That’s LiFePO4…

July 14, 2011

The Lithium upgrade

Some of you who follow my blog might have been wondering where I’ve gotten to this last few weeks. I’ve not been idle on the upgrade front, that’s for sure. In fact I’ve been spending almost all the spare time I can get on my biggest project to date, – that is the much covetted Lithium cell upgrade. After many weeks and about a hundred hours of work, I’ve finally succeeded! :D

As you can see, I have a fair amount of catching up to do on my blog and guide, as I’ve only recently began to start work documenting this. A lot was involved: In addition to designing and building the Lithium (LiFePO4) pack itself, I also had to come up with a way of charging it.

I could have gone for an off-the-shelf Battery Management System (BMS), but I’d heard to many horror stories about people ordering these things only to find themselves little support or documentation to assist them with actually getting it up and working. Instead, I elected to build one from scratch from a much used, well-known and proven design conceived originally by Gary Goodrum and Richard Fechter. The Goodrum-Fechter BMS is not available pre-assembled, but comes as a bare circuit board whose parts must be acquired and assembled separately. This meant a lot of work, but I opted for this route because of the strong support that this design has from the e-bike community, the good documentation that comes with it to help users assemble and test it, and the solid forum support offered by the board’s designers.

As detailed in my previous blog entry, the pack I eventually settled on was a  24s2p, 76.8V nominal, 86.4V charge voltage, built out of 48 Headway LiFePO4 38140S 12Ah screw-tap cells. It weighs just over 19Kg, versus the 48Kg of my SLA bank, so I’ve shed nearly 30Kg!

My 76.8V LiFePO4 Lithium pack

I elected to mount mine vertically. It wouldn’t fit flat as it was too wide, so I sacrificed some seat space as before. The advantage of this arrangement, though, is I can more easily remove it if needs be. There is also just enough space down the side to slide the BMS down alongside it so it is flush with the battery case. Not only is this a convenient arrangement that keeps everything nice and tidy, but the side of the battery box can act as an extra heat sink to protect the BMS from getting too hot.

Here’s the final arrangement. I had to take a chunk out of the seat, but I’m quite happy with the result. My Lyen’s controller is en route back to me, so I’ve only had chance to use it with my cheapo controller, but I’m delighted by the improvement to its all round performance.

The final assembly

The reduced weight and peppier voltage levels means it accelerates and handles better, and doesn’t have to struggle so much with hill starts. I’ve yet to put it to a proper range test, but so far it shows every indication of having better range than it did before. Though the Lithium bank is a slightly lower capacity (24Ah) than its previous 28Ah SLA bank, much of that power was previously wasted hauling the much heavier ‘lead sled’ into motion from stops, and fighting gravity up every incline.

All that remains to be done here is a bit of weatherproofing, to protect the bank and battery compartment from the wet when poorer weather arrives. I might also add a couple of brackets to more firmly secure the pack in place, and prevent it from shaking about on rough roads.

The Charging System

Unlike an SLA battery bank, which can just be series charged with few balancing problems, a lithium bank must have its charge levels very carefully managed. Any power from a charger unit therefore needs to go through a suitable BMS which manages the distribution of the power to the individual cells of the bank.

Most of the work I did involved constructing the aforementioned Goodrum Fechter ‘Zephyr’ unit. The architects of this board simplified the process as much as they could by giving electronic parts lists (BOM) that could be drag-and-dropped into the online shopping cart of the U.S. components store Mouser.

The completed unit is shown below. It has a perspex lid so that the cell-level LEDs are visible and can be monitored if needs be. In everyday use this isn’t really necessary – a single charge-status LED on the end tells you when it’s finished, but it can help identify bad or weak cells if problems develop with the pack. It also has an end of charge (EOC) adjuster that let’s you fine tune the point at which the charger cuts off and finishes the cycle.

The BMS has three cables. The grey Anderson connector and the 24-pin ATX connector go to the battery pack. The first serves the main terminals, and the other manages the individual cells. The black and red Anderson connector is connected to the charger itself. I’ve used good, thick 1mm2 tap wires to keep resistance down and allow for higher charge rates.

The charger I’m using is a BMSBattery EMC-900, which supplies the 86.4V charge voltage at a rate of 9A. However, I’ve wired the BMS up with nice, thick wiring so that I can take advantage of the 20A charge rate that the GF Zephyr is capable of handling, if needs be, and can always just get a bigger charger.

I’m delighted with the improved charge rate, which is almost four times what I got with my 2.5A SLA charger. In my latest test-run, I ran the bike for ten miles and then charged it up again. It was fully charged again in just under an hour, as compared to the 4 hours or so it would normally take under the old SLA system!

Here’s the final arrangement for charging. The BMS fits snuggly down the side of the lithium pack and is connected in between the charger and the battery pack.

The Lithium pack on charge via the BMS

The BMS can also be used as a cell-level LVC cutout system via throttle pass-through connectors which power down the throttle signal if any cell voltage becomes dangerously low. I haven’t got that up and running yet, and am simply monitoring voltages manually with a multimeter for now. I may go with an alternative, and simply use the BMS for charging, rather than having the BMS permanently attached.

