Application Note 5

Characterizing AutoZero Repeatability with the AMETRIX Model 100 Series Picoammeter

Glenn Fasnacht, President, AMETRIX Instruments

gfasnacht@ametrixinst.com

​

How well does the Model 100 Series Picoammeter AutoZero itself? This is a study of measurements taken on multiple units to assess how well AutoZero actually performs.

​

I. Introduction

There are two important reasons for a good zero on a transimpedance based ammeter. First, a transimpedance amplifier based ammeter is touted as having zero voltage drop, unlike the IR based ammeter used in a standard DMM, that can have typically ±0.1 V drop at ± full scale. The other reason is that the transimpedance based ammeter should indicate zero current flow when there is indeed zero input current.​

​

The AutoZero function’s job is to approach these two ideals.

​

Figure 1 is a very simplified version of the current measurement path.

​

The AutoZero DAC adjusts the transimpedance amplifier’s (U1) offset until its output, buffered by U2 measures the same as circuit common. What remains as an input offset is whatever the offset voltage is of U2, which is a very low offset voltage amplifier.

​

​Ideally U2 would have zero offset voltage the DAC would adjust things perfectly and there would be zero input voltage at the Input BNC, and the ADC would measure zero volts, but this is the real world and real world opamps don’t have zero offset voltage, and ADCs and opamps have internal noise sources that corrupt otherwise perfect measurements.

​

Another pair of issues are time and temperature; both of these external forces cause changes in electronic components.

​

A unit calibrated at 23 °C will not remain the same at other temperatures. The change is small and may be acceptable, but for the most accurate measurements the instrument should be re-AutoZeroed if the temperature changes “too much”. The quoted too much is application dependent and up to the user to determine.

​

As stated earlier, components age too. As time passes the stresses relive, stresses induced by component manufacturing, and stresses induced by installing those components onto the printed circuit board, the instrument’s offset may change. Again, AutoZero to the rescue.

​

Since these error sources cannot be totally removed, it is important to characterize them so that the datasheet is accurate and the user has a clear understanding of the level of performance they can expect from their instrument.

​

II. The Tests

Three things will be measured after performing the AutoZero: 

1. The offset current on the 20 mA range 

2. The offset current on the 20 nA range 

3. The input offset voltage at the Input BNC

​

The procedure used is as follows: 

1. A randomly chosen Model 101 was the unit under test (UUT) and allowed to warm up for 30 minutes 

2. The AMETRIX Soft Front Panel (SFP) was used to set up the UUT, perform the AutoZero, and to acquire all current measurements. The SFP was configured for 1.9 samples per second with a 20-count running-average filter. 

3. A 6. digit DMM was used to measure the actual Input BNC offsets. The DMM was set to Slow mode and a 10-Measurement Repeat averaging filter was applied.

​

Three types of tests were run: 

1. One to determine the repeatability of the AutoZero within a single instrument 

2. One to determine the variability of AutoZero among a group of instruments 

3. One to illustrate the short-term stability of the transimpedance amplifier

​

A. Within Instrument AutoZero Repeatability 

The first test was repeated AutoZeros on a single Model 101 where: 

1. The UUT input was left open 

2. AutoZero was initiated 

3. 20 mA range set and after the filter settled the current was measured and recorded 

4. The DMM was connected to the input and the burden voltage measured and recorded 

5. The DMM was disconnected 

6. The 20 nA range was set and after the filter and measurements settled the current was measured and recorded 

7. These steps were repeated 20 times to obtain a 95% confidence level on the result 

​

Note that the number samples confidence level was determined by figure 2

​

Where n is the number of samples and p is the desired probability of success (0.95).

​

The repeatability for this randomly selected sample is considered typical.

​

B. Instrument-to-Instrument AutoZero Repeatability 

The next test involved performing steps 1 through 6 on multiple instruments.

​

C. Short-Term Stability 

Simply acquire data for about ten minutes and plot the results.

​

III. The Results

A. Within Instrument AutoZero Repeatability

Here in figure 3 is a plot of the results.

​

For the blue trace, the 20 mA range, the scale is 150 nA to +200 nA. For the brown trace the scale is 150 fA to +200 fA.

​

As expected, the 20 nA range had slightly more variability than the 20 mA. Also the Input BNC voltage varied slightly.

​

The worst case offset voltage is -10 μV, whereas the Input Voltage Burden can be as high as ±100 μV. 

​

The allowable offset error on the 20 mA range is ±1 μA and the worst case here is -0.07 μA. 

​

The allowable offset error on the 20 nA range is ±1 pA, the worst case error here is 0.14 pA. 

​

Compared to the instrument specifications these errors are insignificant.

​

B. Instrument-to-Instrument AutoZero Repeatability 

Figure 4 shows how a group of instruments respond to AutoZero.

​

The offset for the 20 mA range is almost exactly on the zero line. 

The Input BNC offset voltages are all within 11 μV of zero.

​

As expected the much more sensitive 20 nA range has more variability, but it is centered nicely around zero.

​

C. Short-Term Stability 

OK, I quit after nine minutes ... but figure 5 illustrates the results.

​

No visible slope, only fairly Gaussian noise of 24 fA r.m.s. and a mean of 0.9 fA.