[tiger82]

Our initial Kia Soul EV Pack performed flawlessly and was able to power our Tesla drive unit at up to 375HP at the rear wheels. However, 27kW-hr was not enough for our 20 minute races. We knew this when we put it together, but when we designed the car in 2015, there are no other packs that fit our cost, performance, weight, and size criteria.

Earlier this year, we looked at both the 60 kW-hr Chevrolet Bolt pack and the 2018 40 kw-hr Nissan Leaf pack. We initially decided on the Leaf pack because of the friendly form factor of the modules, but the launch was late which meant that it would be difficult to get a salvage pack in the time frame we wanted. We changed our minds when we saw that Bolts were readily available and the Bolt pack offered us the possibility of using liquid cooling.

The entire Bolt pack is 60kW-hr and 900 pounds. The configuration is 3P96S arranged in 10 modules. Eight have 30 cells, and two modules have 24 cells. Cells are arranged in groups of two so that one surface of each cell is touching a cooling channel. We decided that a 2P96S configuration would be ideal. Nominal capacity of two parallel cells is about 115 A-hr.

We purchased a 2016 salvage vehicle out of California via an IAAI auction and proceeded to strip it down. Unfortunately, when we disassembled it, we saw that the tabs were ultrasonic welded after mechanical assembly. This meant that we had to cut the cells apart and re-join them. The good news was that the carriers that held the cell pairs were symmetrical which allowed us flexibility upon re-assembly. The cooling scheme was also very simple. Aluminum cooling plates sat under the modules and contact with the cell sides was via a simple sheetmetal “C” section.


To fit into the envelope of our old battery pack, we decided to utilize six modules stacked three high. We added an extra cell per module to bring the total cell count up to 96. The module dis-assembly took many patient hours with a Dremel cut-off tool. First the connectors joining the parallel groups had to be cut, then the 3rd cell had to be cut away.


This left a number of cell pair that were already connected, and singles which had to be flipped around and re-connected. We designed custom copper connectors that were CNC cut and bent to re-join the cells via a bolted connection. One benefit of using our own connectors is that we were able to use thicker copper than the factory LG cell connectors.


We did careful testing to make sure our bolted joints had no greater resistance than the factory joints. Voltage drop at the joints was measured under load using an AC power supply.

Because of our different pack form factor, we were unable to use the factory cooling plates. We designed our own that that doubled as mounting plates for the modules. The cooling plates were made by cutting a serpentine fluid path out of foam, then sandwiching the foam between sheets of aluminum. Chilled water runs in the channels The ends of the cooling plates doubled as a mounting platform for our contactors, pre-charge relay, master on/off, and Orion2 BMS. We utilized four cooling plates to cool both the top and bottom of the cell. The factory configuration provides cooling on one side only for most of the cells.

The battery box cover was constructed of flat composite panels constructed of carbon-kevlar cloth wrapped around a foam core. The box needed to be light, yet provide a high degree of ballistic protection. We did a simple vacuum bagged, wet lay-up, using a piece of melamine as the mold. One of our sponsors, Calmar fiberglass, let us use their oven to cure the high temperature epoxy.

Total weight of the pack with all electronics, covers, and mounting straps is about 550 pounds. Overall size including the electronics is approximately 41”x16”x16”. This is similar to our Kia Soul pack, however now we have 13 more kW-hrs!

The cooling plates appear to work very well. We will use them to pre-cool the pack before a race and also to keep temperatures down during a race. During testing we were able to drop the temperature of the pack 10-15 degC/hr. For the 2019 race season, we will implement an on-board AC compressor/chiller system.

For our initial tests, we ran the vehicle on a chassis dyno with no battery covers and used an infra-red thermal camera to make sure that we didn’t have any un-wanted joint heating or cell heating. Max power at the rear wheels was about 10-15% lower than our Kia pack. This will not affect us for road racing as we cannot run the 20 minute races at max power anyways. This translates to about a 9C rate for the LG Chem cells (vs 15C for the SK Innovation cells in the Kia pack).

We did some testing at our local race track and preliminary results showed that we have the extra range were looking for. We were able to turn regenerative braking off which meant that we could drive the car harder into corners. Lap times were accordingly faster. The first actual race with the new pack will be in mid October.

