Increase the Accuracy of Low-Level Voltage Measurements

Electronic products populated with complex circuitry and using lower operating voltages, make it difficult to perform accurate voltage measurements because significant errors can be introduced by noise and drift sources. These sources include thermal electromotive force (EMF), magnetic fields, ground loops, and Johnson noise. Taking steps to understand and minimize these factors is crucial for making meaningful measurements at low voltage (Table 1).

Thermal EMF

Thermal EMFs are thermoelectrically induced voltages and the most common source of errors in low-voltage measurements. These voltages arise when junctions of dissimilar materials are at different temperatures. Typical thermoelectric voltages of some materials with respect to copper are shown in Table 2.

The thermocouple effect generates a voltage that is proportional to the temperature at the junction of the dissimilar metals, said Steven Ross at VXI Technology. Thermal EMFs can be controlled by reducing or eliminating dissimilar metals or selecting metal-to-metal junctions that eliminate thermocouple effects when exposed to temperature extremes. The effects of thermal EMF can be compensated for by taking measurements in a controlled temperature environment where the amount of EMF can be calculated and subtracted from the final result.

The magnitude of the voltage generated by the thermal EMF depends on the type of metals used as well as the temperature difference between junctions, said Tom Hayden, sales support and training manager at Keithley Instruments. This voltage offset adds to the signal to be measured, producing an error in the measurement. To minimize thermal EMFs:

Avoid the use of steel bolts and nuts in fixturing that is part of the signal path, because they could introduce thermal EMFs as high as tens of microvolts.

Avoid iron-copper connections.

Use tinned copper wire and connectors for run-of-the-mill applications to limit thermal EMFs to a fraction of a microvolt.

Only use clean, copper-to-copper, crimped-on connections in critical, low-level voltage measurements, and keep all junctions at the same temperature.

Constructing circuits with the same material for conductors minimizes thermal EMFs. For example, gold and silver have thermal EMF lower than lead-tin by a factor of 10 (Table 2).

Minimizing temperature gradients within the circuit also reduces thermal EMFs.

Just place all junctions in close proximity and provide good thermal coupling to a common massive heat sink. This coupling must take place through electrical insulators having high thermal conductivity. Since most electrical insulators do not conduct heat well, special insulators such as hard anodized aluminum must be used to couple junctions thermally to the heat sink.

Also consider where you place your DMM when making measurements, said Tee Sheffer, president of Signametrics. Keep the instrument away from sources of heat and cold because they can contribute significantly to thermal EMF-related errors. Data and power cables also can be a source of thermal gradients.

Another helpful procedure to minimize thermal EMFs is to operate your test equipment in a constant ambient temperature. Any remaining thermal EMF that is relatively constant can be compensated for by zero controls on the measuring instruments. To keep ambient temperatures constant, equipment should be kept away from direct sunlight, exhaust fans, and similar sources of hot or cold air. Wrapping connections in insulating foam also minimizes ambient temperature fluctuation from air movement.¹

Magnetic fields can generate spurious voltages. Even the earth’s weak magnetic field can generate nanovolts in dangling leads. As a result, circuit leads must be kept short and rigidly fixed in place.

Basic physics shows that the amount of voltage a magnetic field induces in a circuit is proportional to the area enclosed by the circuit leads. The leads must be run closely together and magnetically shielded to minimize magnetically induced voltages.

The effect of magnetic fields on measurement data is reduced by shielding the lines between the DMM and the DUT, agreed VXI Technology’s Mr. Ross. Isolate the measurement instrument and the DUT from external influences by physical proximity or shielding, and minimize line length between the instrument and the DUT.

The types of noise caused by electromagnetic (EM) coupling may be identified by the source or where they appear in the circuit under test, said Keithley’s Mr. Hayden. These include 60-Hz line noise, EMI/RFI, normal-mode voltages, and common-mode voltages.

Magnetic fields generate noise currents by inductive coupling in low-impedance circuits and devices. Electric fields typically produce noise through capacitive coupling in high-impedance circuits and devices.

The most common form of external noise produced by EM coupling is caused by external 60-Hz line noise, continued Mr. Hayden. Line noise is common, especially near fluorescent lights. The noise components these electric fields superimpose on a DC signal often cause inaccurate and fluctuating measurements. To reduce the noise, use shielded, twisted-pair leads, and keep measurement circuits away from these external noise sources.

EMI/RFI noise can be caused by steady-state sources such as TV or radio or impulse sources like high-voltage arcing, said Mr. Hayden. This type of interference may show up as a steady-state offset, or it may result in noisy or erratic readings.

