Weber and Coulomb's MX-5

Post by **coulomb** » Mon, 19 Oct 2009, 17:38

Thalass wrote: Looks good!

Thanks; we think so. The only problem is that we haven't added up the cost of the components yet; scared about what the answer will be.

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

Sell?

Software (and hardware) wants to be free! As in open source, or in this case, open design. People will be free to build their own, and adapt it as needed. We'll put up the Protel files when they are stable; more minor changes are happening.

People will also be free to go blind and batty assembling hundreds of these things themselves

. We don't have the time or inclination to make them ourselves. We'll likely get someone else to populate our boards, although Weber said something about shopping for a frying pan recently, so maybe not.

I suppose if there was a demand, we could organise a big build run and have leftovers to sell. Maybe after version 2; I get the feeling that a BMS has a lot of evolving to do.

Edit: manners

Post by **Squiggles** » Mon, 19 Oct 2009, 19:52

Why don't you have studs in the batteries and fix the straps with lock washer and nut..then put the BMS board on top. No problem with straps and lugs then? Also able to replace BMS boards without interfering with cables.

I am sure we have discussed this before, one of those de ja vu moments happening.

Post by **coulomb** » Tue, 20 Oct 2009, 00:32

Well, the cells come with stainless steel bolts and washers already.

Not a great reason, I'll agree.

Actually, the boards test the impedance of the connection on the positive side. To get the flash rate down to about 1hz requries a pretty low resistance connection. So that's another (admittedly weak) reason to leave the bolts as they are.

Post by **coulomb** » Thu, 22 Oct 2009, 15:03

################

Post by **woody** » Thu, 22 Oct 2009, 16:12

I haven't thought as far in the BMS direction as you lot* have, but was thinking for a while of having huge resistors in the BMS so I didn't need a braking resistor.
This would also allow fast balancing.
But this White Paper you linked kills off the need for fast balancing, so it makes sense just to have a braking resistor (Infrared Hi-Mount stop light?) and little balancers and a smart charger which talks to the BMS or doesn't need to.

* acmotor, SI units, Johny, nevilleh

Post by **coulomb** » Thu, 22 Oct 2009, 18:20

The other cost reduction measure which I forgot to mention is the replacement of one "large" SOT-669 MOSFET transistor (basically the footprint of an SOIC-8 chip) with a pair of smaller and much cheaper SOT-23 MOSFETs in parallel. (SOT-23 are the same size as small signal transistors.) Now that we're only pulling less than 1.1 A @ 3.6 V, this should be quite satisfactory. The combination of low gate threshold and low on-resistance was quite expensive (it had 2.1 mR compared to the 31 mR of the cheaper devices). The new transistors cost 2x26c (qty 5) verses almost $3.

So now we've attacked the "big three" costs: the opto-relay, the 7W resistors (their price seems to be much higher because they're the last few hundred in stock of 6R8 in the small size), and the MOSFET. There is also the PCB cost at an estimated $3, but not too much can be done about that.

Edit: SOT-8 -> SOIC-8

Post by **coulomb** » Thu, 22 Oct 2009, 18:30

My apologies if this posts twice; I seem to have lost this post.

woody wrote: ... was thinking for a while of having huge resistors in the BMS so I didn't need a braking resistor.

Ah, regen; I forgot to mention regen. That's what made us think initially that we had to be hairy chested about bypass current. But in fact, most regen doesn't result in filling the pack, so the pack rarely reaches bypass voltage on regen. The exception is when you live at the top of a hill, or have a long downhill stretch soon after leaving home and charging to full. In those cases, we've decided it's the controller's job to either back off on regen (falling back to mechanical brakes), or to turn on the braking resistor. (The braking resistor has the hairy chest

).

If a truncated regen event results in some pack unbalance, then this just has to wait till the next garage charge, where charging current can be controller much better. We have such low headroom between bypass (3.57 V) and overvoltage (3.60 V) that probably very little balancing is possible with regen anyway.

