For my grandpa's country I am going to install tens of 4-20mA sensors,
which will be routed through some hundreds meters of good cable to the
data logger, at the border of the house. To give some (yes, I know it's
impossible to give total) protection against nearby lightings, etc.. I
thought about adding surge protection.
If I understand it right, I should put one device in parallel with each
sensor, and another device in parallel with each data logger input (i.e.
the two ends of each cable), right?
Varistors, gas-dischargers, etc.. to me it seems that a Transil diode may
be a very cheap solution but a very effective one nonetheless. Am I wrong?
And, being there two in parallel for each cable (one at the sensor and
the other at the data logger input), are they going to false the results
by much? I am aiming at 16bit resolution, with a full scale precision of
If you want to protect the datalogger and the sensor then you will need
protection at both ends. Just using diodes will not give you sufficient
protection, you will need gas discharge tubes as well.
If you want to protect the sensor and the electronics are not earthed, a
parallel connected device will be OK but for the datalogger you will
better with a series connected device, this will give you more
protection and react quicker.
Any true industrial product will not introduce any errors into your
signal, especially as you are using 4/20mA signals rather than voltage
Same protection that permits your telephone CO to operate without
damage during every thunderstorm is also your solution. Remember the
difference between differential and longitudinal mode currents? You
are installing differential mode protection. But destructive surges
are longitudinal mode. A surge voltage is same on both wires
(overhead or underground) as current passes destructively through
datalogger to earth ground. What would a zener diode between wires
see? Zero voltage as thousands of volts confront a datalogger.
How does a telco everywhere operate during every thunderstorm
without damage? Bennison, Ghazi, and Ferland measured surges on
telephone wires during thunderstorms in 1968 and 1969 in IEEE
Transactions on Communications. Hundreds of transients occurred in
each thunderstorm. Yet damage is unacceptable. Protection during
thunderstorms is that routine in every day in every town even 70 years
An application note from one industry professional demonstrates the
wire that enters a structure must connect to a single point earth
ground - either directly or via a protector. In your case, a
protector may be a transzorb, gas discharge tube, or MOV. Protector
is not protection. Protector is a connecting device to protection.
What did Bennison et al demonstrate? Destructive transient is
longitudinal. It seeks earth ground either via a data logger OR safety
earthed before entering a building. Protector must dump a surge into
earth long before it can get to the datalogger.
Essential to protection is single point earth ground. Incoming
wires inside every cable must connect to that same earthing electrode
at the service entrance. No earth ground means no effective
There are two reasons to install protectors: First, you need to protect
the nearby structure from fire caused by the discharge. Second, you
might want to continue working through the discharge and maintain
minimum noise pickup. These goals often conflict with each other.
The goal of fire safety is to limit the damage by grounding the cable
shield frequently. However, this can introduce quite a bit of noise.
Some multi-conductor cables have two shields: One around the bundle, for
grounding everywhere, and one around the pair for grounding locally.
The problem you get in to with long runs of 4-20 mA current loops is
that the local ground potential at each end can be very different. If
you put a surge protector at each end, then one or both are almost
guaranteed to fire with any nearby lightning strike.
My suggestion is to use an isolated differential instrument. Ground
everything to ONLY ONE ground and put the protectors there. Use
whatever discharge devices you can afford. In my experience, what
matters more is that these things get replaced right away.
It's also important for you to use isolated analog inputs. Having one
input fire and not the others will result in a lot of damage unless the
inputs are isolated.
People write books about this stuff. This is just a very quick
overview. I suggest you read up on some telephone company grounding
manuals. They have experience with this sort of thing...
Jacob Brodsky, PE
Thank you for all the replies, they were truly enlightning.
First of all, maybe I should stress that the multiple sensors and the
single, multiplexed data logger, are completely isolated with respect
to the ambient (and the latter is battery powered, and powers all of the
sensors), although the sensors and the data logger aren't galvanically
isolated each other (I am using MIC2982 chip to power on/off each sensor
at will, of course respecting their "warm up" specs, and an ADC with
multiplexer to select the input from each channel, which has its own
sense low-tempco resistor. The data logger has its own memory, which
I will read every n months or maybe I may add a GPRS later).
This "system is isolated from Earth, but components aren't from each
other" probably was already clear, so I'm not going to change a bit,
but I wanted to make it clear since none of the sensor nor the data
logger is connected to Earth ground.
