DIY High voltage probe - safety concerns

[dazz1]

Hi
One of the problems building a HV probe is testing it. 
Aliexpress to the rescue.   Search for a "1000Kv Diy Kit High Voltage Generator 400Kv Dc High Voltage Generator Booster Board Inverter High Voltage Diy Kit Module"

An old car ignition coil could probably do something similar if you don't want 1,000,000 volts. 

Don't ask me if I believe the kit actually produces 1MV but who would know. 

[tautech]

Hi
One of the problems building a HV probe is testing it. 
Aliexpress to the rescue.   Search for a "1000Kv Diy Kit High Voltage Generator 400Kv Dc High Voltage Generator Booster Board Inverter High Voltage Diy Kit Module"

An old car ignition coil could probably do something similar if you don't want 1,000,000 volts. 

Don't ask me if I believe the kit actually produces 1MV but who would know.

What do you want your headstone to read ? 

[Gyro]

Yes, that output voltage spec is pure Chinesium, divide the output voltage by 10 - 30. Those things break down without an external spark gap.

[dazz1]

Hi
There is another feature of these types of HV probe that would limit frequency bandwidth.
That is the loop formed by the probe and the earthing lead.  My DC probe has a generous loop so the earth lead is safely away from the HV bits and pieces. 
If it was a constant earth loop area, it could be compensated out to a certain degree, but it isn't. 

So I think a DIY 25kV HV AC/DC single sided passive probe is likely to be physically limited to around 25MHz to 50MHz or so with reasonably priced off-the-shelf parts. 

The intermediate step between a fully passive and a fully active differential probe is to insert and active compensating amplifier between the probe and the scope.  The more I think about this, the more I think it would be a good thing. 
If the amplifier output was matched to 50ohm, it would remove the cable between the probe and the scope from the compensation network.  With a 50ohm terminator at the scope/meter input, you could use any 50ohm cable at any length. 

[tautech]

Hi
There is another feature of these types of HV probe that would limit frequency bandwidth.
That is the loop formed by the probe and the earthing lead.  My DC probe has a generous loop so the earth lead is safely away from the HV bits and pieces. 
If it was a constant earth loop area, it could be compensated out to a certain degree, but it isn't. 

So I think a DIY 25kV HV AC/DC single sided passive probe is likely to be physically limited to around 25MHz to 50MHz or so with reasonably priced off-the-shelf parts. 

The intermediate step between a fully passive and a fully active differential probe is to insert and active compensating amplifier between the probe and the scope.  The more I think about this, the more I think it would be a good thing. 
If the amplifier output was matched to 50ohm, it would remove the cable between the probe and the scope from the compensation network.  With a 50ohm terminator at the scope/meter input, you could use any 50ohm cable at any length.

34kV 50 MHz 1000:1 scope probes are available.
https://www.pintek.com.tw/customer/pintek/upload/HVP-39-28-18-10-SPE.pdf

[MariuszD]

Yes I have seen that manual before. 

Quote from this document

The 100-MOhm resistor is four inches long which makes the input RC time constant complex.
The distributed nature of the input capacitance is indicated in the schematic.

Yes I read that too.  I interpreted this as  making the input RC a complex impedance, which needs to be compensated out.  The manual is ambiguous.

It is not ambiguous, look at the schematic. They draw multiple parasitic capacitors around the input resistor.
Complex RC time constant doesn't mean impedance is a single complex number. Time constant is a different thing than impedance.
The RC time constant is never written as a complex number.

I suspect that if you give up the wide band, the compensating capacitor can be a single one.

[dazz1]

Yes I have seen that manual before. 

Quote from this document

The 100-MOhm resistor is four inches long which makes the input RC time constant complex.
The distributed nature of the input capacitance is indicated in the schematic.

Yes I read that too.  I interpreted this as  making the input RC a complex impedance, which needs to be compensated out.  The manual is ambiguous.

It is not ambiguous, look at the schematic. They draw multiple parasitic capacitors around the input resistor.
Complex RC time constant doesn't mean impedance is a single complex number. Time constant is a different thing than impedance.
The RC time constant is never written as a complex number.

