Think you do not need surge protection? Take a look at the above double log graph above. As you can see, distance from a lightning strike is the only sure fire way to prevent damage. At a distance of 1 mile, 1 meter of wire will have about 130 volts induced upon it. Of course our CAT 5 cables running up towers are almost always longer than 1 meter. Frequently 20 meters or more. So, that would mean more than 2500 volts induced by that lightning strike a mile away.
The rise time of these strokes has been measured to be around 200 nanoseconds or faster.
How many radios and switches can withstand 2,500 volts on their Ethernet ports?
And if the strike is a tenth of a mile away, that is 25,000 volts showing up on that RJ45.
You probably are somewhat convinced that protecting your data circuits is a good idea or you would not be here. That said, we will discuss the various circuits that can be used to protect telecommunication cables.
The simplest circuit:
This is where data circuit protection started. Originally, any long distance data was hauled in by the local telephone company or perhaps Western Union for telegraph. The original protectors were carbon arc gap protectors. This schematic shows a double gas tube. If the voltage gets high enough the tube fires and shorts to ground. The impulse can come from either side (longitudinal surges) or across the pair (transverse surges). Most lightning is longitudinal, in that it normally does not hit one wire and seek to complete the circuit by going to the other wire, it is induced on both wires and seeks earth ground.
However, almost all the time we are talking about induced voltages. And if one wire has a higher induced voltage than the other wire, there will be a difference between the wires and that difference should not be ignored.
This circuit will actually work for Ethernet circuits. It is slow but it is better than nothing.
So some economy protectors do this (see drawing below) to claim they have a TVS (transient voltage suppression semiconductor) but this is actually worse.
Sure, it has a fast TVS but it will only clamp transverse impulses. It does not allow anything to go to ground. Super cheap design. Totally ungood.
Marginally better. Yeah, you can claim a TVS and a GDT but it is doing nothing for longitudinal impulses. Another cost cutting design.
Now we are starting to get somewhere. TVS1 clamps impulses across the pair. (Transverse, Balanced or "Metallic" impulses) TVS2 and TVS3 take care of impulses on the wire seeking an earth ground return. (Longitudinal, Unbalanced or "Common Mode" impulses). But this is a lightweight system. It will not handle the big impulses.
Now we are cookin' with gas. This has the GDT to handle the big impulses and the TVS components to handle the fast rising impulses. TVS take care of business until the GDT wakes up and shoulders the load. Two deficiencies in this design: 1) It has high capacitative loading so bad for data. 2) It is not an optimal value design for circuits like CAT5 that have 8 conductors to protect.
OK, so high speed data needs low capacitance. Designs started to migrate to designs that used a reversed biased rectifier diode in series with the protection elements. The DB component is a bridge rectifier. It allows data signals to pass until they build up a charge on the output, then they become reversed biased and high capacitance. This is a fast design, low capacitance design, but will not survive large impulses. So whadyagonna do?
Add a GDT of course. This is a common design. It takes care of the initial fast impulse and also takes care of the big impulses. One thing it does not do well is to handle the high currents of the big impulses. That is because a diode bridge is used to steer the impulses. Bridges put 4 rectifiers on one semiconductor die. So all the heat is concentrated on a very small piece of silicon.
Is there a better way?
Of course there is a better way. Use discrete rectifier diodes to steer the impulses. For example, a diode bridge that is use by most surge protection manufacturers has a constant forward current of a half amp and a capacative load of 13 pF. The discrete diodes I use in my designs have a current carrying capacity of twice that figure and a capacative load of half that amount. Double goodness. Much gooder than bridge designs.
(Grammar and spelling nazis, bugger off, you lost WWII, deal with it)
Again, why is capacitance bad? It provides a leakage path, a short circuit if you will, of GigE data signals. This can lead to CRC errors or speed down-shifting.
So how do you improve on this design.... I give some clues on the GigE issues pages but I am not going to give away all my secrets. There are several significant improvements to the design above. Which I incorporate into my products.
The sad thing is, my competitors do not even incorporate all of these 8 design features outlined above. Why?
There is no other reason.