Even at 9A, the whole assembly usually runs cool as a cucumber. The only time it gets even slightly warm is at the end of the charge cycle. This is when the ‘shunting’ occurs that protects individual cells from overcharging and bleeds off overflow from already charged cells through big resistors on the back of the board.

Over coming days and weeks I’ll continue to catch up with documenting the details of the upgrade, including the BMS build, testing, and troubleshooting. I’ll also detail setting up the charger to get the exact voltage I needed to make it work just right with the BMS. In the mean-time I’ll be putting the pack through it paces witha more extensive range test, and also seeing how it performs once I’ve got my sporty Lyen’s controller back in place!

Advertisements

Building a LiFePO4 Lithium Pack

July 10, 2011

The 76.8V LiFePO4 Lithium Pack

Parts

Tools

  • A Phillips head screwdriver
  • Crimping tools
  • A soldering iron and solder (to reinforce crimps)
  • Heatshrink

Why Upgrade to Lithium?

Upgrading to Lithium is fairly expensive, and the process of putting together a suitable pack and charging system is time-consuming and often difficult. So you might ask why such an elaborate upgrade is worthwhile.

Performance

A typical 72V SLA bank composed of six 12V 28Ah batteries weighs around 48Kg. A similar rated Lithium bank weighs 60-70% less. The pack I built ended up weighing in at around 19Kg. At almost 30Kg lighter, acceleration is greatly improved, hill-climbing is less of a struggle, and the lower mass consumes less energy, adding to range and lowering running costs.

Discharge characteristics.

Whereas an SLA bank will show a steady decline from full charge and across a significant range of voltages – gradually ‘winding down’ as it runs out of power, Lithium cells deliver a fairly steady voltage for a long period before fairly abruptly falling to below a safe operating level. This means they have to managed more carefully, but do deliver more even and consistent power throughout a discharge cycle. It won’t become sluggish and fade out half way through, like SLAs do.

Economy

An SLA bank will typically only give you around 500 charges before the batteries need replacing. A well-managed LiFePO4 Lithium bank will give you 2000 charges before it needs replacing. A typical SLA bank will cost around £250 to replace, versus the £650 cost of replacing a similar LiFePO4 bank: Four times the lifespan, but two-and-a-half times the cost. Furthermore, the price of Lithium cells continues to drop, making replacement costs even cheaper.

Designing the Pack

The first stage of upgrading to Lithium is to build the actual battery pack. If you’ve been running off of a 72V system, then it makes sense to have a pack that delivers a similar voltage but preferably a little bit higher. This will give you better performance, but still enable your controller and other bike components to work as it would off of a 72V system. A nominal 72V system operates typically between 68V and 82V  depending on the charge state of the batteries, so our “72V” rated controller and converter does have a certain amount of flexibility in the range of voltages that it can work with, so it makes sense to ‘push the envelope’ a little to get more power.

The LiFePO4 cells I wanted are nominally 3.2V per cell, with a maximum charge voltage of around 3.6V per cell. The cells will charge higher than this, but any more will start to take a toll on its lifespan in terms of the number of charge cycles it will give. 3.6V per cell is the generally agreed to be a good figure to work with by way of the upper limit.

With this in mind we need a bank that is ideally a good “square” number that delivers a voltage slightly better than 72V nominal. In the end I decided upon a 24s2p configuration, that is, a pack comprising 48 cells divided into 24 pairs connected in parallel. Connecting cells in parallel, remember, doubles their capacity while keeping their voltage the same, whereas connecting cells (or cell-pairs) in series multiplies their voltage while keeping the capacity (amp-hours) the same.

The cells I chose were Headway LiFePO4 38140S screw-tap cells. Nominally 3.2V, at 12 Ah each. By connecting them into pairs, my bank would have a 24Ah capacity for the full voltage, and by stringing together 24 in series, I would have a nominal pack voltage of 76.8V, high enough that I could get a little more power, but not so high that my system couldn’t cope with it. The voltage of the bank fully charged would therefore be 86.4V, just 4V or so higher than that of a 72V SLA bank. The pack number was also a good ‘square’ number that gave me a 6 x 8 pack of cells that was of the right dimensions for fitting onto my scooter.

Finding a LiFePO4 Supplier

This can prove a little challenging, as the Western world now seems to be lagging significantly behind Asia in terms of electronic technology and have been remarkably slow to show any interest at all in supplying or reselling EV components or equipment. So like most e-bike controllers and chargers, China seems to be the best place to get them from.

You could go straight to the manufacturers: Zhejiang Xinghai Energy Technology Co., Ltd. , however I found that quotes from their reps were higher than those of many resellers who can be found hanging around on Alibaba.com. EV forum users have frequently gone to EVAssemble to get their LiFePO4 cells. There have been complaints about their delivery times and customer service, however they have proved to be reliable in providing quality cells.

In the end, I ended up taking quotes from a number of resellers, the cheapest of which were from ‘Michael’ of EVAssemble and ‘Michelle’ of a little-known reseller Sports Discovered Ltd. However I found Michael to be so slow and disorganised that Michelle won by default, with a slightly lower quote than the one provided by EVAssemble.