I suspect that if you give up the wide band, the compensating capacitor can be a single one.

The Tektronix manual is ambiguous because there are different interpretations of what it means. 

The Tektronix network tunes different parts of the square wave in the time domain.  I think the Tektronix network tunes out signal reflected from the mismatch with the scope input.   One cap won't achieve that.

My HV scope probe is 1000:1 with 100Mohm input resistance.   The HV circuit I am fault finding has resistors up to 68Mohms, so a 100Mohm probe just drags down the voltage I am trying to measure.  My 1,000:1DC probe has 10Gohm input resistance but there is 100x difference in gain going from a multi-meter to a scope.

I notice on the spec sheet you linked, the input capacitance rises with input volt rating.    This is to be expected as the increasing volt rating requires a physically larger resistor, and that will have more capacitance. 

I have also noticed that AC  HV scope probes tend to be short and fat.  I suspect the length is dictated by the length of the input resistor.  I suspect the diameter is relatively large to accommodate a shielding tube and minimise the input capacitance.

I have also noticed that no modern probes I have seen includes a bare metal earthing ring between the probe and the handle.  If the ring is connected to the earth, then any flash-over from the probe end will be grounded and will not reach the user.   If the earth lead is not connected, then the ring will be at the HV input voltage via the input resistor.   25kV / 10Gohm is 2.5uA.  Probably enough to give someone a zap (like HV static electricity), but not enough to end a life.   

In contrast, an arc-over along the  probe surface, through a plasma gas arc will have close to zero resistance and the peak currents could reach spectacular values.  I am careful to make sure my HV scope probe (rated to only 5kV, no earthing ring) is very clean before each use.  I clean it with IPA at least an hour before use.


One thing hasn't been discussed is the noise contribution of the input resistor.  a 10Gohm resistor at BW 50MHz will contribute 91uV = -20dBV = -60dBkV.  Reduced S/N ratio is the penalty for increasing the input resistance.

[MariuszD]

The Tektronix manual is ambiguous because there are different interpretations of what it means. 

The Tektronix network tunes different parts of the square wave in the time domain.  I think the Tektronix network tunes out signal reflected from the mismatch with the scope input.   One cap won't achieve that.

They use a resistive wire in the coaxial cable to dampen reflections. Like in all other probes. There is nothing ambiguous, but I see that you really need your interpretation to fit even though many arguments are against it.

It all comes down to the amount of work, you assume that you will make a much simpler probe with a comparable results to the Tek's probe by using a worse resistor. If your assumption is not right, you will simply have to create a more complicated probe.

[dazz1]

Hi
I have asked a local PTFE supplier for relevant electrical specs.  Silence.
I asked Ohmite for frequency related specs on specific resistors.  Silence 

I am not going to use PTFE unless I know the insulation properties published by the manufacturer.
Similarly I am not going to purchase multiple expensive resistors to find the one that works.

To build a very high resistance 10Gohm passive probe with a decent frequency response would be difficult.
If I was going forward on this, I would look at a hybrid active, single ended probe using a trans-resistance opamp as the grounded sensor. 
By isolating the probe from the cable and scope input, it would be a lot easier to increase the frequency response.    The performance of the probe would not be determined by the coax.
The accuracy of the probe would not be dependent on the input impedance of the scope or volt meter. 

[RJSV]

   In school I was amazed and struck, by the amounts of 3-D graphical relations, in the construction of HIGH VOLTAGE components, and support structures.  Especially with optics, but the question here is about safe HV probe construction.
   One very visible example is those repeating 'egg crate' style surfaces, that are meant to expand the volts per meter, in relation to surface runs (of current leakage).  Since these things aren't always intuitive I'd have to ask:
   "Is your particular design both experimental, AND a casual hobbyist amateur design ?"

   I know, I can be an A-hole, with rule enforcement, but...
That HV Probe, built at home MIGHT be quite safe,  but that's not the point.   For example;  Do you always have a companion nearby, while experimenting ?