To avoid measurement problems with EMI/RFI, do not try to make low-level measurements near powerful TV or radio transmitters or interference sources, said Mr. Hayden. Shielding the test leads and DUT may be sufficient. But, it may be necessary to use a screen room to sufficiently attenuate the troublesome signals. If this fails, then filtering is the remaining solution, including filtering and noise-rejection functions built into the DMM.

Ground loops occur when the source and the measuring instruments are not connected to a common ground. A voltage difference between the source and instrument grounds will cause a current to flow around the loop. This current will create an unwanted voltage in series with the source voltage.

A typical example is when several instruments are plugged into power strips on different instrument racks. Frequently, there is a small difference in potential among the ground points. This potential difference can cause current to flow and create unexpected voltage drops.

The cure for ground loops is to ground all equipment at one point. The easiest way to accomplish this is to use isolated power sources and instruments, then find a single earth ground point for the whole system.

In any resistance, thermal energy produces motion of charged particles. This charge movement results in noise which often is called Johnson or thermal noise. It is the noise generated by any resistance at a temperature above absolute zero. The power available from this motion is:

P = 4kTD f

where: k = 1.38 × 10^-23 joules/Kelvin (Boltzmann’s Constant)

T = temperature in Kelvin

D f = noise bandwidth in Hz

The Johnson voltage noise developed in a resistance R is:

At room temperature (approximately 300°K), the equation becomes:

Johnson noise occurs in a resistive device and is caused by the thermal motion of charge carriers, restated Keithley’s Mr. Hayden. It has a white-noise spectrum and is determined by the temperature, bandwidth, and resistance value.

Since all voltage and current sources contain an internal resistance, they exhibit Johnson noise. In most applications, Johnson noise is quite low and not a problem. For example, with a 100-W source impedance at room temperature, Johnson noise is much less than 1-µV peak-to-peak at 10 kHz bandwidth.

A typical situation where Johnson noise could be a problem is in low-level measurements using a high-speed analog-to-digital converter at high sample rates. The bandwidth required for such measurements exposes low-level signals to a broad spectrum of Johnson noise.

The input impedance of the DMM also could be a factor in creating Johnson noise, said Signametrics’ Mr. Sheffer. Some lower-end system/benchtop DMMs could have 10 MW in series with the signal path, and this adds significantly to the Johnson noise.

For Johnson noise, measure at a place on the board where the signal is as large as possible, said Raul Martinez, product marketing manager at Tektronix. If practical, filter the signal or use the average or slow-response mode on the DMM.

If you make low-level measurements occasionally, you only need a general-purpose DMM with better-than-average resolution, accuracy, and noise performance specification, said Barry Scott, DMM product manager at Hewlett-Packard. However, if your work keeps you focused consistently at the nanovolt level, then a sensitive DMM tuned for low-level measurements will provide much better results. The nanovolt meter offers many features to improve low-level measurements, including:

More Resolution: By improving the measurement resolution, you effectively increase the DMM’s capability to sense small changes in input signals. The dynamic range also is improved, which means you can minimize range changes and associated settling errors.

Lower Input Ranges: The lowest range on most general-purpose DMMs is 100 mVDC, which means a 1-mV measurement is only at 1% of full scale, where noise dominates the reading. A nanovolt meter extends measurement ranges using special, low-noise amplification circuits. For example, the HP 34420A has a 1- mV range which when coupled with 7½-digit resolution, gives 100-pV sensitivity.

Low-Noise Instruments: Excellent sensitivity is useless if you cannot pull a useful reading out of the noise. Nanovolt instruments are helpful because they are designed to minimize noise from a reading. The best instruments offer high-quality, Teflon-encased, twisted-pair cabling to minimize noise pickup and reduce loop area, and they integrate analog-to-digital technology to reject normal-mode noise that couples into measurements from power-line sources. Filtering further reduces noise, with some instruments providing selectable digital filters that calculate a moving average of 10, 50, or 100 readings.

Low Thermal Input Connectors: Nanovolt meters typically use a specially designed input connector to minimize errors. Typical connector leads are 99% copper and closely spaced to maintain a proper isothermal relationship.

Offset Compensation: This feature removes thermal EMF errors from resistance measurements by modulating the current source on and off and subtracting the measured offsets.

The most important DMM features for low-level measurements include low offset voltage and drift and an autozero capability, said National Instruments’ Mr. Becker. A high common-mode rejection ratio to control ground loops or other common-mode voltage errors also is important. For example, to keep the error 120 dB. Finally, look for a meter that offers normal-mode or 60-Hz rejection to eliminate power-line frequency noise.

1. Catalog and Reference Guide, Keithley Instruments, 1998, pp. A-39 to A-44.

The 5490 DMM measures DC voltages to 1,000 V at 0.025% accuracy and makes true rms AC voltage measurements. Overvoltage protection is 1,100 VDC + AC peak. DC and AC current measuring ranges are from 500 mA to 10 A, resistance from 0.01 W to 50 MW , and capacitance from 10 pF to 50,000 µF. The instrument provides 600 VDC capacitance and resistance protection and frequency measurements from 0.5 Hz to 500 kHz. A 50,000-count resolution LCD provides easy readings, supported by a 34-segment analog bar graph. $675. B+K Precision, (800) 462-9832.

The HP 3458A Multimeter provides user-selectable measurements from 8½ digits VDC at six readings/s to 4½ digits VDC at 100,000 readings/s in 100-ns steps. True rms AC is measured at 50 readings/s with 100-ppm accuracy for input frequencies from 1 Hz to 10 MHz. The instrument features eight menu command keys, a numeric entry for constants and measurement parameters, and keys for 10 user-defined setups. The bench meter also can make two-wire and four-wire resistance measurements. DC current ranges are from 100 nA to 1 A with 1,350 readings/s at 5½-digit resolution with 14-ppm accuracy. AC current measurement ranges are from 100 µA to 1 A with 50 readings/s and 500-ppm accuracy. $6,900. Hewlett-Packard, (800) 452-4844.

The GDM-8055 Digital Multimeter supports voltage, current, resistance, and decibel measurements. It has 0.006% VDC accuracy and 1-µV resolution. The unit measures DC voltages from 200 mV to 1,000 V, ACV from 200 mV to 750 V, DC and AC current from 200 µA to 2 A, resistance from 200 W to 20 MW , and decibels above -78 dBm. Software calibration is performed from the front panel or remote interface. $899. Instek, (626) 336-6537.

The 2015 Multimeter combines a total harmonic distortion (THD) measurement function with a 6½-digit accuracy for voltage, current, and resistance measurements. It uses a 20-Hz to 20-kHz sine-wave signal generator with continuous and pulsed outputs for measuring response and harmonic distortion in the frequency domain. Distortion measurements include THD, THD + noise, and signal-to-noise + distortion. $3,495. Keithley Instruments, (800) 552-1115.

The NI 4060 for PXI and CompactPCI is a DMM for modular instrumentation systems. It features a 5½-digit accuracy for VDC, true rms VAC, DC and AC current, resistance, and thermocouple measurements. The meter has three digital filter settings of 10, 50, and 60 Hz for rejecting power-line noise. It makes continuity measurements by gauging the resistance between the two probes and comparing it to a set value. Trigger-output and trigger-input signals are provided for scanners. VirtualBench-DMM software provides a soft front panel for controlling the meter without the need for programming. $995. National Instruments, (800) 258-7022.

The SM-2020CT DMM features 5½-digit measurement capability and frequency, hardware, and programmable level triggering functions. Frequency and period measurements are made with five digits of resolution for AC voltage and current. The external hardware trigger and the programmable level trigger define the threshold to capture low-frequency waveforms. The unit provides one to 200 readings/s, two- or four-wire resistance measurements, and 300-mVDC to 300-VDC measurements with 1- µV resolution. $995. Signametrics, (206) 524-4074.

DMM Module Uses

FIFO Buffer and DMA

The VM2710A Module is a 6½ -digit DMM with a first in first out buffer and direct memory access architecture that allows you to off-load data from the DMM to the host controller. It provides AC and DC voltage measurements and two- and four-wire resistance measurements. The board performs up to 2,000 readings/s over the backplane and has 256,000 readings of on-board memory. It allows register and message data access. VDC measurements range from 100 mV to 300 V, and true rms ACV measurements range from 100 mV to 300 V. Resistance is

measured from 20 W to 20 MW . $1,900. VXI Technology, (949) 955-1894.

Table 1. Measurement Error Sources

Copyright 1998 Nelson Publishing Inc.

Table 2. Thermoelectric Potentials

Meter Has Frequency and

Programmable Trigger

PXI-, CompactPCI-Based Meter

Offers 5½-Digit Accuracy

Meter Combines THD Function

With Conventional Measurements

Meter Has 0.006% VDC Accuracy

And Software Calibration

DMM Offers 100k Readings/s

And 8½-Digit Resolution

DMMS

Bench DMM Features

0.025% Accuracy

References

Low-Level Features

Johnson Noise

Ground Loops

Magnetic Fields

August 1998