High Tensions
In email, NevilleH mentioned something about Bill Dube (the Killacycle guy) and 1 mm per volt clearance. A little research (thanks Weber) turned up this thread:
SRJC student's Focus-conversion-EV burned to its metal frame
There is an overview of the vehicle
here, but (as yet) no information about the fire.
The actual reasons for this fire aren't clear yet. It will be interesting to read what caused it, but in the mean time, it sparked some other discussions, like these, where it was recommended that there be 1mm separation between conductors for every volt of potential difference between them:
(single posts now)
http://www.diyelectriccar.com/forums/sh ... hp?t=56648
http://www.diyelectriccar.com/forums/sh ... hp?t=56743
http://www.diyelectriccar.com/forums/sh ... hp?t=56758
http://www.diyelectriccar.com/forums/sh ... hp?t=56759
This next one introduces the scary concept of the "plasma daughters" and how "plasma boy" got his name:
http://www.diyelectriccar.com/forums/sh ... hp?t=56775
I actually read the full Plasma Boy article; it's long but interesting (not all of it is relevant to the plasma subject, however. If you do skim it, make sure you don't skip the bit about the marshmallows; hilarious).
Ok, well, where does that leave us?
First of all, although they mention these plasmas being fed by metal vapour, I'm not at all certain that there would be enough to sustain a plasma for minutes. So surely the safe limit would be closer to the sparc in air limit, which is about 30 kV per centimetre (3000 V per millimetre). That's a huge difference, and I'm nowhere near confident enough about the physics to say all those experienced guys must be wrong.
So for now, let's just consider 1 mm per volt and see where it goes in a high voltage conversion like the MX-5. We have 228 cells in series, for a total of some 228 * 3.6 = just over 820 V across the whole pack (if and when we get to series half-pack (Low Voltage, LV) mode; maybe we'll stay with parallel half-packs at "only" 410 V). That requires 820 mm (410 mm for the LV mode) separation at worst case.
We have a creepage distance of 5 mm on our BMS boards; with this figure it's good for 5 V (TTL signals)!! We do have notches in the PCB under the opto isolators, but in some cases, these notches are required to keep the creepage distance as large as 5 mm! (In other words, the air gap is treated as a perfect insulator. But if a plasma starts, air is way worse than most insulators. This creepage distance is mainly to avoid dirt and moisture tracks that can become conductive).
Suddenly, our whole BMS design is looking unsafe. Granted, it would be rare that the full 820 V would appear across a 5 mm gap, but it's almost certain to happen at one point or another, if the pack ever develops a leakage to chassis. So we can't use arguments like that; worst case has to be totally safe. The last thing we want to do is to set back high voltage AC conversions by a decade because we didn't do our research correctly. Plus, we'd rather have the MX-5 to drive than to have it used as a bad example for future conversions.
So what can we possibly do about it? Resourceful as ever, Weber came up with the following "industrial fibre" solution:
Weber email wrote: Transmitter is an IR LED with lens and coupling
http://i-fiberoptics.com/pdf/IFE91A.pdf
Footprint 25x9 mm. Height12 mm.
$3.82 from Digikey
Receiver is a photo transistor with lens and coupling
http://i-fiberoptics.com/pdf/IFD92.pdf
Footprint 25x9 mm. Height12 mm.
$4.01 from Digikey
The "cable" is $19 for a 10 m roll. It's cheaper than the TOSLink audio stuff and is "connectorless". Just cut the cable with a sharp knife at right angles, poke it into the transmitter or receiver on the PCB and tighten the knurled knob.
Believe it or not, Weber had two lengths of this cable in hand; part of a tubular guitar project from another life. It's actually plastic, and can be bent to a minimum radius of 25 mm. Not totally convenient, but certainly doable.
It means that we can eliminate any proximity of high voltage components with chassis, on the BMU boards. (The current design uses 12 V and ground from the auxiliary battery to power the end-of-row circuitry. It happens to dovetail with our existing experiments (thanks, Neville!) with infra red optical connection between BMUs. We've shelved that idea for the time being, as it seems to be too sensitive to sunlight; convertibles are liable to see a lot of sunlight at times!
We will likely have to drive the transmitter end with significant current, either 20, 50, or even 100 mA. This does increase current consumption, but the current will be on for only a quite low duty cycle, except when performing a download, and these are infrequent (and never happen on the road). So we may consider evening out the current load, and drive the optos (which we will still use for the 95% of cases where cells are next to each other) with more than the present ~ 2 mA, so we can use cheaper optos. In fact, they can also be smaller optos, since they will be isolating only 3.6 V at most, so there is no need for a large package with a PCB notch under it. [ Edit: we just realised that if a cell goes open circuit, there could be pack voltage across the optos, so actually the long format optos and the slot will remain. But we can still go to cheaper optical isolators. ] Any time that communications have to leave the immediate vicinity of the BMU, we will use the fibre connection.
Another place where we have to consider clearances in battery boxes where a series string of cells turns around (does a U-turn, if you like). One box for example has two strings of 23 cells (and another string). The string of 23 enough that we will have a contactor separating the two strings, so if a plasma event starts for some reason, at least those two strings will be separated by the contactors dropping out. We believe that the EV200 contactors can interrupt the full pack voltage with full short circuit current (if only a few times in their lives; probably we'd replace them after any such event).
Other boxes however have the limit of 28 cells arranged in two strings of 14 which turn around at one end. The initial plan was to simply cable to two strings of 14 cells. But that means that we end up with 28 * 3.6 = 101 V maximum, separated by less than 100 mm. So we considered wiring the strings with a "long link" the length of the string, so that if you dropped a spanner across the gap between strings, you'd see a constant 50 V plasma event (dangerous enough as it is), as opposed to a 0 - 100 V plasma, depending on where the spanner was dropped. The average case is the same (50 V), but the worst case is better (50 V verses 101 V). But then we have an extra cable in the battery cage, running past all the other terminals on the way. If a plasma event were to start, surely one could not rely on the (double) insulation of the cable to last, so that introduces many more places for the plasma to go, all of them spawning daughter plasmas... shiver.
At one point, we considered putting Andersons at the end(s) of the cable, so that the unfused plasma can be broken [ edit: or prevented, when working inside the cage] . But that doesn't help when the plasma short circuit burns off the cable insulation. We considered yet another insulation barrier over the terminal posts. 25 mm cable conduit (the type with rectangular cross section and a snap-on lid being one side of the rectangle), and it looked promising for a while. However, that didn't leave enough space for the return cable. We also tried sawing round 20 mm conduit in half, making two semicircular sections. These fitted pretty nicely over the terminals, and we may end up doing this if we can think of a quick way of cutting the conduit without a really wide cut (or there would not be enough material left).
In the end, we decided against the "long link", and instead decided it would be best to put "half anderson" connectors (single pole; they don't seem to be available in finger safe versions) at the turn arounds. So this is the "medium link" option
.
[ Edit: The main point of the andersons at the turn arounds is that the battery can be made spanner-drop proof, or much more so, by disconnecting the half-way link before working inside the box. In the event of a plasma, I think it would take too long to get to the connector; better to yank at or chop the cable. ]
This is where a lot of our time goes, instead of actually building the car: figuring out how to make it safe. I guess that's one of the costs of being (almost) pioneers in this area (nods to acmotor and a4x4kiwi, and thanks guys for blazing the trail).
We may well have many other places where we'll have to consider how we could deal with a possible plasma event.
[ Edit: It's occurred to me that this 1 mm per volt thing only applies where it is impractical to fuse the conductors. So with many contactors in series, and a pack fuse, we don't need to worry about clearance inside the controller, for example. Similarly, if a cell opens, the high voltage that appears across a tiny printed circuit board is protected by contactors and the fuses, so a "plama boy" style event should never occur, or at least should blow itself out in a second or so, and not cause a significant fire hazard. ]