   What comments, did your lawyer issue ?  (Snarky, but I needed to go there.)

Home owner's insurance ?  O.k. that was out of line, there, but just sayin.

    Aside from the criticism, though, the topic is an interesting one.  Especially the weird, 1950's space travel looking hardware, for high voltage isolation, of some hand-held gizmo.
There are, besides measurement, applications for a 'Shorting Wand' type of tool, for shutting down HV supplies, motors, etc.

[dazz1]

   In school I was amazed and struck, by the amounts of 3-D graphical relations, in the construction of HIGH VOLTAGE components, and support structures.  Especially with optics, but the question here is about safe HV probe construction.
   One very visible example is those repeating 'egg crate' style surfaces, that are meant to expand the volts per meter, in relation to surface runs (of current leakage).  Since these things aren't always intuitive I'd have to ask:
   "Is your particular design both experimental, AND a casual hobbyist amateur design ?"

   I know, I can be an A-hole, with rule enforcement, but...
That HV Probe, built at home MIGHT be quite safe,  but that's not the point.   For example;  Do you always have a companion nearby, while experimenting ?

   What comments, did your lawyer issue ?  (Snarky, but I needed to go there.)

Home owner's insurance ?  O.k. that was out of line, there, but just sayin.

    Aside from the criticism, though, the topic is an interesting one.  Especially the weird, 1950's space travel looking hardware, for high voltage isolation, of some hand-held gizmo.
There are, besides measurement, applications for a 'Shorting Wand' type of tool, for shutting down HV supplies, motors, etc.

I am OK making an HV probe for myself, but I wouldn't make one for someone else.
I always have someone else in the house when I am working with anything above 48V.
I am somewhat familiar with HV equipment.  I have a short length of 200kV (that's not a typo) power cable in my garage from a past job. 
I am not prepared to use materials (including PTFE) not specifically rated for HV use.  If you study semi-conductor materials, tiny amounts of impurities turn an insulator into something that doesn't insulate.
In electronics, usually HV supplies are low current.  Easy to drag down with a 100Mohm probe.
Designing a AC probe with 10Gohm input is a significant challenge.
The fault that initiated my investigation looking at HV Probe design is rectified.  I am not looking at this problem any more.

[HighVoltage]

My approach to a high impedance DIY high-voltage probe is different.
Separate probe for AC and separate for DC. The AC probe is exclusively a capacitive divider, the DC probe is exclusively resistive divider. Sometimes DC level is known or not important.

Very good, this is a perfect approach !
I have been doing this for many years with great success.
If I compare my signals with the Tek 6015A, they are perfectly matching on the scope screen and all values can be set to match the measurements.

[sourcecharge]

So, I have a need for my high voltage, high impedance probe again.

I was mainly trying to use it for up to 1.5kv but I need the high impedance for the measurement.

I decided that I wanted the output of the probe to be fed into a non-inverting opamp as a high impedance buffer so that the output can be used with any meter, and oscope.

After doing this, I ran into mains line noise trying to measure AC and DC measurements, and I figured out that this was not only noise that I was seeing from the mains but also the signal noise.

Meaning the 10khz that I thought I was measuring was actually only noise that bled through the same way as the mains line noise.

I messed around with noise canceling circuits, but I found that the best way to get rid of it was to shield the probe (steel sheet metal tube), the wire (BNC), and the buffer circuits (steel sheet metal box).

I found that the probe was only usable for DC and low frequency (5hz) AC measurements as the compensation was completely off.

Although it will allow me to measure a ripple voltage, I still would like a HV high impedance AC probe.

I think this is the best approach for DIY:

My approach to a high impedance DIY high-voltage probe is different.
Separate probe for AC and separate for DC. The AC probe is exclusively a capacitive divider, the DC probe is exclusively resistive divider. Sometimes DC level is known or not important.

In my AC probe, the capacitor had only 0.1pF; I suspect that no commercial  HV probe has such a high impedance. Of course, the capacitance of the probe with a 0.1pF capacitor is greater than 0.1pF but still less than 1pF. I can't compete with Tek trying to build a better copy of the P6013A, but I can achieve a higher impedance in a probe that only transmits AC (or DC).