Post by **acmotor** » Fri, 23 Oct 2009, 03:11

Yes, I mentioned in the BMS thread that my BMS final version was shunting 1.5A (2 x 4.7 in parallel driven by bipolar across 3.7V).
Heat was an issue for cont. shunt which is why I was sounding alarm bells at your resistors both in power and location if you recall !

The big PCB copper and 10W ceramics help on my BMS but still I've gone for ~5W (1.1kW across the pack for full pack regen, and balancing that doesn't take a week).

My concern in dropping too low on the shunt current is largely for correct charging as the pack ages. Blind freddy can charge a new pack and get away with it !

I am running regen with no brake resistor on red suzi at present and you would be surprised a how much regen there is and how often I have to put left foot on the brake pedal in the first 5km of driving to limit pack voltage. Yep, brake resistor will be required as you say.

Post by **coulomb** » Fri, 23 Oct 2009, 15:31

We're going to move the components to the underside of the board:

There is actually more room than it shows in the hastily taken photo between the "1 watt" resistors and the vent. They've been moved to the south corners already.

To not have to do the artwork completely from scratch, and as an act of mental masochism, we're going to leave the components on what Protel calls the top side. So we will refer to "red" and "blue" sides rather than "top" and "bottom" now. We've already started talking about north and south ends of the board; north is top according to the Protel view.

This saves us some height that the rather thick inter-cell straps force on us. However, there are some challenges with the high voltage isolation to overcome. Weber has already started the process.

We'll also take the opportunity to separate out the 330R LED resistors from the quad 330R array, so we can make the LEDs less bright. That will save a tiny bit of cell energy as well, particularly for the red LED, which could be using power when a cell is sick and can't afford the extra 24/7 drain. For now, the other two arrays will stay, even though they are very fiddly to work with.

Edit: moved pix together. The second photo obviously before the resistors were put on the blue side.

Post by **Electrocycle** » Fri, 23 Oct 2009, 22:02

are you worried about the PCB compressing over time and loosening the bolt?

Post by **antiscab** » Fri, 23 Oct 2009, 22:52

a spring washer (or 2) would help with that.

Post by **coulomb** » Sat, 24 Oct 2009, 01:07

electrocycle wrote:are you worried about the PCB compressing over time and loosening the bolt?

Well, yes we are as a matter of fact.

antiscab wrote: a spring washer (or 2) would help with that.

Yes, but will it be enough? I find it hard to believe that fibreglass would change thickness by several mm, but apparently it can happen. See also comments in the "Home made BMS" thread re "epoxy creep" here: viewtopic.php?t=900&p=14366#p14366.

But that has nothing to do with mounting the PCB upside down.

We still may end up replacing the PCB "wings" with a pair of ring lugs designed for cables, and soldering the lubs to the PCB. One advantage of this is making the PCB much smaller, yielding many more PCBs per panel, making the boards cheaper. But we decided against that for some reason we decided against it, and now I can't remember why. Sigh.

Post by **Squiggles** » Sat, 24 Oct 2009, 01:21

coulomb wrote: We still may end up replacing the PCB "wings" with a pair of ring lugs designed for cables, and soldering the lubs to the PCB. One advantage of this is making the PCB much smaller, yielding many more PCBs per panel, making the boards cheaper. But we decided against that for some reason we decided against it, and now I can't remember why. Sigh.

I can see a few reasons, extra time, extra cost (lubs would be dearer than PCB), more potential points of failure.

How about you solve the mounting problem by putting studs in instead of bolts and .......

Post by **gpsnettrack** » Sat, 24 Oct 2009, 01:27

Guys,

If you have a non working board and would like us to have a look at it re mounting of components etc, let me know . WE are located in Sumner Park and have developed many boards from scratch for different projects.

Alan

Post by **weber** » Sat, 24 Oct 2009, 06:04

All,

Bruce at Electronic Innovations raised the PCB creep problem with us today, quite independently of our own concerns.

