ElectricSubaru


Last update: Dec 26 2016

Motive Battery Pack

Batteries are stowed in two groups, both in the back of the car. The main battery box is below the trunk floor, where the spare tire normally hides. A smaller secondary rack is located above the trunk floor. If not for the rear sway bar, the main box could have been bigger and held all the cells.

Cells

Propulsion power comes from 50 CALB CA100 cells. The nominal pack voltage is 160V, capacity is 16kWh. The cells can supply 300A continuous or 1000A for 30 secondsPDF. Total weight of all the cells is 370lb (170kg).

CALB CA100 Cell

I carefully sanded the terminals flat. Some were misshapen and few were corroded.

Main Battery Box

The main box is a trapezoidal shape containing 36 cells. The trapezoidal shape fits between the frame rails, and leaves space at the rear corners for passing cable through the side. The box is made from aluminum plate (0.125in thick) welded at the sides and bolted to a flange of structural steel angle (1.5in x 1.5in x 0.125in). The box was fabricated by Johansen Mechanical for $300.

Lower box fabricated Lower box fabricated Lower box flange Lower box flange Lower box painted

The box floor is strengthened by aluminum T-bar (2in x 1.25in x 0.125in) from Orange Aluminum. The box sides are filled with plywood wedges. The wedges are just over half the height of the cells so that they cannot tip.

Lower box internal structure

The box flange rests on two thin strips of MDF, contoured to fit the profile of the top of the frame rails. Six bolts secure the box to the frame rails, three on each side.

Cross section

The box lid is polycarbonate with five straps screwed to the flange. The straps hold the cells in the box in case the vehicle rolls over. The straps are made of wood on the underside of the polycarbonate and a strip of aluminum (1in x 0.125in) on top.

Lower box installed Lower box from outside

I used the Harbor Freight Electric Body Saw to cut the trunk floor. It worked best with the blade teeth pointing up.

Body Saw Body Saw

Trunk Rack

A rack in the trunk behind the rear seat holds 14 cells, the charger, and the battery disconnect switch. It is constructed of aluminum angle (1.5in x 1/4in) and is bolted at the four corners into rivet nuts in the frame rails. Some wood blocks hold the rack at the height and angle needed to clear a hump in the middle front of the trunk floor. The rack is on top of a section of the original trunk carpet.

Trunk rack Trunk rack Trunk rack Trunk rack Trunk rack Trunk rack Trunk rack

The sides of the cells are covered with foam board for insulation from heat given off by the charger.

Trunk rack

The lid is made of aluminum angle (1.5in X 0.125in) with a polycarbonate top. The polycarbonate is between the terminals and the aluminum. A ring of wood around the edge spaces the lid so that the polycarbonate clears the terminal bolts. The lid is held down by two Keeper 85512 ratchet tie down straps, rated at 1200lb breaking strength (each). Two cutouts for power connections are insulated with rubber grommets cut into a 'C' shape.

Trunk rack

Connectors

The connector straps are from EVTV. They are very thick and wide, with approximately the same cross section as 4/0 AWG cable and are available in lengths for both side-to-side and edge-to-edge configurations. They have an oval hole at both ends that has a smaller terminal contact area than would a round hole. Adding a copper washer between the battery terminal and the connector strap improves the amount of contact area. The washers are Dorman 725-002. The straps come with M8x15 stainless bolts. With the extra washer, M8x20 works better.

EVTV Battery Connector EVTV Battery Connector with Copper Washer

The diagram below shows how much the contact area increases by using the washer. The contact area without using any washer is shown yellow. Adding the washer, this contact area increases to the full size of the terminal, shown green. The contact area above the washer is violet, which is much larger than the terminal.

Surface contact

The copper washers that make contact with the aluminum battery terminals are tinned with Tinnit Tin Plate to reduce galvanic corrosion.

Tinnit Tin Plate

What! No BMS?

There is no battery management system. The argument for using a BMS is based on the idea that cells self discharge at differing rates. This is true of lead acid cells, but extensive experimentation and testing performed by Jack Rickard and John Hardy indicates that self-discharging of LiFePo4 cells is negligible. The exception is that a small percentage of these cells have a soft short defect that causes self discharging. I bottom balanced the cells as advised by Jack Rickard. I plan to manual check the cells periodically.

I first charged each cell to 3.5V, then discharged to 2.75V. The charging step ensures the cell is fully formed. The heavy lifting was done by a Junsi iCharger 306B. The final discharging was done manually with a 0.15 ohm 250W resistor, a 1 ohm 50W resistor, and a voltmeter.

Junsi iCharger 306B Power Resistor Multimeter

One of my cells did in fact have a soft short. These defective cells can easily be detected by checking their voltage after discharging. If the voltage drops over night, there is a soft short. The bad cell was replaced.

Bad Cell

Cell Voltages Over Time

In Apr 2015, after about a year of operation the cell voltages were measured at the bottom and top of their range. During this one year interval, the pack was partially discharged about 100 times. The car was driven mostly short trips. Therefore, the pack state of charge generally remained high. Cell voltage measured after discharging and resting a cell is an accurate measure of state of charge. However, during charging the voltage does not accurately indicate state of charge. In spite of this inaccuracy, the voltage is what is used to decide when to stop charging, with a goal of not exceeding 3.5V for any cell. Cell 48 was replaced and so is not included in all data sets.

The bottom end measurements shown in the chart below were taken after the pack was discharged and rested repeatedly until the lowest cell reached a stable 2.75V, the voltage at which the pack was originally balanced.

The bottom end voltages show small variations, which means the state of charge of the cells shifted slightly relative to one another. The cells were originally balanced to within 0.02V but a year later span a range of 0.07V. These differences in voltage correspond to a very small difference in energy. Jack Rickard recommends bottom balancing to within a range of 0.10V (2.75V±0.05V). The changes could be due to self discharging, but that idea contradicts what is known about these cells both experimentally and theoretically. Therefore the changes are more attributable to the cells charging and discharging at slightly different rates due to variations in the cells.

Cell Measurements

The charts below show the top end measurements which were taken while charging, near the end of the charge cycle. The pack was rebalanced to 2.75V and the measurements repeated. At the end of the charge cycle, the charger drops the current from 14A to 5A. These measurements were taken at 8A.

After rebalancing, the top end voltages have a larger range and higher maximum. In particular, the three weakest cells 1, 3, and 27 were further from average after rebalancing. The general indication is that most of the cell's full state of charge tended to synchronize to the rest of the pack during the year. Keep in mind voltage during charging does not accurately indicate state of charge and differences could be explained by random variations and coincidence. Within this sample, however small, the pattern is consistent.

Cells 37-50 are located in the trunk above the floor, near the heat of the charger. Most of their top end voltages are much lower than average, so they may have been affected by heat. I added a second trunk exhaust fan after seeing the data.

Cell Measurements Cell Measurements

One point of interest is not shown in the data above. At the moment before the charger shut off, the maximum voltage on cell 27 was 3.461V before rebalancing and 3.522 after. This 0.06V difference is significant in that it easily makes the difference between keeping under the 3.5V limit and going over.

My conclusion is that the decision not to use a BMS is sound. The original justification for not using a BMS was an expectation that the relative states of charge would not change. The cells did in fact change, however the weakest cells tended to converge rather than diverge. My main concern is to avoid overcharging the weakest cells and it appears the likelyhood of that happening goes down over time.