Component cost at digikey right now for 100x 1600VDC PP (0.0006 DF) 100pf caps are about 20 to 30 bucks.

I'm wondering, what type of caps (and how many) did you use for your 0.1pF probe?

Did you shield that probe and how did it react with or without the shielding?

Is this the reason why the capacitance was higher than 0.1pF?

Did you have to compensate C2 after the C1 was completed?

Did you wind up using a small variable cap in series or parallel with C2?

What was it's final cutoff frequency and rated voltage limit?

I'm interested in knowing how well this turned out because it really does seem like the best option.

[MariuszD]

I needed this probe to build a device, but the probe wasn't a priority, which is why the final version hasn't been completed to this day.

I'm wondering, what type of caps (and how many) did you use for your 0.1pF probe?

DIY capacitor
The input capacitor was a PTFE tube with a metal rod inside and a copper ring on the outside.

What was it's final cutoff frequency and rated voltage limit?

My device was giving 5kVp and the PTFE tube was 1mm thick, so I had a margin of safety. On the 10ns rise time waveform, I didn't see any significant discrepancies compared to the P6013 (25MHz).

Did you shield that probe and how did it react with or without the shielding?

Under favorable conditions, measurements without a shield are possible, but it's better to use one.

Is this the reason why the capacitance was higher than 0.1pF?

Capacitance to shield

Did you have to compensate C2 after the C1 was completed?

The division coefficient needs to be trimmed.

One of the first versions of this probe used a capacitor with alumina ceramic plate, and I observed corona discharges on the plate, this is unacceptable. Later, I filled the empty spaces with silicone grease.

[sourcecharge]

The following electrical characteristics of PTFE come from a datasheet from professional plastics, here in the US.

Unfilled PTFE has basically the best DF of most known materials (<0.0002), and also it's dielectric constant is only at 2.1, both at 1 Mhz.

Unfilled PTFE has a breakdown voltage of 285 V/mil, which puts your probe at about 11kv.

Sourcing unfilled (white, not clear) PTFE is as easy as 1.75mm 3d printer filament tubing if it is truely unfilled ptfe, as there are glass filled and carbon filled ptfe which I'm guessing is either clear or black.

Your design is pretty good (if not one of the best), and I like the use of PTFE as it really is the best material for this.

Just a couple of questions though.

Was the shielding grounded, or was it C2?

If not, how could you differentiate between them if C2 is grounded in the C1:C2 dividing network.

Did you use a seperate grounded shield, with a virtual ground on C2 in your one of you latest versions?

I'm assuming that you used the output to an active adaptor because the impedance of C2 has to be so great that any meter would ruin it.

I'm also assuming that you are using BNC cables for probe outputs.

That would mean if the probe was shielded with a ground tube, the output of C1, and C2 would both require a BNC output terminals with the grounded shielding not connected to C2 which is on a virtual ground.

At these low of capacitances, the BNC cables are now part of the dividing network too, are they not?

IDK.

Alot of the problems faced with a multi series capacitor DIY probe is solved with this design.

I think the virtual ground on C2 would work with a grounded shielding.

Putting an active buffer between the HV and the oscope is benificial too.

[intabits]

of 20kV or more, maybe add a 10MegOhm or 1MegOhm resistor directly as to form a "worst case" voltage divider and hook it up to some (not even HV) cable and test leads.
Worst case would be the GOhm-range input impedance of my HP3456A if it is accidentally in auto-range. The input capacitance might save it, but I better not count on that. Proper grounding beforehand is mandatory in any case.

Yes, do include a worst case bottom divider resistor, selected to give the correct division ratio when in parallel with the 10M DMM input.
If it's in a GOhm input impedance range, you'll just get the wrong reading, but no damage.
See the values shown in the Fluke 80k-40 manual: https://assets.fluke.com/manuals/80k40_6_iseng1000.pdf

[MariuszD]

Was the shielding grounded, or was it C2?

If not, how could you differentiate between them if C2 is grounded in the C1:C2 dividing network.

