Scoping Signals in the 10s of Kilovolts

[madshaman]

Hello,

I'm looking for pointers/advice on how to go about the following:

I'm going to be building a circuit to dump very fast high voltage pulses into a load composed of nitrogen gas (N2 laser).

I would like to be able visualise the pulse in the time domain using an oscilloscope.

I'm not rich so I'm planning to use a Tek7104 I'm restoring (got the mainframe working, yay, and I'm on to the 7A29 vertical amps once I'm back home) using an old Tek scope camera mount retrofitted with a CCD camera.

I don't have enough knowledge or experience to acquire the signal properly but I do know that if I don't do it right, in the very best scenario I'll get a single shot trace that looks absolutely nothing like the actual signal (and I'll probably totally change the system behaviour with my test setup).

In the worst case I'll simply destroy my scope.

I know how to attenuate the signal using a simple voltage divider (and how to do that safely) but I'm positive this will, at best, get me garbage and I suspect this will alter the system's behaviour too much to be of real use.

Even if I do my best to apply what little I know about signal propagation, I'm certain I'll fail and won't even know I'm failing, so a related question would be: is there any way I can verify my test setup?

Any help would be greatly appreciated.

Thank-you in advance.

[ConKbot]

Do you need voltage or current measurements? just a waveform? A pulse current transformer would give current waveforms easily.  If you have an exposed HV node, you could use capacitive pickup to get the voltage waveform, but neither of those would be accurate for measurements.  The current transformer could be if you had a way to calibrate it.

[Jon86]

Just get one of those high voltage automotive ignition probes, they're cheap as chips on ebay.

[madshaman]

Do those cheap as chips probes accurately acquire signals up to 1Ghz?

[madshaman]

Do you need voltage or current measurements? just a waveform? A pulse current transformer would give current waveforms easily.  If you have an exposed HV node, you could use capacitive pickup to get the voltage waveform, but neither of those would be accurate for measurements.  The current transformer could be if you had a way to calibrate it.

Yes, I should be more specific, i want to acquire the voltage waveform, without distortion, phase or otherwise.

An N2 laser self terminates in nanoseconds so the rising edge needs to be very fast and the pulse width needs to be as short as possible.  Any energy delivered to the system after the termination of lasing action is wasted.

I need to visualise the pulses accurately to debug my circuit.

[Fraser]

10's of kV at 1GHz bandwidth....... you will be looking at a very expensive setup I am affraid.

The PICOScope automotive ignition pickup is just a capacitive coupler around the insulated HT lead. It is not an accuarate oscilloscope probe and definitely not good to 1GHz !

You could use a 1GHz bandwidth active Oscilloscope probe with an E field antenna to spatially pick up the radiated signal from your set up but no calibration will be possible. Only waveforms. A 1GHz active probe is also a very expensive piece of kit unless you already own one.

At such high voltages I believe a non contact method may be best to avoid arching and influence on the signal you are trying to capture.

[madshaman]

10's of kV at 1GHz bandwidth....... you will be looking at a very expensive setup I am affraid.

The PICOScope automotive ignition pickup is just a capacitive coupler around the insulated HT lead. It is not an accuarate oscilloscope probe and definitely not good to 1GHz !

You could use a 1GHz bandwidth active Oscilloscope probe with an E field antenna to spatially pick up the radiated signal from your set up but no calibration will be possible. Only waveforms. A 1GHz active probe is also a very expensive piece of kit unless you already own one.

At such high voltages I believe a non contact method may be best to avoid arching and influence on the signal you are trying to capture.

Thank-you, I *do* own both an active 3.5Ghz probe (untested as yet), and two 1.5Ghz passive probes (tested and verified with a 20Ghz sampling scope).  Obviously I nabbed these used.

Yeah, this isn't exactly an easy problem to solve, even with halfway decent equipment to start.

Was kinda hoping someone would say: "Oh yeah, I had to do this at such and such and here's how you do it, and this is why..."

I might browse IEEE to see if any papers sufficiently describe a similar test setup.

