Avoid Common Switching System Problems

• loss of signal integrity,
• shorter than expected life of switch contacts, and
• throughput bottlenecks caused by switching time.

In many cases, you have the solutions to these problems within your reach. Indeed, you can avoid the problems altogether by using switches properly and by using good programming techniques.

Have you ever connected an instrument to a UUT and not seen the test results you expected? After you check for obvious problems such as wrong instrument range and bad connections, you may find measurement problems caused by your switching system. For example, you may find noisier and less accurate test results if you use switches than if you connect a measuring instrument directly to a UUT.

Contact resistance and stray capacitance can attenuate or distort a test signal as it passes through the switching system. In addition, interference from noise can disrupt signal integrity. Instruments can pick up noise from long cable runs, and if you use unshielded or improperly shielded cable with a switching system, you may get electrical noise in the form of RFI and EMI from plant equipment.¹

Test a Switch
You can see the effects of a switch on a measurement by conducting a simple test. Connect a power supply or other DC voltage source to a DMM having at least 3½ digits of resolution. Adjust the voltage until the meter reads 1.000 V. Then, connect the instruments through a relay (Fig. 1). If the relay has even 1 mV of offset voltage, you will see a reading other than 1.000 V on the DMM.

Figure 1. With no current flowing, a switch's offset voltage will cause a voltage drop measured by the DMM.

Because the DMM has a high input impedance, virtually no current will flow through the switch during this test, so the error introduced by the switch will consist of the offset voltage only, without any IR losses. As current through the relay increases, so does the IR loss across the relay.

Specifying Relays
When building a test system, you need to use relays with an acceptable offset voltage. If you process your measurements with a PC, you can sometimes compensate for offset voltages through software. But because the offset voltages can vary with temperature and humidity, you should specify relays with offset voltages sufficiently below your desired measurement resolution.

You also need to evaluate a relay’s contact ratings, resistance, and switching life, and you can take steps to ensure the relay remains within spec. Many relays have metal contacts and aren’t hermetically sealed, so the contacts corrode over time. If you suspect oxide build-up on the contacts, periodically add a cleaning cycle that forces substantially greater voltage and current (up to rated maximum) on the contacts (you may need a dummy load). You need enough voltage—typically greater than 10 mV—across the relay contacts to pierce the oxide that can form on the contacts. You can measure the effect of oxide build-up on contact resistance with a DMM that has a dry-circuit mode, which limits measurement voltage to less than 10 mV. That small a voltage won’t destroy the oxide.

Relay manufacturers usually specify their products for up to 10 million closures without exceeding a low contact resistance value—typically, less than 1W. If a relay wears out prematurely, its contact resistance may exceed the manufacturer’s specs. Often, relays fail prematurely because they’ve had power applied to them before they were closed. That’s called “hot switching.” Hot switching degrades a relay’s contacts, which can degrade the quality of the switching connection and reduce contact life by a factor of 10 to 100 times.

No Way Out
You can’t avoid hot switching in some applications, such as battery testing or power-supply testing, but you can use cold switching when you test passive devices, and thus prolong relay life. To ensure your application uses cold switching, program the test sequence to turn off any voltage or current source before activating the relay.

Even if you take this precaution, the relays could still be susceptible to inadvertent hot switching. In a typical test sequence, a system applies a test signal to the UUT, takes a measurement, and then shuts off the test signal while it switches the test fixture to the next test point or the next UUT. If the system cables have excessive stray capacitance and that capacitance charges to the full test voltage, the capacitances may not have time to discharge before the switching system connects the next channel. Some charge may remain in the cables, which could result in relays being hot switched without your knowledge. To avoid inadvertent hot switching, program sufficient delay time at the end of a test sequence before the system switches to the next channel. The reduced contact damage will result in more accurate measurements.

Unfortunately, lengthening the test time to reduce contact damage also reduces throughput, which directly relates to profits. You’ll need to decide whether you can save time in other areas of the test. If you can’t, you may have to keep the switching times short to maintain the desired throughput levels, thus risking that hot switching could cause relays to fail. You must decide whether the increased throughput is worth the downtime needed to replace the failed relays.

Saving Time
The first place you can look to save time is at your test instrument’s setup and execution times (available in the instrument’s specifications). Some of these times may be programmable, allowing you to make speed-vs-accuracy tradeoffs.

A relay’s actuation time is the theoretical minimum delay that the relay adds to test time. You can easily find that time because switching-system manufacturers will specify it. Calculating additional delays caused by the manufacturer’s control software, software drivers, or your own test program code is more difficult. (See “Software Drivers: Convenience at a Price”)

Cut Through the Layers
You can’t easily reduce latencies introduced by the manufacturer’s control software, but you can dramatically reduce test cycle times by eliminating layers of code in your test program. After you select a switch card with the appropriate relay type and contact configuration, your most important decisions involve test sequence programming. The way you write program code has a profound effect on switching system throughput.

When you use software to trigger events in a test system, you may introduce unnecessary latencies. If your test software runs under Windows, command execution time can vary by up to 50 ms. If your entire test sequence must execute in less than a few hundred milliseconds, don’t use software triggers.

Many switch systems include a memory feature that lets you store channel pattern sequences. Then, you’ll need only a few commands to step the test program through each memory location (channel), and you’ll shorten test execution time.

Typically, you get the best throughput when you combine switching-sequence memory with external hardware triggers. If you use dedicated hardware to trigger the switch system and receive status feedback, your cycle time will approach the minimum established by the relay actuation time.

External triggering typically requires nothing more than short, single-shot digital pulses that initiate a relay action. If you program your system with the proper sequence, the system will quickly act upon those triggers. By comparison, when software triggers an event, your test system must execute a scan sequence one event at a time because the scan list isn’t stored in your switching system. T&MW

FOOTNOTE
1. Switching Handbook, Third Edition, Keithley Instruments, Cleveland, OH, 1995, pp. 3–56 and 3–57. www.keithley.com/cgi/new_cgi/ssform.cgi?tmo.  

FOR FURTHER READING
Buffi, Kevin, “Evaluate Relays for Automated Test Stations,” Test & Measurement World, January 1997, p. 39. 

Sarfi, Tom, “Minimize Switching Time in Scanning Systems,” Test & Measurement World, August 1999, p. 43.   

 Douglas Rathburn is a senior applications engineer at Keithley Instruments where he is responsible for customer support and for writing application software and technical documentation. He received B.S.E.E. and B.A.—Economics degrees from Case Western Reserve University in Cleveland, OH. E-mail: rathburn_doug@keithley.com.