Sorry for wordplay with few topics ago, but problem was as similar as completely different. So....
I had need to measure i and U via certain high-voltage circuit fed by 1kV 25A power supply.
This supply is certain type of SMPS with non-capacitative output, both wires floating. Tester shows no current and even no voltage against the network zero nor AC nor DC.
Thus I brought a Hantek PC-oscilloscope box, very small current sense resistor and 100Meg / 1kOhm voltage divider matrix. Expected values 2...3V in one and 1...2V in other input. Both input commons are together and attached straight to one of PS outputs, as well to PC gnd via USB power cord.
Need to say, PC gave a overcurrent USB switchout sequence, I rebooted and now occilloscope have no a reaction at all. Tried the same oscillo to another PC - not working, tried another oscillo to both PC, and working well.
So... WHAT I did wrong??
More over, If any I need sometimes to make a shot for few kV over the real Earth gnd, I always had used Picoscope or Hantek just attached to laptop in battery regime. Now I am not so sure that SUDDEN high-Voltage rapid change, one needle, may not kill my dear hedgehog.
So options - not rely to the USB gnd connection and use the video-plug gnd as surplus earth be attached to those high voltage common point?? Will it help at all? What a groundworks would be the most useful for this problem? I am rather sure the problem was caused by harsh needle in PS switch-on moment, when at least capacitatively the PC gnd is bound to network gnd, what is detrimental all the ways. I mean, the any metal piece has few pF capacitance over global plane, and for short needles its enough to let in the killer wave?

 

Do you have a drawing of what you are doing? It would greatly help with any attempts to diagnose this.

So, what you are trying to chase is where exactly where either a high current or high voltage path were present. From your description I can think of a few points where the issue may have occurred:
- A leakage capacitor from the mains to the output may have been the culprit - measuring the output wires with an ohmeter would not show this as it blocks DC.
- A faulty resistor on your input divider that caused a high voltage differential on the Hantek inputs. Either the 100M shorted or the 1k open could cause an overvoltage. That or the resistor arced in face of the 1kV.
- A distraction that caused a high current path from the earth GND (on the PC through the Hantek) to any of the high voltage outputs.

Your option of having a floating laptop is always good to cover scenarios 1 and 3 above, but hopeless in 2.

What I would try to do in all cases is first check the present of voltage differences using a high quality DMM (not a chinese cheap but a tier 1 brand) in both AC and DC ranges to try to detect any leakage. After that, insert a very high resistor in series with the output of the power supply to limit its current - something similar to the series incandescent lightbulb that people use when powering up ancient equipment in unknown state.

 

1)OK, rapidly sketchet that, here
2) Now everything are Chineeze, Mastek, Canon, Hamburger, Pepsi, Coca, Marlboro, and no big surprize if Trump once will show he is Chineeze.
3) my DMM has 20 MOhm and shows less than 1 V. Both tests is an routine before any measurement to me.

 

If I understood the diagram correctly, the oscilloscope's input B can receive the full 1kV if there is arcing on the Spark Gap (if by spark gap you mean this).

Also, the leakage capacitor theory may still apply: if there is a difference of potential between the "true gnd" and the desktop PC GND (which is in turn connected to the oscilloscope probes GND), sparks may fly.

In the light of the damage done, I suspect the first scenario is more plausible: 1kV may have damaged your Hantek.

 

The current sense is minor suspect, as it generates the multiple tiny under-electrolyte micro-sparkles having the current in more than safe range with a factor of ten. However, as those was niobium electrode.... who knows the habits of this rare metal.
And second is that PC have eternal handshaking with Hantek, not simply damaged one channel. Thus I suspect the USB interface IC is damaged.
Today I must organize experiment with spare oscilloscope but I deadly fear.
The voiltage difference between PC gnd and Planetary gnd is obvious, it cannot be avoided, especially for the moments when PS is switched on or off. But that PS of Amtek DSA1000/15 have said it manual that any trial to destroy the floating character of load will instantly kill the 5000 worth apparatus. The negative gnd will kill it immediately and positive gnd will kill it if over 300V difference to true gnd (what means always).
Ought use a pedalling electro-station of floating network, then I be able to make a proper grounding

 

Thanks for all supported me at my pain, but received a good news, the Hantek is willing to shift my scope by warranty. So the loss was only the nerves.
Actually, re-positioned all the technique units to be maximally far from any grounded metals, organized in output of PS the 1 Mohm dummy load and winded 2 turns on the steel 20kg weightlifter donough with the power cable as well put on it PC wire`s ferrite core and already dozen experiments was succesfully provided each at 20 minutes long written in session, and it all serves well. Thus, the first alarm was, probably, just the "random error".

 

Thanks for all supported me at my pain, but received a good news, the Hantek is willing to shift my scope by warranty. So the loss was only the nerves.

Congratulations.

Would they insist on that if they read this thread?

 

Already more than week and new machine is working well, .... no, well is not right the word. When at high frequencies I never before hooked this problem, but when I have a one sweep going about 1000 seconds I clearly see that, example 801.st second reorganizes peaks into the 17.th second; the 802.st makes the huge peks in the 23.st second to unexist, but in the 3rd second some previously unknown peaks are obtained and so on and so on. Until the very end the picture obtained in beginning is changed up to non-recognizeable shape.
So, I wrote the question to the Hantek. No response. Then I made a screenshots illustrating the problem and as kindly I can invited them to identify the reason why future may change the far past.
No answer at all. I am hardly changed my mind about their cheap and rather fast scopes. Sorry to tell but this is a piercy example of firms of NONRESPONSIBLE about their own deeds.