Power supply EMP protection

My starting point is to connect a hefty TVS in parallel with the output caps, add a series 100uH (give or take) choke to the V+ line and connect a GDT between its other end and V-. The GDT side will be the PSU output. A kind of a PI filter. Does it make sense?

PSU -*-L-*-hostile_world | | TVS GDT | |

-----*---*----

Best regards, Piotr

This is a really hard problem. If the powered unit at the end of the

100m wire can be fully floating, it gets easier. Although if somebody were to have their hand on it when the EMP occurred they'd get a big jolt.

But, if both ends need to be tied to a local ground, then it is really tough. The 100m wire forms part of a loop, the earth completes the loop. Depending on how close the lightning bolt is, it couples magnetically to the loop, inducing a current.

For signal wires, you can put some resistance in series, which makes the GDT and other protectors work a lot better. Obviously, for multiple amps at really low voltage, you can't have much series resistance at all.

Is there any way you can put the converter at the load end of the wire? The input end of the converter might be a lot more robust against transients.

I've had plenty of stuff I thought was quite robust damaged by NEARBY lightning, and I'm talking about hits that were several hundred meters away! The longer the wires, the more current you develop.

Jon

There will only be this multiwire cable (probably ethernet or something comparable), no earth connection. So I am trying to protect against the surges induced in this cable only. Would using a shielded cable help noticeably? I know what a Faraday cage is, but I don't know how would the FTP-like cat5e aluminum foil shield behave during such a massive dI/dt.

Best regards, Piotr

That's the purpose of the TBUs: provide a really quick series dynamic resistance. +200 fixed ohms in front of the digital circuit, to be sure.

Hence the initial attemp with some inductance.

There will be many devices connected to the wire, like grapes, each of them will have their own protector and a converter. Sometimes an LDO, sometimes a switcher, depending on the load. None of them is supposed to consume more than 300mA and that is well within the existing TBU capabilities. But IMHO it doesn't solve the problem of protecting the main PSU, no matter what voltage it produces.

Exactly, so I am trying to design robustness into the project from the very beginning, not to random hack the network of modules afterwards.

Best regards, Piotr

Google EMP mill spec

e.g. Mil-STD-185 is a PDF

Piotr Wyderski wrote:

Chuckle. That's also what I was thinking.

I think he means either EMC (Electromagnetic Compatibility), EMI/RFI (Electromagnetic or Radio Frequency Interference). Instead of a general acronym, perhaps disclosing which country and which standards the power supply will need to comply.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

No, I mean ElectroMagnetic Pulse protection and I thought it was clear from the description of the problem. Purely a reliability issue, nothing related to any form of compliance. And Mr NoName is right, lightning and nukes have a lot in common, so the military experience might be valuable here.

As I said, it is all about making the damned thing bulletproof, not about pleasing a civil serpent. The country is Poland, but I don't consider our thunderbolts to be any different.

Best regards, Piotr

Don't forget to put something on the ground wire. At least:

PSU -*-L-*-hostile_world | | TVS GDT | |

-----*-L-*----

Pere

Lightning you get a huge current and if you are really out of luck vaporised phone conductors and an inconsolable temporarily deaf switchboard operator. A direct strike even with supposedly certified anti-lightning surge devices on the line between two buildings the current was so great that the thick copper earth had been mad hot and the potential from ground high enough to fry various components.

It was curious that buffered line drivers managed to survive by allowing more delicate components behind them to fail. It is too long ago to remember the details beyond that it took IBM most of the next day to rebuild the mainframe to partially working again and the entire of the next week before BT could get us a fully working phone system.

Nuke EMP you have the problem of sudden movement of electrons and breakdown voltages as the flash passes - countermeasures for that will not necessarily be at all relevant to a lightning strike. Beyond the observation that things inside hermetically sealed metal boxes tend to have a better chance of surviving.

One key thing that you have to remember is that how well the protection circuit behaves depends a lot on how much current your reference ground can actually sink before it ends up floating high. There is a hell of a lot of current flowing in a lightning strike arc. The frozen fractal burn patterns on things that take a hit are very interesting.

The hit on our building pretty much vaporised the downward run of telco cables at the front of the building leaving a 6" stripe of soot on the wall. You can't really survive a direct hit but you can hope to survive nearby hits provided that they are not too close.

--
Regards, 
Martin Brown

Yes. Average is 30kA, I've read. 500Mj and 15 coulombs, they say. Source:

Piotr, common-mode chokes. They pass along a DC voltage, while allowing the common-mode voltage to soar. If you have a string of powered bubbles along a long cable, you can go through a CM-choke at both the input and output of each bubble, so that the high ground voltage developed at the distant end is shared among the series chokes.

--
 Thanks, 
    - Win

The problem with lightnings is the ground bounce. Assume a grounding resistance of only 1 ohm and only 1 kA of current spike, the device potential bounces up with 1 kV compared to the neighborhood.

