (max lifetime)||105Ah (50% DOD)1||100Ah (70% DOD)|
|Charge Cycles (max lifetime)||1000||50002|
|Available Capacity (max power)||158Ah (80% DOD)||115Ah (80%)|
|Charge Cycles (max power)||550||30002|
|100% Discharge?||No||91% capacity after 290 cycles|
|Self Discharge||2% / month||04|
|Can be stored discharged?||No||Yes|
|Volume||25 litres||7 litres|
- Lifeline recommend 50% DOD for max lifetime, and not more than 80%. I have a feeling the lifetime figures might be quite optimistic as Odyssey claim "up to 400 cycles when discharged to 80% DOD" for their AGM batteries.
- EVTV have tested these cells extensively, and here is a manufacturer datasheet. This lists >= 2000 cycles, I'm taking the figures from EVTV because they're from testing, whereas manufacturer datasheets are generally rubbish
- Pricing in € because that's what I paid in. Price for the LiFePo4 cells includes the BMS and Contactor, as I consider these essential. Pricing for the Lifeline battery sourced from here. Both exclude VAT and shipping and prices from around October 2014.
- OK, no self-discharge on the cells but if you have a battery management system that will draw power - I'm budgeting about 20ma, which is about 0.4%/month for this size of pack. For long term storage you'd disconnect this.
So what's my setup? My shopping list is below. A few of my purchasing decisions were based on a presumed 100A maximum current draw - this will be when using the electric start on the outboard. Finding actual measured figures on this is almost impossible - the Yamaha 9.9HP owners manual lists the "minimum marine cranking amps" as 323A, which is obviously not the current that will be drawn from the battery otherwise the cables would be the size of your arm. Best I could find was a the starter for a 9.9HP Evinrude being measured at 32A, and a Mariner 175HP measured at 150A. So 100A ought to do it.
- 8 x Calb CAM72 cells, in a "4S2P" installation - 4 sets of 2 parallel cells, wired in series. Each cell is rated at 3.2V with a 72Ah capacity, so I get 12.8V (give or take) and 144Ah. Cost per cell is about €75 per cell, although inevitably that will vary. The cells came with a plastic outer case and a cover over the terminals, and the case was bolted to hold the cells firmly together.
- Enough bus-bars to connect them - 8 came with the cells, I need 11 in total.
- A pile of M6 Nordlock washers. These were written up elsewhere, they're new to me but look useful when you absolutely, positively do not want your bolts to shake loose.
- A battery management system. I went with the HousePower BMC, because it was relatively cheap, simple, and I could understand it. It's sole purpose is to cut power to the battery when the voltage is too high or too low - nothing else. You'll need 4 x cell boards (remove the fixed ring and solder on a wire for both the +ve and -ve connections, terminated with a crimped 6.2mm ring connector). Total cost was about USD$125.
- A mains charger. I struggled to find one designed to work with LiFePo4, as opposed to just an "also works with Lithium" sticker. Although charging LiFePo4 is actually simpler on paper than lead acid - none of that bulk, absorption, float bollocks, just cut off at a certain voltage - ironically you're going to pay more for it. The Sterling Power PCU1220 delivers 20A and has a configurable cut-off voltage, which is good as its standard LiFePo4 profile cuts out at 14.6V. This is 3.65V/cell, the same as the "alert" level in the BMS, which means the battery cutoff would constantly be tripping. So while it's useful for the initial balance, I'll be running a custom profile at 14.2V when installed to ensure the alarm won't trip. For these features I paid £269, which seems like a good investment.
- Accessories for the Battery Management System. In my case, this is a Tyco Kilovac LEV100 "Contactor" (as best as I can tell, just posh name for a big relay) for £10 on eBay - some standard automotive relays, a DPST switch (this will carry minimal current which compares nicely to the whacking great battery isolation switch you'd normally have to install for lead acid), a pushbutton, and (for testing) a 12V buzzer and the end of a 12V LED strip, wired into the high-voltage alarm circuit on the BMS.
- A desktop PSU. This is an essential item for the initial balance, which is an equally essential step that has to be done, although probably only once. I bought a "Tenma 72-10480" on eBay for about £40, which will put out 3A - enough if you're not in a hurry.
