Hi,
I am measuring amount of water in our well. There are 20 wires of different length. Depending on water level some of them are submerged while others are not and I measure change of impedance. I am using 74HC4052 to multiplex those 20 wires to 6 Arduino pins and I have a 100unF 16V ceramic cap on each wire to block DC current and protect the wires from electrolysis. There is also one "common wire" that is deepest and that has no blocking cap - used for the impedance measuring and to prevent voltage difference on the blocking caps to be too large. You can get some idea from this image:
Everything worked well until a storm not so long ago - after the storm I started to get wrong data. I dissected the device and found two of the 100unF caps damaged (they are short circuit) and two of the multiplexers damaged - one channel is permanently active and the corresponding driving signal also shorted to the channel. The red signals in the image are shorted:
It is not hard to guess the storm applied somehow too much energy to the circuit. But on the other hand only 2 out of 20 caps and 2 out of 6 multiplexers were damaged. So the circuit "nearly survived". Now I need to know more to prevent such failure in the future. What caused the damage? I guess when a lightning strikes nearby large currents flow in the ground and since the ground has quite large resistance large voltage differentials are present everywhere. But how large? Also I would expect those potentials will have a very short duration and quite large internal resistance. But I know so little about this topic so I don't know what exactly to expect. Is adding a resistance in series with the capacitors enough? Or do I need proper TVS device? Which rating? Or was it only a bad luck 16V caps + internal ESD protection diodes of the HC4052 were not enough to protect from the storm? Or is it unlikely such poorly protected device suffers only minor loses during a storm?
Or is there some other danger for such device connected to the earth? Or is the connection to the earth only a minor problem and the damage was likely caused by a current induced in large loop formed by my long measuring wires?
I would be glad for any pointer to information regarding this topic.
Thanks.
There are protection diodes inside the Arduino, but they are low-current (see Figure 22 in the [u]ATmega datasheet[/u]).
Normally, you can add a series resistor to limit the current (and the diodes limit the voltage) but with thousands of volts you'd need a high resistance value, and it still might arc across the resistor. Of course the resistor would need to be inside the case and close to the Arduino, so any current has to go through the resistor.
I have read something about overvoltage protection and I think I am able to do something if I know the requirements. The problem is I have no clue what I face. What caused the failure? HC4052 has "absolute maximum" current of clamp diodes 20mA. Such current charges 16V 100nF cap from 0V to 16V in about 100us. Information on Wiki is a bit ambiguous about one flash duration: they say "average duration is 0.2 seconds made up from a number of much shorter flashes (strokes) of around 30 microseconds". In another part of the page they say "The electric current within a typical negative CG lightning discharge rises very quickly to its peak value in 1–10 microseconds, then decays more slowly over 50–200 microseconds." Now I don't know - it is possible the currents were quite low but caused too large voltage difference over the blocking cap. When the cap failed maybe it somehow damaged the connected multiplexer? In that case protecting the caps with parallel TVS diodes would be enough. But is it possible for a dying cap to damage the rest of the circuit? I would expect the energy stored in the cap is dissipated as heat when the cap fails, not as a current surge damaging other parts...
Or are simply voltages and/or currents too high and damage of the caps is unrelated to damage of multiplexers? In that case I need to implement proper protection on inputs. But what means "proper"? I know there are "step potentials". But how large can it be? Is there a difference between horizontal and vertical distribution of the potential? I understand it is impossible to protect Arduino against direct strike of lightning. But it is very unlikely and I am willing to take the risk. I expect someone somewhere have some guidelines for "protecting against nearby strikes". Something like "there is less than 30% probability to encounter 1kV/m potential per year". But I am unable to find something like that. Also I expect the transient voltages will have quite large "internal resistance" - the current must flow through water and soil which is not good conductor. So hopefully energy needed to dissipate will be "not so high". But again - comparable to "normal" ESD discharge? 10 times more? 1000 times?
I don't think any kind of electronic protection will do squat against lightening!
Your best bet would be to put a well grounded lightening rod some distance away. This should protect against a full on strike, and maybe then some of the conventional protection circutry will have a chance.
Smajdalf:
It is not hard to guess the storm applied somehow too much energy to the circuit.
It certainly sounds like it. There appears to be a link to the storm. But at the same time, it looks like the surge wasn't 'that' bad, because a big hit could mean vaporisation of wires and circuit board .... eg. no board left, carbon imprint etc.
For now, maybe put in a much higher rating capacitor, and monitor the situation during the next big storm.
You are correct that lightning cause voltage differences in things it flows through. It also induces currents in nearby things. A million amps going into the ground nearby can cause inductive currents in your wires.
I think the TVS diode is good. If you can place the TVS just after a protection resistor, that helps. Even if the resistor is a very low value, like 10 ohms.
For the TVS value, you can pick a number equal to your circuit voltage or just a tiny bit higher. A 3.4V TVS will conduct nano-amps at 3.3V but it will conduct many amps at 3.9V.
Another trick I saw on a weather station many years ago was to have every incoming trace on the PCB go past a ground trace. Those two traces had triangles etched so the two triangles pointed at each other with a 1mm gap between the points of the triangles. The sharp point concentrates the electric field caused by voltage and guides the spark to only occur at that one place.