[alm]

If the rise times are in the nanoseconds, then forget any divider you build yourself. You'd be lucky to get 1 MHz bandwidths, or rise times in the microseconds. Assuming you're skilled enough in HV construction, since making a probe for 10 kVs is non-trivial even at DC.

Normally my answer would be a Tek P6015(A), but take a careful look at the frequency derating curve. You're likely exceeding the capabilities of this probe, which I believe is the only one of its kind. The P6015 is often available used for reasonable prices, but note that this needs to be filled with a freon compound to achieve its rated voltage specs, and availability of this stuff is extremely poor because of environmental regulations. That's why they designed the P6015A, which doesn't use freon and is therefore much more expensive used. This probe only has 75 MHz bandwidth, so it's also a little slow for your application. Good luck finding something with more bandwidth and a higher voltage rating, I don't think such a thing exists.

I would probably indeed look at capacitive or inductive pickup. If you want good frequency response, maybe you can fit a well-insulated wire through a Tek CT1? These have a decent bandwidth and are fairly affordable, but you would obviously be measuring the current signal, not the voltage signal. I'm also very doubtful that you'll find wire rated for tens of kV small enough to fit through the hole. Maybe a slower probe like the P6022 would work? Forget about accurate voltage measurement with any of those setups.

[pinkysbrein]

He can build one capacitor which can discharge in the ns time range ... why not build two to form a capacitive divider?

Couldn't you just clear small part of the ground plane then put another piece of dielectric over the ground plane and then put some aluminium foil over that so it covers the cleared part and a larger section of the ground plane? It should have small capacitance to the high voltage plane and large capacitance to the ground plane ... thus forming a capacitive divider?

Of course you'd need an active probe or an in place buffer to measure the voltage over the divider.

[ElectronicTonic]

I was going to suggest the Tek P6015 as well, but certainly not for 1 GHz BW. Although, if you can find one, it would come in handy for calibrating your probe at HV, but at lower frequencies.

What I can suggest is that you watch a series of videos I made about designing and building a 2 kV, 100 MHz scope probe from other scavenged probe parts. It was a long process. I started with just building it, then ran into problems. PSpice simulation really helped, and the finished product worked just as well as a professionally made probe.

From my experience, I have two pieces of advice when designing a HV, high-BW scope probe.
(1) Simulate your design to iron out the kinks. When every pF counts, simulation is key. And don't forget to take into account the inductance of your return ground path, especially at 1 GHz.
(2) If your scope can be situated right next to the HV source you want to measure, then design the probe to be a simple box that plugs into scope input on one end and has HV input directly on the other end. If you don't have a long coax cable between input and output, then your design can be simplified because you wont need to compensate for the coax's capacitance, inductance, or internal reflections.

Here is the first of four videos about this 2 kV probe I made. You might have a look at some of the other videos in my "High Voltage Scope Probes" playlist.

[Rufus]

I don't see it being necessarily that difficult providing what you are measuring has plenty of drive.

A resistive probe which is just a resistor feeding 50 ohm coax has inherently high bandwidth. A 5M resistor will give you a X100k probe. The problem is getting 5M of resistors that can stand 10s of kV and the stray capacitance across them.

You can compensate for stray capacitance with a capacitor to ground forming a capacitive divider with the same ratio as the resistance. A possible trouble is just 20fF of capacitance is about 25k at 1GHz and for 10s of kV that is a significant load, however, does the capacitance/stray of what you are driving make 20fF look significant?

[madshaman]

I was going to suggest the Tek P6015 as well, but certainly not for 1 GHz BW. Although, if you can find one, it would come in handy for calibrating your probe at HV, but at lower frequencies.

What I can suggest is that you watch a series of videos I made about designing and building a 2 kV, 100 MHz scope probe from other scavenged probe parts. It was a long process. I started with just building it, then ran into problems. PSpice simulation really helped, and the finished product worked just as well as a professionally made probe.

From my experience, I have two pieces of advice when designing a HV, high-BW scope probe.
(1) Simulate your design to iron out the kinks. When every pF counts, simulation is key. And don't forget to take into account the inductance of your return ground path, especially at 1 GHz.
(2) If your scope can be situated right next to the HV source you want to measure, then design the probe to be a simple box that plugs into scope input on one end and has HV input directly on the other end. If you don't have a long coax cable between input and output, then your design can be simplified because you wont need to compensate for the coax's capacitance, inductance, or internal reflections.