 

[Nuts&Volts]

Awesome post!! Love the detail and Love that you made a big modification pretty simple. Are busbars you made hanging off the tabs or are they mechanically supported too?

So your max power limitation is battery voltage sag or temperature? I’ve wondered what the power limit of the Bolt cells are


Sent from my iPhone using Tapatalk

 

[tiger82]

There is no further mechanical support behind the copper cell connectors. When bolted together, the assembly forms a very rigid structure.

As far as max power, that would be governed by the LG cell chemistry and the internal resistance. Remember, we are only using 2 parallel cells instead of the 3 for a stock pack. I would suspect that a stock pack would be able to put out more current than ours.

Although battery heating is a problem for us, the internal resistance of a Li-Ion cell actually decreases with temp. The maximum battery allowable battery temperature during a race is a somewhat arbitrary self imposed alarm limit.

 

[BuildMore]

Amazing work, you're an inspiration!

 

[tiger82]

Just an update on our experience with the modified Bolt pack in our Tesla Cobra Race Car.

The pack was finished in the fall and we ran several practice, qualifying, and race sessions. Despite the water cooling, the pack ran much hotter than our old air cooled Kia Soul pack. To work around this, we tried pre-cooling the pack using ice water and running ice water through our cooling plates but were unable to keep the temperatures down to acceptable levels (<70degC) at the end of a 20 minute race. On the positive side, we had enough range to finish our 20 minute races.

Over the winter, we built a single cell test bench. We used a resistive load sized to totally discharge the cell (4.25 to 2.75 volts) over a 20 minute period to simulate race conditions. We ran dozens of tests to see if we could improve the cooling. We tested the Bolt style indirect top and bottom cooling plates, Direct water cooling plates on the cell sides, and 100% immersion in mineral oil. We found many interesting results.



We found that the stock .040” aluminum plates that sandwich the cells cannot transfer heat fast enough for our application. (This was verified via heat transfer calculations). This is despite using top and bottom cooling plates (the stock Bolt arrangement only uses a single plate). We had much better results with direct water cooling plates (similar to Chevrolet Volt) between each cell. We also found that pre-cooling the pack was absolutely the wrong thing to do. We found we could increase voltage, usable energy, and decrease end of test temperatures by pre-heating the pack!
Immersion cooling also provided very good results due to the very even cooling provided. However, we discarded this concept due to the extra weight and liquid tight enclosure requirements.

As a result of the tests, we were encouraged enough to continue working with the Bolt pack for now. The compact size and light weight of the cells were major decision factors. There was some consideration to using two Gen2 Chevrolet Volt packs as many in the DIY community have already done, but we were not thrilled about the extra weight (about 200 pounds), and the very low energy density which would make fitting the cells into the car very difficult.

We designed and built a series of direct water cooling plates that are attached to each of the stock Bolt cooling fins. They are made up of laser cut serpentine water passages in a .125” piece of aluminum plate sandwiched with an aluminum skin. Overall thickness increase for each cell is only 3/16”. This will make our pack 6 inches longer and 50 pounds heavier which is an acceptable trade off.





Externally, we are adding an inline heater from a Tesla Model S. Current strategy will be to pre-heat the pack to about 40C, then only turn on cooling when the pack temperature starts to rise about 50C. Initially, our cooling will consist of a radiator and ice bath. We will also test the use of a Tesla AC compressor/chiller at some later date.

First test day is in mid April, and the 1st race of the season is in mid May.

 

[Yabert]

Really nice details! I had no doubt that there is no miracle around the high energy density of the chevy Bolt battery
The power density is way lower than some other cells found in salvage electric car.
Thanks also for the nice details about the Kia soul battery. Impressive spec.
I hope you will be able to find a better / simpler battery in the near future to continue to beat all those gas burner. You will be the perfect example that the ICE are obsolete and the inspiration for others racers.
Continue your good work.

On a side note, do you think / try to repalce your 96S configuration by a 108S configuration to have a bit more peak power still by draining lower amps from the cells?
Is this is possible with the Tesla unit?