I ended up ordering 50 cells (so I would have two spares), and specified the number of holders, and 2-cell/4-cell connector plates to be included.  I strongly advise anyone undertaking this projecy to ask for spare plates and holders. The plates can be ruined by shorting accidents, and sometimes vary in quality (a couple of mine were slightly warped). The holders can have their tabs broken easily, especially if you are drilling little holes in them to mount perspex siding like I did.

My order from Michelle came to around $1,040 all told for 50 cells, with connecting plates and holders. This included delivery by FedEx. My delivery – consisting of three big boxes – arrived within 6 days of payment. It was all professionally packaged with all the right numbers of the requested bits included. Importantly, they had been ‘discrete’ on the import documentation in detailing the value of the goods…

The packages had seven boxes like this one, with my cells all snuggly packaged in little foam compartments.

The holders and connecting plates were included as requested. The holders are plastic brackets that go onto the ends of the cells, and let you fit them together like Lego into any size or shape of pack that your want. The connecting plates let you cross-connect them to get the configuration of bank that you want to deliver the desired voltage and capacity. They are simply screwed onto the ends of the cells using the little bolts that come attached at either end.

Once everything was unpacked, it was time to set about putting together my bank. After measuring up my bike, I had settled upon a 24S2P configuration (24 two-cell pairs connected in series), forming a grid of 6 x 8 cells.

A big, completed pack looks quite complex and a little overwhelming, but it’s quite simple when you break it down into its constituent parts. The following four cells, connected at the bottom by a square, 4-cell plate, form two parallel cells in series.

Below, the same pattern is extended through eight cells, forming a quarter of what will eventually be the complete pack. You can think of this assembly as four ‘cells’ (parallel cell-pairs) in series, with the direction of current alternately going up and down from left to right.

At the end of the row, the series ‘bends’ round to go in the other direction back towards the left again.

The assembled bank looks like this, with the master positive front left, and master negative front right, the nominal voltage of the full bank being 24 x 3.2 = 76.8V.

The most tedious part of the job, though, is yet to come. Lithium packs need to be carefully balanced using a  Battery Management System (BMS). This means that each of the 24 cells needs a tap wire to run to an external connector, which can allow a BMS to keep the cell voltages properly balanced during charging. I decided to use 1mm2 copper core wire as this would allow for higher rate charging and minimise resistance in the system.I used a sequence of four colours, blue, green, yellow, and white to make it easier to keep track of which wire went to which terminal and prevent misconnections. These wires, though, wouldn’t fit in the 25-pin serial connector that I was planning to use, so instead I settled for an ATX connector of the type typically used by computer power supply units. The connectors and pins are available from specialist PC component retailers.

Figuring out how long all the tap wires would need to be for each terminal needed a little thinking about. I started from the standpoint of all the wires threading their way to the positive terminal of the bank. Then I added 240mm for a final length that would run off of the bank and to the ATX connector. Each pin on the connector would correspond to the positive terminal of each consecutive cell, starting from pack’s negative pole and going upwards. So the first pin would lead to the positive of the first cell (+3.2V), the second, to the positive of the second cell (+6.4V), and so on. The final pin would connect to the master positive, reflecting the full bank voltage.

Once I’d cut my wires to the required length, I crimped each of them to a male ATX pin, secured the connection with solder, and clipped it into the relevant hole of the male ATX housing. It’s advisable to have a few pins spare, as it’s easy to damage or oversolder them.

Once the connector was completed,  I set to work on the other ends: Each was crimped and soldered to a ring connector. I put a short length of heat shrink down each wire before I soldered on the connector, so that the joint could be insulated and further strengthened with this afterwards.

The connectors were then simply secured to each cell pair using one of the screws at either end.

I wanted the final unit to be safely housed, so I had two pieces of perspex cut to mount on either side to protect the connecting plates that carried the charge. I carefully drilled small holes in some of the ‘legs’ of the cell holders. I started with a small bit to drill a guide hole, then drilled it out with a slightly larger, 2mm bit.

I drilled corresponding holes in the perspex and secured it to the legs on the cell holders. I kept the green backing film on this until I was properly finished.

In addition to the cell tap wires, a BMS also requires two, thicker main terminal wires to be connected, through which it charges the main bank. For this I used 6mm2 wire crimped to an Anderson connector. The wire wasn’t thick enough to fill the crimps for the Anderson, so I used 10mm lengths of thick copper strands to fill it out to allow a nice, solid crimp.

The final result is a LiFePO4 bank that is ready for installation. The tap wires and main connectors are for the BMS to manage charging and for this or another unit to deal with LVC (low-voltage cutout) at the cell level. The main power wires for the bike can be added secured to the main bank terminals via another pair of ring connectors. This is made easier by the fact that each cell pair has two screws, so there is not need to pile multiple connections onto a single bolt head.

Ready to go, the 76.8V LiFePO4 Lithium pack

The next stage is to assemble a suitable BMS and charger combination to manage charging and LVC warnings. These can be bought commercially, but because I wanted to be able to repair and maintain my own, I elected to build one from scratch to the well-known Goodrum-Fechter design. Details of this in the next section…