Here is the first of four videos about this 2 kV probe I made. You might have a look at some of the other videos in my "High Voltage Scope Probes" playlist.

Thank-you very very much.  I was also thinking that I'd have to become a *lot* more familiar with probe design and probing techniques.  If the circuit was within the voltage range acceptable to the 7A29 I think I could use a short bnc cable soldered right on to the board.

It does look like I'll need an approach with the scope right next to the circuit, but it's a scary prospect as the circuit itself is design to deliver a very large amount of energy into the load in a very short amount of time; I can imagine all that energy dumped into the scope.  I'm fairly certain there would be fireworks.

Thanks again for sharing.  I won't be able to watch your video until I'm somewhere with a decent bandwidth connection.

[madshaman]

I don't see it being necessarily that difficult providing what you are measuring has plenty of drive.

A resistive probe which is just a resistor feeding 50 ohm coax has inherently high bandwidth. A 5M resistor will give you a X100k probe. The problem is getting 5M of resistors that can stand 10s of kV and the stray capacitance across them.

You can compensate for stray capacitance with a capacitor to ground forming a capacitive divider with the same ratio as the resistance. A possible trouble is just 20fF of capacitance is about 25k at 1GHz and for 10s of kV that is a significant load, however, does the capacitance/stray of what you are driving make 20fF look significant?

The circuit will have enormous driving ability.  Picture a Marx generator, but instead of spark gaps, imagine high voltage bjts in avalanche mode.  For the capacitors, I have a large number of 1200V capacitors with the lowest internal resistance and inductance I could find.

I plan to wire the whole circuit point to point, and I anticipate it's going to be a nightmare laying it out to balance avoiding inductance while preventing arcing.  One of my ideas for prototyping it is to embed most of the circuit in paraffin wax.

The load itself is a tube of N2 gas at 30 torr at the lowest.  I'm going to start the tube length at 20cm and the gas will between two copper electrodes, one a flat strip, the other will be a 45° wedge "pointing" at the plate.  This arrangement comes from an IEEE paper which evaluates the effectiveness of different electrode shapes for an N2 laser.

The electrode spacing (wedge edge to strip) will be about a centimetre.  However, I will be experimenting with different spacings as well as different gas pressures.

The "tube" will be composed of polycarbonate.

The gas itself will present some inductance and the electrode arrangement will of course form a series capacitor.  I don't anticipate having much in the way of parasitics.

To answer your question, I think I'll need to complete construction to measure the load's capacitance but I'm fairly certain it will exceed 20fF.

None of my ideas are new, in fact I'm basically just ripping off design ideas from various papers.

[TerraHertz]

You have a very similar design problem as with something I'm working towards. And no, I haven't really come up with a waveform capture solution yet either.

You might find some useful information here:
http://w140.com/tekwiki/wiki/P6015

It occurs to me that the P6015 with its Freon liquid insulator was probably designed before the availability of silicone oils. Also I don't yet understand why other high voltage insulator oils aren't suitable. Why Freon? Perhaps it's necessary to use a non-polar molecule, otherwise energy from the electrostatic field goes into rotational movement of the molecules - which won't be a constant drain over time - thus causing signal distortion.

The central problem with dividing a high voltage fast waveform down, is that whatever resistive structure you use will have unevenly distributed capacitive and inductive parasitics, thus forming a horrible RC/RL 'distortion box'.
And yet for high voltages you _have_ to use large geometry, high value resistances to avoid arc-over.  Or in general exceeding the dielectric breakdown strength of the resistive material, or even its substrate.

For very fast pulses perhaps a purely capacitive divider could work. But that has to avoid field variations, propagation delays in the electrode structures, and so on. I've heard somewhere there is a HV probe physicists use, that is mostly a pair of concentric metal spheres, insulated with Freon. But don't have any more info on it, not even the manufacturer.