- A last-resort fuse. Everyone seems to recommend "Class T" fuses, but I couldn't find one under 200A. I want to blow at 100A so got a 100A ANL fuse. Remember to spend time laughing at the "display fuses cases" for these on eBay, as they're apparently used by the lads who beef up their car stereos - "phwoar, look at my big shiny fuse". I bought a Blue-Sea 5005 holder because it was made of a plastic that won't melt when things go wrong (unlike most of the boy-racer products), and some Blue-Sea fuses as they're ignition protected - i.e. they won't catch fire when they blow. Both more expensive than I would have imagined, the fuses alone are about £12 each from Aquafax.
- Cabling accessories - some 25mm² tinned copper cable, some 6mm and 8mm lugs and a Hydraulic Crimper. Hydraulic things are cool, and at £25? That's about the same as 1m of cable. Bargain.
Down the line this will be added to:
- A Solar Power MPPT charger - this has to be designed for Lithium, and I'll go for either a Genasun GV-10 or GV-Boost, depending on the output voltage of the panels.
- A shunt and a custom circuit board. Lead-acid batteries have a voltage that is linearly proportional to their charge state, so you can derive one from the other. LiFePo4 doesn't, so the simplest way to measure charge is "coulomb counting". A shunt will give you an indicator of current - to turn over a 9.9HP outboard I'm budgeting up to about 80A in short bursts, so have gone for a 50A shunt, coupled with a microcontroller that keeps track of the current in and out of the battery.
And here's the process.
- Get yourself off to see Peter, CALB's European reseller, who I'd highly recommend, or your seller of choice if you prefer. I'll admit I wasn't 100% sure what I was going to get in the box, but they turned up in with the cases, covers, bolts and busbars (order as many as you need). Pack of gum for size reference.
- Step one, fit the busbars and wire on the HousePower "cell boards" - 4 of these, one for each pair of cells. The red LED will wink reassuringly when done - actually I suppose it could be a warning, I haven't read the instructions yet.
- Once done carefully wire everything else in. Here is my setup - haphazard looking, but it follows the layout below (you can drag and zoom this - hooray for SVG)
- Charge the pack as full as you can with the main charger. One of the cells will trip the high voltage alarm, that's fine. At this point the HousePower cell-balancing guide suggests a few options, but I would suggest the only one that is even remotely practical is to set your desktop PSU for 3.55V, connect it up to one of the cell pairs, turn it on and walk away. I wasted days trying to drain the cells and then going back to the main charger - the fact is if you're putting 20A into the cells, you can't even come close to balancing them with the BMS. After my initial charge I found I needed about 24-48 hours per pair of cells - the current started at 3A and gradually dropped to < 100mA at which point I'd move it to the next pair of cells.
Install everything in a box. I made a pretty sturdy box out of foam and glass with
a recessed area
for the control board, relays etc., and embedded some copper bars under the glass
on the top side
which I drilled and
tapped, so there are no exposed connections to the battery itself. Photo shows the
fuse, relay, BMS
board, buzzer, contactor and shunt (not wired in yet - awaiting my own battery management
Cabling is 25mm² for the primary circuit and and 6mm² from the charger. Forgive the
connector, it's temporary.
If you're making your own box like this, don't forget to put in some means to secure it - I can bolt it down to the shelf with a couple of M6 bolts which thread into nuts embedded in the bottom of the case.
Conclusions at time of publishing
So far, none. It's been wired in since about end of February 2015 and is taking charge from the mains charger and delivering power on demand. It's shut down when unattended at low voltage (before I put the charger in, and I left it running the main computer on the boat for about a week) and at high voltage (when I set the charger to 14.4V rather than 14.2V), and both times I came in to find the battery alarm on and everything disconnected as required.
However I don't have any proper measurement of charge status yet, or any long-term figures. I haven't hooked it into an alternator or solar for charging, and the most I've tapped it for is about 10A: I need to test starting the engine, which means buying an engine. So these tests will come over time, and I'll update this page when it does.
Followup 1: 5th July 2015
My battery controller has packed up. I'm a bit annoyed about this, it suddenly started going haywire with the main contactor flicking on and off rapidly. It could be a bad connection somewhere, but I couldn't identify where visually or with a multimeter. I'm also not terribly impressed with the design of the system - I remember a joke in Readers Digest as a kid about a car with a single red warning light in the middle of the dashboard, explaining that the experienced driver will normally know what is wrong. I didn't find this very funny then and I still don't now - there is no diagnostic information coming out of this board other than a buzzer. Inevitably I have designed a replacement system which has the key features I want - integrated coulomb counting, complete configurability of all levels and thresholds, and a USB interface. I'm ironing out the kinks but it's looking good so far. Will document it once it's had a few months shake down.
