0.05µA-100µA current conducting and relay switching - is it achievable?

[Visher]

Hello everyone, I would like to ask a question about sub microcurrent conducting and switching. I am developing pretty sophisticated tester of a medical device which, by my decision, should be sensitive enough to detect sub-microamp currents.

There are two main problems in such a design that I have recognized - I need to carry those currents on a relatively long path back and forth to a switching and measuring frontend (let's say about half a meter total, at least) and I need to switch them in various configurations.

To solve first problem I decided to use proper coaxial cables with SMA connectors and use shield as guarding. I was thinking about triaxial, but someone adviced me that would be an overkill (that one would aim nano/pico-amp currents). The shield would be at signal's potential, tied at one end to signal with full short and at the other with a resistor (to signal, again). What do you think about that?



The second problem is relay switching. I have already found out a solution to route everything on a PCB with minimum distance, using relays in that kind of form factor (long distance from coil to contacts, contacts along with body, tall slim case):

https://www.tme.eu/en/details/v23092a1024a302/

However I have some doubts about contacts, especially that one says "Minimum contact load" is 10mA at 5V.. I'm sure it doesn't mean that it will not conduct currents below 10mA. Is it about contact current cleaning or something? Should I worry on long term? Also, the contact material seems to be an issue, I found out that probably the best achievable solution (mercury wetted reed relays are too expensive..) are those with pure silver contacts coated with thin gold layer - if no large currents is applied and there is no arc, the gold layer should last for longer time. Do you have some experience with microcurrent switching? I made some tests with miniature relays and microcurrents, and it was fine, but maybe I'm missing something..

Thank you

[Marco]

That's not a shielded cable, that's just a thicker cable. Interference couples to the "shield", doesn't get shunted anywhere and couples to the core.

Can you really not put some electronics in the head by the way? This seems a bit archaic.

PS. as far as I can see you can switch the current with simple 2n7002s. Atto Ampere gate leakage and drain source resistance only has to be large relative to the burden resistance, doesn't need to be peta Ohms.

[Visher]

That's not a shielded cable, that's just a thicker cable. Interference couples to the "shield", doesn't get shunted anywhere and couples to the core.

I get my knowledge from this video: (and few other PDFs, decided not to use active guard, just simple one):
https://youtu.be/xWeL9c7Vfy0?t=82
and I tried to apply this in practice. The problem which I seek for is leakage current through an isolation rather than EMI. By the way, I can wrap all of these coax cables into shielded cable profile, no problem thus.

Can you really not put some electronics in the head by the way? This seems a bit archaic.

Do you mean this problem with longer distances? I have a requirement to place measuring device away from D.U.T, this test is pretty restrictive about arrangement. Somehow you may think I'm creating an "isolated earth" (with it's own ground potential, floor, and D.U.T. on it..)

[Marco]

The problem which I seek for is leakage current through an isolation rather than EMI.

Now it leaks to the shield, which is directly connected to the core ... what does the shield get you? Only the actively driven shield can work. if you want less leakage from passive means get a cable with more PTFE.

I have a requirement to place measuring device away from D.U.T

Why? It's not like a fault at the measurement side won't still feed voltage directly back to the DUT. If you distrust your ability to provide proper insulation to the body from the couple of Volt supply,  then adding a couple of wires shouldn't increase your trust by much.

If it's just about keeping things small with high dynamic range, as I said you don't really need relays. Integrating a multirange transimpedance amplifier in the head doesn't need to be big.

[Visher]

Now it leaks to the shield, which is directly connected to the core ... what does the shield get you? Only the actively driven shield can work. if you want less leakage from passive means get a cable with more PTFE.

Is it about the top resistor I have drawn on the right side instead of left side, or the whole idea is wrong?

Why? It's not like a fault at the measurement side won't still feed voltage directly back to the DUT. If you distrust your ability to provide proper insulation to the body from the couple of Volt supply,  then adding a couple of wires shouldn't increase your trust by much.

If it's just about keeping things small with high dynamic range, as I said you don't really need relays. Integrating a multirange transimpedance amplifier in the head doesn't need to be big.

About the first paragraph, I have to re-think that. The second paragraph.. Well, this might be irritating, but I have to satisfy these requirements without providing any design doubts. So I decided to carry those currents somewhere away and there do the switching and measurement..

