Is Your Accuracy Being Degraded?
by Paul Lantz and Tee Sheffer, Signametrics W.
Test systems almost always involve multiple stimulus and measurement instruments for testing several points on the DUT. This implies the use of some sort of switching subsystem. If you think switches are simple and straightforward to select, this article may change your mind.
Every switching system involves trade-offs. Sometimes commercial requirements override technical specifications with unpleasant results. Even some technical requirements may conflict with each other and must be worked around very carefully.
We frequently talk to frustrated customers who have put together systems that combine a 7½-digit DMM with a switching system that supports only 4½-digit accuracy. In a remarkably large number of cases, they had never thought of the switching system as a possible source of error. In other cases, they knew there was a possible problem but couldn’t pin it down or control it.
Once you understand the need to select the right switching module for your electrical requirements, you will see that one size does not fit all applications. You need to design your switching system based on what you need it to do. This may mean the use of more than one type of switch in your system.
A Cautionary Example
We encountered a painful example working with a large medical devices company. The company had contracted with an integrator to provide a system for testing the reliability of cables used in pacemakers. The requirements were deceptively simple: measure the resistance of several wires while they were being flexed and twisted. The system needed to detect resistance changes of 20 m
The integrator chose a PXI switching card and 6½-digit PXI DMM from a major supplier based on a commercial alliance agreement rather than technical considerations. The resulting system exhibited readings that drifted randomly by 200 mΩ rendering it useless
The integrator did not have the experience to resolve the problem, and the supplier did not provide technical support. The end-user was frustrated because he had a useless and expensive system, project delays, no explanation of the problem, and no solution.
Eventually, the end-user called Signametrics Customer Support. We explained that the problem was caused by inappropriate hardware in the system. We guaranteed that a Signametrics SMX4032 Instrumentation Switching Card and an SMX2064 7½-digit DMM would do the job or he could send them back for a refund. The system performed well, and the customer now is happy.
How could this happen? Easily, when you realize the many variables in this situation. The end-user is a mechanical engineer who trusted the integrator’s judgment. The integrator is skilled at software and systems integration but didn’t have a measurement background. The original salesperson did not have enough instrumentation knowledge to understand the requirement. The original equipment vendor did not stand behind his product.
When you specify the switching system, it is tempting to make some assumptions that may prove troublesome later. Here are some common misperceptions:
• Switches are ideal and simple devices that route signals between various test equipment. Not true. This will lead to perplexing outcomes.
• A matrix lets you connect any input to any output. If the requirements change later, you can handle it by a simple software change. Partially true. But there is a heavy price that goes with it.
• If you specify a switching system that can handle the maximum voltage and current, it can handle all of the input requirements. Not true.
• If you need to measure very low voltages or resistances, it’s not a problem because a general-purpose switch or a matrix can do it well. Not true. The switch is designed to do whatever its specifications indicate. If something isn’t specified, such as thermal offset voltages, you don’t know what you will get.
• You need to make some accurate resistance measurements. A four-wire connection eliminates all errors in the switching path. Not true. It reduces some of the errors, but not all.
Matrix vs. Multiplexer
One of the first decisions is whether to select a matrix or multiplexer type of switching topology. The matrix is very versatile. It allows you to connect any input to any output. If your requirements change later, it’s just a software change. And the simplified circuit diagram looks so clean.
Unfortunately, if you try to implement a large switching matrix using one relay per crosspoint, you will have a huge number of relays. Physical size and cost quickly get out of control.
The cost-effective way to implement a matrix is to use a switching-tree type of architecture. But the tree architecture forces each connection to go through several sets of relay contacts, degrading system accuracy significantly. This degradation can take the form of increased leakage currents and thermal offset voltages from the relays. There also is considerable risk of welding a relay contact with a matrix since it does not offer the protection of a multiplexer, where all relays automatically open prior to closing a selected channel.
A multiplexer is more straightforward. It switches 1-of-n inputs onto a sub-bus using a break-before-make sequence. Then it uses a group switch to switch the sub-bus into a measurement card. If many inputs are involved, you can add more scanners with more sub-buses. The number of relays stays reasonable, so the size and cost also are reasonable.
Some multiplexers allow you to switch 1-of-n inputs to 1-of-4 outputs (Figure 1). Many also provide some software tools to simplify control of the switch card. There can be some leakage issues if you have high impedance inputs, but there is very little degradation of low-level signals.
Figure 1. Signametrics SMX4032 Instrumentation Switching Card
General-Purpose, High-Density Switch
The instinctive tendency is to select a high-density switch. This does make sense for many applications.
It may not be obvious by reading the specifications, but there are several measurement functions that will be considerably degraded. RTD and thermocouple measurements are the most severely degraded because high-density switches have high thermal offset voltages. Counter to a widely held belief, using a four-wire resistance measurement method does not eliminate them. A secondary factor is the tendency of high-density switches to pick up more noise due to the proximity of PCB traces and relays.
In summary, a high-density switch can pack a lot of functionality into a small space, but it is not suitable for accurate measurements of lower-level signals.
Instrumentation Quality Switches
Instrumentation switching means switching with very low voltage offset errors, low contact resistance, low noise, and low cross-coupling. If you are measuring resistances below 10 Ω, such as in continuity tests, you need contact resistances less than 1 mΩ and thermal offset voltages less than 1 µV. This may seem obsessive, but anything more will quickly degrade the accuracy of your system.
High-density general-purpose switches generally are not specified with instrumentation-type specifications. This means that the manufacturer has not characterized them for that application.
Why do low thermal EMFs and low contact resistance matter? If you look at the specifications of a 7½-digit DMM, you will note that the instrument is accurate to a few microvolts. Even a good 5½-digit DMM is accurate to about 50 µV. If you put your signals through a switch that introduces 100 µV of error into the signal, you have a system that is accurate to about 4½ digits.
Table 1. Error Contributions From Various Switch Configurations
Resistance measurements must be made at a fairly low voltage, usually less than 2 V, to control heating and avoid turning on semiconductors connected to the DUT. This means that the errors introduced by a switching system affect all resistance readings, not just low resistance readings.
The most instructive way to look at these errors is as a percentage of the resistance being measured, as shown in Figure 2. It depicts the error contribution of various switching card types on resistance measurements. It assumes standard multimeter test currents and test voltages of less than 2 V. The sawtooth results from decreasing the test current on higher resistance ranges.
Figure 2. Error Contributions From Various Switch Configurations for Resistance Measurements
As shown in Figure 2, your switching system could make your resistance measurements much less accurate than you expected. However, there are some things you can do about it:
• Many DMMs have an Offset-Ohms function that partially corrects for this error. There is a significant increase in both measurement time and noise.
• Some DMMs use 10 mA of excitation current to measure low values of resistance. This provides a respectable 10-fold reduction in the error for resistances below 20 Ω where they are the worst.
• Contrary to popular belief, using a four-wire resistance method does not change these errors. Four-wire measurements only eliminate errors caused by test-lead resistance in the measurement path; they do nothing to reduce voltage errors in the measurement path.
Where Do These Voltage Errors Come From?
The physics of electromagnetic relays makes voltage errors almost unavoidable. Electromagnetic relays depend on a nickel-iron alloy element to move and complete a circuit when a magnetic field is turned on. That nickel-iron element is soldered to a copper circuit board.
The connection point between nickel-iron and copper makes an unfortunately good thermocouple that generates about 40 µV/°C of temperature change. This can be controlled if both ends of the relay are at the same temperature. Many high-density and matrix switches are designed with other priorities, and there can easily be a few degrees of temperature difference across each relay. Consequently, the voltage errors appear.
What is the maximum voltage that the switching system must handle? Voltage ratings are pretty clear. Just allow some margin for voltage transients. If the switches must handle power line voltage, safety requirements come into play.
What is the minimum and the maximum current? Relays that can switch more than 2 A use contact plating materials that do not work for dry-contact switching. In other words, they do not work well unless there is a significant current flow through the contacts. If you need to switch high currents and very low currents, you need two kinds of switches.
High-frequency isolation and low attenuation require special switching topology and transmission line connections. This is quite specialized.
Does the software driver for the switch card support the topologies you need? The available software can make your job considerably easier. Or it may not.
Switching arrangements that you are likely to need include the following:
• Single ended for high density.
• Differential for higher accuracy and low noise.
• Four-wire for accurate ohms measurements.
• Six-wire for in-circuit ohms measurements.
Some switching cards come with a driver package that makes all of these configurations easily implemented on the fly. An on-board microcontroller makes it efficient and fast.
The switching system is one of the most important parts—and in many cases the most expensive part—of your test system. If the switches are carefully chosen, you can achieve the full accuracy of the instruments in the system.
About the Authors
Paul Lantz, who has been involved with test and measurement since 1972, is vice president of engineering at Signametrics. Before joining the company, Mr. Lantz spent several years with Fluke in a number of engineering positions including project manager and principal engineer. In addition to holding three test and measurement-related patents, he received a B.S. in electrical engineering from Seattle University and an M.S.E.E. from New York University and has authored several published technical articles.
Tee Sheffer is the president and founder of Signametrics. He earned undergraduate and graduate degrees in electrical engineering from the University of Washington. From 1977 until 1990 when he founded Signametrics, Mr. Sheffer was a senior staff engineer and project manager at Fluke. He holds 10 patents in the area of test and measurement and has authored several technical articles.
Tee Sheffer is the president and founder of Signametrics. He earned undergraduate and graduate degrees in electrical engineering from the University of Washington. From 1977 until 1990 when he founded Signametrics, Mr. Sheffer was a senior staff engineer and project manager at Fluke. He holds 10 patents in the area of test and measurement and has authored several technical articles
illustrates how much error can be introduced when you measure different signals with various qualities of switches. It shows that an instrumentation switch introduces less error than the 7½-digit DMM while a typical high-density switch has errors equivalent to 5 digits or worse.