Designing and building an expandable lithium battery system
|Test fixture takes shape|
After some arm twisting (Ray Holan bought me some batteries), it looks like the time to get serious about a real Better Battery.
Somehow a bunch of 20AH A123 cells ended up for sale from China (they say Made in the USA)at prices that make this an economical alternative to using NIMH.
Thanks to Peter Perkins early work with these cells, we know that they are more than capable of the 100A discharge and 50A charge that the Insight can dish out or require.
This blog will follow the process of testing the cells, designing an enclosure that allows active heating or cooling of the cells. developing a modular BMS system and finally integrating it into the car.
The blog starts on the bottom of the page.
How do the big guys do it
After the initial experiments,and getting a better feel for the characteristics of these cells, I did some searching for how these type cells are cooled in real world applications. Both of the pictured automotive systems have liquid cooling of the pack. A commercial vehicle that can be sold in the 50 states needs to operate over some big extremes in environment. The guys that have seen their Insight or Civic start on the 12V starter in winter ,can also be the ones that could get the inside of the car to 120F in a parking lot in summer, so to design a pack, one needs to decide if they are willing to risk battery damage at the extremes, or do we want to provide active cooling heating into the system.
|How do the big guys do it?|
So lets throw out some ideas:
The AC can generate cold, and the heater can generate hot, so an air fluid heat exchanger right in the AC/heater ducting with separate insulated fluid handling system to a heat exchanger along the sides of the pack as in the bottom photo, or possibly a ducted path for the heater / AC output to the pack blowing over the extended fins of the thermal spacers.
The two cells/ heat exchanger plate probably works fine when you can send chilled fluid through a heatsink on the plates or through the plate it self like the volt.
For an air cooled system, a plate between each cell would provide the best solution. The problem comes when you have a car at 120F and everything in it including the air and battery is at the same temperature. The opposite when the car and batteries are frozen at 10 below zero.
In a garage, AC power could keep the pack in the zone, but then you are in trouble if the car is not home and you need to overnight in cold climates.
The cells are rated to operate to -22, but based on the changes I see at 40F, I would not expect much performance in below zero temperatures, and to wait for the car to warm up and then the warm air to heat the pack, and you may be at your destination before your in the best temperature zone?
(Posted 6/29/2012 by mikey)
Hot and cold cells compared.
I charged both cells, then put one in the sun, and one in the freezer, and then once the temperatures stabilized, as quickly as I could, placed them in the test fixture and ran a medium 50A discharge to see what happened. The cold cell had a much bigger voltage drop that got better as the cell self heated, and the hot cell actually dropped in temperature before turning around and rising towards the end.
|hot and cold cells compared|
Looks like the cells like between 90-110F as their best performance zone.
Ran another run where I let the cells drop right out, and we see evidence that the self heating gets worse as the cells start dropping that last 2V.
(Posted 6/29/2012 by mikey)
Higher thermal sensitivity look at the cell heating still shows no dents
The thermal images are in a special JPEG format that stores the ratiometric thermal data for each pixel, so reprocessing the image with different thermal scales and ranges is possible.
|A better look at the cells for effects of the dents|
I reduced the thermal range represented by the color scale to show any thermal gradients across the cells, and still no sign of the dents, so I don't think it effects the cell performance at least in the limited samples I have.
The images is really 19,200 individual IR thermometers, so there is a lot of data there.
The test fixture photo with a few more thermal points
My dinner of some Polish Pierogies, the hot microwave, and a scary picture of the wife.
(Posted 6/27/2012 by mikey)
Painted cells to increase emissivity
Put 3 coats of flat black paint on the cells to make the cell surface more of an emitter, so I can get some good thermal uniformity photos.
|> 100A discharge yields 21AH|
The discharge was at the highest so far at 103A, and as the photos show. we had some red glowing load bands. Even the #10 lead wires got up to over 100F.
The cells did even better at the 103A than they did at 68A, so I suspect that warm is better than cool for getting the most energy out out of the cells.
Next test will be to freeze the fixture and see what the differences may be.
Pretty impressive cells for sure.
(Posted 6/27/2012 by mikey)
The world in IR Cool
My new instrument arrived today, and the 4 hours I had to wait before turning it on were like hell.
|Some new eyes take a look at the thermal world|
Finally the first look, and 2 hours later I finally pointed it at the A123 cell. The biggest factor in accuracy with IR imagers is the emissivity of the surface. Aluminum, Copper shiny steel and most metals are more reflective than emitting, but only a thin layer of rust, anodizing or paint is required to make the surface emit the correct black body radiation.
So I will paint the cells until the reflection is minimized, and run some more thermal test to see if the dents in the cells are a different temperature than the areas not stretched.