Followup 2: 14th August 2015
My new battery management system went in yesterday and is finally working, although with caveats. The design above needs revision.
As described, the shunt will not measure the power used by the BMS itself. This turns out to be important as the contactor I'm currently using, the Tyco Kilovac LEV100, draws a whopping 460mA to activate its coil. This is a vast amount of power. The coil generates a lot of heat even when the current is zero; when combined with the current required for a full charge at 20A, the temperature of the BMS enclosure got up to 54°, which is not ideal - I had to keep bumping up the threshold for the high-temperature alarm!
When daytime sailing I'd probably have my primary computer, associated controllers, transducers and VHF on, which together draw about 1700mA, so the contactor would take 20% of the total power draw of the system. This might be OK in an electric car, but not if you're trying to conserve power. This again illustrates that a BMS for sailboats has a very different set of requirements than one for electric cars or solar storage at home.
To fix this:
- Move the shunt immediately after the fuse, and power the BMS after the shunt - the shunt now measures everything in and out of the battery.
- Swap the contactor for something else. Options are a latching relay, which I'm uncomfortable with as if it (or the BMS) fails, it may fail in the on state. Option two is a solid-state-relay, which has negligible "coil" current but which will generate a lot of heat at high current. Option three is to do away with it altogether...
- As designed the contactor will cut the battery from everything else, but it won't isolate the various circuits from each-other. This is bad - imagine it kicks in while you're accepting charge from an outboard motor. The battery is isolated, but the motor and your house system are still connected - filthy alternator power ravishes your delicate electrical system and trashes everything, except your safely isolated battery. Solution is to do away with the contactor, and instead ensure each circuit is on a separate relay which are all cut (and isolated) when a shutdown condition occurs.
- The cell balancing system I've currently designed is resistive, i.e. it dumps excess power when charging through a resistor. This is poor system. For one, it's dumping precious power and makes heat, and dissipating more than about 500mA is difficult. I recently discovered active balancing with a flyback converter which looks promising, although it will mean abandoning my single-wire cell loop.
On the plus side, my BMS talks over USB so I can easily hook it into my computer as you can see above. This was before I'd measured the battery state of charge, which is why it's 0%. The power through the shunt is measured at about 8000hz, so I can very accurately count coulombs to record the state of charge.
It's also talking to the cell management boards so I can measure cell voltage and temperature. This turns out to be very important as it will tell me which cell is playing up and whether the system is properly balanced. Turns out it's not - see the graph on the right. Hooray for raw data!
I hope this will partially explain why when I measured a full charge/discharge cycle on the battery I got about 76AH, which is way too low. Missing the 0.5A for the contactor as described above will throw this off by another 5%, so for now take that measurement with a large pinch of salt.
Followup 3: 7th September 2015
Last week we turned over the engine for the first time. The Yamaha 9.9 has a 600W starter motor (50A at 12V), so if we assume a large margin for losses in the process, turning the motor over drew probably in the region of 60 - 70A. We only ran the starter for a few seconds to verify the wiring and measure the current (inconclusive due to setup errors as it turned out).
A week later the battery was dead, and although the Battery Management System I'd designed would have informed me of this, ****ing British ****ing Telecom had seen fit to pull the plug out of my internet connection at the build site then make me wait two weeks for them to fix it. So no idea when or at what voltage the battery shut down. On inspection I found one - just one - of the 8 cells had dropped to 2.4V, which should be impossible as a) this is below the minimum voltage at which the management system would have shut down, and b) the cells are wired in four sets of two pairs, with each parallel pair having the same voltage. The cell was one of the two that made up "Cell 3" in the PDF in my last report, which was the one that triggered the low voltage during discharge.
I'll skip the process of how we got there, but here's what we think happened.
First the setup. Each cell has a positive and negative terminal, 12mm in diameter and tapped with an M6 thread, and made of aluminium. On top of these terminals are rigid copper bars 2.5mm thick which link both the two cells in the pair and the sequence of four pairs; the center of each bar has a thin (<1mm) layer of insulation. The bars are fastened to the battery terminals with an M6 machine screw in A2 steel, locked tight with a Nordlock washer - I think these are steel with a zinc protective coating. When I set the cells up, I bolted them together in pairs, gave each pair an initial full charge, and they've been paired ever since.