[Marco]

Is it about the top resistor I have drawn on the right side instead of left side, or the whole idea is wrong?

The idea of a shield isn't wrong, but it needs to be driven or grounded. A grounded shield will give you predictable leakage, a driven shield will get you minimal leakage.

As it stands the resistor is irrelevant and the insulation between the shield and core is irrelevant. You effectively have just a really thick cable with only the outer insulation being functional.

PS. I'd just use twinax with a grounded shield, good enough.

[Visher]

The idea of a shield isn't wrong, but it needs to be driven or grounded. A grounded shield will give you predictable leakage, a driven shield will get you minimal leakage.

As it stands the resistor is irrelevant and the insulation between the shield and core is irrelevant. You effectively have just a really thick cable with only the outer insulation being functional.

PS. I'd just use twinax with a grounded shield, good enough.

Thank you for discussion, now I wonder for what I'll go in. If money and PCB space is not an issue, I could implement active guard and have leak-free measurements. Or compensate for them for every single measurement.. Maybe I'll go for the first one.

Talking about the switching problem, how can I reliable switch such a low voltages and currents as few µV and 1000x less µA with a mosfet transistor? Is it simple as that, when I apply voltage to gate, it creates a "resistance tunnel" of miliohms, regardless of Vds and Ids?

[tszaboo]

Just use analog switches. That is what they are made for.
Something like an ADG633 gets you 0.2nA leakage. And it is characterized.

[RobK_NL]

You might want to read Keithley's Low Level Measurements Handbook

[Visher]

Just use analog switches. That is what they are made for.
Something like an ADG633 gets you 0.2nA leakage. And it is characterized.

I wasn't aware of that kind of ICs, so thank you. However there's one think I didn't mentioned - if something fails, there might be high voltage on those switches, dozens of volts. If a relay is rated e.g. 30 VDC, for short time it will withstand hundreds of volts, current there is limited by measuring device, max. 5-10 mA.
In case of ADG633 - it might blow up And protection diodes to protect that have few µA leakage current..

You might want to read Keithley's Low Level Measurements Handbook

Definitely will read that, thanks!

[tszaboo]

Just use analog switches. That is what they are made for.
Something like an ADG633 gets you 0.2nA leakage. And it is characterized.

I wasn't aware of that kind of ICs, so thank you. However there's one think I didn't mentioned - if something fails, there might be high voltage on those switches, dozens of volts. If a relay is rated e.g. 30 VDC, for short time it will withstand hundreds of volts, current there is limited by measuring device, max. 5-10 mA.
In case of ADG633 - it might blow up And protection diodes to protect that have few µA leakage current..

You might want to read Keithley's Low Level Measurements Handbook

Definitely will read that, thanks!

You place a series resistor on the line to be protected (current limit), and a BAV199 to limit the voltage to the power supply rails. If the temperature is sane (around room temp) it is not going to leak anything.

But of course you can just use a relay.

[duak]

Reed relays are also a common solution to this problem because their contacts are sealed in a glass tube with an inert gas to prevent oxidation.

A note about minimum contact load: the relay is not guaranteed to reliably conduct less than 10 mA over its lifetime.  Some relays also give a minimum voltage spec while others give a minimum power spec.

I was once asked to figure out why a product would often indicate a spurious fault after a few months in service.  Turned out the original designer put two relay contacts in series in a safety loop.  The contacts were rated for 10 mA and 12 V minimum and were in a 12 V circuit with a 1K0 load.  Meets spec, right?  Not quite; because the contacts are in series, the voltage divides between them  arbitrarily and does not meet minimum spec.  I added another 1K0 resistor to the floating node between the two contacts to establish a minimum current for the first contact and the problem was solved.

Cheers,

[Visher]

Reed relays are also a common solution to this problem because their contacts are sealed in a glass tube with an inert gas to prevent oxidation.

A note about minimum contact load: the relay is not guaranteed to reliably conduct less than 10 mA over its lifetime.  Some relays also give a minimum voltage spec while others give a minimum power spec.

I was once asked to figure out why a product would often indicate a spurious fault after a few months in service.  Turned out the original designer put two relay contacts in series in a safety loop.  The contacts were rated for 10 mA and 12 V minimum and were in a 12 V circuit with a 1K0 load.  Meets spec, right?  Not quite; because the contacts are in series, the voltage divides between them  arbitrarily and does not meet minimum spec.  I added another 1K0 resistor to the floating node between the two contacts to establish a minimum current for the first contact and the problem was solved.

