Challenging nanoampere current measurement technology

Measuring low currents is very delicate. Clever analog design techniques, the right components, and equipment can help.
Key Points
Measuring low currents faces physical and noise limitations.
Early mechanical meters could resolve currents in the femtoampere range. JFETs and CMOS amplifiers are suitable for measurement. To measure femtoampere currents, the current needs to be integrated into a capacitor. Integrating devices can measure femtoampere currents and provide a 20-bit output. Thousands of applications require testing circuits with low currents, the most common being the measurement of the photocurrent generated by a diode when illuminated by light. Several scientific applications, such as CT scanners, gas chromatographs, photomultiplier tubes, and particle and beam monitoring, require the measurement of low currents. In addition to these direct applications, manufacturers of semiconductors, sensors, and even wires must measure extremely low currents to characterize their devices. Measurements of leakage current, insulation resistance, and other parameters require consistent, accurate measurements to establish data sheet specifications.

But few engineers understand that a device's data sheet is a contractual document. It specifies the device's performance, and any objection to the device's operation is attributed to the data sheet's specifications. Recently, a customer of a large analog IC company threatened legal action against the manufacturer, claiming that the device he purchased was operating at currents far above the sub-microampere rating specified by the company. The ultimate cause of the incident was that although the PCB (printed circuit board) assembly house had properly cleaned the boards,
the assembly personnel left fingerprints on critical nodes when handling the PCBs. By being able to measure these tiny currents, semiconductor companies can prove that their devices are operating properly and that the leakage currents are coming from dirty PCBs.

The difficulty in measuring small currents comes from the various interferences that can interfere with the measurement. This article will discuss two breadboard circuits that must deal with problems such as surface leakage, errors caused by amplifier bias currents, and even cosmic rays. As with most circuits, EMI (electromagnetic radiation) or RFI (radio frequency interference) will introduce errors, but at these low levels, even electrostatic coupling can cause problems. When the currents to be measured are small, in the femtoampere range, the circuits are susceptible to more interference. Humidity can change the value of capacitors, causing large surface leakage; vibration can create piezoelectric effects in circuits; even small temperature changes caused by room fans can create temperature gradients on the PCB, causing false readings; room light can also degrade measurement accuracy, and light from fluorescent lamps can enter the transparent end of a detection diode, causing interference (Reference 1).

If you want to determine the performance of a crystal oscillator, you need to accurately measure small currents. Jim Williams, a Linear Technology scientist and a longtime EDN contributor, demonstrated a circuit he designed for a customer who needed to measure the root mean square (rms) current of a 32kHz watch crystal (Figure 1). One of the difficulties with this measurement is that even the 1pF capacitance of a FET probe can affect the crystal's oscillation. Specifically, one of the goals of the current measurement is to size the low-value capacitors used for each crystal. A further difficulty with this measurement is that it must be measured accurately in real time at 32kHz, which rules out the use of an integrating capacitor. The signal is a complex ac signal that the system designer must convert to an rms value to evaluate.

Williams said, "The rms operating current of a quartz crystal is important for long-term stability, temperature coefficient, and reliability." He said that the demand for miniaturization introduces parasitic problems, especially capacitance, which makes accurate detection of rms crystal current more complicated, especially for micropower types of crystals. He explained that the measurement can be made using the high-gain, low-noise amplifier in Figure 2, combined with a commercial closed-core current probe, and an rms-to-dc converter to provide the rms value. The dotted line in the figure shows the test circuit for the quartz crystal, which demonstrates a typical measurement situation. Williams uses a Tektronix CT-1 current probe to monitor the crystal current, which only produces minimal parasitic loading. The coaxial cable feeds the probe's 50Ω to A1, and A1 and A2 get a closed-loop gain of 1120, an additional gain above the nominal 1000 to correct for the CT-1's 12% low-frequency gain error at 32.768 kHz.

Williams studied the effectiveness of this gain error correction for a sinusoidal frequency (32.768kHz) using seven sampling groups from a Tektronix CT-1. He reported that the device's outputs were all within 0.5% of 12% for a 1mA, 32.768kHz sine wave input current. Although these results appear to support this measurement scheme, Williams still thinks it is worth pointing out that the results are from Tektronix measurements. He said, “Tektronix does not guarantee performance below the specified -3dB, 25kHz low-frequency roll-off. A3 and A4 provide a gain of 200, so the total amplifier gain is 224,000. This number produces a 1V/mA scaling factor at A4 for the output of CT-1. The LTC1563-2 32.7kHz bandpass filtered output of A4 is fed to A5 through an LTC1968-based rms-dc converter, which provides the circuit’s output.” Williams explained that the signal processing path consists of an extremely narrow-band amplifier that is tuned to the frequency of the crystal. Figure 3 plots typical circuit waveforms. According to Williams, the crystal is driven at the output of C1 (upper trace), producing an rms crystal current of 530nA, shown as the output of A4 (middle trace) and the rms-dc converter input (lower trace). "The spike you see in the middle trace is the unfiltered component from the parasitic path in parallel with the crystal," he said.[page]