Yes, use spark gaps, TVS, series resistor, the works - ultimately you can't prevent lightning damage, but
being more robust means surviving more events that aren't direct hits.
Opto-isolation is useful for limiting the scope of damage, but ultimately an optic fibre coupled opto-isolator
is needed when the voltages get really high, since 10mm gap isn't much. Potting up optoisolators can
improve the performance as flash-over risks are reduced.
Or wireless comms can be used.
Note that the most important danger is fire - where the wires come into a dwelling is the point you
want to protect in depth lest a nearby strike is routed direct into the house.
ChrisTenone:
I don't think any kind of electronic protection will do squat against lightening!
I would agree, I had a customer that suffered a nearby lightning strike.
No electronics or wiring were actually hit, but the discharge enduced enough voltage in just about every bit of nearby wire that it zapped lots of stuff, telephones, monitors, keyboards, LAN connections etc.
I've seen a masthead uwave datalink which copped a strike.
Returned for repair under warranty ! (oh really?)
Most of the aluminium casting had been melted, and the PC at the bottom of the mast had been fried.
So I suppose it all depends on what you mean by 'nearby'.
Various tricks may help , but nothing guaranteed... in particular attaching cmos gates to susceptible external bits seems highly dubious. Resistors? diode clamps? etc etc Perhaps nested earthed enclosures with layers of protection?
You can't fight gigajoules with any techniques I know.
Also, do you need continuous reading? If not, a simple way to avoid the danger is to leave the circuit as it is and replace the wires/components when damaged. Couple the measuring circuit only when you need to measure.
SADs work best (Silico Avalanche Diodes).
They have a flat response to the standard 5x20 uS pulse used for testing surge supressors. This type of equipment is expensive but it works. You did not state whether your power supplies ran off AC/DC converters or batteries.
Paul_KD7HB:
A nearby lightening strike is similar to an EMP burst, except lightening may contain several discharges. Research how to protect from EMP.
Paul
Sorry, I have to enter pedant mode, "lightening" is the act of making something lighter, "lightning" is an electrical discharge in the atmosphere! Noone has seen darkning though
No one has suggested burying the wires in trenches, as deep as possible. And separate the wires in the trench from each other. Conduit made for burial may be used.
In my limited experience with lightning I've formed two conclusions;
Nothing to my knowledge will survive a direct lightning hit.
For a nearby lighting strike the considerable energy directed into the ground dissipates in a near radial pattern resulting is a voltage differential of some number of volts / meter that drops the further you get from the strike zone.
To simplify the problem I would look look at it in with some simple assumptions ( perhaps not 100% correct but simple enough to move forward)
The voltage differential from the sensor(s) to the base circuit is a function on how close the lightning strike is. For a practical number I would use 100 to 200 V of ringing AC + some DC.
I would assume the current capacity is very high.
So your circuit needs to be able to survive the above voltages. OR have some sacrificial component that is easy to replace.
Sorry I don't have an exact circuit for you but I hope this somehow helps you find it.
That's why tge use SADs gor surge protection.
They clamp the voltage at the specified clamp
voltage but have the capability to conduct large currents. (see Joslyn datasheets ).
You need suppession at the main feed for the building and then at the local feed and last at the DC supply (input & output).
Sorry for not responding for long time - I wanted to write "proper answer" and a lot needs to be said.
I am sorry for the click bait title - I do NOT try to protect the device from direct lightning strike. I know it is nearly impossible to protect from it and I believe it is also nearly impossible for it to happen - after all our house with a lightning rod is nearby. Also as far as I know a lightning never struck "near". There used to be a charred tree nearby (~200m) likely from a lightning but it is many years ago. But on the other hand I cannot be 100% sure about this - maybe the lightning hit a nearby house's lighting rod or something else nonflammable and strikes in 100m radius are common?
@ JohnRob: thank you for the link to the document. I hoped for something like that - some (very rough) numbers to know what is likely. Do you (or someone else) have a link to some "more technical" and "less advertising" document?
Why do you assume current capacity is very high? The linked document as well as Wiki says 30 000 amperes is current of the lightning. It is roughly 1A/m^2 at 100m distance. Provided there were 2m of water in the well and it has 0.5m diameter only 1A wanted to pass via the water - and only part of this wanted to pass via my wires?
@ falexandru: the continuous measurement is the reason for the device. Disconnecting it is not an option.
Is the moment lightning strikes the only moment a harm to the device may happen? Meaning is it enough to be prepared "only" for very short pulses of high voltage and/or current? Or are there other less dangerous potentials?
My idea of lightning strike is that (positive) ions accumulate in the earth but since they "have nowhere to go" they are evenly distributed. When lightning strikes a lot of negative ions enters the earth at one spot "eating" the positive ions at the spot. So nearby positive ions "rush to fill the gap" quickly restoring even distribution. But maybe during the storm the ions slowly constantly shift around "anticipating" the strike - each time passing through my poor circuit slowly wearing down the protection diodes until something breaks?
