Weber and Coulomb's MX-5

Post by **coulomb** » Sun, 06 Sep 2009, 04:40

Our last two EV days were spent on battery racks. Here is the bottom frame with some dummy runners for the cells to rest on, and three cells:

We left 1 mm spare on each side of the cells, and it all just squeezes in. The white styrofoam is for inserting the thermometer; obviously it won't be there in a finished rack. It is all 25 x 25 x 3 mm angle steel, apart from the long edges of this rack which are 30 x 30 x 3 mm (because it is so long), and some 25 x 3 mm flat bar that will go across the racks for strength. (Edit: added more dimensions).

We used a special tool to space the verticals 3 mm off where the base rests on the bricks:

You can see it in use on the right. We didn't continue the verticals down to where the rack rests on the bricks, to allow room for welds.

Speaking of special tools, we found this tool useful for holding pieces of angle at 90°:

We modified the tool so there was clearance for the welding nozzle. (Edit: moved text)

A weld before sanding back with a flap disk (edit: was lap wheel):

Here is a test fit below the boot:

The red flap at the top of the picture is some fibreglass at the side (edit: was "very back") of the car. So that's one of the low points that could scrape a driveway or speed bump. (Edit: You can see some scrapes on it now.)

Some more pictures of the completed outer frame, with the lid friction fitted:

Post by **woody** » Sun, 06 Sep 2009, 04:49

Let me be the first to say, "Nice Rack!"

Post by **weber** » Sun, 06 Sep 2009, 18:46

Thanks Woody.

The outside dimensions of this cage are 1092 x 379 x 231. It's a tight fit, but we need every cubic millimetre to fit the 208 (or more) 40 Ah cells, as low as possible in the little MX-5. Basically, if we don't have to bend something to get it in, or bash it with a hammer, then maybe we could have made it a bit bigger and fit another cell.

This is the single biggest cage of cells, of the 7 we have planned for the MX-5. It has room for 65 cells in two rows of 23 and one row of 19. It goes the full width of the vehicle under the boot, where the muffler used to be, and more.

Here's a screen shot from a 3D model we made using the free version of Google Sketchup.

The (zipped) Sketchup file RearBatBox.zip shows 30 x 30 x 3 mm angle everywhere. We sent it to our engineer and he said we could reduce it to 25 x 25 x 3 everywhere except the 4 long outside pieces, provided we also wrapped a band of 20 x 3 flat around the middle of it, as Coulomb mentioned.

Most of it hangs _under_ the chassis rails, but because of the need for a breakaway angle for dips and driveways it slopes up towards the rear of the vehicle where it does take away some boot space. That's why one row is missing 4 cells. It interferes with the chassis rails in the rearmost row and so two cells are replaced with low-height dummies underneath each chassis rail. And the lid will be modified with cutouts that fit around the chassis rails. It will eventually have brackets welded to it to mount it to 6 existing bolts. Four at the rear and two at the front.

The "dummy cells" may end up housing the ELV breakup contactors and fuses for this cage.

The cage does not hang any lower than the sloping plane passing through the lowest points of the existing structural members that surround it.

Unfortunately access to the cells will require lowering the whole cage out from under the vehicle before the lid can be removed, although some will be visible from under the boot carpet through clear polycarbonate. When full of cells the cage will weigh about 120 kg so it will have to be jacked in and out.
[Edit: Changed "25 x 3 flat" to "20 x 3 flat"]

Post by **weber** » Sun, 06 Sep 2009, 19:16

That screen shot shows the cage as viewed from the left hand side of the vehicle. i.e. the right side of the image is the rear of the vehicle. The orange things are the chassis rails where you can see 4 of the mounting points. The other 2 will be in front, low down, to the wishbone hinge bolts on the rear subframe (not shown).

MIG welding is such fun! Once you get all the conditions right it's a breeze.

Mind you, there is a hell of a lot to get right. Voltage, wire size, wire speed, welding speed, type of gas, gas flow rate, torch angle, viewing angle, two-handed torch-support, no galvanising, clamping the joint, getting it all square, no draughts. What else have I forgotten? And when you use different thicknesses or shapes of stock, or butt joints instead of lapped, the first _four_ of those change.

But once you get it right, it's such a buzz just gluing stuff together with liquid metal. What power.

Post by **coulomb** » Wed, 23 Sep 2009, 20:58

What's wrong with this picture?

Holy hole in the donut, Batman!

In fact, there isn't much donut left, they're mostly hole. In the circled positions, the hole has partly "eaten out" another hole.