As you can see from Williams’ circuit, even with the integration technique, it is difficult to measure nanoamps. This problem is particularly difficult because the measurer must make the measurement in real time. There are further complications, such as the fact that this ac measurement requires a bandwidth of 32 kHz to capture a lot of energy in the oscilloscope current waveform. Williams solved these problems with a sensor. The Tektronix CT-1 sensor (Reference 2) costs a whopping $500, but without a good sensor, Williams would not have been able to recover the signal from all the noise. In addition to good sensitivity, the CT-1 has a 50Ω output impedance, which results in a lower noise signal path than a high-impedance output. Another important principle demonstrated in this example is the importance of limiting the bandwidth of the signal path. Williams built a narrowband amplifier chain to remove any noise from frequencies that are not of interest. Finally, Williams used good low-noise design principles in his circuit. Important nodes were connected open, leakage paths were minimized, and with a 50Ω source impedance, the LT1028 is probably the lowest noise amplifier available from any manufacturer.


Femtoampere bias current

Paul Grohe, an applications engineer at National Semiconductor, provides another great example of measuring tiny currents. Several years
ago, National Semiconductor decided to sell the LMC6001, an amplifier with a guaranteed 25 fA bias current, which meant that the company needed to measure the bias current of each device to verify the specification. The test department was unable to provide test equipment during the planning phase, and all the circuits had to be mounted on a standard probe card. Grohe and colleague Bob Pease built a proof-of-concept setup to demonstrate the feasibility of resolving small test circuits as low as 1 fA (Figure 4). Many books and discussions use an integrating capacitor to measure small currents (Reference 3). The idea is that a small current can charge a small capacitor, and you can read the voltage to infer the current. In some cases, the current is external from a sensor. In this case, the current is leaving the amplifier's input pin. Figure 5 shows a simple schematic circuit in which the amplifier is measuring its own bias current.

The reality of measuring small currents is much more than what is shown in the figure. First, Grohe could not use the device itself to measure its own bias current. If he tried to use the device itself as an integrator, he would not be able to correct for the effects of a socket and other leakage associated with the test setup. To do this, a separate low-bias current device is needed as an integrator (Figure 6). Using a CMOS LMC660 amplifier, the bias current is kept to less than 2 fA. With this technique, Grohe can simply remove any DUTs (devices under test), and the integrator can measure its own bias current, as well as the leakage current of the test socket and the PCB on which the integrator is mounted.

Figure 7 shows that Grohe did not insert the DUT into a socket and none of the pins were in contact with the PCB. To minimize leakage, Grohe made only the two power pins as long, separate sockets that were not mounted on the PCB. Likewise, he connected the pins to be tested to a socket and a 2-inch suspension wire and connected the pin/socket combination to the input of the integrating amplifier. To prevent the DUT from running in an open loop, Grohe soldered the two sockets together and bridged the output pins that were floating in the air. Air movement can bring charged ions and cause false readings, so Grohe enclosed the entire DUT in a shielded, copper-clad box.

The next problem was choosing an integrating capacitor. Initially, Grohe felt that the best choice of capacitor might be an air-dielectric capacitor, so he made two large flat plates measuring 4 inches by 5 inches to serve as integrating capacitors. This capacitor was the size of the second copper-clad box that would house the DUT. Using a large capacitor proved to be a bad idea. The large surface area provided a large target for cosmic rays to generate ionic charge that could affect the measurement (Figure 8). Grohe then tried to minimize the size of the capacitor while still using a good dielectric. He stumbled upon RG188 coaxial cable with Teflon insulation. A 2-inch length of this cable provided 10 pF of capacitance for the integrating capacitor (Figure 9). As an added benefit, the outer braid acted as a shield. Grohe connected it to the low-impedance output of the amplifier. With this capacitor, the density of cosmic rays was only about once every 30 seconds. Grohe made 15-second integrating measurements, eliminating the effects of the rays by taking five measurements. Later, Grohe abandoned the single-shot measurements. Any source of ionizing radiation (including old watches with radium dials) will present radiation problems. Note that Grohe pried the input of the amplifier up to avoid leakage from the PCB.

Before taking a measurement, you need to reset the integrating capacitor to zero. Using a semiconductor switch is impractical because most analog switches introduce leakage current and 5pF to 20pF of capacitance. Capacitors also have varactor effects, where the capacitance changes with applied voltage, further complicating the measurement. To minimize these problems, Grohe used a Coto reed relay. He knew that the coil could couple to the internal reed when the relay opened, so he specified that the relay be electrostatically shielded. But to his dismay, there was still a large jump in the measurement when the relay opened due to charge injection. You can also think of a reed relay as a transformer, with the reed assembly as a single-turn winding. This behavior showed that the electrostatic shielding had failed to prevent interference, and the voltage generated by the magnetic field at the high impedance end of the circuit caused the charge injection. The relay did not open immediately, and the pulse needed to charge the coil created a significant current injection in the split second before the relay opened. Grohe minimized this problem by determining the absolute minimum voltage swing required to operate the relay. This way, the relay would pull in at 3.2V and release at 2.7V. He used a set of resistor taps on an LM317 adjustable regulator to control the output between these two values. He chose not to energize the relay with the full 5V, which reduced the jumps in the integrator output and made it repeatable. He then eliminated the jumps by injecting a small current into the second-stage gain amplifier.