Page 5 of 68

Weber and Coulomb's MX-5

Posted: Sat, 11 Apr 2009, 14:46
by coulomb
With 224 cells and a braking resistor coming on at 780 V, we would be limited to an average of only 3.48 VPC, which might be OK for regen.

We will limit the low side VPC to 2.5, so we'll actually have 520 VDC min for 208 cells, or about 367 VAC. We're hoping that the controller can dish out about 167 A RMS to make this 112.5 kW electrical. Oops, more like 200 A, with the power factor. Edit: max is 234 ARMS.

We'd prefer less battery sag, which is now the main reason we're looking at China HiPower cells, which seem to come in "high discharge" versions. We'll test a 40 Ah cell against a Thunder Sky 40AH cell, and scale the results to 30 Ah, and hope it scales reasonably linearly. We may yet decide on 40 Ah HiPower cells, to minimise the sag at high power. Unfortunately, that means three sets of battery packing calculations while we wait for that cell to arrive (probably a month away).

Edit: 224 volts -> 224 cells; 750V -> 780V; not enough -> might be OK

Weber and Coulomb's MX-5

Posted: Sat, 11 Apr 2009, 15:51
by woody
EVPST are from then click on LiFEPO4 cells. The high discharge ones have low internal resistance so you end up with a very stiff pack (0.02V drop/ C). They won't get down to 2.5 volts.
My work with them is all based off the specs, I haven't found anyone who's tested them.

I've said 224 because that means higher AC voltage under load which means higher peak power for AC due to extending the constant torque range higher.

208 may be the limit because of the same reasons you state...

Weber and Coulomb's MX-5

Posted: Sun, 12 Apr 2009, 03:32
by coulomb
Not much to show today; we need to sort out where the 208 40 Ah cells are going to go. We did the usual thing of making a simple cardboard model of the motor, and some battery footprints.

Image    Image

To figure out the front to back balance, we have a spreadsheet of distances from the rear axle to the various centres of mass, the effect that they have on the rear and front axle loads, and some totals. It's been somewhat surprising. Recently Weber decided the only way was to have the controller in the boot, but that doesn't fit well with batteries, and of course there would be long wires to the motor. So we'll start another spreadsheet for having the controller at the front.

We're nowhere near our required total of 208, but we think we'll be able to find a way to put them all somewhere. We'll also have a look at these EVPST batteries (thanks for the link and the idea, Woody), though I suspect we won't be able to afford decent range with them.

Weber and Coulomb's MX-5

Posted: Sun, 12 Apr 2009, 05:06
by woody
Hi Guys,

I started this post as some sort of encouragement, I hope you take it that way...

Zeva put 40 LFP160AHA cells in: 240kg / 147Litres, leaving the whole boot empty. 16 under the boot (59 litres), 24 under the bonnet (88 litres).

You're going for 208 LFP40s which is 312Kg / 211 Litres, only 64 Litres more, a 40cm cube :-)

Also you're keeping a gearbox and your controller could be significantly bigger, will be a tight squeeze with only 150L of boot space to start with :-(

Industrial Controllers aren't small 45L (Danfoss 5027-5032), 60L (Telemecanique 45-75kW), 73L (Danfoss 5042-5062), 99L (Danfoss 5072-5102).

Maybe you'll need one of these bad boys for your luggage: Image Image


PS: Cortina has 460L boot :-) "You could fit 3 bodies in there" says De Niro.

Weber and Coulomb's MX-5

Posted: Sun, 12 Apr 2009, 06:29
by coulomb
We're also keeping air conditioning and power steering. It sure looks like something has to give.

I wonder if we could fit three rows of 16 on top of the boot, with those luggage holders...   Image

Your controller could be bigger (than Zeva's)
Our controller is huge: 800 x 310 x 300 mm (roughly; we hope to lop something off the 800 when we remove the rectifier and inductor).

Weber and Coulomb's MX-5

Posted: Sat, 18 Apr 2009, 23:16
by weber
Thanks for the encouragement Woody. We think we can fit 208 TS-40s now, without taking up any boot space. The breakthrough came when we realised we could put about 40 of them (in 2 rows of 20) poking through the existing parcel shelf in the space that remains when the soft-top is folded. So the new parcel shelf will be about 160 mm higher in the front half as compared with the back half which will still be at the original level.