I plan to wire the whole circuit point to point, and I anticipate it's going to be a nightmare laying it out to balance avoiding inductance while preventing arcing.  One of my ideas for prototyping it is to embed most of the circuit in paraffin wax.

What's wrong with oil? One BIG advantage: it's 'self healing'.  Secondly you can drain it off to work on the circuit.

I'm nowhere near beginning construction on the HV part of my project, but I'm expecting to be using coaxial transmission lines operated in an oil bath. Suitable oil yet to be determined.

I do have a Tek P6013A probe, though it's utterly inadequate for the project needs. Amusingly, it came with the large arc burn marks you see in the pic. Something vaporized some of the diecast metal of the probe compensation box, grounding to the scope frame. It doesn't seem to have done the probe or compensation box circuitry any harm, but I bet it gave the operator a surprise. They hadn't made any attempt to clean it up either. I filed the worst of the burnt metal jaggies off, and it works fine - within it's specs.
Hopefully the other electrode for those arc burns wasn't the operator.

I also have a 7104 and 7A29 modules, but a) the mainframe is currently dead, and b) really it's useless for single shot waveform capture. The microchannel plate screen is for beam intensity enhancement, not a storage tube. Coupling it to a CCD sensor sounds like a good idea! Did you try any fast single shot captures yet?

The HP 54120T 20 GHZ digitizing scope I have - same problem, in that it can't capture single shot waveforms. Plus the very amusing +/- 2V MAXIMUM voltage limit on the 50 ohm sampler inputs. Or you destroy the GaAs input circuitry. Oh right, I'm sure I can get that working safely nearby something generating multi-kilovolt impulses - not. Well theoretically it's possible with good shielding/grounding, but as the sampler modules for the 45121T are no longer made, there's not much margin for trial and error.

What might be usable: the Tek 7912AD, which is a really weird and cool thing. Lots on the net about it. But I don't like my chances of getting one at all, let alone getting it operational. I did get a couple of 7A29P plugins so far though, just in case.
The 1985 Tek catalog lists 'destructive testing' as one application for the 7912AD. Which um... as I understood it, it was developed for use in nuclear weapons testing. So yeah...

Hey, if anyone has a 7912AD they don't want, I'd be delighted to give you some freed-up space in return for it.

[pinkysbrein]

What high voltage transistor are you using BTW? I wasn't aware those could non destructively avalance.

[TerraHertz]

Speaking of Tek P6015: ebay 370929628416

Don't think I've ever seen a seller provide such crappy pixelated photos before.
Whether that means he's hiding some details, or is just an idiot, who knows?
If the latter it might be a good deal, as the stupidly bad pics will scare off some bidders.

Shame I'm completely out of cash atm, or I'd try to buy it.

[alm]

It occurs to me that the P6015 with its Freon liquid insulator was probably designed before the availability of silicone oils. Also I don't yet understand why other high voltage insulator oils aren't suitable. Why Freon? Perhaps it's necessary to use a non-polar molecule, otherwise energy from the electrostatic field goes into rotational movement of the molecules - which won't be a constant drain over time - thus causing signal distortion.

They redesigned it with silicone isolation, look at the P6015A. Replacement of the freon in the P6015 has been discussed extensively on the TekScopes list, but is non-trivial. It requires a boiling point not far above room temperature, so just a few drops of liquid ensure a sufficiently high vapor pressure within the probe. The probe will slowly leak, and you don't want this to compromise the isolation. Weight might also have been a consideration, filling the space with oil might make the probe very heavy. Obviously the breakdown voltage will need to be high enough. And the dielectric constant will need to be fairly close to the freon they used, since it forms a capacitor that's taken into account in the compensation network. I believe people have considered and tried various coolants, I'm not sure if any of them was really successful.

About buying a P6015 on eBay: note that without freon the maximum voltage is severely derated, I believe the specs will be fairly close to a P6013 then. There's a reason why P6015s without freon are much cheaper than P6015A probes. And note that even with freon the P6015 won't even come close to handling tens of kV pulses with ns rise times.

