(Wall of text warning)
A long while back I picked up a GAY MTA memory voltmeter as mentioned in
https://www.eevblog.com/forum/chat/show-your-multimeter!/msg1426415/#msg1426415Neomys Sapiens was kind enough to scan the original manuals to me, which I uploaded to ko4bb:
http://www.ko4bb.com/getsimple/index.php?id=manuals&dir=06_Misc_Test_Equipment/GAY It came with a standard Lemo triax connector to RG58 plug lead, which it claimed was suitable for use up to 1kV peak. The main purpose of the meter is to capture transients on both high and low voltage AC and DC rails.
Ing. Guido Gay was clearly a clever chap, combining a fast analogue sample and hold with a comparator and discrete D/A to capture transients of <50ns without sag in a 1970s design, but safety standards have changed a lot and I wouldn't be happy risking putting mains into that clip lead and connector, even if it has a 1kV range, particularly with a 1Meg resistive input.
Reading the manuals revealed that the instrument was originally also supplied with a 10kV peak X10 probe with 10Meg input. Time to build a probe then - one that needs to provide accurate resistive and capacitive attenuation.
The first challenge was finding a matching nearly unobtanium Lemo triax connector (triax because input low is isolated from the case, even though the input cable is coax). I considered modifying the existing lead to either connect to the probe via 4mm jacks or BNC connector, however the first would be subject to capacitance variations from the plug leads and the second would leave the exposed BNC shell connected to input low. Luckily I found an ebay seller in Germany offering nice clean used Lemo triax plugs (Size 1S for reference) for not-silly money.
The original probe was 'large cylindrical probe with tip' style with input low clip lead, hardly ideal for monitoring supply rails. I decided that Shielded 4mm jack sockets would be much more practical. A plastic case is necessary for insulation of the floating probe.
I was keen to ensure safety at maximum peak voltage so high voltage resistors and capacitors are essential. I found some TE HB series 3Meg 2W 7.5kV 1% 100ppm/'C resistors at Farnell. Three of these in series provides the required 9Meg resistance.
The AC attenuation was a little more tricky - no opportunity for a trimmer cap, so it had to be right first time. I noticed that the manual indicates that the probe cable length was "calibrated", ok employ the same strategy then. The MTA input is specified as 1M / <50pF. By measuring the capacitance of the existing lead unconnected and connected I found that the input capacitance of my unit is actually 43pF (here's where my trusty old Marconi TF2700 RCL bridge comes into it's own - by carefull twiddling of the loss balance control it can measure 43pF in the presence of a 1Meg parallel resistor). I therefore needed to provide an input attenuator capacitor of one-ninth of this input capacitance plus a sensible length of RG58C/U (102pf/m). Three 47pF capacitors in series provides 15.7pF (x9 = 141pF), close enough. I found some Murata DEC1X3J470JC4B 47pF 6kV caps, these are SL(JIS) ceramic which is a class 1 type (minimal voltage coefficient). Putting each the three caps in parallel with each of the three resistors provides equal voltage sharing.
I avoided a PCB for minimum leakage and capacitance and opted for air-wired. The resulting structure, braced by the capacitors is very sturdy, no danger of accidental shorts. The resulting assembly measured 16.1pF (TF2700 again). To achieve an accurate capacitive divide ratio, I just needed to adjust the RG58 length (plus Lemo connector )to ((16.1 x 9)-43.1)pF = 102pf, convenient. I cut the RG58 to about 1.1m and then worked back until I hit 102pF, almost exactly 1m.
The photo shows the internal view of the probe. It's quite compact for the peak voltage rating. I've insulated it carefully with additional plastic shims to increase creepage and clearance to the case seams and sleeved the screw pillars for the same reason. The RG58 inner and braid are also carefully separated to keep the capacitance stable. I've been all over the probe with my 10kV insulation tester, checking all the seams and fixing screw heads. I also tested the RG58 outer and Lemo connector shield to 3kV, comfortably above the 1kV rating of the MTA's input low to case ground capacitor rating of 1kV. I reinforced the RG58 insulation between the outer of the RG58 and Lemo connector with a good thickness of Kapton tape.
The final result is a 10Meg <15pF x10 attenuator (original probe spec <20pF). I tested it against the plug lead on DC and AC sine wave up to 100kHz and also function generator fast edge square wave signals and the divide ratio correlation appears accurate. The only limitation is continuous power dissipation - 10kV continuous would cause 9W dissipation in the attenuator (Something to be very aware of when making high voltage dividers!). I have set myself an upper limit of 3kV rms, this drops the maximum dissipation to 300mW per 3Meg resistor, way more than I need and doesn't compromise the 10kV peak capability.
I hope this proves useful, maybe even interesting, for others needing to construct similar AC/DC dividers. Thanks for reading this far!