Beginner level: DIY an accurate picoammeter (measure picoamps!)

[SilverSolder]

Sounds like a plan!

[NeverDie]

For doing detailed millivolt measurements on an oscilloscope (and maybe a DMM as well), I'm now convinced that the type of cabling used is absolutely critical.  I've reached that conclusion from just now trying a number of different cables and noticing the differences.
 WRT to common mode noise, the difference is just night and day when using BNC cables vs using regular banana-plug type cables.  So much so that I'm surprised digital multi meters don't have BNC jacks on them as standard equipment.  In fact, I just now did a search to see if any of them do, and I couldn't find a single handheld portable DMM's with built-in BNC connectors. At best all I could find were BNC to banana plug adapters.  Only some of the high end lab bench multimeters seem to have BNC connectors.  The closest I could find in hand-held were combo multimeter-oscilloscopes.  I guess for DMM's common mode noise may simply get averaged out?

If you all already know what I mean, then great.  If not, and you'd like me to put together some show and tell pictures for compare and contrast, let me know and I'll post some to illustrate just what I mean.

[SilverSolder]

For doing detailed millivolt measurements on an oscilloscope (and maybe a DMM as well), I'm now convinced that the type of cabling used is absolutely critical.  I've reached that conclusion from just now trying a number of different cables and noticing the differences.
 WRT to common mode noise, the difference is just night and day when using BNC cables vs using regular banana-plug type cables.  So much so that I'm surprised digital multi meters don't have BNC jacks on them as standard equipment.  In fact, I just now did a search to see if any of them do, and I couldn't find a single handheld portable DMM's with built-in BNC connectors. At best all I could find were BNC to banana plug adapters.  Only some of the high end lab bench multimeters seem to have BNC connectors.  The closest I could find in hand-held were combo multimeter-oscilloscopes.  I guess for DMM's common mode noise may simply get averaged out?

If you all already know what I mean, then great.  If not, and you'd like me to put together some show and tell pictures for compare and contrast, let me know and I'll post some to illustrate just what I mean.

DMMs should be more or less immune to common mode noise.   One of the big issues when making sensitive measurements is to avoid ground loops at all costs.

It would be a great idea to draw a schematic how the measurements are made, including the test instruments and power supplies in the diagram, and paying attention to which of them are connected to Earth.

[NeverDie]

DMMs should be more or less immune to common mode noise.   

At least under present circumstances, what I'm seeing doesn't seem to be in complete agreement with that.  My 500,000 count DMM does tend to settle if given enough time, but any movement on my part throws it off again. 

Let's compare/contrast measurements using BNC cabling versus banana plug cabling:

If I turn off averaging and BW limiting, so as to give the noisest picture on my oscope, here is what my oscope shows both with nothing plugged into it and with a quality 6 inch BNC cable plugged into it (open circuit, so nothing on the other end):


Now, here it is with the same settings, but with DMM probes plugged in instead:


Increasing the voltage/reticule shows that it's more than 100 times worse:

and from the wavelength it becomes clear that it's mains common mode noise that's a strong component of it.

I do think it maybe does make a difference if trying to get single digit picoamp resolution out of the picoammeter.  Admittedly, the testing environment can probably be improved yet further, but for now it just is what it is.  Less dramatic compare/contrast was done by the tinyCurrent guy.    Here is the tinyCurrent noise when using the BNC connector:

and here it is using leads:


Well, just to be certain, after I get the lunchbox testing enclosure set up and connected to the picoammeter, I'll try doing the picoammeter measurements using a DMM both with regular DMM cabling and also with using coax connected to the DMM via a BNC adapter.  I'm pretty sure I'll see some kind of difference, but I'll wait until then before I reach final conclusions.

The very first thing I'm planning to test with the picoammeter is this Vishay load switch: 
https://www.vishay.com/docs/66597/sip32431.pdf
which claims to have a leakage current of just 10pa.  My conjecture is that, given the environmental noise, it would be very challenging to accurately measure in the single picoamp digits with just a DMM and DMM probes without the benefit of the new test enclosure and upgraded cabling.