Anyway, I understand now that I have to use a device such as this one:
which combines the strenght (5KA) of a gas discharge tube with the
quick reaction of two bidirectional tranzorbs, integrated in the same
package. Thus my sensors protection would become:
sensor 4-20mA power ------*-----*--------- to data logger 4-20mA power
Tranzorb | |
Earth Ground ---*----Gas
Tranzorb | |
sensor 4-20mA signal ------*-----*--------- to data logger 4-20mA signal
after some hundreds meters of cable, each data logger input would get:
from sensor 4-20mA power ------*-----*-------- datalogger 4-20mA power
Tranzorb | |
Earth Ground ---*----Gas
Tranzorb | |
from sensor 4-20mA signal ------*-----*-------- datalogger 4-20mA signal
thus the GDT/TVS hybrid will divert to Earth longitudinal currents/voltages
(well, I'm not really sure what "longitudinal" means here though: common
mode?), so that a really high (billions of volts?) common mode transient
doesn't try to reach Earth through the datalogger (which *currently* was
thought to be completely isolated from Earth, and battery powered. What if
I mount it 30 cms above Earth? why would a transient "want" to pass through
the data logger, when it is quite isolated from Earth anyway?).
I guess I better still keep ALSO the unidirectional 24V tranzorb to protect
from possible "transversal" overvoltages, since the two bidirectional TVS's
would do it too but are rated a too high voltage.
But, what if in the site where I have to place the data logger and/or each
sensor (water level sensors) there's no ground connection? Should I bury e.g.
one meter of copper wire to create the ground? Some sensors will be immersed
under 100+ meters of water.
Also, one thing that worries me is that, being the sensors put far away
with respect of each other (its like a "star" system where the data logger
is in the center) a lightning somewhere will create a big difference of
potential in a sensor versus another opposite sensor, and the cables will
bring this difference right into the data logger.. thus in theory I would
also need to "gas discharge / tranzorb" (meant as a tense here) each wire
with each other wire, where the combinations would explode.
Finally, I understand that the SL1122A200 will be totally ineffective if I
don't connect it to a low impedance, true, Earth ground. But there's none
currently.. so what would be a good way to do it? Burying how much ground
Thank you very much for all your support, each opportunity like this one is
great to learn new things.
Common mode or longitudinal mode: point is that current on any or
both wires will even pass through that datalogger to obtain earth
ground. When discussing these currents, resulting voltages can make
linoleum, wood, concrete, and wall paint electrically conductive.
Even if all inputs have galvanic isolation, still, that only becomes a
capacitor to longitudinal surges. Isolated to AC electric is not
isolated to surges. Everything becomes conductive - including the
best electrical insulator: air.
All appliances (including datalogger) contain internal protection.
So that internal protection is not overwhelmed, earth that Semitron.
An earthing connection that is as short as possible ('less than 3
meters'). Same electrode that everything else (including AC power to
the datalogger) uses. If the datalogger uses two wire AC, then it is
not isolated from earth - when discussing transient voltages and
currents. What is isolated for human safety is not isolated for
transistor safety. Why? Different voltages define same material 'as
and as not' conductive.
Again, refer to that application note:
the building (ie datalogger) and antenna tower (sensor) have
earthing. So that sensor is not harmed, a Semitron or something
equivalent (see also Sidactor and see warning below about its
operating voltage) earths sensor end of 4/20 ma wires. As long as the
sensor and its wires remain at voltages same as earth beneath, then
sensor is not damaged. With a protector on sensor end, then surges
from the earth or surges via wire (from other sensors) are not
Telco does same thing. Lightning to any other building in town is
equivalent to a lightning rod connected to their $multi-million
computer. Telco does what you are doing - earths wire as it leaves
that other building and again earths wire as it enters their
building. Earths via a 'whole house' type protector - because
destructive surges will find earth ground even destructively through
isolated (ungrounded) equipment.
Currently Telcordia (the standards organization for all telcos) is
upgrading GR-1089 - their standards for surge protection. For
example, gigahertz ethernet cable new standard requires an earthed
protector from each wire to earth ground AND no protector between
twisted pair ethernet wires. A protector between wires is for surges
that typically are not destructive which is why new standards do not
call for a protector between wires. Standards are quite blunt about
what is necessary for communication wires (such as ethernet): a
protector from each wire and short (less than 10 feet) to a single
point earth ground.
