Lighning protection

Instead of the depletion mode FETs which are kind of expensive people often use PTC.

Be careful with the TVS. Those things have a large capacitance and if your data rate is too high you'd see marginal signal integrity or it won't work at all. One way (if available) is to use a helper voltage that is bypassed and limited. You could, for example, run a small current through a TVS or zener to keep it up at xx volts and then run a diode from the signal line into that. Same for the negative-going side. That way there is only the small capacitance of two diode which remain reverse-biased at normal operation.

A common mode choke can't hurt either. Mostly that's needed anyhow to pass EMC.

--
Regards, Joerg 

http://www.analogconsultants.com/

I would humbly suggest an RC right after the spark gap, maybe instead of the fets. One of my rules is that no pin of a semiconductor should have a direct path out to the world. The gap might let through a lot of spike before it ionizes. And I suppose it could fire and launch a picosecond edge along the trace to the fet.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

I'd add common mode choke + C after the GDT to slow down the hit. Use a sectional-wound choke instead of bifilar and you'll also get some differential-mode filtering. Lightning strike is usually manifests as a common mode signal.

LC+GDT+TVS-combination with common RS485 transmitters has survived quite well on a rocky hill with plenty of close strikes and bad grounding with

200m wires. No depletion MOSFETs there. Multiple ethernet switches have died even with 10m cables on the same hill and being further away from the antenna mast receiving the strikes.

Direct strike would of course be an another matter :-)

--
mikko

You might check the relevent sections of IEC61000 and IEEE/ANSI C62 to figure out just what level of transient or surge immunity that you intend and what exposure level is targeted.

For a specific level, on a specific equipment port type, there are intrinsic costs for predictable (and demonstratable) level of immunity.

I'm not sure the term 'lightning protection' will be an accurate enough description of the threat level, or whether data lines can be considered alone, without other ports/packaging being considered.

You'd also need to know what was being protected, at the physical and performance level. Are all RS485 isolated ports the same internal schematic? I doubt it.

RL

The spark gaps can pass brief peaks as high as 600V before firing, which would tend to make a mess of whatever's on the other side of the PTC, though. It needs to drive double-terminated twisted pair (58 ohms-ish), so the resistance has to be pretty small, maybe 10 ohms per side at most. The FETs I had in mind are IXYS IXTY08N50D2, 500V, 5 ohms, about

78 cents @ 1ku. They might avalanche a bit with really bad transients, but should survive. Whether they're worth the bother, I'm not sure yet. I can probably reduce the ON resistance by enhancing the gates, but that would require a bunch more parts to ensure that when a transient hits, the gate enhancement goes away without punching through the oxide in the process.

For the test board I'll probably put a few things in parallel and try them out. Junior Tesla time. ;)

The trouble is that a lot of the places where these gizmos are used don't conform to any sort of electrical code we'd recognize over here. Things like ductwork being part of the ground return for big squirrel-cage motors, so I can't put a good number on the required transient resistance. I'm guessing 5kA, 20/1000 us plus maybe an amp of follow-on current.

Right, thanks. I'll probably put the TVS across the supply and put diodes to there. It's 200 kb/s RS485 (RZ initially, for convenience), so I have a bit of leeway on the reactance.

Roight. I'll probably use another of those 150 uF coupled inductors per pair. That's the value I wound up with for the flybuck--turns out that I can control the sag pretty well by putting a resistor from V_in to the feedback divider, so primary-side regulation should actually work. (Without the resistor, the coupling coefficient had to be 0.995 or more for it to have a ghost of a chance.)

Thanks

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

The problem is that double-terminated RS485 takes over 80 mA to run it, so the resistor would have to be pretty small, maybe 5-10 ohms per side at most. That would be about a 60-A transient, worst case, which will require a pretty beefy resistor. It would also make it difficult to use low-capacitance diodes to the supply--I'd need 1N5823s and not BAT54s.

