Current Leakage of connectors

[sean87]

Hi guys,

Can somone kindly explain how to eliminate or limit the current leakage resulted from connectors?

For my project, I need an edge connector, something like : http://uk.farnell.com/fci/10046971-003lf/connector-card-edge-vert-2x35/dp/1568682

I read somewhere that if insulation resistance of a connector is for example 1G ohm, if you apply 1 volt you lose 1nA in leakage. How this works and how to prevent that? Is it only possible to limit it by using a higher resistance connector (which theoritically it will result in less currents being leaked). I also have heard of Guarding...but no idea what it is it about.

Please let me know if my question is not clear.

Thanks.

[Simon]

what is your application ? you are asking to defy ohms law and the material properties. your first step is possibly a larger connector with more plastic between the terminals. Also check that the spec is strictly 1GR or maybe the spec is > 1GR in which case things are less clear and possibly better than you think

[sean87]

My Application is to make a test setup for measuring capacitor leakage current. My idea is to design a PCB that will host up 10 capacitors on each side.  and this PCB can be inserted in this edge card connectors like :


Then I am going o connect these cards to a low-current switch matrix and the output to a picoammeter.  The capacitors will get charged using a voltage source through series resistance. The goal is to measure current leakage down to 1~10 pA

I already tested this setup with single capacitors, with no PCB or edge connector inside a alminum project box. I could measure around 1pA. But I need to consider speed aswell, so using a PCB and connecting multiple capacitors with switch to the picoammeter will be quite fast, but the offset current due to connectors and stuff like that are preventing me from success!

[Conrad Hoffman]

The picoampmeter should have close to zero ohms input impedance, thus the signal line will be near zero volts. If you simply shield it with a zero potential shield, that should take care of things. What you need is guarding, nothing more than a trace on each side of the signal line, connected to ground, to insure that no voltage sources can communicate with the signal line. The problem is that you'll only have half or less the number of contacts on your connector because every other contact needs to be a ground connection. Basically, if you can draw a line on any surface, from the signal line, to any voltage source, you need a zero volt guard separating the two. Guarding can also be used with non-zero signals by buffering the source and then driving the guard trace or shield such that there's always zero voltage difference to the signal. Ohms law says that with zero voltage, the current through a resistor should be zero.

[alm]

Even though burden voltage might be low, impedances tend to be fairly high because of the low currents involved. Something like 100 kOhm on the lower (eg. 2 nA) ranges.

A driven guard might be necessary for lowest leakage. Typically low leakage circuits would use different connectors (eg. BNC or even Triax) and would use specially cleaned PCBs to reduce leakage. I'm not sure what kind of leakage you can expect from a basic card-edge connector and standard PCB handling procedures. Guarding won't help against leakage through the connector insulation.

The Keithley low level measurements handbook (available as a free download on their website) might be a good read.

[Simon]

I think for such a precision the op is expecting a bit too much convenience. physical design is probably key here

[ejeffrey]

Even though burden voltage might be low, impedances tend to be fairly high because of the low currents involved. Something like 100 kOhm on the lower (eg. 2 nA) ranges.

100 kOhm impedance is nothing.  If your insulation resistance is only 10 MOhm, the current will divide 100:1 between the picoammeter and the insulation.  Unless you are operating at high temperature / high humidity or have poor design you should easily be able to get 100M - 1 Gohm, which will reduce your error due to leakage below 0.1%.  Driven guards are usually only needed for high impedance voltage measurements, or to reduce capacitive leakage in AC measurements.

[sean87]

Thanks all for the inputs.

I finished building up the test setup.

I measured the offset voltage, it is way more than pico range, somewhere around 5nA.

I also readed some stuff on guarding topics. But I have no idea how to implement guarding in this case? Can you please explain a bit more...my brain stopped working after hard work on the shield and pcb design for hosting DUT capacitors.

Here are some pictures. Please let me know if you need more info.

[Zad]

Can you not just calibrate the leakage current out?

[Conrad Hoffman]

IMO, there's no way to implement guarding on that. Guarding is a physical issue and has to be designed in from the start. Wash everything in whatever's necessary to remove flux, then wash in deionized water and bake dry. Live with whatever leakage is left.