Did you use a seperate grounded shield, with a virtual ground on C2 in your one of you latest versions?

It was done very simply. The shield and capacitor C2 were connected to a common ground. The capacitance between the C1 second plate and the and shield, and the load capacitance required a reduction in C2.

I'm assuming that you used the output to an active adaptor because the impedance of C2 has to be so great that any meter would ruin it.

Ultimately, it's better to use a high-impedance buffer. I was only interested in high frequencies, so I connected my probe to the input of the factory made 1:10 probe. C2, at nearly 100pF(with 1:10 probe capacitance), loaded with a 10MOhm resistance, limits the bandwidth from below to 160Hz.

At these low of capacitances, the BNC cables are now part of the dividing network too, are they not?

I was experimenting with buffers for the active probe. Fast operational amplifier, noninverting, with 50 Ohm cable matching

I think the virtual ground on C2 would work with a grounded shielding.

Putting an active buffer between the HV and the oscope is benificial too.

I don't see a reason to use virtual ground, but buffer is beneficial

[sourcecharge]

The reason why I ask about a virtual ground is because of the active adaptor's input offset and the possiblity of limiting the common mode capacitance thats grounded, due to the BNC cable on the non inverting input.

The high impedance non inverting buffer (HINIB) of the active adaptor also have parameters of input bias current (Ib), input offset current (Ios), and input offset voltage (Vos).

3 seperate parameters, all contributing to the total offset.

I'm not perfect with the calculation, but B2spice is saying that if the Ib and the Ios are 25pA each, then the total current that is calculated for the offset is 37.5pA.

Let's say for instance that impedance of C2 is 1Mohms, even if its much higher.

That's 1Meg ohm x 37.5pA which equals 37.5uV offset extra from the Vos, so that precision 1uV Vos opamp for example, now has 38.5uV offset on the output.

The impedance of C2 is much higher though, so the offset is going to be much higher too.

In fact, the impedance of C2 is so high that the HINIB of the active adaptor requires a resistor from the noninverting input to ground in order to not fry, which is now our R2.

R2 will have to be used in the compensation equation: R1xC1=R2xC2, but it can be anything you want, although I'm not sure what it should be in relation to C2's DF or if it matters.

So this offset on the input of the HINIB is there if you ground C2 with R2 grounded as well.

The reason why I say to use a virtual ground is to null the total offset.

In fact, if you are using a quad package or same opamp that came from the same lot#, it's possible to null the total offset using three opamps to null the first.

It works like this:

A HINIB as the first of three extra opamps, has it's output to C2 and R2 as the virtual ground.

Instead of rail splitting and using a pot to manually adjust the offset on the bias of the HINIB, which can vary due to rail voltages and temperature, two other opamps are included.

The first of the two opamps, starts off with as a HINIB with a resistor from the non inverting input to ground that is 2x the impedance of C2 and R2.

An extra pot can be included to this resistance, for nulling out the final Vos.

This creates the uV (or mV) offset needed to feed into an inverting buffer with a gain of 1, and biased to ground.

The output of the inverting buffer is fed into the virtual ground buffer's non inverting input, and nulls the total offset of the 1st opamp that is affected by C2's, and R2's high impedance due to the input bias current (Ib), input offset current (Ios), and input offset voltage (Vos).

I'm not sure if using a virtual ground though will limit the C(cm).

That's why I asked you about it.

Here's the schmatic that shows the difference with and without the virtual ground using the TP-17 (It's a cheap, mid ranged precision single opamp):

Of course this can be solved by using a jfet or cmos opamp as the first HINIB that has a typical 0.5 to 1pA Ib or Ios with low Vos, but those that have the bw and 36V Vcc max are more expensive.

Beware though, while doing a search on digikey (or where ever you shop), some opamps may say 5pA Ib, Ios, typical, but they can swing wildly to a max of +/-500pA.


Edit:

Just wanted to add that if the Ib, Ios, or Vos is opposite polarity, the resistor could be only 1x the C2, R2 impedance. 

