Device Output Protection from External High Voltage

[daveyk]

I work on industrial UT instruments.  When calibration them, a very expensive Rohde & Schwarz RF STEP attenuator can be used.  I want to protect its attenuation pads from an external "pulser" input in case of an accident hooking it up to an instrument or board set.  That pulser can be -300volts, 10ns Rise Time and anywhere from 30ns to 500ns in width.

For brief periods of time, the Agilent 33250A generator can survive the pulser, and the Siglents survive too, but modern attenuators seem to vaporize easily internally.

For certification I use HP355E and HP355F attenuators, with custom control and custom support software I wrote and they are very good.  You can hit them with one of those "pulsers" and they "Thank you, give me more", not so much a R&S with a $5K repair bill when a pad vaporizes, if they will even support the older ones any more.

I am thinking back to back 6.2v Zener Diodes across the output and then to ground and then a very light fuse in series with the output in a pomoma box.  I am not worried about what that does to the bandwidth.  All calibrations (tweaking) are done at 5MHz, sinewave, 1vpp.  Any loss or BW attenuation would most likely be relative.  What I am thinking it that the ZENERS would not be fast enough to protect against the pulse.

What components could be used that react very fast for output protection?

Thanks

Dave

[Jay_Diddy_B]

Dave,

what is the amplitude and frequency of the desired signal?

Regards,
Jay_Diddy_B

[David Hess]

What is the maximum pulse amplitude that must be passed?

If back-to-back 6.2 volt zener diodes would otherwise clamp at a suitable level into what I assume is 50 ohms, then a diode bridge biased with 100 milliamps can operate as series protection with much higher bandwidth than needed to support "only" a 10 nanosecond transition time.  Performance is only limited by diode recovery time, breakdown voltage, and feed-through capacitance into 50 ohms.

Tektronix used this sort of circuit to protect 50 ohm inputs of the 1 GHz 7A29 vertical amplifier and 350 MHz 485 oscilloscope.

[daveyk]

Thank you.  Do you have a schematic clip of what that looks like?  It's not the rise time, and perhaps not the -300+v pulse that destroys.  Some of the UT instruments have what is called a Square Wave pulser.  I suspect when "accidents" happen (caused by interruptions), it could be that you have a -300v, 230ns wide pulse, at a PRF of up to 2000Hz,  being accidentally injected in to the attenuators.  The old HP355E/HP355F don't have a problem with that.  They have 2 watt precision thru-hole resistors inside them, although there micro switches handle it too.  The newer Agilent 8494G/8496G Attenuators have their attenuation pads vaporize instantly.  The Rhodes and Schwartz are easily destroyed by those pulses too ($5000 Repair Bill).

I am typically extremely careful around my attenuators, but accidents to happen.  I am especially concerned to protect the Rhodes and Schwartz.  I also don't want to add too much insertion loss (<1dB).

In the old days with Wavetek 166 Generators, I remember back to back 22volt Zeners were soldered across the output BNC to ground.  It seems to me that that only provided partial protection.  What ever diode clamping is done needs to be fast.

[daveyk]

A colleague suggested such as in this attachment.  I think it is a great idea. but it would probably add ~6dB insertion loss.  I could handle that with my software, but the program I will use the R&S attenuator with is not under my control.  Maybe a replace the series resistor with a fast blow fuse, but I doubt it would ever blow.  I would have to do some bread boarding experiments if I want to try a fuse. I don't know if a fuse would blow faster than an attenuator pad.

[David Hess]

Thank you.  Do you have a schematic clip of what that looks like?

The two examples from Tektronix which I mentioned are shown below.  Details are available in the respective service manuals.

Note that the supply voltage does not need to be higher than the input voltage however the voltage rating for the diodes does need to be higher.  Also this circuit works equally well to protect inputs or outputs; it is bidirectional.

[nctnico]

I work on industrial UT instruments.  When calibration them, a very expensive Rohde & Schwarz RF STEP attenuator can be used.  I want to protect its attenuation pads from an external "pulser" input in case of an accident hooking it up to an instrument or board set.  That pulser can be -300volts, 10ns Rise Time and anywhere from 30ns to 500ns in width.