[NeverDie]

Anyhow, switching topics, I plan to use this as the passthru into the lunchbox test enclosure:

I like that this type of passthru can be anchored to the wall of the enclosure with four bolts.  A problem I've been having with the non-bolted screw-on type BNC connectors is that they tend to come loose after connecting to them a few times, and then the resulting twisting can put the solder connections inside in jeopardy.

Then for anything I want to test inside the test enclosure, I'll wire it up with its own personal one of these:


so that it can be easily connected to the passthru when its ready to be tested and then disconnected and removed from the enclosure when its testing is done, thereby making room for whatever comes next in the testing queue.

 

These parts won't be arriving until Wednesday at the earliest, though, so in the meantime I may wire up some less elegant temporary connections so that I can get started.

[SilverSolder]

"Common mode" refers to the same signal being present on both leads, exactly the same.   A DMM or scope is pretty good (but not perfect) in eliminating that.

What you are seeing appears to be plain old "differential mode" noise, caused by magnetic and electric fields in the room.   

If you think about it, the leads from your DMM, if you touch the probes together,  make up a big loop - essentially, like an inductor with one winding.   Any magnetic field that passes inside the area of the loop will be "captured" as a signal in the leads.

A coaxial cable avoids most of that problem by creating the smallest possible area between the two leads...  but it is not perfect, obviously.

It can be really hard to get a quiet low level signal in a typical home environment.  Just yesterday, I was trying to measure a low level signal and couldn't figure out why the spectrum analyzer noise level was so high...   until I unplugged the laptop charging on the bench, and the noise dropped significantly!

[magic]

open circuit, so nothing on the other end

That's your problem right here.
Short the other end or connect it to an opamp configured to output 0V (or anything else) and you will see that shielding makes little difference.

[NeverDie]

A coaxial cable avoids most of that problem by creating the smallest possible area between the two leads...  but it is not perfect, obviously.

I have a different model of it than that: think of the coax as joining two faraday cages.  Then the internal (non-grounded) wire effectively remains on the inside of the cage and is protected from noise that way.  At least that's how it should work when I connect the picoammeter to the lunchbox test enclosure using coax cable.  The metal shell is ground in both enclosures, which is connected to the coax sheathing.  Then both enclosures and the coax sheathing become one and the same ground, and effectively, one and the same enclosure, with a narrow tunnel (the coax cable) joining the two enclosures.  When tied together this way it's like a single dogbone shaped enclosure.

[NeverDie]

open circuit, so nothing on the other end

That's your problem right here.
Short the other end or connect it to an opamp configured to output 0V (or anything else) and you will see that shielding makes little difference.

When I short the two leads together I get:


So, the 60hz mains noise did disappear, but it appears there remains a lot of other noise.  This seems consistent with silversolder's big loop theory.

[SilverSolder]

open circuit, so nothing on the other end

That's your problem right here.
Short the other end or connect it to an opamp configured to output 0V (or anything else) and you will see that shielding makes little difference.

When I short the two leads together I get:
(Attachment Link)

So, the 60hz mains noise did disappear, but it appears there remains a lot of other noise.  This seems consistent with silversolder's big loop theory.

Nice, but I can't take credit for Ampère's law (part of Maxwell's theory of electromagnetism)!   

According to Maxwell, the amount of voltage induced in a loop by a magnetic field that goes through it is dependent on the area covered by the loop and the strength of the field.

It works the other way round too:  the bigger the loop area, the more magnetic field a given current will send into the surroundings!

This is a really neat trick to have in mind when you design circuit boards or cable layouts of any kind:  keep it tight!  Always keep the return current next to the outbound current.  That way your project becomes less prone to interference from the outside, as well as less prone to sending noise into other things nearby. 

This is also the reason twisted pairs of wire work so well:  all the little loops end up canceling out.

[NeverDie]

This is also the reason twisted pairs of wire work so well:  all the little loops end up canceling out.

 I thought the theory behind twisted pair was that it helps only if you look at the voltage difference between the pair. At least, that's how ethernet does it.  Does it really confer any advantage if not doing that?