(Posted 6/26/2012 by mikey)
heatsink vs no heatsink comparison
I set up an interesting thermal comparison. I made two 0.061" thick aluminum plates that are the same size as the cells, and two that were 2" longer and wider.
|A-B comparison of cell with heatsink and cell without|
I placed the same size plates top and bottom of one cell. and the larger plates so the extra 2"of plate was off the bottom and one side of the other cell. the top and bottom plates form a heat sink area where I can blow air.
As expected, the cell with extra fins was only half as hot as the one with only the plates, it stayed under 85F. What I did not expect was that the cool cell dropped off noticeably quicker than the warm one. will reverse the heatsinks and do another run, to see if things change cells as final confirmation.
**** Results of moving the heat sink to the other cell was as expected, the cool side being changed also changed the cell that dropped out first almost exactly the mirror image of this trace. Temperature matters. Where is the sweet spot between longevity, temperature, capacity
(Posted 6/25/2012 by mikey)
Same discharge with huge heat sink on one side of cell
We saw 100F at the end of the first 62A discharge on the cells with no heatsink. I did the same 62A discharge, but this time instead of insulation, I used a huge finned aluminum heat sink,with a lot of air running over it. the idea is to see how well the heat generated in the discharge was transferred out of the cell with one side kept at 74F.
|even a monster heat sink cant get the heat out from one side|
I was a bit disappointed to see the temp on the back side was only held to 90 degrees instead of the 100. This would lead me to believe that the heat sinks would work much better if they were on both sides of each cell rather than every other one?
The ambient temperature probe was getting heated by the discharge load. will move it to a better spot in the next test
(Posted 6/24/2012 by mikey)
In preparation for the second discharge I tried charging three different ways.
|high medium and low charge rates compared|
First I charged at 34A till the first cell hit 3.55V
On turning off the charge, the voltage immediately dropped to only slightly more than the starting voltage.
I next charged again at ~7A, and the pack voltage again rose to the 3.55 and dropped out, only to show the voltage again drop to a lower value, but higher than the first rebound.
Finally I set up two constant voltage supplies for 3.53V, and let the charger stay connected until the current drops to near zero.
After removing the charger, the voltage stayed at the 3.53V
Hi current charge gets you to the target fast, but then you have to hold the target voltage until the current drops to few ma as the cell fully charges.
Probably better to approach it more slowley, at a more moderate rate.5-8A may be a nice number.
(Posted 6/24/2012 by mikey)
First discharge graph
Well the fixture seems to work. I think that we have a test system.
|First recorded discharge, first test ot test fixture|
Some uncertainty as to if I was fully charged at the start, but we have a respectable 18AH
The temperature rose immediately and this was at only 68A, so I want some active cooling, based on just this test.
Next I will charge back up, and re-run the test but this time with an infinite heat sink right against the cells instead of the insulation.
The violet cell started at a lower voltage, but soon after the discharge started it was higher, and it remained higher for the rest of the discharge.
Wonder what that means?
(Posted 6/23/2012 by mikey)
Cell tester is modified and lithium test fixture takes shape.
Got out the soldering iron, and made some modifications to the cell tester. Had to disable the other 4 cell channels,and have special calibration valuse for the lithium cells in the software.
|Getting set up for some testing|
Added max V charge stop in the test software.
had to add a header on the cel voltage amp where I can plug in some gain changing resistors, to rescale the amplifiers.
Set uo two temperature boards dead center on each cell, and mounted the board on some soft foam mounting tape, so it sits just a bit above the surface of the plywood, made two thin aluminum flashing rectangles for the cells to sit on, so the whole cell temp is better represented by the single temp probe. The cells need pressure on them to assure intimate contact between the plates and the electrolyte saturated spacers, and since the purpose of the test is to see the thermal characteristics of the cell, I opted to use a 3/4" foam/3/4" plywood pressure plate, with a 65 lb block of steel as a weight. This should allow us to first test the non cooled cell, and to clearly see any heating effect, as the aluminum plate will show the overall cell back side temperature. Once we see how the cells behave at 100A, and see any heating, we can try different cooling plate thicknesses on the top side and hopefully see the effect of active cooling in the bottom thermal monitoring zone, which should simulate the cell to cell juncture in a 2 cell stack.
At high charge currents, I see some temperature measurement errors, so I need to do a bit more work on the amplifiers.
In all it is already quite useful, and hopefully I can get us a discharge graph and AH measurement soon.
(Posted 6/23/2012 by mikey)
Is using this asking for a problem?
In the 56 cell batch of pouches, two of the cells showed evidence of getting hit with an object that dented them right through. I will use the two dented cells for the first test, and see if his area gets hotter, or if the cell is degraded in any measurable way, but common sense would make one not want to bury this in a pack of nice flat cells.
|Is this going to be a problem?|
(Posted 6/21/2012 by mikey)