This is right behind the occupants' necks and so will need to be well restrained re the 20 g's frontal collision. Other battery racks might be allowed to bend severely, provided they don't detach from the vehicle. But these can't be allowed to move. Our engineer has OKed this and suggested some means of restraint.

Also, these cells (in fact all cells) must be sealed off from the cabin since if they were overcharged or overheated due to equipment failure they might vent corrosive gas or vapour.

These cells will be the biggest contributor to raising the height of the center of gravity, so we'll put them everywhere else we can first.

Weber and Coulomb's MX-5

Posted: Sat, 18 Apr 2009, 23:51
by woody
weber wrote: We think we can fit 208 TS-40s now, without taking up any boot space.

That's a bummer, I was hoping you'd be the Guinea Pigs for the EVPST cells :-)

Weber and Coulomb's MX-5

Posted: Sun, 19 Apr 2009, 00:07
by weber
Some weeks back I was talking to our MX-5 parts recycler (Richard Larsen of MX-5 Plus) about the electric conversion and I mentioned that, compared to the original ICE, we expected to have a lot more torque from our electric motor but max rpm would be a lot less.

He said, "So you need a taller diff, about 3.6 to 1". (The existing is 4.3 to 1)

I, having previously calculated a number very close to this, was impressed and said, "That's right. Pity they don't exist.".

He said, "But they do.".

It turns out that they only exist in Australian delivered 1.8 L MX-5's from 2001 to 2005. No other country ever got MX-5's with 3.6 diffs. In fact they are 3.636... (= 40:11).

Some owners find the 3.6 diffs too tall for their ICEs and trade them for 4.1 diffs, so the 3.6s are available to us. They are also much stronger diffs, physically larger, which is just as well. Richard said that when people add bigger engines or superchargers to our model of MX-5 they typically come back the next weekend for a new diff.

I said "What about the gearbox?". He, and his mechanic, both said, "No, the standard gearbox is very tough. No problem there".

Weber and Coulomb's MX-5

Posted: Sun, 19 Apr 2009, 02:54
by acmotor
Now that is a really useful option to have up your sleeve ! 4.3 or 3.1
Nice to have feedback on gearbox strength too.

Tell me, what is the history behind calling it a 'taller' diff ?

Weber and Coulomb's MX-5

Posted: Sun, 19 Apr 2009, 03:39
by weber
That's 4.3 or 3.6, only a relative change of 1.18 or 0.85 depending which way you look at it, but still worthwhile for maintaining near original top speed and near original gear-change points in km/h. Which raises the dilemma of whether to make the tacho (and possible audio feedback device) lie about the motor revs and instead tell the driver what the ICE revs would have been.

I think the "taller" metaphor is based on a taller person having longer strides where a stride is an engine or propshaft revolution and the distance covered in metres per propshaft revolution is greater for a diff with a smaller number.

Weber and Coulomb's MX-5

Posted: Fri, 24 Apr 2009, 04:33
by coulomb
The first of two cells to be tested against each other arrived yesterday. The other is not a stock item and will be a few more weeks. It's a Thunder Sky LFP-40AHA. It measured 3.30 V before charging or discharging, having been shipped from China via the USA, and could have been in the USA for months.

We used this old PC power supply to charge it:
You can see the back end (bottom right corner) is mostly cleared out. I had fitted a linear pot for voltage adjustment. I had modified it to be a 12+ V only power supply, so today we modified it back to being a 5 V only supply. It was charged at up to 17.6 A, which is a little under 0.5C.

My multimeter only goes up to 10A (safely), so we calibrated a shunt made from a few turns of single strand copper wire on a strip of ceramic tile. We found that if the two multimeters agreed at 10 A, they would be 5% out at 5 A. I thought it must be the multimeter's shunt getting hot (it had some three times the voltage drop of our home made shunt), but the results seemed to indicate the opposite. When we immersed the shunt into water, the difference vanished, and we could get the two multimeters to track very well by adjusting the shunt with a clip lead, moving the clip lead up or down to adjust the resistance.