We've pointed out the error to Batch PCB, and they will look into this and send us a new prototype board. In the meantime, we can actually build prototype boards, and use very large washers to connect to the cells

It's almost like the PCB makers ignored the drill file altogether, and just guessed the hole size to leave only a thin donut. Or perhaps our drill file didn't merge correctly with the others from the same panel.

Edit: interestingly, the vias are unaffected.

Post by **evric** » Wed, 23 Sep 2009, 21:10

What about the BIG holes!?

Post by **coulomb** » Wed, 23 Sep 2009, 21:31

A test fit of cells in the first battery rack:

While with most other cells will rest on pairs of angle bar arranged into an inverted T (┴ made up from two angles pieces like ┘└), we're really pushed for space in this rack, so we made up some ┴ pieces from angle and flat, i.e. ┘-). This saves 6mm of space; that's what the loose angle at the top of the picture is for.

The next job is to decide the final height, and see how many cells we can fit. It looks like we might lose a cell or two because of these clamps on the rear sway bar:

The cells will almost touch the sway bar. It looks like we'll be lucky and only lose one cell. A suitably shaped "dummy cell" will clear the bracket yet maintain clamping force over most of the cells' surface area.

Post by **coulomb** » Wed, 23 Sep 2009, 21:50

evric wrote: What about the BIG holes!?

They're meant to be 300 mil diameter (7.62 mm, to clear the M6 bolt with a little slack). To test the boards as delivered, a very large washer will do under the board, but on top it looks like we'll have to solder some strips of metal to avoid shorting to components.

Weber checked the drill file before and after the board was sent off, and it sure looks right.

If anyone "speaks NC", perhaps they could explain the meaning of the M commands at the start? The rest is easy to read: T1 is the 0.3" drill (the largest), others are 40, 32, and 20 mils (1/1000").

M72
M48
T1F00S00C0.3000
T2F00S00C0.0400
T3F00S00C0.0320
T4F00S00C0.0200
%
T01
X021125Y0085
X-004125
T02
X0024Y0019

(much left off obviously)

Post by **coulomb** » Thu, 24 Sep 2009, 03:53

Ah, the M72 means imperial units, per these pages:

http://www.apcircuits.com/resources/inf ... codes.html
http://www.apcircuits.com/resources/inf ... ample.html

As far as I can see, our files are fine, although we do specify a drill feed speed of 00 inches per minute and speed of 00 RPM. Presumably 00 means "default" or "don't care".

Post by **acmotor** » Thu, 24 Sep 2009, 04:28

or OOversized holes

How do you plan to hold the cells down ?

Post by **coulomb** » Fri, 25 Sep 2009, 05:11

acmotor wrote: or OOversized holes

Heh, good one acmotor.

How do you plan to hold the cells down ?

This has changed a few times. The latest is that we don't attempt to keep the cells from moving around inside the rack in a 10/15/20G crash; the racks keep the cells in, but not necessarily immobile.

The lid of the rack has (inverted inverted) T bars matching the inverted T bars that the cells rest on. Of course, these end up between the cells, or between the extension of the face of the cells since the bars poke down about 22 mm and the space above the cells is 20-25 mm. We will have rubber glued to the T bars that will rest/press onto the shoulders of the cells. The cells will also be sitting on rubber, to take up individual variation in the positions of the cells or inverted T bars.

Post by **coulomb** » Fri, 16 Oct 2009, 04:40

We populated a prototype BMS board today:

A big thank you to the kind people at Electronics Innovations for allowing us to use their hot air gun, microtip soldering iron, and for their sage advice on how to mount surface mount parts, and how to repair the inevitable stuff-ups.

As you can see, BatchPCB redid our boards and this time they came out correctly. Unfortunately, they could not tell us what was different this time. One thing Weber noticed is that the slot for the opto isolator (which we know they can't route) was missing even from the silk screen overlay.

Only one of the four bypass resistors was installed, and on long leads, so we could still access the parts underneath.

We found only one stuffup: the push-button switch was wired the wrong way around, so it was always closed. As a result, the error latch didn't latch. I managed to bodgy the switch on at 90° so that it worked. We also noted the need for slotted terminal holes, so that it will work on both SE cells (60 mm) and Thunder Sky cells (64 mm). They are so close, and yet... irritating.