The hotter it gets the faster it creeps. It isn't that it creeps whole millimetres. It doesn't have to. The thing is, for a low resistance connection, a battery terminal bolt needs more torque than that required to compress your typical split helical spring washer. So as soon as the PCB has relaxed enough that your split spring washer starts to open up even the tiniest bit, your tension is too low. Then you get the whole -- high-resistance makes hot joint which makes faster creep which makes higher-resistance -- thing. Positive feedback with negative consequences, such as a fire.

Then Warrick at Electronic Innovations gave us the answer.

The correct solution for clamping a PCB to a battery terminal is the same one used to connect IGBTs and SCRs and DC bus capacitors etc. directly to PCBs inside variable frequency drives.

Namely a Belleville washer, also known as a disk spring washer or cupped washer or conical washer. They provide a much greater load than a split helical spring washer. A wave or star washer might have a similar benefit.
http://en.wikipedia.org/wiki/Belleville_washer

Now that I reread the link Coulomb posted above, I see that I already mentioned Belleville washers and star washers in the "Home made BMS" thread some months ago. The difference now is, I have independent confirmation that they are a good solution.

So we won't be using lugs, or lubs, no mater how highly regarded lubs may be as typographical errors.

Squiggles,

You can see that Rod Dilkes recommends your stud-with-two-nuts scheme for cable lug connections with his BMS modules on pages 6 of his installation manual:
http://www.evworks.com.au/store/datashe ... lation.pdf

Except the BMS module is on the bottom, and it's for a completely different reason, as you can read in the text at the top of page 7.

But we can't afford that much extra height above the cells in the MX-5.

Alan,

That sounds like a generous offer, but I'm not sure exactly what the offer is. We have a second prototype board that is not populated. Please explain further. Off list if you prefer. Thanks.

Post by **gpsnettrack** » Mon, 26 Oct 2009, 05:17

Sorry guys, what I meant was, if you have a board that you have fried , by accident of course, we could have a look at how it could be populated another way, just like a independent look.

If you have a shopping list of parts you want or need just email me a list, we may have them at work already just sitting around.

alan@netcard.com.au
alan@gpsnettrack.com.au

Post by **gpsnettrack** » Mon, 26 Oct 2009, 05:41

Post by **coulomb** » Tue, 27 Oct 2009, 04:42

The panel has been delayed by a week, so we've made full use of the time to make many changes:

Some changes over the last week or so:
* We decided against the resistor arrays after all. They are so small that they are difficult to solder, plus they are actually difficult to source with 1% tolerance. It actually works out cheaper with individual resistors, anyway (neglecting assembly cost). It allows us greater flexibility with component values as well. It was a great design exercise, though.
* The board has been shrunk to the height (edit: as seen on the screen, really width in real life) of 4 5-watt resistors jammed against each other, and as small as possible to get the best tiling. The LEDs will be mounted above unplated slots in the PCB. They look so cute that way; the fibreglass is somewhat translucent.
* Many labels are in text on the blue layer now (will be on top). We won't bother with the expense of a second component overlay.
* We have several routed slots on the board for high voltage isolation.
* The external connections have been moved around to avoid being under the bypass resistors
* The bypass resistors will be on the other side of the board to the main components
* The large expensive FET is replaced with two smaller, much cheaper FETs
* LED resistors increased from 330R to 1K2 to reduce brightness and current draw (they were painfully bright up close)
* We are back to two monitor resistors again, but only one will be populated on most boards
* The expensive but cute push button has been removed altogether, replaced by a pair of pads designed to be shorted with a screwdriver. One of the pads will be a bypass resistor lead.

While we've won a few millimetres by turning the boards upside down, we will lost most of that because the resistors still need a few mm of air beneath them. Also, even with 1.09 A of bypass current, the heat eventually gets to the thermistor and the board gets an over temperature alarm just from the heat of the bypass resistors. That doesn't happen if the resistors are off the board by about 2 mm.