[pinkysbrein]

You can compensate for stray capacitance with a capacitor to ground forming a capacitive divider with the same ratio as the resistance.

The shunt capacitance is generally not the problem, as you say it's easily compensated for, the problem is the distributed parasitic capacitance across the length of the resistor to ground ... without a ground shield it will be variable and can't be compensated for, you can only drown it out with enough capacitance. With a ground shield you can compensate, but now the ground shield adds a lot of capacitance ... no free lunch.

Thinking in femtofarads is hopelessly optimistic.

[Rufus]

the problem is the distributed parasitic capacitance across the length of the resistor to ground ...

I didn't think the divider resistors needed to be anywhere near a ground for the OPs application - he only really has one point to probe, however, you are right the self capacitance of the divider resistors would be very significant.

[pinkysbrein]

madshaman, do you have ieeexplore access? "DC to 1 gigahertz multikilovolt voltage probe" describes a self-built probe, self built by master student with access to a metal shop and a network analyzer though.

[alm]

I wonder how they tested that. Many high-voltage components will have a voltage coefficient, so frequency response at 1.5 V may not be equal to response at 15 kV.

[madshaman]

madshaman, do you have ieeexplore access? "DC to 1 gigahertz multikilovolt voltage probe" describes a self-built probe, self built by master student with access to a metal shop and a network analyzer though.

I do, I'll add it to my file cabinet, thanks!

wrt high voltage avalanche transistors, afaik, any BJT will avalanche but exceeding a given peak current will destroy the device (by a method which I fully don't understand yet) as will normal thermal destruction (too much power dissipated in too short a time).

I'm away from my home machine and out of the country so I don't have access to all the papers I collected for this project.  From memory, one paper (not from my ieee file cabinet :-( ) described a general guideline for choosing candidates for avalanche mode transistors based on Vcbo being close to Vceo).  The paper was evaluating regular (non hyper-expensive Zetex avalanche -specific) BJTs for avalanche operation.  The authors chose a set of BJTs to test noting that some worked well, others did not.

My plan is to evaluate a small number of high voltage BJTs driven to avalanche breakdown via different capacitor arrangements (modulating the peak current and total energy delivered) and graph my results.

Digging back into my digikey orders, I ordered the following parts for this purpose when I was deep into the topic.  I made these choices to the best of my knowledge mixed with a bit of intuition, and I honestly don't completely remember everything I considered (ever spent half a day pouring over data sheets?):

http://www.onsemi.com/pub_link/Collateral/BUH50-D.PDF
http://www.onsemi.com/pub_link/Collateral/2N6515-D.PDF  (PNP)
http://www.fairchildsemi.com/ds/KS/KSC5502D.pdf
http://www.st.com/web/en/resource/technical/document/datasheet/CD00219617.pdf

I ordered large numbers as I figure I'll end end killing many of them.  I also plan to experiment with operating the [simple one transistor] circuit at temperatures down to ~= -70º C to see what effect this has.  I also have a large number of lower voltage BJTs for prototyping.

I might face total failure, but I figure that as long as I document all my observations, I'll learn something and will gain experience with this kind of circuit.  I know I have an enormous amount to learn here.  I also know this will be a huge endeavour as there are so many aspects to the circuit I want to build.

Here's one paper that might be of interest to you although not directly linked to my particular application:

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=5386197

Also, if you put "avalanche bjt" or "avalanche transistor" into the IEEE XPlore search box you'll see many articles on the topic, describing many topologies and also some papers on the development of high voltage BJT specifically designed for high voltage operation.

Although my electronics is half-way decent, academically I come from a computing and information science background so my in-depth knowledge of transistor operation at the physical level (beyond the basic theory) is sorely lacking; I'm hoping to rectify this through this project and I would greatly appreciate any help or pointers to facilitate increasing my knowledge.

P.S.

My apologies for not being able to provide as much information as I could from my home machine, honestly, connectivity sucks here (with my SO who's doing international development work in East Africa) and I had to hold my laptop out the window to access IEEE and my digikey account properly, which is very annoying.  I'm glad to see so much interest in the topic and I'll certainly add more when I return home in December.