The rest of the board worked fine, although the temperature threshold was a bit high at 76°C, and the bypassing was at about 3.60 V, compared to the designed 3.57 V. (Edit: had bypass and overvoltage confused; overvoltage was also a little high at 3.65 V). We'll have to see if that's due to the design or due to component tolerances. The error latch reset was also up a bit at about 3.95 V, due to the voltage drop across the 330 ohm resistor that we neglected in the calculation. That will hopefully prevent some nuisance resetting. (Edit: was nuisance tripping.)

The LEDs are surprisingly bright, especially in subdued light:

With the bypass resistors installed, the board isn't as impressive:

One surprise is that the ceramic beads end up on the outside of the pigtails, not centred on the donut:

Note how one of the beads is touching R6. We can just move the donuts for the resistors in a few millimetres.

There is room under the bypass resistors for real 1W resistors (the ones we got are supposed to be 1W, but I don't think they'll take the voltage). Another thing to fix.

Another slight problem, more difficult to solve, is that there isn't much clearance over the wires to the daisy chain and monitor wires:

Edit: replaced last two blurry photos with sharper ones

Post by **acmotor** » Fri, 16 Oct 2009, 05:16

Good progress.

I still see a problem with radiant heat from the shunt resistors causing problems with the PCB components. Mostly temp sensor.
Also the mechanical stability of the resistor mounting in a vehicle vibration situation.

A thought.... forget the ceramic standoffs 16 x 220 = 3520 of them !

Make another pcb roughly the size of the four shunt resistors footprint with plated through holes at the resistor lead positions to match the main pcb. Solder the resistors firmly to that as a module and then fit the module to the main pcb at the desired standoff height.

The 8 legs will keep everything in place without needing the ceramic standoffs and the little pcb will offer thermal insulation between resistors and main pcb.

The exposed resistor legs will have a lot higher heat disipation although you may chose to drop heatshrink over them for that little less electrical exposure. Maybe not... too much fiddly work.

The little pcb can have a cutout for better access to the opto/monitor pads or just the added height may do the job.

You may go for a hole/cutout for the LEDs and reset sw. ?

Anyway, just a thought.

Did you plan screw terminals or soldered wires for these pads ?

Post by **weber** » Fri, 16 Oct 2009, 19:50

I'm sending off the artwork for the next prototype today. So if anyone has any suggestions that may affect the artwork, now is the time.

Thanks for your suggestions already, acmotor. We plan to use soldered wires for the alarm daisy-chain and voltage monitor connections.

Raising the alarm due to hot bypass resistors could be considered a feature, not a bug

, provided it isn't too sensitive.

You may have a point about the ceramic standoffs. And they are 6 cents each. It all adds up.

Our plan for mechanically stabilising the resistors is to put blobs of neutral-cure silicone between them, as mentioned above. We can't afford any more height (in the battery locations identified in the MX-5).

Post by **Johny** » Fri, 16 Oct 2009, 19:59

I know I'm rough, but.
It looks to me like there is enough room below the bypass resistor above where "Weber and Coulomb" is on the overlay to put a blob of silicon and not have too much trouble (if you have to remove it).
So, blob there then blob in between resistors. Use a jig (piece of ply) to mount the resistors at a common height. If these boards are exposed to too much vibration to cause that to fail, then the cells are as well.

Other than the two points you have made, switch and elongated cell holes...no suggestions here on layout changes/additions.

Post by **coulomb** » Sat, 17 Oct 2009, 20:55

Yes, I think that those ceramic standoffs are out. Thanks for all the suggestions.

Here is the board with a link and a pair of worst-case terminal lugs:

It looks good from on top, but from the side, there is a problem:

The edge of the board wing just barely clears the start of the bump in the links. But it looks to me that we'll have to make the "wings" as thin as possible to clear the heatshrink tubing, despite the extra routing that will be needed, and even then we'll likely have to hit some of the heatshrink tubing on the links with a Stanley knife.

Edit: The reason for the extra routing is that at present, the boards form a single tile tessellation of the infinite plane (I hope I got that right

), so that each row can be separated from the next with a single square wave pass of the router, leaving almost no wasted board (only at the edges, obviously). That's assuming that the panelliser doesn't use a naive bounding rectangle fitting algorithm.

Edit2: More befitting spelling. Edit3: more spelling. It seems that BEC do use a naive panelliser, so we may have to Protel up a whole panel. Sigh.

Post by **acmotor** » Sun, 18 Oct 2009, 01:15

That's a long way of saying it doesn't fit right !

Those copper straps will need some attention to remove oxidised surface and some zinc terminal paste between them and the ali battery post/nut.

