Hi all,
right now I am totally demoralized, but I want this to be the day
I learn something and improve myself (I'm a self-taught hobbyst,
not an engineer, but with very high goals and made some really
advanced designs, and right now I am working (as I posted some time
ago) on a insect-like robot system solar-powered and permanently
connected to the Internet via GPRS, thanks which each robot swaps
informations with the others and help each other to reach a common
goal, e.g. to go from point A to point B of the forest (I own an
enduro bike (KTM 300 XC-W) and I know some places in the forest
where no human has been before) on a territory where just having a
GPS (like each robot has) is not sufficient, as you find crevasses
here and there, or hills that are too step for the weak robots to
climb, so the explorators keep in touch will each other, sharing
information, asking for help (it's my duty to get in in this case)
and to work together to reach the goal.. while everything gets
monitored into my home server).

Anyway, the story that is causing me trouble belongs to the past:

I have designed a datalogger for my grandpa, to monitor the levels
of his wells, temperature, humidity and so on, but expecially the
level of about twenty wells. I know nothing about geology, but my
grandpa is a geologist, and as the zone is landsliding he wanted
to find a correlation between the landslide and the levels (or
variations of) water in each well. Things that he has been doing
when he wasn't retired, and are not enough of interest to the local
government to get funds for. So he asked my help and I've been glad
to come in this project and do what I could, the dataloggers.

Anyhow, I designed the datalogger, built 5 of them, and placed them
around the territory. The most tricky part was that the piezometers
could run hundreds of meters of cable before reaching a datalogger,
and there could be several piezometers (all interfaced via 4-20 mA)
at opposite positions each other.

This has made me afraid of lightnings (I imagined that the two very
distant piezometers reaching the central datalogger may carry very
high tension), so I decided to add a lightning protection thereafter,
lightning protection is a matter about which I knew even less than
about geology (I'm much more for digital electronics). Moreover, the
decision was taken on the field, else I'd have the time to study it
well and choose the right components. When I went there to install
the dataloggers, I thought that the sensors would be all placed very
near to them, I discovered too late that in most cases this wouldn't
be possible at all.

The dataloggers do NOT have isolated channels, when my dsPIC turns
a piezometer on to read and record the data (every 60 minutes, else
the piezometers are always powered off to save batteries), in my
datalogger the A2982SLW IC powers the + side of the sensor (placed
hundreds meters away), and the return current goes, through a high
precision (further calibrated via software) shunt and a 10K limiting
resistor, straight into one of the 7 channels of the ADC, the ADS1243.

Since all piezometers are galvanically isolated each other, I never
though this could pose any problem.. of course until I heard that the
piezometers had to be installed hundreds of meters away each other. :(

I knew that in case a lighting hit a datalogger, a sensor or a cable,
that part of the system wouldn't survive, but anyhow I installed the
best thing I found there at a local shop, gas discharge tubes, those
with 3 terminals (I connected, at the datalogger side, the + and -
wires to the GDT left/right pins, and I connected the central GDT pin
to earth through a 1.5 meters copper pile vertically buried into the
terrain. The bad note is that I couldn't find any GDT rated at less
than 200V.. if I had found one at, say, 30V, I'd have been pretty
confident that it would have served its scope well, but as I had to
install 200V ones, I had to hope that the extra tension induced by
a lightning had to rise very very fast to get the GDT in condution
immediately, and had to fall equally fast, to avoid consequences from
the 30..200V unprotected range.

Sorry if I use wrong terminology, I'm used to the Italian language,
as it's my native tongue, and maybe the translation I get in my
dictionary is not technically correct.

All dataloggers, like I'm used to do with electronics I've to leave
on environmentally harsh places (whereas the IP66 box) or that could
be possibly stolen, were all potted into ROBNOR PX700K epoxy resin:
http://it.farnell.com/jsp/search/productdetail.jsp?sku=537330&_requestid=341872

The system has worked wonderfully for days, the system was calibrated
to gain further precision (although the shunts were IIRC less than 1%)
and all verifies made by measuring manually the depth of the wells
reported/proven that the system was extremely accurate.

