2010-11-27
Picoammeter
I've been doing more ion chamber fiddling of late, in particular simple capacitive charge integration experiments. Such work really teaches you how everyday items we typically consider insulators are actually quite conductive. Along these lines I had a need to measure very tiny currents and/or very large resistances. This picoammeter was thrown together quite quickly to fill that need.
Building such "electrometer" circuits was once quite a challenge, but we live in the age of integrated FET op-amps with sub-femtoampere bias currents. I have a small quantity of LMC662 op-amps, while dated they are still extremely good - bias current wise. The LMC662 has an typical input bias current of 2 fA which means we can use it quite effectively at pA levels basically ignoring its bias current.
This circuit is the simplest ammeter that could possibly work. Recall that op-amp outputs swing such as to minimise the voltage difference between their differential inputs. By using a very large resistance in the feedback look we can make an ammeter that can measure exceedingly feeble currents with essentially no burden voltage. The input looks like a short-circuit at frequencies within the loop bandwidth - basically an ideal ammeter.
There are several implementation details with such a circuit, mainly the very reason this device was constructed in the first place - almost everything is slightly conductive. Normal PCB material, solder rosin, finger prints and many connector insulations are quite conductive at even the nanoampere level. The capacitor directly in the feedback loop must be very high quality, ideally air, teflon or glass dielectric. I used a polystyrene capacitor which I tested using the op-amp for leakage. It was difficult to tell if it was leaking more than the reed relay I used to reset the test circuit, so I considered it good enough for the application. The connection to the op-amp input pin is made in mid-air, no PCB is used and everything was cleaned with isopropanol and dried with the heat gun to minimise leakage. The cheap BNC connector is probably the leakiest thing in the circuit. During testing I shed an eyelash onto the project and its leakage railed the op-amp - keep everything clean! The only other special component is the 1 Giga Ohm resistor. I ordered several some time ago at moderate expense for just this kind of project. It is probably the most difficult part to source. Shielding is absolutely mandatory for a circuit like this, the input impedance is enormous; electrostatic effects and mains/RFI will swamp your signal unless you use shielding.
As the op-amp can swing about +/- 4 volts the circuit can measure +/- 4000 pA, but the digital LCD module I chose has a bare FSD of 200 mV (limiting us to +/- 200 pA). I did not shunt it to match even 2000 pA direct-read, preferring to keep the 100 fA resolution. I also did not provide a nulling mechanism. The offset of the op-amp is very good but not perfect, I simply subtract (or ignore) the 300-400 "fA" offset. You may wish to implement nulling on yours especially if you add an expanded range amplifier to give 10s of fA or better resolution.
The op-amp input is completely unprotected. I was being lazy, and will no doubt pay for it at some point. You could add a resistor to help protect it, either inside the loop or outside and recompute the V/A. When I blow mine the first time I'll probably add one, so far it has proven quite robust. At the same time I'd probably add a slope-adjustment to allow calibrating out the resistor tolerance for better precision, the resistor is only 5% IIRC. The device is a crude instrument, but it gives me a reasonable way to approximate very large resistances and extremely feeble currents. For more versatility you might want to add switched resistances into the feedback loop to give you other range options. If you put the switching on the low-Z output side as a divider it need not be extremely high quality.
Note that I set mine up to read +ve values for electrons into the connector. This is unconventional. As you can see in the image above a +ve bias on an ion chamber gives a -ve reading on the early prototype. You can swap the differential inputs at the LCD module if you want it to read conventional current. Note also that some LCD modules do not have fully differential inputs with respect to their supply and can not be used without a DC level shifter.
5 comments.