[NeverDie]

Here's the stopgap version of my lunchbox test chamber:



and here is what should be a fully functional test setup, with the picoammeter directly connected to the test chamber (as previously described above):



Now all I need to do is put a DUT inside the lunchbox, solder its output connections to the BNC connection wires, close the lid, and take measurements!   With this setup, I expect the measurements will be completely shielded from outside noise and interference. 

[SilverSolder]

This is also the reason twisted pairs of wire work so well:  all the little loops end up canceling out.

 I thought the theory behind twisted pair was that it helps only if you look at the voltage difference between the pair. At least, that's how ethernet does it.  Does it really confer any advantage if not doing that?

Yes, it does.  As the wires twist, you can think of it as a whole bunch of little loops connected in series, but with opposite polarity.   So when a large external magnetic field passes through the loops, each one cancels out its neighbor.  It totally works,  you can test it the same way you did your probes (i.e. short the end of a twisted pair, see what you get compared to an open loop). 

It is worth using twisted pairs everywhere inside a sensitive experiment, to add that extra little bit of silence.  After all, an aluminum box doesn't stop magnetic fields...


Looking at the voltage difference between a pair of wires is a separate technique that also reduces noise, but it is always better to avoid noise in the first place!

[NeverDie]

I've decided to make an easier first test.  I'll test the S-5470, which is a current detecting switch that should turn-on at 700pa nominal (possible range is 520pa to 880pa):  https://www.ablic.com/en/doc/datasheet/photo_ic/S5470_E.pdf  I'll have it light an LED when it detects the threshold current, and then measure with the picoammeter to see what the current actually was.  That should be an easy measurement target to hit.

Then, its "typical" current consumption is around 10 to 20pa, which makes for a smaller measurement target, but it still should be easy for the picoammeter to measure.

So, we'll see if it the picoammeter's measurements are in agreement with both measurement targets.

[NeverDie]

After all, an aluminum box doesn't stop magnetic fields...

Should I maybe switch to a ferrous metal box for the picoammeter?  The lunch box already is.  The tradeoff is that aluminum is more conductive than iron, so I'm not sure which is the more important parameter to optimize.

[Gyro]

In practice it's not going to make any difference. It's more down to practicalities  - for instance, your lunch box is probably solderable, whereas Aluminium wouldn't be.

It looks as if you have good metal to metal contact on the lid too.

[SilverSolder]

I should have said that aluminum does block magnetic fields for higher frequencies - but it won't block "DC" magnetic fields, or low frequency ones (like mains).  That's where using twisted wires and keeping everything tight on your PCB board (minimize loop area, again) helps you a lot, even inside a box, and doing the two things together is normally enough to beat this problem into the ground.

[SilverSolder]

Another test that might be fun to try, is to see if the uCurrent and the picoammeter agree (put the two in series and see what they say...  for sure, the same current flows through both of them if they are in series, right?).

I would guess that at around 1,000pA,  the two should be getting close, but that is just a wild guess... 

[NeverDie]

Another test that might be fun to try, is to see if the uCurrent and the picoammeter agree (put the two in series and see what they say...  for sure, the same current flows through both of them if they are in series, right?).

I would guess that at around 1,000pA,  the two should be getting close, but that is just a wild guess...

Good idea.  If I put an additional BNC connector on the lunchbox (making a total of two BNC connectors on the lunchbox instead of just the one that's there now), then I can put the uCurrent Gold inside the lunchbox and export its measurements using the additional BNC connector.  That way the uCurrent Gold, which has an unshielded enclosure, can become better shielded.  Maybe in this way the uCurrent Gold can even deliver the "superb accuracy" that's advertised for picoamp currents and that  I quoted in the very first post of this thread.  Also, it will then be possible to directly compare the uCurrent Gold's measurements with the picoammeter measurements, since I'll have access to both readings outside of the lunchbox test chamber at the same time.

Of course, this approach assumes that the uCurrent Gold doesn't throw off noise or  interference that might screwup the picoammeter measurements.  I don't know whether it will or not, but if we try, then we'll find out.

[magic]

So, the 60hz mains noise did disappear, but it appears there remains a lot of other noise.

