High Voltage sub nanosecond switching

[Sander]

I’m looking for a part that can switch 4000 volts into a 50 ohm load within a nanosecond. The pulse duty cycle should not exceed 1 microsecond (80 amps for at maximum 1us). I was thinking of a mercury relay or a hydrogen thyratron, but the ones I have seen seem to slow or lack the specifications. Has someone used, or know of a part that is capable of this? Maybe a high power VHF/UHF transmit tube?

[richard.cs]

It is fast for a thyratron, they work by avalanche ionisation so I can't see how you'd get the switching time. A UHF transmit tube would be better in this respect, you should be able to slam it on fast enough :-)

You don't say what your repetition rate is but if you can pull the A-K voltage down to a hundred volts or so you're at 8 kJ anode dissipation per pulse, that doesn't sound too unreasonable. The 80 Amps is going to be interesting, I think you'll struggle to find one that specifies a peak cathode current that high *but* for such a short pulse maybe it'll be ok, meerly depleting the cloud of electrons surrounding the cathode without actually requiring all that much cathode emission. Is there 80 microcoulmbs of charge floating around the cathode? I have no idea.

No idea about the mercury relay, It seems like it should be pretty fast.

[Sander]

The repetition rate is 100Hz @ 2kV and 60Hz @ 4kV.

[Marco]

Does the trailing edge have to be that fast or just the leading edge?

The Russians make plenty of fast rise time pulse generators, I think they generally use SCR or IGBT followed by a magnetic switch to get to ns rise time and then throw it at a SOS to get it sub-ns.

PS. probably not the best time to order anything from Russia though.

[Circlotron]

Yeah, if the trailing edge can be slower turn off then as you suggest, a mercury wetted relay but discharging a cap into your 50 ohms. Plenty of available current and no problem snubbing the relay contacts because the current will be zero when the contact open.

[T3sl4co1l]

You can also use planar triodes (including many commercial UHF parts) in transmission line structures:
http://www.slac.stanford.edu/cgi-wrap/getdoc/slac-pub-3546.pdf

Or avalanche transistors; http://icecube.wisc.edu/~kitamura/NK/Flasher_Board/Useful/research/RSI03066.pdf

[Alex Eisenhut]

Would a trigatron work at that low a voltage?

Eh but I don't know what the duty cycle would be...

[MagicSmoker]

Pretty sure you are going to need to use a "pulse forming network", and probably with hardline coax. Jim Williams (of Linear Technology fame) wrote some good app notes on this a few years ago.

[edit: mistakenly attributed the late Jim Williams to Analog Devices... doh.]

[David Hess]

A trigatron built into a transmission line is what occurred to me as well.

[Alex Eisenhut]

http://www.siliconinvestigations.com/KRYT/Krytron.HTM

there we go. Another glass solution, but NLA...

http://www.tubecollector.org/kn22.htm

[tautech]

The repetition rate is 100Hz @ 2kV and 60Hz @ 4kV.

Sounds like HVDC transmission line stuff?
http://en.wikipedia.org/wiki/HVDC_Inter-Island
Check out the thyristor switching hall halfway down the page. 

[Marco]

HVDC converters don't have to switch below a ns.

Mercury wetted relays/contactors are probably the cheapest shot at getting below a ns, it's a bit of a gamble though because you don't know the internal inductance of these things ... it's not what they are designed for. Also very high jitter and less long term reliability than fully solid state solutions.  If he wants a square pulse using a piece of HV coax as a storage capacitor would indeed be the way to go.

[Sander]

That’s exactly it. I want to dump the energy stored in a piece of coax into a 50 ohm load and AC couple that pulse to a mains connector of a device under test, like the INS series from NoiseKen:

http://www.noiseken.com/uploads/photos0/131.pdf

They use a mercury relay, but the ones I can find are specified to only 3 amps. I could order one and hope it won’t explode…

Krytrons are probably exactly what I want, but I don’t have a DeLorean DMC-12 and I don’t know any Libyans so they might be hard to get.

The avalanche transistor solution would be excellent, easy to get cheap parts but it can’t handle a pulse width of more than 15ns. I need 50ns – 1us pulse width.

The solution with the planar triodes might work, but it will exceed the specified current by a order of magnitude. Even if I replace the final stage with a GI-39B I still exceed the specs 5 times.

[Alex Eisenhut]

The only other thing I can think of is an IGCT but they won't get you the PRF you want.

But you'll be able to power the commuter train for a small town.

So there's that.

What about a triggered spark gap? A DIY Kryton, get a spark gap tube, and wrap some fine wire around it and use a photoflash trigger circuit ?

[Marco]

They use a mercury relay, but the ones I can find are specified to only 3 amps. I could order one and hope it won’t explode…

I doubt a 1 us pulse with no need to extinguish an arc will do anything to it, I think it dumps orders of magnitudes more power into the mercury when it's breaking a contact of 3.5kV 60 Hz.