The end of charge came up quickly, with the power supply feeding 9.1 A at 3.492 V, so we realised the need for Kelvin connections to the actual cell. (Even 600 mm of 16mm² cable dropped some 0.06 V, so we had to move the multimeter measuring the cell right to the cell terminals.)

For discharging, we used a bigger version of our current shunt:
Edit: That muck isn't from heat or anything, it's just glue from the old tile.

While the ceramic is ideal for holding the hot wire, it took a few moments to figure out a way to attach thick leads to it. We ended up filing 3mm notches into the ends, and using large washers to hold a 16 mm² cable lug, and soldered the wire to the lug.

Here is the discharge setup:
You can see the cell at top, a 63 A breaker, and the 16 mm² wires to the load in the ice cream bucket.

Behind the cell is an infra red thermometer, to keep an eye on the temperature of the cell, and also of the water. The water ended up at about 60°C. The yellow multimeter is reading the voltage drop across a part of the load that we calibrated with my 10 A multimeter. (This is no longer in circuit). We had to fiddle a bit to get a spot that wasn't on the edge of the ceramic, or underneath. It's 10 milli-ohms, so the shown reading of 328 mV indicates 32.8 A. The long grey multimeter is showing 3.188 V across the cell. The wires to this meter are carefully covered in tape, since were were naughty and didn't fuse those wires.

We only went to 33% DOD (attempting to break the cell in, as some say is important). We want it to perform its best (and the China HiPower cell at its best too, of course), for a realistic evaluation of which is better for our application.

Edit: mm^2 -> mm²
Edit: first plot:
I'm rather rusty with Gnuplot; haven't used it much since the Roulette days (don't ask Coulomb about roulette!). So I can't remember how to display more than one column from a data file overlaid. So for now, it's just voltage.

Note the almost ruler-straight discharge from 4 minutes to 25 minutes (there are actually 6 data points there, it looks like 2). The cell started at 3.339 V, shot up to 3.276 in the first minute, then settled to 3.315 V an hour later.

The current was initially 33.2 A, then stayed around 32.7 to 32.8. So that's just over 0.8C for this 40 Ah cell. The temperature rose from 23.3°C to 25.0°C at the end. Not much excitement at such a low discharge current, and such a low depth of discharge.

Edit: 30% DOD -> 33% DOD (Depth Of Discharge).

Weber and Coulomb's MX-5

Posted: Fri, 24 Apr 2009, 05:53
by acmotor
Image I hope these construction techniques will not extentd to the 600V battery pack Image
9 out of 10 for creativity though !

Weber and Coulomb's MX-5

Posted: Sat, 25 Apr 2009, 06:08
by coulomb
Our current planned battery layout:


It's not quite to scale, and not terribly accurate (e.g. controller in the elevation view).

Weber and Coulomb's MX-5

Posted: Sun, 26 Apr 2009, 04:19
by weber
Here's the Dodgy Brother's charging setup for our testing of a single LiFePO4 cell. 24 A max (0.6C for the 40 Ah cell).


A recent addition is a kelvin sense wire (the skinny fused wire on the right) to improve voltage regulation at the cell terminals.

Still using the water-stabilised DIY shunt today. But a real one is on the way.

Weber and Coulomb's MX-5

Posted: Sun, 26 Apr 2009, 05:08
by coulomb
We used a different method of attaching 16mm² cable to the shunt today. We drilled two holes with a special bit designed for cutting ceramic and even glass; it worked very well, reasonably quickly, and made very clean holes.


The holes were used to anchor small cable clamps. We soldered directly to the thick cable using a portable butane soldering iron (often together with my 60 W Weller soldering iron; those cables suck the heat really well).

We also soldered the thick cable direct to the PCB of the power supply.

You can just see a loop of wire at the top of the ceramic. That's the end of a loop of wire we twisted together to adjust the resistance (calibrating it at 10 A using a multimeter). Unfortunately, the twisted wire doesn't conduct well enough without soldering, so we ended up soldering, checking, adjusting, checking, etc. Yesterday's adjustment with a clip lead (moving up and down the length of the wire) worked much better. When the final point is known, solder a short medium thickness wire from there to the thick current carrying wire.