The key point is that if all 4 op amps have about the same offset, you can use their own offset to null the total offset. 

[MariuszD]

Now I understand that you want to compensate for the offset resulting from the polarization currents. Resistor methods were used for bipolar amplifiers where the current Ib was the base current, making it large (nA) and repeatable, while the current Ios was small and resulted from the asymmetry of the input stage. In FET amplifiers, the Ib current doesn't flow into the gate; instead, it's the difference in current from the ESD protection diodes.  Therefore, it can vary significantly for both inputs, even flowing in opposite directions. Ios>Ib is not surprising. Diode leakage currents increase exponentially with temperature, so they can be many nanoamps at 125°C. This is why the maximum from the datasheet can be so high. However, the probe will not be used above room temperature.
Your compensation method should be tested in practice. A simulation won't tell you if it works.

I needed a fast probe, so I was only considering amplifiers >50MHz. These amplifiers have an offset >200uV. Above 70MHz, it's difficult to find an offset smaller than 2mV. That's why I wasn't expecting microvolt accuracy. It is known that a capacitive probe does not transmit DC, so if anything appears at the output, we can consider it to be an amplifier offset or a leakage current from the input capacitor. We can even block the DC component at the buffer output.

In fact, the impedance of C2 is so high that the HINIB of the active adaptor requires a resistor from the noninverting input to ground in order to not fry, which is now our R2.

Without R2, the circuit will saturate and won't work.
Due to the polarization current, a resistor R2 is necessary, the disadvantage of which is that it limits the bandwidth from below. That's why I'm considering bootstrapping to increase the effective value of this resistor.

R2 will have to be used in the compensation equation: R1xC1=R2xC2, but it can be anything you want, although I'm not sure what it should be in relation to C2's DF or if it matters.

The capacitive divider will not be compensated (R1 non existent), but that doesn't matter because the bandwidth limitation will be <200Hz. In theory, a capacitive divider has no upper bandwidth limit, but it is known that the stray inductance of real capacitors will prevent "microwave" operation
DC can be transferred using a separate low frequency resistive probe if needed. DF of PTFE should be low.

The reason why I ask about a virtual ground is because of the active adaptor's input offset and the possiblity of limiting the common mode capacitance thats grounded, due to the BNC cable on the non inverting input.

I didn't plan for the cable to be before the buffer because if the cable is after the buffer, it can be matched to 50R.
I don't know what the parameters of the capacitive divider directly connected to the cable will be. The RG-58 50R cable has about 100pF/m, but I'm worried about reflections. I'd have to try. Maybe it's better to cannibalize the cheap probe and get a lossy cable from it.
I came up with the idea that I could suppress reflections with a resistor inserted between the capacitive divider and the cable, an example from the simulation is shown below. The total capacitance of C2 is 200pF, of which 100pF is the cable. It even works somehow, but I don't expect this simulation to accurately represent reality. Beside, a 30ns ringing decay time is not a good result, my primitive probe without an amplifier and cable was better.

I don't see a capacitive divider or cable in your diagram, so the details are unclear to me.

I'm not sure if using a virtual ground though will limit the C(cm).

I don't quite understand. I was only considering single-ended probe. I don't see anything that justifies introducing a common-mode/differential-mode distinction. Do you mean the amplifier's internal capacitance?

The key point is that if all 4 op amps have about the same offset, you can use their own offset to null the total offset.

Amplifiers in a single package do not have identical offsets. And there's no reason they should. Beside, we don't know if the observed offset is a real asymmetry in the input transistors or the remaining inaccuracy after the manufacturer performed offset trimming.

[MariuszD]

My working prototype was a capacitive divider 1000:1 connected to a commercial 10:1 probe (probe input impedance of 15pF||10MOhm) which allowed me to avoid the problem of reflections in the cable and loading the divider with a capacitance greater than C2.  The improved version with a buffer can't be worse.