After 3 days, one even with a violent temporal, I left the system
perfectly running with a sense of satisfaction, although the lighting
protection was deeply worring me anyhow, although it had survived at
least for the first (even violent) temporal.

I have left my dataloggers for 6 months alone, my grandpa country
is in another, distant region, and because of some problems of my
own I couldn't go back there until now, although I had planned to
go back after 15 or 30 days maximum. :(

Well, yesterday I returned there, and I found 3 out of 5 dataloggers
dead. :(

One has the tantalum cap burned, a clear sign (to me that designed
it) that the input channel brought high tension through the A2982SLW
to the main power, and thus burned the tantalum cap (and the rest).
The other two had no tangible sign of damage, but all is buried into
the epoxy as I said.

Of the two dataloggers that survived, one had some channels nonworking,
but it's probably due either to the piezometer broken or a broken cable,
unfortunately I had to take the dataloggers and bring them back home to
my lab, and I didn't have the time (or the mind) to verify if the sensors
are ok, still there, etc..

But two data loggers survived.. and the firmware worked flawlessly (not
even one WDT reset was necessary, etc..).

Sorry for the long post.. here my questions go:

1) is there any way I can dissolve this epoxy resin now that I could
learn something by seeing which components burned? (the only visible
one is the tantalum cap accross the batteries, just in one datalogger).
I used ROBNOR PX700K epoxy resin:
http://it.farnell.com/jsp/search/productdetail.jsp?sku=537330&_requestid=341872

2) as although I love my grandpa I don't have the time, expecially now,
to redesign everything from scratch, but only to improve on the existing
design and rebuild 3 new dataloggers, I'd also like to keep the existing
PCB's I have if possible. This means that the datalogger remains, in
substance, hundreds of meters of cables that run, through a shunt and a
10K resistor, directly into each of the 7 channels ADC pins. A weak design,
BUT could it work if I improve on the external lighting protection instead?

3) the GDT's.. even now that I looked into a suitable, low voltage model,
I cannot find any GDT that goes down to ~30V. Needless to say, adding low
voltage tranzorbs/transils to a 200V GDT doesn't make me feel safe, as they
could prevent the GDT's from triggering, and/or fail short, or work only once.
The best I came as idea so far is to keep the datalogger as is, and to design
an additional pass-through for each channel, independently powered, where
a mosfet senses voltage coming from the datalogger (thus if the channel is
on) and thus activates a relay that puts the channel truly online (through
those hundreds of meters of cable). This way, except for about 150ms each
hour, the datalogger channels would remain galvanically isolated (relays
are rated even 5000V if I'm not mistaken) from the cables and piezometers
and even more from each other, and (unless I get a lightning strike into
that short 150 ms window) the 200V GDT's would be certainly adequate to
protect the relays. Do you think this can work as a pass-through solution?

4) Should I add GDT's (and copper piles) even near the sensors, or at the
datalogger side it is enough?

5) I wish the cables were externally conductive (shield) but they are not.
Should the cables's be, too, connected to GDT's and copper piles every, say,
100 meters? Or, again, surge protection (to earth) at the datalogger side
is enough?

Anything obvious that I miss? I have Googled for 2 days and nights, read
a lot about surge protection and lighting arrestors, but I still have the
feeling that I don't understand this matter.

And to quote someone else, the problem is that just because you survive a
lightning strike today doesn't mean that it was the same condition as the
lightning strike that took out everything last month. It takes a long time
to prove a solution in the field especially if it's not raining, and here
where I could test it, it doesn't rain often, especially in this season.

The dataloggers haven't costed me much in terms of money, they have costed
in terms of time but it was the pretext to learn a lot of things I hadn't
had a chance to look indepth before (dsPIC, SPI, and more).. what really
makes me sorry is that my grandpa, who's not really young, has lost
6 precious months worth of data because of my technical limits, and he
has always had the highest regards for my achievements.

Thank you very much for your time.
Mario

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist