4mm binding posts leakage to ground.

[Alex Nikitin]

4mm binding posts commonly used on measuring equipment, including metrology level meters, calibrators and references. The isolation leakage to ground is not a problem for a typical use in an average multimeter and even less so in a power supply etc. However if you chase ppms it might easily become a serious concern. While 1Gohm sounds a lot, it may introduce an error of 10 ppm for a 10K resistance and 1000ppm (0.1%) for 1M resistance. And if you are looking for a 1ppm error at 1M level, a minimum leakage resistance acceptable would be above 1 Tohm. That is the main reason I've decided to place this thread in Metrology section.

I have means to measure these kind of leakage resistances and my Keithley 617 electrometer routinely set up for measuring leakages to ground for up to +/-100V test voltage. The self leakage of the measuring system, including a 1m coaxial cable and connectors, is below 5fA at 100V, or >20000 Tohm = >20 Pohm.

A little while ago I've bought a number of good quality (Tektronics spares, most likely Superior Electric BP21) binding posts from a forum member, with the idea of building some useful switching boxes, mostly to use with my HIOKI DM7275 which has only one input so gets abused with constant cable switching. My usual binding post of choice is Pomona 3770 Cu-Te however they are expensive, at about $15 each, so I had some hopes that a cheaper alternative would work for me in a switching box and other not so critical places.

The Pomona 3770 series uses polycarbonate for isolation, but the Superior Electric BP21 uses nylon, so I did expect leakage to ground to be larger, however I didn't realise how much larger it would be! My measurements on a post mounted in a metal box gave figures up to 30nA (!) for 100V or about 3.3 Gohm. The Pomona 3770 leakage measured at about 10pA at 100V or 10 Tohm, not perfect but 1500 times better (and generally good for most purposes you might use a 4mm binding post. I did also measure leakage to case ground for some of metrology items I have in the lab. The SR1010 resistors arrays measured from 100pA at 100V (1Tohm) for 100K set up to 1.8nA at 100V (~55Gohm) for 10K set. The Fluke 742A-10k measured 31pA or respectable 3.1Tohm leakage resistance to ground.  So I thought I've wasted my money on these Tek posts!

After some consideration I've decided to check if it is feasible to add insulation to Tek posts, using some PTFE washers, however I've found that suitable sizes are not available from stock. To order a custom size would be prohibitevely expensive. Nevertheless, I kept looking till I've found a cheap spacer from Rosenburger (available from RS), made from Polystyrene to isolate panel mounted BNC connectors, with perfect dimentions to take a 4mm binding post from Tek/Superior. Come next day, I've mounted a Tek binding post using a couple of these spacers and got a very respectable leakage - below 30pA or 3 Tohm. On par with the Fluke so I was reasonably happy and started to assemble the switching unit. After I've mounted all 9 posts I've measured the leakage and was somewhat surprised by the big differences between posts, the worst ones measure over 30pA, the best - under 2pA. So I've found a softer place on the wall and banged my head a bit on it, while another 20 spacers were washed in IPA. After thorough drying the spacers and re-mounting the posts, the leakage went down to 10-30fA on all nine, making roughly a 1000 times improvement. My head deserved the punishment as I should've washed the washers before assembling the lot first time!

Below are some photos, now I am not entirely happy with the switch I've used (an old ceramic-based Soviet military type) as it is the leakiest part in the whole assembly (around 3-5pA or 20-30 Tohm) however it is still better than the Fluke, so I'll live with it! The switch box allows for two 3-wire lines to be switched to one output (plus the OFF position), and two BNC connectors switched in the same way to one BNC output. The next test is to see how bad the thermal voltages are in this setup, with brass contacts.

Cheers

Alex

P.S. - it is worth mentioning the environmental conditions for all measurements mentioned above, the room temperature was 25C, relative humidity 40% .

[BILLPOD]

Good Morning Alex, VERY nice work.

[Stray Electron]

  Yes, that's a nice bit of detective work!  I'm glad to see some real numbers and some real measurements for projects like this.

   That rotary switch appears to be a copy of the American made (Switchcraft? brand) that were made in the 1960s.  Have you tested any of those for leakage?

[EC8010]

Interesting stuff. To save further head-banging, you did scrub the switch wafer front and back with flux cleaner then rinse it with copious isopropyl alcohol, didn't you?

[Alex Nikitin]

  Yes, that's a nice bit of detective work!  I'm glad to see some real numbers and some real measurements for projects like this.

   That rotary switch appears to be a copy of the American made (Switchcraft? brand) that were made in the 1960s.  Have you tested any of those for leakage?

Thank you, as I've mentioned, the switch I use (4way 2 positions) leaks about 3-5pA at 100V. I have a similar switch 3W3P and it is considerably worse due to a somewhat different design with a single phenolic ring for all moving contacts (see the photo attached), it leaks about 5nA  .

Interesting stuff. To save further head-banging, you did scrub the switch wafer front and back with flux cleaner then rinse it with copious isopropyl alcohol, didn't you?

