So, what you want to design is a Go / No Go - tester. In this case you only need to be precise at the threshold of 20nA leakage current, which simplifies the situation.

I would experiment with the following idea:




You see a static circuit, without any chopping scheme. You clamp on the diode of interest (DUT), press the button to apply the reverse voltage and within less a second you have a reading of 1.00V at 20nA leakage current. A reading of < 1.00V means "Go" and >= 1.00V means "No Go".

When not using a chopping scheme, parasitic bias currrents, leakage currents and offset voltages of all involved parts must be negligible. This means that you must be careful when choosing the parts. In order to suppress ESD I would only use a "passive" circuitry, which avoids transzorbs, SMD-varistors and Schottky diodes, but only uses caps to diminish the ESD pulse voltage. Ceramic X7R caps show very low leakage currents, but transzorbs, SMD-varistors and Schottky diodes suffer from leakage currents which can be in the range of what you want to measure and beyond.

There are ultra low leakage diodes, like FD333, FDH333, etc. though. Couldn't those be used for ESD protection?
No, they are way too slow, their turn-on time is in the µsec range and don't offer any ESD portection at all. Just the opposite, they do also suffer from ESD and can be damaged by it.

The idea of using ceramic caps to suppress ESD is based on the assumption, that a body capacitance of about 100pF is discharging during the ESD event, according to the "human body model". So, a voltage of 15kV sitting on the 100pF body capacitance is transformed to a voltage of about 70V, when being discharged into a 22nF cap.

That's the theory. But don't take the theory too literal. The caps can suppress ESD, but they do also suffer from this torture sometimes and can quit this with an increase in leakage current... So, don't rely too much on this ESD protection. It's much better to take care, that during the testing the operator will follow the well known ESD handling procedures. Then, nevertheless, the caps can be very helpful.

Ok, since we use ceramic caps to suppress ESD, and these caps need to have a minimum capacitance, the shunt resistance, across which the leakage current causes its voltage drop, must not be too high, because it forms a time constant in combination with the ESD protection caps. When choosing a too high shunt resistance, you must wait many many seconds until the reading has settled to its final value.

Another reason why the shunt resistance should not be choosen too high is, that the whole circuit becomes more and more susceptible to hum and noise then.

For these reasons the shunt resistors are 1MOhm, 1MOhm and 1.5MOhm here. The first 1M resistor (from the left to the right) in combination with the two 22nF caps forms a low pass filter (pi-filter) to suppress ESD and to filter out hum and noise. The corner frequency is about 10Hz, so that it even helps to suppress 50Hz num and its harmonics.

The 2N3904 acts as a low cost ultra low leakage protection diode to limit the input voltage of TLC277, when the leakage current is very high or during shorts betweens the input clamps. The base collector junction of 2N3904 shows a leakage current of only about 200pA! This "diode" limits the potential at this point to Vcc + 0.7V. The following 1M and 1M5 resistor form a voltage divider and limit the input voltage of TLC277 to less than 70% of Vcc, to be surely within the allowed common mode input voltage range.

The TLC277 finally has a gain of about 33 and provides an output signal of 1.00V at 20nA leakage current.

The TLC277 is a low cost CMOS operational amplifier with remarkable specifications:

Maximum input offset voltage: 500µV
Maximum input bias current: 60pA
Temperature coefficient of input offset voltage: 1.8µV/°C
Input offset voltage drift: 0.1µV/month

Another advantage of TLC277 is, that the circuit can be powered by a simple +5V supply (except for the test voltage of DUT, of course), so that it can directly be connected to your microcontroller circuit (ADC, etc.).


Error analysis:

For properly distinguishing between 19nA and 20nA leakage current, the sum of all parasitic errors of circuit must be less than 1nA equivalent leakage current:

Let's assume, that we have a non-selected TLC277 showing maximum input offset voltage of 500µA and a selected 2N3904 (for <=200pA leakage current). What is the error then?

500µV input offset voltage means an equivalent leakage current of 500µV / 1.5MOhm = 333pA.
Adding 200pA leakage current of 2N3904 and 60pA input bias current of TLC277 makes about 600pA. Or by other words, a DUT having a leakage current of 19nA, can appear to have a leakage current of 19.6nA with this circuit.
To achieve highest precision, the choose of precise resistors is needed, of course. So, use +-0.1% tolerance types for 1M5, 6k8 and 220k resistors. The additional error due to these manufacturing tolerances is less than 0.1nA.
These errors should be acceptable and can be furtherly diminished (see below).


How to select the parts?

Connect a low ohmic voltage divider to the negative input "-" to simulate a leakage current of 20nA. A 150k and 2k2 resistor connected to +5V will do the trick. At the output of TLC277 1.033V should be emitted now. The exact value doesn't matter. The 2N3904 and any DUT should be disconnected from the circuit at this moment.
Now, connect the 2N3904 and observe the output voltage. The output voltage must not change more than 3mV. If more, take another 2N3904.

With this setup you can also select a proper TLC277 for lowest input offset voltage. For this, have the voltage divider still connected to the negative input and leave the 2N3904 unconnected. Now take a "handful" of TLC277 and insert one after the other into the circuit and record the measured output voltages. After having inserted all the TLC277, take the average of all recorded readings and insert that TLC277, which comes nearest to this average.

Some last hints:

1. Leakage current measurements are difficult! Moisture and dirt can totally ruin your measurement! So, do not touch anything by hands. Keep it as clean as possible.

2. Leakage current of pn-junctions heavily depends of junction temperature. So, don't touch by hands neither the DUT nor any parts of the circuit coming in thermal contact to the DUT.
Take note, even hot lamps can radiate enough heat to heavily increase the temperature of DUT and involved parts.

3. pn-junctions are suscpetible to ESD! Handle them with care, if lowest leakage current is of interest!


Kai