 

[tiger82]

Yes, we've thought about extending the serial string. The question is at what point will the Tesla controller stop working because it believes there is an error condition? As soon as we are out of the pits and the race starts, this will not be a problem as the voltage will soon drop under load. Adding 12 cells may be within acceptable tolerances. We will probably temporarily wire in some spare cells and try a dyno test later in the year. We also need to verify that the Tesla chargers that we are using (with Damien's control boards) can charge the higher voltage pack.

 

[boekel]

Thanks for the update, and thanks for sharing the results! very interesting.

 

[remy_martian]

Nicely done & thanks for sharing your testing results.

A couple of questions. When you did the immersion cooling testing, did you separate the cells to enable liquid contact with the full surface of the celll or use the Bolt's original C-channel?

Also, on your 3 piece cooling plate sandwich, what method did you use to join the plates together to make them both mechanically robust and leakproof?

 

[x.l.r.8]

That was my next question, I have Damien's board in a 72A gen 3 charger, if i decide to go past 96S at what point am I going to hit errors. I have Toms BMS so it will run additional slave boards and I have a few extra so I could add 6-12 additional strings.

 

[Hollie Maea]

How much sag do you get on this pack under load?

 

[tiger82]

The immersion cooling test was done with a single cell pair. They were not split, but the separator foam between the cells would have allowed fluid to contact the insides of the cell.

The cooling plates were glued together using Black Permatex silicone. Each cooling plate was then pressure tested to 5 psi (working pressure is about 2 psi).

Voltage sag was high. We were seeing 40 volts when pulling 200kW out of the pack. If you use the full pack (3 parallel cells), sag should be less

 

[joekitch]

hmm, been wondering why so few people seem to be looking at Bolt modules for conversions, i guess needing to make your own cooling plates adds complexity, weight, and packaging volume, and the lack of power density isn't great if you wanna GOFAST on the street.

or am i misinterpreting this, and people really should be using Bolt modules instead of Volt modules for their conversions?

 

[TheLifeHacker]

Truly appreciate the details. We are doing a 911 conversion utilizing the Chevy Bolt pack with a Tesla drivetrain. We went back and forth with Tesla battery modules v. Chevy Bolt. However, we found a sweet deal on the 2017 Bolt pack that we couldn’t pass up. I hope we made the right choice. I’ll post more after testing.

 

[Frank]

Was it hard connecting the coolant passages together? Did you use the Permatex here also? Thanks for the detail.

Sent from my SM-T380 using Tapatalk

 

[TheLifeHacker]

In regards to the Chevy Bolt Pack tension rods. Do you recall amount of torque that you utilized or the original torque?

 

[tiger82]

We have found that the best procedure for assembling the pack is to use minimal compression. With too much compression, we can see decreased voltage isolation. It appears that surface currents can be transferred from the mylar cell pouch to the metal covers. If you are not modifying the pack as we did, then the interlocking plastic carriers should set the maximum amount of compression possible.

 

[tiger82]

Quick Update: We have been running the pack this summer with our new individual cooling plates. They appear to work well. We are able to keep pack temperatures down using only a radiator. We also have an in-line ice box, but have not had to use that.

Pre-heating the pack certainly helps performance for the first session of the day. We are still not happy with the amount of voltage sag with the pack though as it limits the usable capacity of the pack. For 2020, we are thinking of going to a 3rd iteration of battery pack that can meet our criteria of high C rate, high capacity, reasonable weight, and water cooling. We are looking at Tesla Model 3 modules as one of the possibilities, or a complete custom pack of our own design using 21700 cells.

 

[dwolvin]

That is an amazing build log and design! Thank you for sharing, can't wait to see what you come up with after the lockdown.

 

[CantDecide]

This is incredible work, I'm impressed you put so much engineering effort into this!

I have a quick question about the conversion from 3p to 2p configuration on the cell groups - If I understand correctly, you split the modules into separate cell groups, then cut one of the parallel cell pouches off each cell group, so the groups went from three to two pouches. Then you took the cut off singles pouches and paired them back together to make more 2p groups?

If I understand that correctly, what was driving factory in doing all that work instead of leaving them as 3p groups? Was it just space constraints, or more surface area for cooling?

Thanks!