There are not much problem, if the device is only connected to the mains. However, if the devices also has a connection to some other networks (say telephone, CATV etc.) grounded at a different grounding o. There will be kilovolts between the local bounced ground and remote stable ground. This can cause a lot of damage, even if optoisolators are used.

Using fibers solve this kind of problems.

With devices all connected to local ground will also survive, if equipotential bonding has been used between local devices so that all devices bounce up without big ground potential differences.

Using fibers or at least optoisolators also help against nuclear EMP. It also helps using EMP protected rooms (lot of fingertock in the doors, using long 2 m high and 1 m metallic walkways working as waveguides, cutting out much of NEMP energy below 75 MHz due to wavequide cut-off).

There's no way you can power anything requiring 8 V 3 A through an ethernet cable. especially 100m of it. POE specs something like 250 mA at 48 V, which will deliver 12 W max. MAYBE, maybe you could deliver 12 W to a load 100 m away with the POE scheme, where the power flows over two pairs of wires. But, 8V/3A is just not going to work with those tiny wires.

I don't think the shield will do anything at ALL for the EMP case. The magnetic field just goes right through the shield (and induces the same voltage on it, too.)

Jon

No, nearby lightning is a kind of EMP, although over a smaller domain than the one the nuke guys fret over. If you get a lightning bolt within a few hundred meters of a place with electronic gear connected by wires, you will likely have damage. Ethernet ports are galvanically isolated, which helps a lot, but they can still get popped by a nearby strike. Other stuff that is NOT isolated, and may have loops set up between mains wiring and, for instance, phone wires, cable TV/Internet, water pipe grounds, etc. can lead to large circulating currents and damaging voltages.

Another thing is ground voltage gradients, when the lightning hits a tree a couple hundred m away, you can get thousands of Volts across a house, from the mains service entrance on one side to the phone or cable service on the other side.

Jon

Ha! This happened to me once. A bolt hit the power pole at the end of the street in our development, and you could plot the damage-vs-distance...

First house it melted the copper wires

Next house, the wires survived but the appliances didn't

Next, appliances were OK but electronics fried

We were #6 or so down the line. Fried the ethernet port in my hub (this was long ago) that happened to have a long cable in it, plus serial and parallel ports that had cables in them, plus the printer. Anything with short (or no) cables survived.

Even today, I have lightning protection in my outdoor thermostat as older revisions didn't, and occasionally a strike in the region would be enough to fry the electronics it was wired to.

Someone mentioned optical isolation. Please consider this as an excellent start. Your task, should you decide to accept it, is to design the light guides and absorption surfaces that have enough power. Do it right by design:

Permanent Certainty by Design is possible.

Well, that makes it easier. Ethernet is transformer-connected, the transformers will saturate and all that gets through in a big surge is some capacitive common-mode. A scattering of spark gaps or neon lamps will take most of the sting out of a nearby ligntning strike.

There's folk who put sparkplugs across the AC lines. Replacing plugs is easier than rewiring the house (don't use resistor plugs, though)..

Do you try to protect the power supply or the load device(s) at the other end of the cable or both ?

2x1.7 ohms. If the load can live with 5 V, i.e. 3 V voltage drop, 4

Consider feeding the 100 m cable with 50 Vac or 60/75/120 Vdc to remain within ELV range and use _isolated_ converter(s) at the load(s) to feed the actual load(s). Much thinner cables can be used.

So there are (multiple?) communication lines to other devices. Are all these powered through the same 100 m cable or some powered locally.

If the other devices are powered locally from local mains, I would _strongly_ recommend using isolated ports, such as galvanic isolation with RS-422/485, Ethernet or 20 mA current loop, which typically can withstand 500 .. 2500 V common mode voltages. There can be all kinds of ground bounce issues around the network, so good common mode rejection is easily . only thing left is handling the differential overvoltages, when you can forget most of the common mode issues.

If data communication is between two buildings with separate mains connection, I always recommend using fibers between buildings, while ordinary galvanic isolation within a building with a single mains connection is OK.

These simple rules work well even at large industrial sites even in the tropics with daily thunderstorms. Previously, sales persons objected about the cost of isolation, but I haven't heard complains for years from sales persons or end customers, since the system just works :-).

Both. They will be mostly low power devices, just the PSU is common and the aggregated load can go to 3A. The load-side protection is much easier due to much lower currents and perfect coverage by the ready-made protector families. 3A is well beyond this comfort zone.

They will all be 3.3V, so 4.7V drop is allowed, and most of them will be located closer to the PSU, not at the very end. But your calculations are correct, the ethernet cable must be excluded.

Good idea, but the 50VAC PSU would still need to be protected.

Just one CAN and one RS485.

All of them are powered from the cable, there is no external world grounding/local supply involved. The loads are mostly sensors.

Already done that way. :-) The only remaining part is reliable powering the whole caboodle.

Best regards, Piotr