These guys seem to be pretty cheap BTW.

What worries me is that these things are relatively big (10 cm). The electrode seems to be a relatively big chunk of copper so inductance isn't going to be too bad ... but an impedance matched coaxial assembly it ain't.

[T3sl4co1l]

Does it have to be square (or squarish) pulses?  IEC 61000-4-4 uses exponential-ish pulses.  I don't know offhand if there are other, more demanding, tests (MIL specs perhaps?!).

The avalanche transistor solution would be excellent, easy to get cheap parts but it can’t handle a pulse width of more than 15ns. I need 50ns – 1us pulse width.

Why can't you switch a TL with avalanche transistors?

Typically, they stay switched on for several microseconds (recombination time).  The usual example (single transistor) uses a transmission line; I should think it would work with many in series too.

Tim

[TerraHertz]

I seem to recall that firing of the explosive charges used to compress a nuclear weapon core had electrical requirements somewhat similar to this. The multiple detonators all had to be fired with sub-nanosecond timing accuracy, so it became a matter of generating a very fast high voltage pulse, and coupling that to all the precisely measured lengths of coax leading to the individual detonators.

Of course in that case there was no pulse repetition requirement.

I think somewhere I had saved an article on the technology used, but can't find it now.

[Paul Moir]

Krytron, previously discussed.  Dismissed due to following disadvantages:  delicate, inefficient, export controls, angry Libyans showing up, having to collect pinball parts for exchange. 

[Alexei.Polkhanov]

The Russians make plenty of fast rise time pulse generators, I think they generally use SCR or IGBT followed by a magnetic switch to get to ns rise time and then throw it at a SOS to get it sub-ns.

I was looking for complete HV pulse generator for a while now. Best I was able to find are broken and require hard to find replacement parts.
If you know any supplier that sells them please share!!

Here is the link to a document describing "modernized" design of a sharp HV pulse generator. 

URL: http://www.lle.rochester.edu/media/publications/lle_review/documents/v133/133_07_Solid.pdf.

I am trying to build down-scaled version of it right now for time of flight laser 3D scanner.

No tubes or Krytrons - I bet you can see those only in spy movies and in North Korean radars these days  Pulser stalk by itself is pretty old idea but with good MOSFETs and right cores you can get very good performance.

Here are parameters from that doc.

[David Hess]

I like the design.  The linear layout reminds me of a distributed transmission line amplifier while connecting the primaries in parallel and the secondaries in series reminds of how an Ft doubler works.

Transformer coupling is not suitable for long duration pulses.

[eurofox]

Maybe this application note is interesting 

[Marco]

If you know any supplier that sells them please share!!

http://www.moose-hill.com/technolo.htm#Technology%20Transfer (might be a zombie website)
http://www.megaimpulse.com/
http://eng.iep.uran.ru/naudep/imp/razr/razr_2.html (academic, but they seem to have sold some of their devices to other academics)

There's also multiple companies making Pockel Cell drivers which are HV and sub ns :
http://www.fidtechnology.com/apps/fds/fds2-1M101B201.html
http://www.kentech.co.uk/index.html?/&2

Here is the link to a document describing "modernized" design of a sharp HV pulse generator.

Hard to get sub-ns with transformers ...

It wouldn't let me download the PDF BTW. Had to to view it in google and move it to google drive to download it. Maybe only for US'ians?

Why would you need a HV pulse for a ToF laser scanner?

[David Hess]

Maybe this application note is interesting 

This application note was one of the first things I considered mentioning but avalanche pulse generation has severe limits on pulse duration.

[Alexei.Polkhanov]

I used design based on Avalanche Transistors similar to one described in app note that eurofox linked. It is similar to what Jim Williams described in one of his app notes - it also uses regular germanium transistors (like 2N2369) in avalanche mode. Zetex claims that their transistors are designed specifically for avalanche mode, so you don't have to pre-select ones that work from large set and plus they can work reliably in avalanche mode for a long time. I have read somewhere that when almost everyone stopped making  germanium transistors they found a factory somewhere in Russia that still makes them and bought it. Now they charge something like $12 / transistor.

Many laser scanners work in phase shift mode - laser diode sends continuous wave signal and distance is measured by phase shift from reflected one. TOF is on other end uses short pulse and its accuracy really depends on how sharp is the leading edge. One way of doing it is to send sharp and short pulse (~200V) to laser diode. It should be short enough not to kill the diode. Another way is to get laser working in continuous mode and then make pulse by turning the Pokels Cell on/off - in that case you need HV pulse. These are not all design alternatives of couse.

I think it would be cool to build a 3D scanner that can scan entire mountain. Artillery range meters can easily handle distances like 80km, so it should be possible to make 3D scanner that capable.

[gdewitte]

Marx generator? http://en.wikipedia.org/wiki/Marx_generator