Weber and Coulomb's MX-5

Posted: Sun, 03 May 2009, 02:20
by coulomb
We progressed to higher discharge currents today. First, a 3C (120 A) discharge. Remember that TS cells are supposed to be able to be discharged continuously at 3C:
As you can see, the voltage dipped only to 2.9 V, then to 2.85 V, then actually increased all the way to the end of the test (to 2.914 V @ 113A). The discharge current averaged about 115A, just a bit shy of the 120 A we were aiming for. The test terminated at about 33% DOD, since this is a new cell, and we are trying not to discharge it beyond 33% for the first few cycles. The temperature ended up peaking at just over 30°C, or just over 10°C rise. It certainly looks like TS cells can handle 3C comfortably when fully charged.

The load was bifilar wound to cope with the extra current. The temperature of the load water at the end of the test was in the fourties Celsius.

What we are more interested in though is 6C, or 240 A. This is about where we estimate our peak motor current would be, and we'd like the cells to dish out that current for at least 20 seconds for acceleration.

So we beefed up the cables and the load (now two bifilar wound loads in parallel). We decided that the 3C load will come in handy when the other cell arrives, so we made another similar load and paralleled it. This one had to be somewhat lower resistance, because the battery voltage will be lower (we estimated close to 2.5 V), and also the resistance of the rest of the circuit, including the shunt, becomes significant. Here is the setup:


Note the pairs of 16mm² cable; each is rated to carry 63 A continuously (as is each half of the breaker). So that's OK for long 3C tests, or short 6C tests (hopefully).

I was wondering what to do if the breaker couldn't interrupt the load. I had visions of this:
But then Weber pointed to the number "6000" in a box on the breaker, indicating that it could interrupt 6000 A at 48 V. I was a bit more relaxed about things after that.

That's a 200 A shunt in the clear container at the front right. It got hot enough (some 60°C) at 115 A that we decided it could do with water cooling as well as the load.

The cell is wrapped in some thick rubbery stuff (I think it was a garden kneeler). The clamp is just to make sure that the cell doesn't expand, and to make sure that the digital thermometer makes reasonable contact. The insulation of the foam will hopefully be worst case, and about the same as an infinite set of batteries of the same temperature surrounding the cell.

The two-layer load: [Edit: added later]

To make sure that the new load had about the right resistance, we initially put the two loads in series (around 1.5C load, or 60 A), and adjusted the new load so that it dropped the right proportion of resistance. We wanted it to be 115/125th (so the new one would pull 125 A for a total of 240 A), and also 2.5/2.9ths, since the battery voltage would be lower. The fine maths was somewhat wasted, however, and we ended up taking off half and even full turns to get the current about right. Here is a graph of the first discharge:

Edit: Where the current (red) goes off the bottom, it goes to zero.

The upper graph is a zoom of the first part of the lower graph (over the actual discharge). Time is in seconds. Some of those currents were actually averages of the earlier and later currents; we weren't as organised as we became later. As you can see, the cell voltage is barely above 2.5 V, but the temperature is around 25°C. We watched the temperature for a while to see what would happen; it kept rising for about 20 minutes, but not by very much.

The next three discharges were only a few minutes apart, to better simulate real driving conditions. Sorry about the very wide image:
You can see that we got bolder on the third discharge; although it doesn't look it, the cell voltage remained above 2.5 V, so we continued the test for 70 seconds. Here is the detail around the long test:
Note how the voltage actually rises, presumably as the internal temperature rises. (Most of the temperature figures during this discharge are interpolations; we didn't figure out that "blank divided by 10" isn't a blank according to Excel at that stage. Blanks get ignored; blank divided by 10 gets treated as a zero). The voltage actually rose again at the 70 second mark, but we'd exhausted our boldness at that point.

So the conclusion so far is: Thunder Sky cells handle 6C discharges pretty well when warmed up a little. (Warmed up a lot if you live in Antarctica). The voltage sag is somewhat severe, though.

It will be interesting to see how the "high discharge rate" China HiPower cells compare.