[sean87]

IMO, there's no way to implement guarding on that. Guarding is a physical issue and has to be designed in from the start. Wash everything in whatever's necessary to remove flux, then wash in deionized water and bake dry. Live with whatever leakage is left.

Thanks Conrad, Although a bit dissapointing to hear it is not possible to implement guarding on my setup. Would you kindly give a tip, if I am going to build the setup again how can I put the guard in it?

[sean87]

Can you not just calibrate the leakage current out?

The picoammeter has a REL option, but the problem is the offset I get is not always the same. One measurement rule is repeatability...so I think I have to minimize this offset from test setup to minimum value possible. I guess my only way is to implement guarding but I wish I knew how...

[Conrad Hoffman]

Guarding can be a big PITA and from the replies here you may not need it. If you rebuild/redesign it's just a matter of following every surface from signal lines, to any source of contaminating voltage, and placing a trace or shield such that the contaminating voltage can't reach the signal line over that surface. The first issue might be dealing with the connector- every other pin will have to be a ground pin. In fact, one might better select another type of connector scheme entirely, maybe some Samtec or pin/header solution where every pin is separate and the PCBs have little guard rings around every signal line. You can't have leakage around the connector body if there is no connector body! IMO, you'd be better off determining how bad the problem is, and what you really need, as a full guarded system will probably be larger and more expensive. Also consider time. As the equipment gets older and accumulates a film of schmutz, the leakage will get worse. It will also get worse under high humidity conditions. That's where a properly designed guard system really shines.

[sean87]

Thanks agai conrad.

Take the following circuit (just idea)



If we consider switches are low-leakage switches with <20 fA rated leakage, What would be the most important part in the above circuit to consider the most danger to effect the measurement?

What I thinkk myself, is the charging circuit, that is from DC Source to S6, then to R1-R5 and then to C1-C5, I should not care about the leakage, after all I am going to measure the leakage when switch S6 is closed.

Then the only current path would be from the capacitor which it's respective switch is closed to connect to Ammeter (In my case a picoammeter) And no need to say that only one of the S1-S5 switches can be close on any time.

So do you think I am right? Should I put the care only in the circuit after the capacitors? or charging circuit (Source, switch and serier resistors) also should be considered (well the switch S6 is closed after capacitors are charged). But I dont know yet that siwtch S6 should be closed during the measurement or NOT!!!

[Conrad Hoffman]

S6 and one other would have to be closed to complete the circuit for the ammeter and make a measurement. No circuit loop=no measurement. I'm not sure you've grasped the idea that guarding relates more to physical construction than circuit function. All one can say from the schematic is don't allow leakage between any voltage source, and the input of the ammeter. Exactly how to do that depends on the physical surfaces that you have to guard.

The other issue is that this is a rather brutal circuit to connect to a picoammeter. No doubt the meter is protected, but I'd want some switching to insure that the meter never sees charging current. Even though it's limited by the 1k resistors, that's still a lot to hammer the meter with. It's just good practice, even if the meter is robust. Maybe a switch to short the meter input to ground until everything stabilizes.

[sean87]

Yeah you are 100% right about the switches and protecting the ammeter from current surge when circuit starts. I'v tried to put your notes into following circuit:



The idea with this one is when circuit starts, all the even switches are closed (s2,s4,..) and odd switches are open. Odd switches are connected to picoammeter and ammeter ground is connected ot source's ground. Then for measuring each capacitor's leakage, its direct to ground switch will be open and its switch to picoammeter will be closed. After few mili-sec I'll take a reading (This wait time I think should be taken into account because of switch's bounce and stuff like that).

Do you think this is a better idea?

Thanks again!

[Conrad Hoffman]

That seems like a much better idea. Now all the caps can be fully charged and you can test any cap under any conditions you want. You could also implement the same thing using SPDT (is that right?) switches if desired, figuring that a cap should be connected to either ground or the ammeter. I guess it depends on exactly what test you want to run. What you've shown is the most flexible. Generally one does a soak for some specified time, then the leakage measurement. With a volt meter you can also easily turn this into a dielectric absorption tester.