For brief periods of time, the Agilent 33250A generator can survive the pulser, and the Siglents survive too, but modern attenuators seem to vaporize easily internally.

For certification I use HP355E and HP355F attenuators, with custom control and custom support software I wrote and they are very good.  You can hit them with one of those "pulsers" and they "Thank you, give me more", not so much a R&S with a $5K repair bill when a pad vaporizes, if they will even support the older ones any more.

I am thinking back to back 6.2v Zener Diodes across the output and then to ground and then a very light fuse in series with the output in a pomoma box.  I am not worried about what that does to the bandwidth.  All calibrations (tweaking) are done at 5MHz, sinewave, 1vpp.  Any loss or BW attenuation would most likely be relative.  What I am thinking it that the ZENERS would not be fast enough to protect against the pulse.

What components could be used that react very fast for output protection?

I'd use high-speed zeners like the ones used for ESD protection on ethernet lines. Because your test signal is LF and the pulse you want to protect against is HF you could use a ferrite bead and a capacitor. You'd have to choose the values so that they don't load the signal too much at 5MHz.

[Jay_Diddy_B]

Hi,

The schematic shared by Daveyk contains the same circuit, the diode bridge circuit (D714, D717):




However, it is not in a position to protect the external step attenuator.

I suggest, given the expense of the attenuator repair, I suggest the OP considers designing a building a step attenuator optimized for this application, that understands the R&S commands.

Regards,
Jay_Diddy_B

[tggzzz]

I work on industrial UT instruments.  When calibration them, a very expensive Rohde & Schwarz RF STEP attenuator can be used.  I want to protect its attenuation pads from an external "pulser" input in case of an accident hooking it up to an instrument or board set.  That pulser can be -300volts, 10ns Rise Time and anywhere from 30ns to 500ns in width.

You need to characterise both the device you need to protect (power, voltage, current that it generate and can tolerate) and also whatever generates the pulse (voltage, impedance, etc).

The protection circuits for a 10kV ESD strike to a USB pin are rather different to those for a mains transient applied to a signal generator.

Once you can draw a schematic of the relevan bits of both devices, you will be able to simulate what happens with, say, LTSpice. That should give you more hints about what is important, but be very aware that the details matter and may not be included in the simulation.

[nctnico]

I work on industrial UT instruments.  When calibration them, a very expensive Rohde & Schwarz RF STEP attenuator can be used.  I want to protect its attenuation pads from an external "pulser" input in case of an accident hooking it up to an instrument or board set.  That pulser can be -300volts, 10ns Rise Time and anywhere from 30ns to 500ns in width.

You need to characterise both the device you need to protect (power, voltage, current that it generate and can tolerate) and also whatever generates the pulse (voltage, impedance, etc).

The protection circuits for a 10kV ESD strike to a USB pin are rather different to those for a mains transient applied to a signal generator.

Once you can draw a schematic of the relevan bits of both devices, you will be able to simulate what happens with, say, LTSpice. That should give you more hints about what is important, but be very aware that the details matter and may not be included in the simulation.

Yep. The layout will be critical. The signal will need to pass the pins of the protection device. The protection device should not be connected to stubs. Same for grounding of the protection devices. A solid ground plane will give the best results.

[Elasia]

+1 for proper routing and planes.. thats just as important as the protection cap / mov / tvs diode or whatever you end up using

[daveyk]

I have a field job week, but once back, I need to experiment.  I have a Pomona Male BNC - Female BNC box.  When I get back, I need to experiment, perhaps bread-board some protection.  I will go to Digikey and find the lightest (smallest wattage) 50 ohm resistors I can find (SMT most likely).  50 Ohms to ground would represent an attenuation pad that blows.  I want to see what it takes to blow it unprotected (I suspect a 300v square wave pulser at about 125ns width running at 1KHz PRF).  If I can blow that resistor, then I can experiment with protecting it.

Dave