has reached the end-of-life and is now read-only. Please see the EOL announcement for details

Signal line with and without surge protection. The surge energy is reduced by an order of magnitude with a small Gas Discharge Tube.

· · Web · 1 · 2 · 7

In comparison, the same signal data line, with a small Thyristor Surge Suppressor for protection. Its response time was a lot quicker than the Gas Discharge Tube, the large 200 V surge was clamped to just 100 V within microseconds.

But the speed the TSS got destroyed by excess power dissipation was just as fast... It immediated failed with an open circuit, and after that the full 100 V surge went into the load. Tubes are more robust than semiconductors, news at 11... (To be fair, it was only a small data-line TSS, large power-line TSS chips can often compete with GDT).

Modern fast-acting, low-current SMD fuses are remarkable. This is the same surge, same data line, but with a fuse installed. It interrupted the surge within just 100 microseconds! It was so fast that my oscilloscope's timebase setting didn't even capture that properly.

Of course, a fuse in itself is useless, if the load is high impedance, or if the surge is a 8/20 spike, it won't do anything. But it looks pretty promising for limiting the long-term energy dissipation inside protective devices for a constant overvoltage surge.

The problem of testing surge protection circuits is that the test is destructive. You need to remove and solder SMD parts onto the board non-stop to keep the test going... You only get one chance at a time, and if didn't set your oscilloscope's trigger properly before you fire the surge... :blobcatknife:

Overshoot and ringing artifacts when a fuse disconnected the protected circuit in a surge. Ironically it made the input surge to overshoot, boosted it from 200 to 250 volts, and arguably made the surge more destructive.

I think what was happening is that the pulse capacitor and the wiring's parasitic inductance formed an LC resonator. The instantaneous current was 10+ A because of protective clamping to ground, then the fuse suddenly interrupted it. This was enough to excite even the tiny wiring inductance.

The second experiment showed the Thyristor Surge Suppressor attempted to clamp 200 volts to ground, but eventually exploded due to excess power dissipation, and the remaining 100-volt surge to enter the load.

But when paired with a fuse, TSS's performance is just phenomenal. It clamped a 200-volt surge down to a tiny 26-volt, 500-nanosecond bump. Magic!

It's just magical to see a fuse to blow in 750 nanoseconds. But I guess it's what happens when you pass a current over 100x of the rating across it.

Tested again with a fuse and just a regular TVS diode chip instead of a telecom TSS chip. The result is quite similar, 200-volt surge is clamped to just 28 V within 500 nanoseconds.

Conclusions: TVS diodes can be effective even for a long duration surge, as long as you use a really fast fuse to limit the total energy before the TVS explodes and fails open. The TVS itself can still be destroyed and fail short, but both are replaced in a repair.

I suspect the thermal coefficient of the fuse's resistance also provided an unintentional form of series current limiting during a surge, which is great news.

High voltage power supply - "Who needs a bleeder resistor when you can use just a screwdriver to discharge it manually with a bang?" (famous last words).

Redid the "blown fuse" test with a proper timebase. 200-volt surge, 50-ohm load, fast-acting 250 mA, 1206 SMD thin-film fuse (Littelfuse 466). The fuse opened within 68 microseconds after the surge. This was my last fuse...

MFW I fired the surge generator, probably destroyed the chip-under-test within milliseconds, and then noticed my oscilloscope was in STOP mode... :blobfacepalm:

The difference between a Gas Discharge Tube and a semiconductor surge protector - if you see sparks from the tube, that means it just started working, but if you see sparks from the semiconductor, that means it just stopped working...

They should make a Gas Discharge Tube surge protector with a Keepalive electrode... This should either dramatically improve the surge response time of a GDT, or make the turn-on voltage even lower than 60 V.

Though, nobody in the real world wants such a Frankenstein device that keeps drawing meaningless power other than me...

> It is also not a good practice to hold a GDT in its glow region as this will significantly reduce the life expectancy of the device. In this condition, significant heat can be developed on the electrodes that can damage the special emission coatings and cause premature failure of the tube. - Bourns application note

Apparently I'm not the first one to come up with this thought, and this is actually a pretty bad idea... :blobcatgiggle:

Just repeated the same surge test on a 250 mA polyfuse. The fuse didn't do anything to the surge, as if there was no fuse. When repeated again with a gas tube connected in series, the fuse arced itself spectacularly, with a bright flash and a loud bang. ⚡💥

Conclusions: A real fuse, even as small as 1206 SMD, can interrupt a 200 V surge within 750 nanosecond, but a similar polyfuse is useless for surge protection - its slow response is only useful for steady-state overcurrent, and its max voltage and current ratings are just miserably low.

BTW, it was funny that researchers abused the original Littelfuse to make the earliest thermistor-type, DC-substitution RF & microwave power sensor in WW2. You basically bias the fuse with DC very close to its rated current, couple RF power into the fuse, and put it in a Wheatstone bridge. Any incoming RF power causes a decrease of DC power to keep the bridge in balance.

It was really fragile and could be burned out easily by excess power, even a transient (the sensor was literally just a fuse!) The response time and impedance matching was also bad, requiring adjusting an RF tuners to find the maximum point for each measurement.

I need to try replicating it someday with Littelfuse's modern 0402 SMD parts...

One problem responsible for the bad impedance matching in this makeshift "Littelfuse" sensor was parasitic inductance. Some World War 2 researchers went as far as removing the fuse glass enclosure, extracting its filament, and soldering it directly in a circuit to improve its RF performance - Fun (???) RF hacks you can learn in MIT Rad Lab books.

With modern SMD fuses, this problem no longer exists. I should be able to get better results. BTW, funny to see the company still exists and keeps making the best fuses after nearly a century... here's an idea: what about keeping a source of ionizing radiation in or next to the GDT? I believe this was done on some thyratrons to improve response time. Yes, many older GDTs used the same technique, though it's rare today as people don't like radiation and/or regulation paperwork. One ITU-T source said modern gas tubes have better performance that older radioactive ones already.

I think one can also use optical radiation, place a light bulb near the tube should improve performance slightly (Not sure if LED works too due to narrow spectrum). It's a well-documented issue that GDT firing voltage decreases when it's removed from the chassis, creating misleading test results. The standard procedure is testing tubes in a dark container, sealed for at least 24 hours.

@niconiconi that's interesting. got any links to info on their optical effects? I wonder if it's a wavelength cutoff thing dealing with the photoelectric effect and the work function of the metal electrodes.

with your high speed applications, how much of a deal is leakage current? ITU-T K.99. Surge protective component application guide – Gas discharge tubes See page 7 (PDF p13) and Appendix 2, page 20 (PDF p26). No information on its physical mechanism though, just a description. In modern devices it's only a minor effect, around 20-30 V, and I don't think really worth doing in practice. in fact radioactivity was its initial fix).

Digital signaling is pretty robust in terms of drive strength. Line drivers are designed for 50-ohm loads, leakage less than 1 mA is nothing. The problem is capacitance, even 5 pF is unacceptably high. Most GDTs can have C = 1 pF and they're one of the very few suitable protectors in high-speed and RF circuits.

Sign in to participate in the conversation

the mastodon instance at is retired

see the end-of-life plan for details: