12 Considerations for Improving Integrating Voltmeter Throughput

Time is money. This phrase takes on a new meaning when it is applied to product test or data acquisition. Rephrased, higher throughput is money saved. Throughput is a measure of the time it takes to test a product in production or the time it takes to scan a set of channels in a data acquisition system.

Poor throughput in production test adds cost to the product. Poor throughput in a data acquisition system means a full set of channels cannot be scanned before the scan must be repeated. In both cases, improving throughput with the current equipment is preferable to purchasing additional test equipment.

The backbone of most test systems is the integrating voltmeter. An integrating voltmeter, often referred to as a DMM or DVM, measures the physical and electrical parameters of the device-under-test.

Although an integrating voltmeter is the foundation of most test systems, it also is one of the least understood instruments. When programming an integrating voltmeter, you encounter concepts such as auto range, number of digits, auto zero, trigger delay, number of power-line cycles (NPLCs), and aperture time. Understanding these concepts is fundamental to improving the measurement rate of the integrating voltmeter.

How an Integrating Voltmeter Works

Figure 1 is a simple block diagram of an integrating voltmeter. Each block performs a function that takes time. Understanding these functions and the time associated with each allows you to optimize the throughput of an integrating voltmeter by using only the functions necessary to make a required measurement.

The attenuation switch is the first block a raw input signal sees. Integrating voltmeter input terminals are connected to the attenuation switch. Voltages higher than about 10 V must be attenuated by a resistive divider before measurement. Voltages that do not need attenuation are routed around the attenuator by a slow mechanical switch.

The second block is the auto-zero switch. It allows the meter to self-calibrate its “zero” every measurement, if desired. The auto-zero switch accommodates per-reading calibration of the offset of the following amplifier stage and the integration stage.

After the auto-zero switch, a programmable gain amplifier raises the voltage from the attenuation stage to a normalized internal voltage. The amplifier’s output voltage is routed to the integrator block. Integrators used in today’s voltmeters have evolved from the dual-slope integrators of the ’60s, but still perform the same function.

At the beginning of the integration, the output of the gain amplifier is connected to the integrator. At the end of the integration time T, the output of the amplifier is disconnected from the integrator. The output of the integrator is then measured relative to its value at the start of integration.

There are several different integrator designs and techniques for measuring the change in their outputs after time T, but they all still integrate for time T. After one or more integrations, calculations are performed, and a reading is sent to the display or the bus connected to the computer.

In a scanning system, a relay closes, and a signal is routed to the integrating voltmeter. If the integrating voltmeter is left in the default setting, the following steps take place. After the channel relay closes, the integrating voltmeter waits for several milliseconds for the signal to settle. After the wait time, the meter determines if it is set on the correct range.

If the meter’s range setting is too high or too low, the mechanical relay in the attenuation block and the amplifier gain may be changed. After waiting for the settling of the relay in the attenuation stage and settling of the gain amplifiers, the integrator is started. At the end of the aperture or integration time T, the integrator is turned off, and the change in its output is determined.

After integrating the signal, the input to the gain amplifier from the attenuation stage opens, and the zero switch closes. The zero is measured by the integrator using the same integration time. The process of measuring the input terminal followed by measuring the zero switch is called auto zero. The processor then crunches the integration value, the value of the zero integration and the calibration and range coefficients, and presents the data to the display or the bus.

A DVM also measures resistance and AC voltage. These are done with additional circuits. Measuring ohms entails turning on high-impedance current sources, routing these sources through the scanning system, and making multiple measurements. AC Vrms is measured using additional circuits that convert the AC signal to a DC level. Standard AC measurements on all meters take a long time relative to other measurements.

The default power-up settings for an integrating voltmeter are selected to give the most accurate readings. Measurement throughput, however, is not optimized. When high throughput is desired, as needed in production test or data acquisition applications, the low-level functions of the meter must be controlled. Use the following check list to help optimize your system performance.

1. Turn Off Auto Zero

Auto zero improves the offset drift of several measurements over time intervals of seconds or minutes, but it also doubles the measurement time. Offset error comes from two sources: the gain amplifiers and the integrator. During initial calibration, the static offsets from the integrator and the input amplifiers are determined; however, there still is offset drift due to minor temperature changes.

The largest source of offset temperature drift is the integrator. If the meter is making several measurements in a short period of time at a stable temperature, turn off the auto zero. Only then will a zero be taken on range changes and integration time changes. All other measurements will use the last zero measured for that range and integration time.

Most meters only store the zero for the current range and the integration time. A few meters keep a table of zeros for various ranges and integration times and do not need to perform a zero after a range or integration time change. To keep the zero fresh, force an auto zero by a command at the beginning of a scan. For many measurements, turning off auto zero is a quick, 2×-speed improvement.

2. Minimize Range Changes

Ranging of a voltmeter takes time and involves a combination of the attenuation and amplifier stages. The attenuation stage has a slow mechanical relay, while the amplifier stage uses faster, solid-state switching. Every time the range is changed from a high attenuated range to a low range, time is spent switching a mechanical relay.

When making measurements on several channels that require range changes, group the channels by range. This minimizes the time spent thrashing the mechanical ranging relay and taking another auto zero.

Minimizing mechanical range changes also increases the life of the instrument. At minimum, keep the signals that need attenuation separate from the signals that do not need attenuation.

3. Check Auto Ranging Time

Auto ranging for integrating voltmeters may be fast or slow depending on the design. Traditional voltmeters start integrating the normalized voltage from the gain amplifiers. If the integrator overflows, then the range setting is increased, and the measurement starts again. If the integration ends with a voltage that is lower than 10% of full scale, then the input range is reduced, and integration starts again. For these older designs, auto ranging is slow.

Some modern meters use a flash analog-to-digital converter to measure the input voltage to the gain stage before the integration starts to determine what range should be used. This method of auto ranging is very fast and may be faster than the time it takes a computer to issue a command to set the range.

To determine which type of ranging your voltmeter has, consult the specification sheet. If the auto-ranging speed is faster than the time it takes your computer to issue a new range and for the range to change, then leave the meter on auto ranging but keep your channels grouped together by expected value.

4. Check Integration Time

Voltmeters integrate a voltage over a set period of time. The integration time is adjustable on all programmable meters. Some meters allow the time to be set to only a few values; others allow the time to be set to any value. If the integration time is set to 1 or more whole NPLCs, 50 Hz or 60 Hz, then the effect of power-line noise is minimized. Since the default mode of an integrating voltmeter provides the most accurate answer, the power-on default mode of most meters is an integration time of 10 line cycles (200 ms for a 50-Hz line and 166 ms for a 60-Hz line).

Most integrating voltmeters also couple measurement resolution to integration time. A shorter integration time usually yields less resolution. When measuring high-level signals that do not need high resolution, such as a battery or a power supply, reducing the integration time will shorten the time spent on the measurement.

However, when changing the integration period, time usually will be spent on another zero cycle. When changing integration time, make several measurements at the new integration time before changing the integration time again. Low-level signals, which have power-line noise or need low-level resolution, should be made with 1 or 10 power-line cycles integration time.

5. Use the Native Data Format

A modern microprocessor does all of the measurement calculations in binary form. However, default IEEE 488 bus traffic is alphanumeric. The meter must convert the binary representation to a larger alphanumeric representation and send the alphanumeric across the I/O system to the computer that, in turn, converts the alphanumeric back to a binary representation.

Your test-system computer may be a Pentium II. But the I/O processor in the meter, which converts the binary to alphanumeric, drives the display, monitors key strokes, and talks on the I/O to the computer, has less processing power than most modern kitchen appliances. Set the data format for bus traffic to one that is native to the meter, and let your Pentium II do any format conversion.

6. Evaluate Command Times

Modern meters use English-like command structures that must be transmitted from your computer to the meter and then evaluated by the processor in the meter. Different computer I/O, command languages, and the meter’s processing power do not allow a general statement to be made about command parsing times.

Run a set of experiments to see if you are over-controlling the measurement setup. Wire up a series of unused channels alternating a short and a safe high voltage. Scan the channels with auto ranging on and again with programmed ranges. Which is faster?

Perform the experiment again with all shorted channels. Try scanning the shorted channels while changing integration times. Create a table of times to program various operations so you can make smart programming decisions.

7. Minimize Function Changes

Changing from function to function, DCV to ACV or ohms, is painfully slow unless your meter is designed for high-speed function changes. Order your channels so you only use each function once during a scan.

8. Do Not Measure AC Volts Directly

Converting an AC voltage to rms using the analog electronics in a meter is slow but accurate. If your meter can digitize the voltage and if the period and frequency content is known, digitize multiple periods of the waveform, and pass the data to your computer to let it compute the rms value. A few meters will perform the digitization and calculation as a function.

9. Turn Off the Display and Keyboard

The display and keyboard use the same processors that are used for measurement control or I/O to the host. Turning off the instrument display and the keyboard frees up processing power for measurement and I/O, which improves throughput.

10. Use Constant Current Sources to Measure Resistances

Resistance measurements require turning on and off current sources and multiple integration times. When measuring resistors that are not part of a circuit, such as thermistors, RTDs, and strain gages, use a constant current source to excite the resistors, and measure the voltage across each resistor. Several resistors may be put in series until the compliance voltage of the current source or the common-mode voltage of the meter is exceeded.

11. Minimize Measuring Reference Junction Temperatures

For thermocouples to be used correctly, the temperature of the isothermo terminal block where the thermocouples are connected must be known. Most systems hide the measurement of the isothermo terminal block from the user. Almost all isothermo terminal block temperatures are measured using a thermistor or RTD. These measurements are resistive, therefore slow and disruptive.

Large temperature measurement systems have several isothermo terminal blocks, and the default setting measures the isothermo terminal block temperature after measuring a set of thermocouples. Randomly jumping from terminal block to terminal block may inject multiple ohm measurements into your scan list.

To minimize the scan time, measure the temperature of all of the isothermo terminal blocks before the scan. If possible, avoid changing function to make the isothermo terminal block temperature measurement (use a bridge or constant current source). Then measure all of the thermocouples, and perform the math on your PC to convert to engineering units.

12. Evaluate Trigger Delay

When a voltage is measured through a switching system, the voltage source has to charge up the capacitance of the analog bus common to the switching system. If measuring a high-level, low-impedance signal such as a battery, very little settling time is needed. The trigger delay can be set low for these signals. However if the source is high impedance, then the delay needs to be greater.

Signals that need a longer delay should be identified and segregated. Treating all signals equally will only slow down the test system.

The measurement speed of an integrating voltmeter may be increased by a factor of 2 to 3 just by controlling the voltmeter settings. Different models of voltmeters are optimized in different ways. Experiment with several settings before selecting a measurement philosophy.

About the Author

Robert Leiby, a member of the technical staff in the R&D Laboratory, began his career with Hewlett-Packard in 1979 and has worked in R&D, production engineering, and marketing. He received a B.S.E.E. from Purdue and an M.S.E.E. from Colorado State University. Hewlett-Packard, Measurement Systems Division, 815 14th St. S.W., Loveland, CO 80537, (970) 679-2866.

October 1998