Separating the resistors by about 1.5 mm had little effect on the eventual temperature, whether the resistors were hard against the board or not. So that's one reason to squeeze the board size down.

Edit: uploaded image. Thanks, Andrew!

Post by **Squiggles** » Tue, 27 Oct 2009, 21:08

Why not drill a large hole under the LED.

Post by **coulomb** » Tue, 27 Oct 2009, 23:17

Squiggles wrote: Why not drill a large hole under the LED.

Alas, a hole large enough to clear the square protrusion under the clear dome would wipe out too much of the pads. A slot of 65 thou width is less than the 90 thou spacing between pads. BEC drills are actually metric, with half millimetre step sizes, so a 2mm would be too small, and 2.5 mm (98 thou) would be too big. Routers can step much smaller sizes than drill step sizes. We're hoping that they have a (say) 50 thou router bit that can make a 65x60 thou slot (plus space for the semicircles at the ends).

I actually used a 2mm drill for the current prototype, and wondered why I had such a hard time getting the LEDs to sit still while soldering. It was because the 2mm hole doesn't clear the square protrusion. The LEDs are springy little blighters, and weigh 1.5 poofteenths of a gram

.

Post by **Squiggles** » Tue, 27 Oct 2009, 23:40

I can remember seeing SM LEDs that were made for the mounting arrangement you need....can't find them in a catalogue though

Post by **coulomb** » Wed, 28 Oct 2009, 02:21

A fuzzy picture of the LEDs showing the pedestal in red:

Post by **coulomb** » Wed, 28 Oct 2009, 02:37

For completeness, the latest circuit.

And a closeup of some of the gory detail:

Note mirror image writing on the blue side; that's the side that will end up being "up". One of the LEDs (D3) can be seen in the bottom left corner. As well as its actual pads, it has its outline, part name ("D3"), the diode symbol showing where the cathode is, the "T" which is the marking on the actual diode for the anode, "R" to indicate to the installer that it's red, and on the back of the board "ERR" to indicate to the end user that this is the latched error signal. Phew!

The resistors stack up nicely; the actual resistors only take up about half the space taken up by their silkscreen overlay and part numbers.

The parts have been renumbered since rev. 31, so there are no more gaps caused by the hectic changes.

Post by **coulomb** » Fri, 30 Oct 2009, 05:46

Back to battery racks today. We needed to make these cut-outs in the rear battery box lid:

Somehow I lost or forgot to take the photos of the cage with the cutouts; that would have made it more obvious. Now the cage is clamped in place, ready for next EV day.

From above the boot in a test fit, and from the side:

We think that this is the final position of the rear battery rack. It's a little lower than we wanted (i.e. the ground clearance won't be spectacular), but it will still be higher than the original rear tie-downs:

Edit: the rear tie-down is normally on the other side of the panel it is temporarily bolted to in the photo. We are unlikely to be able to retain the rear tie-downs. Their original position marks the original rear ground clearance.

Edit: this is the sort of image I wanted to post:

Post by **coulomb** » Fri, 30 Oct 2009, 06:01

The question mark on the "structural member" above is because it really only holds up the back of the boot and the rear bumper bar. This was pointed out by Jeff Owen, who owns a red Honda City electric, visited today and kindly lent a hand and some advice. Thanks, Jeff!

So we could possibly cut into that member, instead of cutting the lid around the member. However, we'll leave that option for the next MX-5 conversion, after we get to know our engineer a bit better. Other options suggested by Jeff include a sheet metal box with reinforcing that replaces the member altogether, or a sheet metal box or rack under the boot, held up by bolts through the chassis member. The member could be reduced in height to get the clearance we need. Apparently, members can be cut and a new piece of metal is welded on the bottom, leaving its strength largely unaltered, and probably plenty strong enough for the back of the car.