Appreciate the circuit: a constant current source (not a voltage
source) creates destructive surges. Your concern is not voltage.
Your concern is current. From first year EE: voltage will increase,
as necessary, to conduct that current (which is also how pre-1970s
cars created 20,000 volts on spark plugs). If attempting to stop a
surge, then voltage will increase as necessary to overwhelm that
"dam" (galvanic isolation, the isolated datalogger, or spark plugs).
Protection can never stop that current because even 4 kilometers of
air could not. Protection means shunting (conducting, diverting,
clamping) current to earth so that near zero voltage occurs on
datalogger or sensors. This shunting so that component or datalogger
protection is not overwhelmed.
Even with an isolated datalogger, voltage will increase as necessary
to conduct to earth ground. Again, all electronics have internal
protection. Earth a protector so that current will not overwhelm that
protection. That datalogger sitting on a table remains a potentially
destructive path to earth - via table top.
Why did Ben Franklin invent the lightning rod? Because wood (in
church steeples) is an electrical conductor (like a tabletop) and
because that electricity cannot be stopped using isolation. Best
isolation - 4 kilometers of air - did not stop it.
How would I create an earth ground? Datalogger is AC powered? Then
the same 3 meter earth ground rod required electric power source must
also be connected 'less than 10 feet' to the Semitron earth ground
terminal. Ground wire must have no sharp bends, no splices, not inside
metallic conduit, separated from all other non-earthing wires, and
remain independent of other earthing wires until all meet at the
At the sensor end, a one meter ground rod would probably be
sufficient. But grounding electrode must contact soil below the frost
line. If wire is deep enough, then a one wire solid and bare copper
wire could be laid in the trench as an earthing electrode. That wire
should be solid, copper, and 6 AWG or larger. Water is not considered
sufficiently conductive. Earthing electrode must be in earth -
either under water or located where wire enters water. If this is
water that people swim in, then that earthing should also connect to
anything adjacent that is electrical and earthed (ie overhead
Appreciate a lesson from scuba diving. When lightning struck the
surface, then divers floating free were unaffected. But those
touching the bottom were connected to the better conductor - and felt
Generally, 4/20 ma sensors operate at voltages less than 30 volts.
Your Semitron conducts nothing until voltages exceed 200. Can your
sensors withstand 200 volts applied between any one wire and all other
wires (or its body) without damage? Best check those specs for
longitudinal mode voltages. Would a 60 volt protector be a better
choice? Semitron is the right idea. But its operating and sparkover
voltage may be excessive. Telephone appliances can withstand 600
volts without damage. Semitron is for telephone appliance
protection. Consult sensor voltage specs. That protector voltage
must be below internal sensor protection voltage (max voltage between
any two points).
Some first dig a hole, drive an earthing electrode into hole's
bottom, then use a 6 inch plastic pipe to form a 'manhole'. Earthing
connections (and maybe sensor) can be inspected, protected, and
repaired. Connection points being the most frequent failure point and
where diagnostic testing is best conducted. A cap atop that pvc pipe
means the whole assembly can be buried or covered so as to be
transparent to casual observers and so that connections are protected
from earth and water; and yet be easily accessed.
Which is why there are two distinct names for the functions (Shield and
Screen). On things like Type 23 frigates the shield was grounded at each
and every bulkhead it passed through while the screens were grounded only
at the instrument rack ends. However, Grounding of Shields is a matter for
individual installation analysis to suit the environmental conditions
appertaining to the installation.
What I use as standard. It precludes consideration of using equipment from
the likes of National Instruments for fast data aquasition in many of the
environments I deal with because of their lack of isolation and their
propensity to poor-man's differential input (use of two inputs combining to
provide the differential input).
Galvanically isolated, Energy Managing, Surge Clamped inputs and outputs for
everything is fairly normal practice for me (analogue and digital). In
short, attention to detail (such as ensuring enough terminals for each and
every conductor, decent signal earthing scheme, decent protective earthing
strategies, isolation strategies, termination management, design for safe
testing) is very important.
Not only the tel-co's but also the Oil and Gas Industry, and the Railway
Industry have large amounts of information on this topic.