Edit: added more graphs and text

Edit: Expanded the vertical axis on most graphs

Edit: shunt -> load

Weber and Coulomb's MX-5

Posted: Sun, 03 May 2009, 04:45
by coulomb
Amongst all that discharging, we obviously also did some charging. Here is a graph of one of them:
The temperature is very flat; charging seems to be somewhat endothermic. (Not only that, but it seems to absorb heat Image ). The temperature rose a fraction of a degree just as the voltage started rising rapidly at the end of charge. The voltage rises quite linearly for most of the charge. With a better charger, the cell might have charged a bit quicker towards the end, with a sharper cutoff of current only when the voltage was much nearer the CV limit (in this case, 4.00 V).

Weber and Coulomb's MX-5

Posted: Mon, 04 May 2009, 02:32
by coulomb
The cell is preparing for some low temperature tests:


It's one of those Peltier effect portable fridges, running off an 18 V centre tapped transformer, half bridge, and capacitor. I'll leave it overnight and see if enough coolness seeps into the cell; at the time of the photo, it had cooled less than 1°C.

It may also be used (after all other tests are completed) for accelerated aging tests (using its heating facility).

Weber and Coulomb's MX-5

Posted: Tue, 05 May 2009, 18:51
by Johny
These tests and documentation are really great guys. I don't know about others but I don't post here because I don't want to break up the exciting, ongoing theme, but I'm always reading and learning from these tests.

Weber and Coulomb's MX-5

Posted: Tue, 05 May 2009, 19:58
by coulomb
Thanks for the kind words, Johny.

Here are some cool results (I mean, results for when the cell is cool Image ):

Started at 12.8°C and 3.36 V. With a few initial checks, the cell was up to 12.9°C.

t=0 V=2.46, t=5 V=2.41, t=10 V=2.40 (234 A initially) Final temp 13.0°C.

It was about this point, with me watching the timer, voltage, current when I could, and making sure nothing was melting or boiling, it dawned on me that these numbers are all less than 2.5V, the Thunder Sky minimum. Image So I stopped the test just after the 10 second mark.

Soon after (with a quick test), the cell sags to 2.43V.

After that, I had several attempts to see if the voltage was over 2.5 V yet. It never got there. Each test was less than 2 seconds, just enough for the digital multimeter to settle down and get a reading. Each reading is about 30-90 seconds from the last:

13.2°C 2.45V 232A -> 13.3°C
13.4°C 2.46C 234A -> 13.4°C
13.6°C 2.45V 234A -> 13.7°C
13.8°C 2.45V ?A still 13.8°C
Short term rest voltage is 3.32 here
14.0°C 2.462V ?A

I have some more results from this morning, where the temperature started under 10°C. As you might imagine, the results were even worse, but I didn't take the right piece of paper to work to post those results now.

My conclusion so far is that for our MX-5 application, where we think we need 6C for 20 seconds or so, is that the Thunder Sky cells are only just capable in warm weather, or when warmed up with normal use for a while. However, I think a pack with lower internal resistance would offer more power (e.g. sagging to only 2.9 V instead of 2.5 volts) and last a lot longer. I can imagine that a lower internal resistance pack would average say 10°C lower in operation, and hence last about twice as long (per the Arhennius (sp?) law).

Weber and Coulomb's MX-5

Posted: Wed, 06 May 2009, 01:17
by weber
Yeah, thanks Johny. You (and James Carlon of ABB) made my day. Everyone should feel free to interrupt this thread with praise. Image

Or criticism.
coulomb wrote:I can imagine that a lower internal resistance pack would average say 10°C lower in operation, and hence last about twice as long (per the Arhennius (sp?) law).

I remember the spelling by thinking of him as a heinous pirate "Arr". Two R's and one N. "Arrhenius".

I think the flaw in the above argument is that it would typically only be 10°C lower for maybe one hour in every 24. The rest of the time both types of cell would be near ambient. So there's no way it's gunna double the life.

Weber and Coulomb's MX-5

Posted: Wed, 06 May 2009, 02:14
by coulomb
coulomb wrote: I have some more results from this morning...

