DIY DC-10MHz optical-fiber-isolated scope probe

[davekni]

The difference in behaviour between a rounded and a sharp breakout point is somewhat expected but it is certainly nice to see this experimentally validated.

Yes, roughly as expected, but I didn't know ahead of time what actual voltage would be for either breakout.  Also as expected, voltage after breakout is similar.

1.05 to 1.1 sounds reasonable for a current ratio.  Effect on resonant frequency will be a bit less since current into coil capacitance doesn't pass through all the coil's inductance.  I haven't tried to model or measure it more accurately.

Thank you for the thoughts about negative arc charge.  I presumed it would be something more about the breakout point surface, electron emission work function or secondary electrons emitted by electron or ion impact from adjacent arc air.

This might be one of those cases where it would be advantageous to build your receiver into a small metal box  with a BNC on the side so it plugs directly into the scope BNC - no cables.

Yes, that would be a clean solution.  My mechanical abilities are minuscule and slow.  That's why I build with milk crates, foil, etc.  I'm planning to just replace coax cables on the two problematic receivers.

For power switching how about a small magnet and a reed switch? No need to open your shielding then.

Interesting idea.  Given how infrequently these optical probes are likely to get actual use, I'll probably not bother with any cleaner solution.

[Uspring]

I presumed it would be something more about the breakout point surface, electron emission work function or secondary electrons emitted by electron or ion impact from adjacent arc air.

Thinking about this a bit more, that might very well be. Negative corona seems to have an advantage over positive one due to the possibility of photoeffect on the cathode surface, which isn't possible for positive corona. A negative charge cloud might build up during lower voltages due to corona and then vanish, when electrons are generated by hot plasma. That is much more efficient in freeing electrons and differences in polarity due to surface photo effects become insignificant.
So the jump in charge might not be due to the onset of corona but due to its end, i.e. replacement by plasma. I've plotted a sliding average of 10 data values from your rounded point breakout charge data. 10 data points correspond to roughly a period, so this will suppress the TC frequency.

It's quite an amazing jump, considerably more than a full cycle worth of charge going into and out of the breakpoint. For the sharp wire breakout, the charge jump is about 10 times smaller. It would be interesting to know the value of your bleeding resistor.

[davekni]

It's quite an amazing jump, considerably more than a full cycle worth of charge going into and out of the breakpoint. For the sharp wire breakout, the charge jump is about 10 times smaller. It would be interesting to know the value of your bleeding resistor.

Yes, I was surprised at the jump, which is why I posted about it.  The bleed resistor (probe input divider resistance) is 30.7k.  With 2uF charge accumulation capacitors, time constant is 61.4ms.  Due to the small amplitude, I had the scope on AC-coupled to remove final opamp DC offsets.  That time constant is 16ms, so the more significant one here.  Still longer than the apparent time-constant in charge data.  (Designed this setup for my DRSSTC, so these SSTC charge values are comparatively small, requiring 1mV/div for the breakout channel scope setting.)
Thank you for the possible explanations.  BTW, as you can see in your low-pass-filtered data, total charge was consistently reversed to slightly positive by the end of the first half-line-cycle 8.3ms later.

[Hydron]

Great to see this going - analogue optical fibre based is definitely a more accessible setup than what I did before!

I'm interested in why you're using 3 channels - are you measuring secondary-probe, probe-toroid, and probe-breakout currents? I only had 2 channels available (though I could actually put all 4 in the topload, but without any left for primary/secondary-base currents), so I had to find out toroid current by subtracting breakout current from the secondary current flowing in.

Good idea measuring current across a capacitor btw - will need to have a go with this myself (I'm finally gonna be in the same country as my big coil again in a few weeks, and have been prepping a UD3 controller etc to be ready for some testing of my own )

[davekni]

I've started a new thread for continued discussion of top-load scope measurements:
https://highvoltageforum.net/index.php?topic=1950.msg14588#msg14588
New more-accurate data are posted there.

I was going to be done with this, but 7% error was too troubling.  So I spent another 1.5 days or so experimenting.  First, changing to RG59B/U coax for two probes (to match the original one) solved the receiver issue.  No need for aluminum foil there - just the little copper-foil boxes around fiber receivers.  Still 7%, though.  Just eliminated foil.

Grasping for straws a bit, I even checked for any possible issue with electric fields modulating attenuation of the plastic optical fibers.  Nothing there.  Considered that perhaps the steel housing for the three transmitters had such a thin skin depth at 103kHz that it wasn't a reasonable shield.  Covered it with copper tape joined with solder.  No difference.  Tried again placing aluminum foil over the entire top-load center connected to top-load.  That makes it worse!  That bit is still a puzzle.

What did help was bending the little foil shields over each probe a couple mm to better cover the ECB.  The breakout channel (channel 3) has always been relatively clean compared to the other two (secondary coil and top-load).  Noticed that channel 3 shield placement was slightly different.  Bending the other two to match reduced total error from 7% to 3.5%.  Haven't been able to make any further improvements, however.  Error appears to all be transmitter shielding related, but adding more shielding often makes no difference or makes it worse.  Still a bit confused about this.  Trying really hard to fight my perfectionist tendencies and be happy with 3.5%.

    I'm interested in why you're using 3 channels - are you measuring secondary-probe, probe-toroid, and probe-breakout currents? I only had 2 channels available (though I could actually put all 4 in the topload, but without any left for primary/secondary-base currents), so I had to find out toroid current by subtracting breakout current from the secondary current flowing in.

Yes, I spent some time studying and analyzing your results.  Thank you for posting.  Your work is what inspired this project.  The reason for 3 channels is to verify accuracy.  3 provides redundancy, allowing me to sum them and see how close they are to zero.  That is where I have the remaining 3.5% error.  Without this redundancy, I might not have noticed any error and not known shielding improvement was needed.  My very first test (before any shielding improvements) summed to 15% of top-load charge.  Definitely would not have wanted to continue presuming accuracy when reality was 15% different.  (Error appears to be roughly split between secondary and top-load charge, both off by ~1.5-2% now.  Error is at the same phase, so not canceling as they should, yielding ~3.5% total error.)