[madshaman]

You have a very similar design problem as with something I'm working towards. And no, I haven't really come up with a waveform capture solution yet either.

You might find some useful information here:
http://w140.com/tekwiki/wiki/P6015

Thanks for the link!  What are you working on?

For very fast pulses perhaps a purely capacitive divider could work. But that has to avoid field variations, propagation delays in the electrode structures, and so on. I've heard somewhere there is a HV probe physicists use, that is mostly a pair of concentric metal spheres, insulated with Freon. But don't have any more info on it, not even the manufacturer.

Yeah, ideally I'd run into someone who's actually had to build such a thing :-(

What's wrong with oil? One BIG advantage: it's 'self healing'.  Secondly you can drain it off to work on the circuit.

Not opposed to HV transformer oil of some sort.  The reason I immediately thought of paraffin is that:

a) it's an excellent insulator
b) it's a solid a lower temperatures and I can avoid accidentally spilling it all over my lab ^^' and I figure it's much less messy to work with
c) I can go buy it in large quantities pretty much anywhere and I've used it before

(it also makes a decent shield for high speed neutrons for when I'm older and crazier and start work on that Farnsworth reactor ;-))

I do have a Tek P6013A probe, though it's utterly inadequate for the project needs. Amusingly, it came with the large arc burn marks you see in the pic. Something vaporized some of the diecast metal of the probe compensation box, grounding to the scope frame. It doesn't seem to have done the probe or compensation box circuitry any harm, but I bet it gave the operator a surprise. They hadn't made any attempt to clean it up either. I filed the worst of the burnt metal jaggies off, and it works fine - within it's specs.
Hopefully the other electrode for those arc burns wasn't the operator.

llol

I also have a 7104 and 7A29 modules, but a) the mainframe is currently dead, and b) really it's useless for single shot waveform capture. The microchannel plate screen is for beam intensity enhancement, not a storage tube. Coupling it to a CCD sensor sounds like a good idea! Did you try any fast single shot captures yet?

No, not yet.  Ironically, I've got my mainframe working but repairing the 7A29 modules I have is still on my todo list.  I've gotten as fas as buying the Tek camera mount I'm going to retrofit but that's all; I don't anticipate too much trouble, the only thing requiring real work will be rigging up the circuit to trigger it at the right time.

The HP 54120T 20 GHZ digitizing scope I have - same problem, in that it can't capture single shot waveforms. Plus the very amusing +/- 2V MAXIMUM voltage limit on the 50 ohm sampler inputs. Or you destroy the GaAs input circuitry. Oh right, I'm sure I can get that working safely nearby something generating multi-kilovolt impulses - not. Well theoretically it's possible with good shielding/grounding, but as the sampler modules for the 45121T are no longer made, there's not much margin for trial and error.

A tinkerer after my own heart, guess what 20Ghz sampling scope I have too ;-).  I've been drooling over the 50Ghz test set, but I cant justify paying $2000 for one (which is the best price I've seen so far), I don't really need it and imagine my joy if the thing turned out to be fried.

What might be usable: the Tek 7912AD, which is a really weird and cool thing. Lots on the net about it. But I don't like my chances of getting one at all, let alone getting it operational. I did get a couple of 7A29P plugins so far though, just in case.
The 1985 Tek catalog lists 'destructive testing' as one application for the 7912AD. Which um... as I understood it, it was developed for use in nuclear weapons testing. So yeah...

I'll check it out, I love weird s**t like this .

[ElectroIrradiator]

Is this a commercial development project, or more of a hobby experiment?

If the latter, then there may be ... other methods ... than an avalanche transistor to consider for your pulse generator, depending on how critical the generated pulse waveform is.

[johansen]

if I'm not mistaken, the peak KVAC rating for the P6015 of which i have one:
is derated for continuous signals only, and the issue is the probe arcing over internally, not some kind of voodoo.

so as long as you've got less than 40KVdc applied to the probe for less than 100ms and less than 10% duty cycle and you're below 8000 feet, you will do fine.



(i had to build my own compensation box. it has 9 feet of mystery oscope lead on the order of 30 ohms per foot.)