A nice Ethernet transformer that will just saturate and disconnect the silly transient would be nice. I suppose I could Manchester-encode the RS-485, but I'd like to stick to vanilla as far as possible. There are enough SCADA security vulnerabilities without putting the fire alarm on the Internet, which would be bound to happen eventually if I used Ethernet.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

Hi, Mikko,

Thanks.

I'll see your hillside and raise you a monsoon thunderstorm, long rusty steel ducts, and no electrical inspectors. ;)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

}snip{

^ Shouldn't that be 'micrometer'? ;-)

joe

I wish. The design requirement is "It has to work through repeated thunderstorms in an old wood/steel/cinderblock/mud/thatch/whatever building in the hills of Bangladesh, with 200 metres of rusty steel ductwork that may be carrying mains current, and very few grounds."

The standards probably have a lot of wisdom in them, though, so I'll check them out, thanks.

Yup. I'm going back and forth with the customer on that point. Making it really bulletproof (10 kA, 20/1000 us) looks like costing about $6 for the suppression components plus another $3 for an ISO3080 transceiver. That's a fair amount, but cheap compared to a service call way out there. Anyway, there's only one data pair to worry about.

I have the luxury of designing the whole thing myself, so yes, they are. The control panel side won't be isolated, but all the far-away ones will. The remote boxes will be bolted to the ductwork, which simplifies the RF problem some.

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

Nah, these ones are about 10 feet tall. 10000H, 150 uF. ;)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

Check out Bourns TBU's and associated parts rather than faffing about with depletion MOSFETS.

Hah :)

joe

thats been one of my rules as well

I express it this way

Don't connect copper to silicon without carbon in-between.

M

That is a small fortune. If you want to spend this much there may be better solutions but you'll have to inquire about capacitances:

Installation in China. Me: "Why did they disconnect PE for the whole machine?" ... "Because some other electronics in there didn't work well enough with PE so they snipped it".

200kbit/sec is fast enough to start watching capacitances. No way this would work dirctly with a TVS. Provide lots of capacitance on the supplies and then check whether all teh parts can stomach the upper tolerance limit for the TVS when it comes on. Usually they can't and then you'll need a shunt regulator bleeder. I had to do that a lot. Engineers first thought that was voodoo but not anymore after the field failures almost vanished.

Then, provide at least a small resistance between the clamp diodes and the IC signal pins unless the IC can withstand 3-4V more than you clamp. This is to avoid the on-chip ESD/substrate diodes being burdened with the lion's share of the jolt. Good RS485 transceivers can stomach a lot more though.

[...]

--
Regards, Joerg 

http://www.analogconsultants.com/

With capacitance or Leyden flasks, shouldn't it be liters?

Oh wait, litres, of course.

--
SCNR, Joerg 

http://www.analogconsultants.com/

fiber media converter.

They make them for serial as well as ethernet. Complete isolation between end points, and distance limiations for serial and interference is no longer a problem. The only issue is price and where to plug in the power packs for each side although they do make rack mount concentrators as well.

I have been working with industrial communication systems for decades also in tropics with daily thunderstorms. I have used the thumb of rules:

1.) you can use non-isolated RS-232 for equipments in the same room 2.) use galvanic isolation (2 kV) RS-422/485 in the same building 3.) use fibres between buildings

After my customers also understood these rules, I haven't heard complaints by the end users in the tropics

Rule #1 for lightning protection: There is no such thing as a ground potential, the best that you can get is equipotential bonding !!

Contact Jeff Dunnihoo, Pragma Design, Inc. He lives and breathes ESD

snipped-for-privacy@pragma-design.com

He can also test your final boards to make sure they are not going to die in the monsoon.

--
Chisolm 
Republic of Texas

Thanks, #2 is more or less where I was going--isolated power, isolated RS485. The system has a number of sensor pods powered by a small concentrator box nearby (bolted to the same duct, cables 6 feet maximum, no isolation). The isolation link goes from the concentrator to the control panel on the ground floor, up to probably 500 feet away.

The customer doesn't want the box ground connected to the long wiring, which is right, but also doesn't want it connected to the board ground, which ISTM is wrong, both for safety and EMC. Ground everything early and often, except when you can't, and then filter the daylights out of it.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net