Post by **weber** » Sun, 18 Oct 2009, 03:59

Johny wrote: I know I'm rough, but.
It looks to me like there is enough room below the bypass resistor above where "Weber and Coulomb" is on the overlay to put a blob of silicon and not have too much trouble (if you have to remove it).
So, blob there then blob in between resistors. Use a jig (piece of ply) to mount the resistors at a common height.

Excellent idea. Thanks Johny. What's rough about it? A blind man on a galloping horse would never tell the difference.

Post by **Squiggles** » Sun, 18 Oct 2009, 04:01

Remember to use neutral cure silicone!!

Post by **coulomb** » Sun, 18 Oct 2009, 16:18

I did an RFI test today. An AM radio certainly picks up the ~500 Hz wail at close range, but at 1 metre, on or off channel, it's not audible at all. There is no pickup whatsoever on FM.

Of course, 208 of them switching away unsynchronised might be a different story.

The wail changes frequency depending on the duty cycle; the frequency gets higher with higher duty cycle, peaking at about 1 kHz with 1 A (about 50% duty cycle, presumably), then decreases again with duty cycle.

Only when it's nearly turning off (around 3.59 V) does the sound degenerate into a series of clicks, which are then visible on the yellow LED as flickers. It takes a minute or two for self discharge plus the current of the increasingly lower duty cycle LED to bring the BMS to its no-bypass state. So that should look interesting at the end of charging.

With 2 amps charge, the battery voltage settles at 3.65 V, per my cheap multimeter.

I tried putting various capacitors across the cell, with little effect. Interestingly, putting a 1000uF or so from the top of the bypass resistors to the negative terminal did change the frequency of the sound. The only component between the top of the bypass resistors and the positive terminal is the fuse. I checked this by connecting directly to the ends of the fuse: one end has no effect, the other has an effect. It looks like the fuse has non trivial impedance.

In fact, it drops 66 mV (per my cheap meter; the other one is in the Prius with 230 V across it) at 2 A bypass. About 33 mR. It's scary when just paralleling the fuse with the multimeter changes the tone audibly

. Perhaps that's more capacitance than the presumably 2 MR or so resistance of the meter. Shorting the fuse causes the cell to settle a about 3.62-3 V compared to 3.64-5 V. We might investigate some physically larger SMD fuses, or even through-hole.

Post by **Squiggles** » Sun, 18 Oct 2009, 16:21

Could the fuse be inductive?

Post by **gpsnettrack** » Mon, 19 Oct 2009, 04:02

Havent been on here for a while, doing other things, but I hope you have a free Discount Card from Jaycar

Alan

Post by **antiscab** » Mon, 19 Oct 2009, 09:29

if you wander into one of their shops and give them ur abn number, then theyll give you a trade card.

Matt

Post by **coulomb** » Mon, 19 Oct 2009, 15:17

Squiggles wrote: Could the fuse be inductive?

Possibly a little. But the fuses have a rated voltage drop (at rated current, I assume) of 77 mV. The thing is, other fuses are worse! Common 20x5 mm glass cartridge fuses have a max voltage drop of some 200 mV.

So we're going to stick with this fuse, and just move the sense wire (it goes via an 82k resistor) to the unfused terminal via a thin track with a short even thinner section (8 thou) near the positive terminal.

Low and behold, the multi hundred hertz oscillation disappears, and the fastest frequency is about 1 Hz now. At 2 A, it's on for something like 8 seconds, and off for a fraction of a second (bypass current is about 2.1 A).

The board has changed a fair bit this weekend:

The main thing is it's smaller overall, both the width of the wings, and the central postage stamp as well. That's to accommodate Sky Energy style straps on Thunder Sky cells, and so that cable lugs can exit from under the boards at more angles. The terminal posts will have routed oval holes, 7x9 mm, to accommodate SE or TS cells. Tracks are rearranged near the thermistor to encourage thermal conduction to the thermistor. The TVS could be replaced with a 1206 3W zener (though we'll likely stick with the Transient Voltage Suppressor). Components like the switch and MOSFET had to be rotated to fit. Little details like the LEDs have diode symbols for orientation, they have labels, etc.

We're lucky to be able to piggy back on a company's panel, so we'll have a new prototype board soon.

We currently don't get any parts from Jaycar.

Edit: update image to rev 20.

Post by **Thalass** » Mon, 19 Oct 2009, 15:22

Looks good! Are you planning on selling this bms setup once you're done?