[BILLPOD]

One of these should be used, regardless of which probe is chosen:   https://www.amazon.com/Enespro-Universal-One-Piece-Electrical-AGHS-U6/dp/B0D1SJ47V8/ref=sr_1_2_sspa?adgrpid=1241349616540143&dib=eyJ2IjoiMSJ9.4D9XaCNcEItaxhgQpPZqZbfeNRMy0hARrFyg4kJQEXmAGlM-TwsStJctHU4hxV-X2dr2YoIxjn1PMBimrcXdufrUlBrZEdmaXQ3k489iUwRPfagXOjFlMqgXr5NxxjW5oDs1Vi--5SgZjc0Uuh6sY-DJj-USSpuUSqAGOA-_XBAitvpDAm-jfhcKGqgXanY_bXGBZjsj5YS7PjPEX0hrXke1HIsKmKQULr6jQDVLSFwF-cZcCt2eMzwR1QKd_uch-Ay__k1iJVxX26Zos_UoKz-IBd9T3F1TbsU4auhPyy4.ONRJ504l8IXA4Y0LXqU5jfW1S2MdL9JowRLIDEE92no&dib_tag=se&hvadid=77584480973536&hvbmt=bp&hvdev=c&hvlocphy=105222&hvnetw=s&hvqmt=e&hvtargid=kwd-77584596179876%3Aloc-190&hydadcr=20635_13426491&keywords=lineman+hot+stick&mcid=e59332a8ba3b3414941008e56fd76e88&msclkid=02a5290da45e1ecf91a041c1a98d3e57&qid=1758203536&sr=8-2-spons&sp_csd=d2lkZ2V0TmFtZT1zcF9hdGY&psc=1

[oz2cpu]

here is my own DIY high voltage and fast 100:1 scope probe

https://webx.dk/oz2cpu/hi-volt-probe/hi-volt-probe.htm

free gerber files, free 3D print files, free schematics it is all there

[dazz1]

here is my own DIY high voltage and fast 100:1 scope probe

https://webx.dk/oz2cpu/hi-volt-probe/hi-volt-probe.htm

free gerber files, free 3D print files, free schematics it is all there

I would suggest adding a flange to the 3D print to prevent your hand sliding down to the pointy end where all the voltage is located. 

[dazz1]

I needed this probe to build a device, but the probe wasn't a priority, which is why the final version hasn't been completed to this day.

I like this design but I think it needs to be partnered with a DC probe with matching performance.

I am trying to diagnose an HV fault at present where the circuit is producing a kV spike where should only be DC.  The capacitor probe would detect that.  In addition, the voltage is flickering at about 100Hz to 200Hz.  Probably too low for the capacitor probe to detect. 

My existing commercial DC probe does not show the spike or the flicker. Its frequency response (unmeasured) is too low.  So I am thinking that this capacitor probe needs to be partnered with a DC to (say) 1MHz probe.  The frequency response is likely to be dictated by the available HV high value resistors, like those made by Ohmic, plus the performance of the electronics signal processing. 

[dazz1]

Hi
I have read on this thread how some have used, or proposed, a range of materials to make HV probes.  People making these choices do so at their own risk, and that is fine.  I am conscious that people may follow without fully understanding the risks.  Basically it comes down to materials properties.

The properties of insulators are greatly affected by impurities included at manufacture.  Silicon semiconductors work because of impurities added to pure silicon.    The manufacture of plastics is a chemical process and all chemical processes include impurities.  Virgin PTFE is a great insulator, but if it is made in an anonymous factory where lowest cost is the priority, it is unlikely to be suitable for HV applications.  I wouldn't bet my life on it.  It would be advisable to buy PTFE from a reputable supplier who can tell you who manufactured the product.  You can then trace the origins of the product see what properties/quality the product has.  If they don't quote HV performance, at least get food grade ptfe where there should be less impurities than ungraded product.

None of the DIY probes I have seen attempt to test the limits of their probes.  It wouldn't be too difficult to set up a test rig, based on a car ignition coil, to test a design.  Of course any probe tested to arc-over should probably be thrown away.  Arc-over is likely to cause damage, even if it is not visible.  Testing would provide data on safety margins.

So I am saying that materials selection and origin is really important.  Testing to failure is important to measure the safety margins.
Keep it safe.