And here they are. Starting temperature, voltage before load, voltage after load:
9.6 °C 3.33 V 2.32 V
9.6 °C 3.31 V 2.33 V
9.6 °C 3.30 V 2.93 V then immediately 2.33 V
9.6 °C 3.30 V 2.7X V then immediately 2.33 V
9.6 °C 3.30 V 2.41 V then immediately 2.33 V, finally 9.7°C.

So it looks like at 10°C, a max-power acceleration spurt will last less than a second. At that point, hopefully the controller can gracefully back off, and provide 90% or 70% or 50% of what is requested, to keep the cells above 2.5 V.

Weber and Coulomb's MX-5

Posted: Wed, 06 May 2009, 02:52
by Thalass

I'm considering buying TS cells, though at this moment I'm unsure of how many I would need. IIRC my (very) rough estimate pointed at 180-something of the smallest AH rating (40AH, I guess). But I am unsure of whether that would be enough capacity for my minimum range requirement.

I'll keep watching this thread for more info. I shouldn't have much trouble keeping the cells over 15 degrees, even in winter here in Perth. I could always do what they do in Canada, and have a block heater plugged in before I go.

Weber and Coulomb's MX-5

Posted: Fri, 08 May 2009, 03:19
by coulomb
Thalass wrote: Praise!
Blush!    Image Edit: and thanks! Image
I'm considering buying TS cells, though at this moment I'm unsure of how many I would need. IIRC my (very) rough estimate pointed at 180-something of the smallest AH rating (40AH, I guess). But I am unsure of whether that would be enough capacity for my minimum range requirement.
We've found almost the opposite problem. For adequate range (say 80 km), we'd possibly only need 20 Ah cells (x 208 cells per pack). But we need the 40 Ah cells, and would like a little more, for peak discharge current. That's why we are considering the China HiPower 30 Ah cells. Others are considering 20 Ah and even 10 Ah very high discharge rate cells.

What range are you wanting?

Weber and Coulomb's MX-5

Posted: Fri, 08 May 2009, 03:44
by coulomb
We've had enough babying of the cell at 30% DOD; today we went from almost full (just a few two second 6C discharges) to 2.5 V.


The little bumps in the current are where some of the hot water was replaced with cold water. The resistance of copper is amazingly sensitive to temperature; just stirring the container the load is in had an immediate effect: the current would shoot up 2 A (only 1%, but quite visible on a digital multimeter). Edit: it would then fall back the 2A and more over the next few seconds as the temperature around the wire rose again.

Note how the voltage initially falls, as you expect, then rises. I believe that what is happening there is that as the temperature rises, the "inner" voltage of the cell is decreasing, but the internal resistance is decreasing faster. The inner voltage of the cell is decreasing for two reasons: the state of charge (SOC) is reducing, and the temperature is increasing. My guess is that the internal temperature rise is fairly linear until right at the end; the recorded temperature takes a while to rise because of the thermal resistance and thermal capacity of the cell and its case, and the thermal properties of the thermometer.

Eventually the fall in internal voltage overtakes the reduction in internal resistance, and the terminal voltage falls.

A rather interesting thing happens immediately that the load is removed. The voltage starts increasing slightly, and continues to increase till the end of the test (when the measured temperature starts decreasing). What's causing this voltage increase?

I think it's because the internal temperature is decreasing, immediately after the load is removed. We don't see the measured temperature decrease for another 15 minutes, but I suspect that's just the thermal lag. As the temperature increases, the internal voltage changes, just as it does for lead acid. Lower temperatures cause a higher internal voltage. With this amount of lag, I'm wondering if it's worth temperature compensating the over and under voltage thresholds at all, and even if it's worth measuring the temperature of individual cells. On balance, I think yes, though it may be worth implementing some crude form of temperature modelling (predicting the internal temperature based on current loads and the measured temperature, and possibly ambient temperature).

The average discharge current was 122.6 A, and the discharge duration was 19.5 minutes, for a total used capacity of 39.85 Ah. Considering that we'd used up perhaps 20 seconds at 240 A, which is a staggering 1.3 Ah, this cell did well. I guess I should do a proper discharge test from absolutely full to determine the actual capacity.