Cheers,

Thank you for this note, I'll definitely be aware of this problem. Also, I have asked my co-workers about that and I was told that low currents at low voltage do not create a good "physical-electrical" connection. It's like it leaves a place for oxidation and does not remove these unwanted products.
I'm thinking about solution for this problem to often apply meaningful current like 50mA. Let's say for every 1s of switching and measurement of µA currents at µV voltages, there would be 5s of 50mA @ few volts to "prepare" contacts, keep them in good shape.
What do you think about that?

And about those reed relays, the problem is form factor.. I also have doubts about contact's signal interference with reed's coil EMF - I know it's a DC, but what if it's not, a little bit of voltage ripple, thus current, thus EMF?

[Gyro]

The only elecromechanical relay that is capable of switching 'dry' circuits reliably is a Mercury wetted one. Obviously availability is an issue these days (though I think some are still manufactured for very specific purposes).

Applying a higher current, as you indicate, is known as a 'wetting' current (relating back to the above wetted relays) and was used in Telecom relays.

Yes, you can pick up induced emf from the relay coil - fast switching edges are best avoided. Some signal relays include an electrostatic screen between the coil and reed capsule that can be grounded. The other issue is thermal EMF (thermocouple effects) this can be caused either by heating the contacts directly (your 50mA) or thermal disipation from the relay coil itself. Bistable relays are recommended for this purpose.

[Marco]

In case of ADG633 - it might blow up And protection diodes to protect that have few µA leakage current..

"Overvoltages at Ax, EN, S, or D are clamped by internal diodes. Limit current to the maximum ratings given"

Max current is 20 mA, lets say you have a 100V input ... so lets see what happens if you put a 50K resistor in the supply lines. Idd is 1 uA max, that should work, but might need a tiny bit of capacitance on the supply so it doesn't tank too much while switching. Especially if the current is unidirectional I'd still consider the 2n7002 approach though, easier to predict exactly how it works.

As I said before, don't overestimate the isolation you need. Only the control input isolation needs to be very high Ohmic, the pass through isolation only needs to be large relative to your burden resistance.

PS. Bourns has some off the shelf solutions for active series current limiting by the way, TBU series.

[duak]

Visher, if the relay you are using has a specified minimum load current then it is not the correct type.  It probably has silver alloy contacts that oxidize or corrode.  Some relays have gold flashing over the silver alloy that allow low current switching.  However, higher currents will quickly vaporize the gold and then low current switching is compromised.  It's an interesting idea to electrically clean contacts but I think it's probably easier and better to get the appropriate switch, semiconductor or mechanical.

Gyro pointed out mercury wetted relays but reed relays with appropriate characteristics are available.  For example, Pickering Electronics in the UK makes an instrumentation grade for thermocouples: http://www.pickeringrelay.com/applications/thermocouple-switching/   Pickering also refers to mercury wetted contacts but I haven't looked into that.

I don't know enough about the application to say if the reed relay presents any other problems.

Cheers,
   

[Visher]

Visher, if the relay you are using has a specified minimum load current then it is not the correct type.  It probably has silver alloy contacts that oxidize or corrode.  Some relays have gold flashing over the silver alloy that allow low current switching.  However, higher currents will quickly vaporize the gold and then low current switching is compromised.  It's an interesting idea to electrically clean contacts but I think it's probably easier and better to get the appropriate switch, semiconductor or mechanical.

Gyro pointed out mercury wetted relays but reed relays with appropriate characteristics are available.  For example, Pickering Electronics in the UK makes an instrumentation grade for thermocouples: http://www.pickeringrelay.com/applications/thermocouple-switching/   Pickering also refers to mercury wetted contacts but I haven't looked into that.

I don't know enough about the application to say if the reed relay presents any other problems.

Cheers,

I do read about contact materials from Polish relay manufacturer's publication, page 15, good amount of knowledge about why different materials are used in relays:
https://www.relpol.pl/en/content/download/4172/520268/file/e_basic%20information.pdf

So I had same conclusion as you posted about relay contacts, I even posted it in thread post:

I found out that probably the best achievable solution (mercury wetted reed relays are too expensive..) are those with pure silver contacts coated with thin gold layer - if no large currents is applied and there is no arc, the gold layer should last for longer time.