Yes, however this didn't change the measurements for the switch 

Cheers

Alex

[TimFox]

A known problem with Nylon insulation is that it absorbs a measurable amount of water from ambient moisture, up to maybe 2%.
https://www.plastics.toray/de/technical/amilan/tec_003.html
Polycarbonate and acrylic are much better.

[Alex Nikitin]

A known problem with Nylon insulation is that it absorbs a measurable amount of water from ambient moisture, up to maybe 2%.
https://www.plastics.toray/de/technical/amilan/tec_003.html
Polycarbonate and acrylic are much better.

Yes, it is a known problem, fortunately the solid PS (polystyrene) has one of the lowest water absorption coefficients, so the spacers help in that respect as well.

Cheers

Alex

[EC8010]

I have a similar switch 3W3P and it is considerably worse due to a somewhat different design with a single phenolic ring for all moving contacts (see the photo attached), it leaks about 5nA  .

What were they thinking of? There appear to be far more different designs of ceramic wafer switches than I ever imagined. I look out for them at radio rallies and now have five incompatible wafer designs and mechanisms. Mostly, they vary centre distance between mounting pillars, but even when they have 1/4" shafts (not guaranteed), the thickness of the flat varies, making it necessary to pop a mechanism into the mill to make a shaft fit a wafer.

I've successfully used polished clear Perspex (methyl methacrylate) for low leakage. Be warned that (like all plastics) when you drill a hole in Perspex, the resulting hole is smaller than the drill. If you get "eee eee" noises when tapping Perspex, it's because you needed to use a larger drill than recommended tapping size.

[mendip_discovery]

Handy info on insulative materials,
https://web.archive.org/web/20190422012506/http://www.kayelaby.npl.co.uk/general_physics/2_6/2_6_3.html

Thanks @Alex Nikitin

I have not gone too deep into the measurements of large resistances other than to play with some insulation testers. However, I would be interested to know how much the leakage you are getting affects measurements at 1G, 100M, 10M etc.

[TheDefpom]

Perspex, a trade name for Acrylic, proper name PolyMethylMethAcrylate (PMMA) is also hygroscopic, meaning that it does absorb moisture, which will certainly increase leakage.

[Conrad Hoffman]

FWIW, I had some leakage problems with the banana jacks on an HP null meter. It turned out that the aluminum panel had corroded a bit and somehow managed to creep or dust it's way to the center conductor. Very slight leakage, but enough to notice and measure. I'm now much more careful about cleaning panels, connectors and hardware before assembly and preventing corrosion. I suspect some clear conformal coating might not be a bad idea, maybe Krylon clear spray.

[edpalmer42]

I've seen recommendations for a different style of binding post for use with high voltages (picture attached).  They're constructed so that there's no chance of internal arcing to the chassis.  Would this style have lower leakage than the two-piece style that Alex used?

Ed

[Alex Nikitin]

I've seen recommendations for a different style of binding post for use with high voltages (picture attached).  They're constructed so that there's no chance of internal arcing to the chassis.  Would this style have lower leakage than the two-piece style that Alex used?

Ed

The leakage depends mostly on material, cleanness and temperature/humidity, as these posts on the picture most likely use Nylon for isolator, the leakage might be substantial. 4mm binding posts should not be used above 1000V anyway. From purely mechanical construction point the traditional design is much sturdier in my view.

Cheers

Alex

P.S. - I run the switch overnight with copper wire shorts on both inputs and the output connected to the HIOKI DM7275. The offset has drifted from about -50nV to about -150nV overnight with about 1.5C change in temperature, and does not visibly change when I switch between two (shorted) inputs (so the change is roughly below 50nV). On the graphs below the vertical lines indicate the point of switching between inputs, there is a small (0.1-0.3mV) spike due to the contact interruption. The first graph is zoomed in for 1h so the switching is better visible. Vertical scale is 100nV/division, 1uV full span.

[EC8010]

Further to the issue of switch leakage, I used a Keithley 6517B to measure the leakage from wiper to chassis of four different types of ceramic wafer switch with 100V applied inside an earthed conductive enclosure:

CRL (USA): 7pA had ceramic spacers too
S402 (USA): 30pA clean
AB "H" Type (UK): 10pA thicker ceramic than the CRL
OAK (UK): 16pA but very dirty

It seems that we can expect a leakage of around 10pA at 100V from a clean ceramic wafer, corresponding to 10 teraohm. The S402 appears to be a different (and inferior) type of ceramic. The CRL switch was salvaged and had been cleaned very carefully whereas the others were as found at radio rallies.

[Alex Nikitin]

Yes, that seems to be the limit. If I need to do a fA level switching, I have some really nice reed relays for it with only 2-3fA leakage on the open contact at 100V, and these can be operated with only a permanent magnet moved in and out. For switching of a voltmeter input with a <20pA input current, such as my Hioki DM7275 or HP3456A, these ceramic wafers are barely OK.

Cheers

Alex