Well, that's the second problem: capturing a lot more bandwidth than you need That LMC cirucit probably won't output more than a few kHz at best due to parasitic capacitances in the opamp.

I'm not saying not to shield. It's harmless, it may even help. If there is something I would be worried about, it's the output cable picking up mains electrostatic field and passing it into the shielded enclosure, if the output cable isn't driven with very low impedance but has some protection resistor in series or whatever. So yeah, coax is better or no worse.

Twisted pair does absolutely nothing for electrostatic fields, by the way.

I don't think you need to be worried much about magnetic fields. All the shielding I used was a grounded juice box (and it wasn't even closed perfectly tight)

[Gyro]

Another test that might be fun to try, is to see if the uCurrent and the picoammeter agree (put the two in series and see what they say...  for sure, the same current flows through both of them if they are in series, right?).

I would guess that at around 1,000pA,  the two should be getting close, but that is just a wild guess...

Good idea.  If I put an additional BNC connector on the lunchbox (making a total of two BNC connectors on the lunchbox instead of just the one that's there now), then I can put the uCurrent Gold inside the lunchbox and export its measurements using the additional BNC connector.  That way the uCurrent Gold, which has an unshielded enclosure, can become better shielded.  Maybe in this way the uCurrent Gold can even deliver the "superb accuracy" that's advertised for picoamp currents and that  I quoted in the very first post of this thread.  Also, it will then be possible to directly compare the uCurrent Gold's measurements with the picoammeter measurements, since I'll have access to both readings outside of the lunchbox test chamber at the same time.

Of course, this approach assumes that the uCurrent Gold doesn't throw off noise or  interference that might screwup the picoammeter measurements.  I don't know whether it will or not, but if we try, then we'll find out.

Good idea (to compare the two).

Remember that the uCurrent Gold has its input negative and output negative terminals connected together, so you need to look carefully at how you connect them in series in the context of the screened enclosure.

The uCurrent does use autozero opamps, whose switching does generate some input noise, but this will be be high enough in frequency not to show up.

[SilverSolder]

So, the 60hz mains noise did disappear, but it appears there remains a lot of other noise.

Well, that's the second problem: capturing a lot more bandwidth than you need That LMC cirucit probably won't output more than a few kHz at best due to parasitic capacitances in the opamp.

I'm not saying not to shield. It's harmless, it may even help. If there is something I would be worried about, it's the output cable picking up mains electrostatic field and passing it into the shielded enclosure, if the output cable isn't driven with very low impedance but has some protection resistor in series or whatever. So yeah, coax is better or no worse.

Twisted pair does absolutely nothing for electrostatic fields, by the way.

I don't think you need to be worried much about magnetic fields. All the shielding I used was a grounded juice box (and it wasn't even closed perfectly tight)

At least it is easy to see if you are getting any magnetic interference:  Short the leads at the far end...  what you see then, is going to be magnetic in origin (assuming the CMMR of your instrument is reasonable!  --edit:  and assuming there are no ground loops!).

I constantly get surprised by how much magnetic noise there is nowadays, when we are surrounded by switch mode power supplies.

[NeverDie]

It looks as if you have good metal to metal contact on the lid too.

Yup, I confirmed by probing with a DMM that the metal lunch box lid has continuity to the rest of the metal lunch box.

[NeverDie]

If there is something I would be worried about, it's the output cable picking up mains electrostatic field and passing it into the shielded enclosure, if the output cable isn't driven with very low impedance but has some protection resistor in series or whatever.

Yes, good point.  That does seem like a possible weakness.  To mitigate I'll use coax to connect the uCurrent Gold's output to to the external multimeter.  I'll also check to see whether the picoammeter measurements become more noisy after I connect the multimeter to the uCurrent Gold's output by doing before-and-after measurements  with the picoammeter (i.e. before connecting to the uCurrent Gold's output and after connecting).

I'm very open to alternative approaches though.  If someone here can think of a better way, then I'm happy to do that instead.

[NeverDie]

I received the passthru's today, so, as discussed above, I upgraded the testing enclosure to use two of them:





And I have a termination cap to cover one of the passthru's when it's not in use: