BL> How long is it since you built DRIP? It must be five years! And
BL> I told you so!

RM> Nine years, at least. That's the oldest file I can find.

 Gee! 

BL> microchips is an ordinary resistor (they're all spiral cut
BL> nowadays) with a physically small capacitor to ground at the
BL> micro end. What value resistor and capacitor depends on the
BL> speed of DRIP (or whatever), and the slower the better. Ideally
BL> I used 10K and 2.2uF electro, but on a telephone dialler I
BL> dropped that to 2.2K and 0.1uF. Little electros are really good
BL> at absorbing sparks. The blue 0.1uF only last for ten sparks or
BL> so.

RM> The serial interface runs at 2400 baud. I think I'd prefer a
RM> series resistor and a pair of back-to-back shunt
RM> [zener-in-series-with-a- diode]'s to local ground. This should
RM> bother the serial data a bit less.

 No, no, no! Diodes and varistor-thingies are useless, they just
explode. The capacitor is *actually* a self-healing spark gap. At 2400
bd, I used an 0.1uF little-blue-thing capacitor and 2.2K. The 0.1's
tend to blow up too, but at least they protect the interface.

 The voltage you get across these things during a fast spark is
surprising, but if you keep it under about 200V, the IC usually
survives. A 1uF electro is best... about 30V with a slow front edge.

RM> Yes. I'm familiar with lightning protection in radio huts on
RM> mountain tops. Step 1 is to install a good ground just outside
RM> the hut. After that, connect that ground to a metal plate that
RM> all the coax cables pass through (through lightning protection
RM> modules that absorb high voltages on the coax inner
RM> conductors).

 We've been through this before... For normal electronics, I use the
"let it all jump up" approach, and forget about a solid earth. With
a transmitter that *has* to have a low-impedance connection to the
antennas, it's different. When you try to tie things down, all you
achieve is enormous current surges.

 Inside the box, everything should jump up together, and all
connection to the outside (except one) must be isolated to limit
currents (and consequent voltages). Usually, the *one* connection is
the mains.

RM> Then, inside the building, connect that metal plate to the
RM> mounting racks of the radios, and to the 240V electrical
RM> ground. Outside the building, install a ground at the foot of
RM> the antenna tower (sometimes the tower foundations will be
RM> adequate, but not always) and connect the coax shields to this
RM> earth as they leave the tower to go to the hut. Make sure that
RM> the highest point of the antenna tower assembly is the tower
RM> itself, and not an antenna.

 Yes... but that's not practical in a normal house.

 There are two distinct approaches to surge protection: the low
impedance high current shunt; and the high impedance high voltage
"jump-up" approach. 

 When we first put transistors into TV, we were blowing them up all
over the place. The picture tube arcs internally, and the rising edge
of the 20KV spark is in the nanosecond range. We tried your way first,
isolating the picture tube and shunting the spark, but it doesn't
work. The physical size of components alone creates capacitance that
gnerates enough current to blow them up. As soon as we went to the
high impedancejump-up approach, it got a lot easier (and cheaper). We
were the only one using that approach, and it's interestign that 30
years later, *everyone* uses our approach for TV.

 When I was faced with micros running remote stations in a lightning
strike region (Ohio in the USA is famous for it), I used a similar
method and it worked. The beauty of it is that you don't have ot do
anything outside. Let it all hang out, let the lightning come, and
simply isolate it. It means that the lightning finds its own high
current path, and the resulting voltages lift the *whole* electronic
box maybe 5KV above actual ground *outside* the box, but less than 30V
inside.

RM> Protection for Drip would be similar. At the Drip end, series
RM> resistors and bidirectional zeners to Drip's local ground. 

 And it didn't work...

 In my experience, zeners (and varistors) are useless. During a
spark, you get 1KV across the bloody things.

RM> At the PC end, ditto, although defining the "local ground"
RM> could be fun - I think I'd use the pc's grounded metal case,
RM> rather than the "common" connection in the RS232 cable.

 You can get kilovolts between the case and the PCB... I've actually
measured 5KV across a 300mmm length of braid. When I sparked in the
dark, I could see hundreds of tiny sparks run along the braid. The
*best* low impedance was a brass *strap* (0.5mm x 10mm).

 The approach that works, is to forget about tying things together
solidly, except at the end. Once it survives the spark *then* you add
the straps. My icemaker had all the isolating resistors at one end fo
the PCB, with the capacitors returned to a nice thick earth, and then
one *thin* earth running off to the rest of the PCB and the micros.
The mains also coame in at the isolated end, and a single strap
connected that thick earth to the case. It worked. Never lost one to a
lightning strike, and the one it replaced blew up every time.

Regards,
Bob


             

--- BQWK Alpha 0.5