Lithium Iron Phosphate Battery Testing
Ian Hooper, 8 October 2007

Introduction

Lithium Iron Phosphate (LiFePO4) is generally considered to be the most promising new battery chemistry for use in electric vehicles. Although energy density is somewhat lower than Lithium Cobalt type cells (like those used in laptops and mobile phones), LiFePO4s are far more robust and offer much longer cycle life, in the order of 2000-3000 cycles. This makes them a much safer and more economically viable option.

The most common brand of LiFePO4 batteries for EV use is probably ThunderSky, who produce a range of large prismatic LiFePO4 cells between 40Ah and 1600Ah. Unfortunately due to the inherent design of the large prismatic cells, they typically have trouble providing more than about 3C of continuous current (e.g the 40Ah cells can only put out about 120A continuous). This is quite acceptable for vehicles of modest performance such as commuter vehicles, but high performance electric vehicles really need a pack which can sustain around 10C continuous.

The Contenders

Five different cells were tested, as follows:

The Test Equipment

The Monitoring Software

Monitoring Software under MacOS X Click here to download the source code

This is just a quick logging application I put together for controlling the battery cycle sequence, logging and graphing the voltage, current and temperature, and calculating overall amp-hours and watt-hours in and out of the battery. The main graph area shows voltage in white, current in cyan for current in to the battery (i.e charging) and orange for current out (i.e discharging), and temperature in red. Below the main graph is the controls, status indicator, realtime feedback of voltage, current and temperature, total amp-hour and total watt-hour in and out of the battery. Left of the graph is a CSV log of the voltage and current, in case I want to process the data in an external application, and a log of the cycle results for continuous cycle tests. The software was developed using Cocoa/Objective-C in XCode 2.4.1 under Mac OS X. Feel free to grab a copy of the software (left) for your own purposes. Disclaimer: Use at your own risk, no guarantees it won't make your computer explode, don't sue me, etc :)

Results: Discharging Performance

Since I am looking to build a high performance EV, discharge performance was the primary aspect of interest to me for these tests. (Cycle life is also very important, unfortunately it takes a REALLY long time to rack up 1000s of test cycles on a battery!) The following table shows discharge curves for the various cells at various rates:

	PHET PE-1150	Huanyu	Valence 18650	Valence 26650	A123Systems M1
1C
3C
5C
10C

Notes:

• The resolution of the ACS754 current sensor when used with the LabJack U3 was not brilliant (hence the noise on the current plots in blue), so amp-hour and watt-hour totals may not be super precise. But, they should still be a pretty reasonable representation and comparison.
• The rows are matched by approximate C rates only - since discharging was done with a resistive load and some batteries have more voltage sag than others.

Comments:

• Wow, those A123Systems cells have such flat discharge curves! Most impressive.
• The measured capacity of the Huanyu cells was a little disappointing.. I'm not even sure what they're meant to be rated at, but perhaps they sent me old samples.
• Some of the cells showed pretty significant heating at the 10C rate (my temp sensor was only calibrated up to 50°C so I can't say for sure how high they got)! If they were Lithium Cobalt cells I might be concerned, but LiFePO4s are supposed to be pretty resilient to high temperatures. In fact, LiFePO4s put out the most power when they are around 60°C. So I'm not too concerned by this level of heating, they're not about to explode or anything.
• The Valence cells both had a strange voltage taper at the end, which makes me suspect 2.0V discharge cutoff I was using might have been a little low. Something like 2.2V might be more appropriate. (You'll notice I set the cutoff to 2.2V for the 10C test on the Valence 26650, and it does look more "correct".)

Results: Charging Performance

The main reason it is useful for cells to handle high charging current is actually regenerative braking (aka "regen"). The fastest rate of charge for vehicles without regenerative braking is usually a fraction of 1C, since single-phase chargers are limited to around 3kW and most EVs have packs well above 10kWh.

Most EVs use series DC motors and don't have regenerative braking, however I believe it will become an increasingly important feature in the years to come as it improves efficiency and range by about 10%. So, it's worth seeing which cells can handle high rates of charge associated with powerful regen..

	PHET PE-1150	Huanyu	Valence 18650	Valence 26650	A123Systems M1
0.5C
1C
2C

Comments:

• I tend to think that a shorter Constant-Voltage phase (the second part of the charge, where the current is falling) is a good indicator of how happy a cell is at a given charge rate. At the 1C mark, the PHET and A123 cells seemed to be showing the best charge curves. Interestingly, the Valence 18650 is the clear winner at 2C, being the only cell able to maintain a constant current for the majority of the charge time.
• The big surprise was the A123 cell, which seemed to really struggle at 2C after such a solid performance at 1C.

Conclusion

Based on the results of the testing I have done, the cells I would currently recommend are the PHET PE-1150. They generally performed well in the discharge tests (other than significant heating - which isn't too much of a concern), and a close second to the Valence 18650s in the charge tests, being able to sustain 2C charge for a significant portion of the charge. (For what it's worth I have also found PHET to be a very professional and helpful company, and the fact that they will sell direct in single EV sized quantities is a big plus!)

Both the Valence and A123Systems performed well and seem like great cells, but their poor availability to the public (and/or high cost) makes it hard for me to recommend them. As a side note, the threshold for a better price bracket on the Valence cells was about 10 vehicles worth, so some kind of group buy might level the scales a bit.

The A123Systems cells are without a shadow of doubt the best cells tested when it comes to discharge performance. By far the flattest discharge curve and lowest cell heating. So if you really need lots of power, they are the way to go - but at 1.7x the price of the PHET PE-1150 cells per watt-hour, you'll either need deep pockets or generous sponsors!

The Huanyu cells performed OK in general and the cells are significantly cheaper, but their lower capacity unfortunately negates the cost-per-cell advantage and you'd end up with a heavier pack for the same capacity.

January 2008: Follow up testing

Since completing this original round of testing, many other manufacturers were brought to my attention. Please click on the links below for further test results:

• 4 January 2008: K2 LiFePO4 Battery Testing
• 29 January 2008: Headway LiFePO4 Battery Testing

Comments

There are no comments for this page yet.