Despite being connected by thick copper bars, on measuring the pair I found one cell at 3340mV and one at 2460mV. It's clear the copper wasn't actually touching both the terminals of the cells in this pair - either as a result of some sort of thermal expansion and contraction due to the large load, perhaps a fraction of a mm shift in the cells as the battery moved around, or perhaps they had never touched at all.
The copper bars are completely rigid and although they were bolted down firmly, the layer of insulation in the middle along with the edge of the battery case could have kept the copper bar slightly suspended and prevented it from making a clean connection to the terminal. The two terminals would still have been connected through the steel bolt and the washer, and we measured the resistance of the washer at 10Ω, I suspect due to the coating. So there was some electrical conductivitiy, but there should have been at least 100 times less resistance.
So what have we learned?
Quite a bit, as it turns out
- The rigid copper bars supplied by the manufacturer are not a good choice to connect cells together. The folk at EVTV use tinned semi-flexible copper straps, which are a good idea but hard to source. An alternative would be 1mm thick copper washers between the battery terminals and the copper bars, to ensure the bars will only be in contact with the terminals, not the battery body.
- The interconnection between cells must be done through copper connected directly to the terminals. Do not rely on the bolt or the washer, as the conductitivity of these is several orders of magnitude worse. In fact the bolts should really be electrically isolated from the battery terminals to prevent galvanic corrosion between the aluminium terminals and the steel bolts; I'll be changing the A2 bolts to A4 steel and coating them in Loctite before reassembling the battery. Their only purpose should be to clamp together the various copper bars and terminal connectors on the wires.
It is not enough to have one cell-monitor for each group of parallel cells. If a disconnect
like this happens within a group of parallel cells, the results will depend on which
cell-monitor board is on.
- If the disconnected cell is the one being monitored, the cell will never show any change in voltage as it is not part of the circuit. The other cells in this group could drop below their minumum voltage, or be charged above their maximum, resulting in a dead cell or, in an extreme case, fire.
- If the monitored cell is not the one that is disconnected then the battery will appear to be working as normal, but the missing cell will mean a severely reduced capacity. Identifying the cause will be almost impossible without a serious imbalance like this occuring.
The bottom line is to be aware that this failure mode exists in any BMS with one cell monitor per group of parallel cells; and this is a very common design.
- Initial full charge of each cell should be done per cell, not per group of cells. Although the group of parallel cells should in theory equalize to the same voltage, this process will take a much longer time than you might imagine. The reason is the voltage difference between the cells will typically be very small: if the fully charged cell is at 3380mV and its lesser-charged partner at 3330mV, the 50mV difference means very little current will flow from one to the next.
What happens next?
Well, first, this pretty much rules out me getting on the water before winter, which is a pain in the arse. The reason is I now need a complete redesign of the battery system, as some other things came up during testing:
- A shunt isn't a very accurate way to measure current. After turning over the engine, the current draw on the system should have returned to its previous level, but was in fact showing itself to be about 2A different. I put this down to the change in resistance due to the heating of the shunt resistor - it's not good enough, so I'll be switching to an ACS756 hall-effect current sensor, which I've tested before with some sucess.
- The more reading I've done, the more I'm convinced constant cell balancing is a must, even for systems that are relatively lightly loaded like a house battery on a boat. LiFePo4 cell balancing is a field that is advancing every few months - there are now about a dozen different techniques I'm aware of. I'll be attempting to implement one of the best of these (a cell-to-cell balancer using a quasi resonant LC converter and boost converter, which is as complicated as it sounds), in the next design. This should result in very, very high efficiency during balancing.
- The rectifier on a Yamaha F9.9 is regulated - I was unclear on this, but bought the service manual and it looks like this is the case with any Yamaha outboard with electric start. Output is regulated at 13V regardless of engine speed, which is 3250mV per cell. This means the engine alternator will be able to prevent the battery from going completely flat, but will not be able to add any significant charge to the battery at all. If I'm to charge the batteries to above about 25% while at sea, I will have to use something else: solar, wind, towed impeller or methanol. Solar was always on the cards, but I was putting it off - no longer an option.
All up, a bit of a shit week, but finally accepting I'm going to miss the water this year might allow me to slack off the pace a little.