Anyway, thank you for confirming that. Unfortunately, I'm still thinking about slim relay form factor and I couldn't find gold plated relays with silver contacts. Best one so far I found is that one:
https://www.tme.eu/en/details/apf10212/miniature-electromagnetic-relays/panasonic/

but it should be in APF103 version, which has gold plated AgNi contacts. Unfortunately, not available right now, Panasonic is out of stock. Delivery time - 24 weeks..

Why such a form factor? I need like 40 of there relays. Their purpose would be to switch small current sources in different configuration. With standard relay form factor, that would result in pretty big PCB area. But with there relays like Panasonic, I can do something like that:



There configurations are: current from A to B, from B to C, from C to A and so on.. and it doesn't end at E

Coil circuit is separated from switchted signal, there is some space between contacts to make signal guarding with additional traces and vias. I see a lot of advantages, but the problem is - these relays are unavailable in gold plated contacts version.

At this point I have to stop writing this post, thank you for discussion, I don't want to take your time anymore because it's my work and I shouldn't post in on internet forum more than I'm doing it right now. Also, it's not kind to make use of you to do the job for me. I got the knowledge from you, the Keithley's PDF, and now I have to work on my own to pack it into something working. For sure I'll notify you how I did and how it's performing.

Thank you again everyone!

[Marco]

You can use an optical relay too, say a TLP222. The 50 odd Ohm on resistance shouldn't be a problem, neither should the >350 MOhm off resistance.

[BrianHG]

Just use analog switches. That is what they are made for.
Something like an ADG633 gets you 0.2nA leakage. And it is characterized.

Have you considered something like MUX36S08, MUX36D04, low off leakage + wide 36v analog range.

[floobydust]

I don't see any mention of the voltages being switched.
Contact potential is usually +/-10uV to +/-65uV.

I've used Keithley 3700-family reed relay matrices, for test fixtures, with 0.5nA-2nA offset and contact potential +/-10uV to +/-65uV.

[exmadscientist]

Panasonic PF series is not a signal relay -- it's got "power" right there in the name in the datasheet. Yes, the gold flash will be better than the non-gold-flash, but it's still not a great choice. Panasonic is also not the only relay manufacturer around: Fujitsu and Omron have good offerings too.

For somewhat less demanding tasks (switching 1mA RMS, also in a test fixture) I have had good success with Panasonic GN series and Fujitsu FTR-B3 series. A few of the Omron series are also very good but last time I looked they were more expensive than the others to no particular benefit for that application.

[Visher]

Panasonic PF series is not a signal relay -- it's got "power" right there in the name in the datasheet. Yes, the gold flash will be better than the non-gold-flash, but it's still not a great choice. Panasonic is also not the only relay manufacturer around: Fujitsu and Omron have good offerings too.

For somewhat less demanding tasks (switching 1mA RMS, also in a test fixture) I have had good success with Panasonic GN series and Fujitsu FTR-B3 series. A few of the Omron series are also very good but last time I looked they were more expensive than the others to no particular benefit for that application.

I have found two more relays in this form factor with gold plating and much lower minimum contact current:
https://pl.mouser.com/productdetail/omron-electronics/g6dn-1a-l-dc12?qs=sGAEpiMZZMtSzCF3XBhmW4H60f5%2F4KqZEpGeTSlrKPE%3D
https://pl.mouser.com/productdetail/te-connectivity-oeg/pcn-112d3mhz000?qs=sGAEpiMZZMtSzCF3XBhmWxSICMMNssF%252bwJuNy3aPmWc%3D

Available on farnell and mouser. TE Connectivity's one seems to be better in this scenario, it has minimum recommended contact current of 1mA. I would periodically apply higher current to clean these contacts and then for a short time do the switching of µA.
Also I'll investigate your proposals later.

I don't see any mention of the voltages being switched.
Contact potential is usually +/-10uV to +/-65uV.

I've used Keithley 3700-family reed relay matrices, for test fixtures, with 0.5nA-2nA offset and contact potential +/-10uV to +/-65uV.

Current shunt is 1 kOhm, so 1µA -> 1mV