Switching Considerations When Testing High-Speed Digital Devices

The ever-increasing speed of digital devices creates new challenges for the ATE system designer. For the problems relating to the rise (or fall) time and impedance matching of high-speed digital signals in ATE, the best solutions include a blend of wisdom from the analog and RF world to ensure the test design is successful the first time.

In many ways, testing analog signals in ATE is more forgiving of error than that of a digital signal. The mere term digital means either black or white with no gray area or, in testing terms, pass or fail.

Normally, the switching of signals to test equipment is required in ATE applications to maximize the effectiveness of expensive test equipment, provide test repeatability, and increase UUT yield. VXIbus provides excellent testing speeds, small packaging, and a good selection of instruments and switching modules from multiple vendors.

Measuring newer high-speed digital signals in an automated environment most likely will require the use of a switching system that is nearly transparent to the signals being routed from the UUT to the measurement instrument. Many VXIbus modules now provide the sources, switching to and from the UUT, and the measurement of these signals. Routing a high-speed digital signal from a source instrument to the UUT is equally important as the signal from the UUT to the measuring instruments.

The Transparent Switch

The ideal switch used in a VXI-ATE application has infinite bandwidth and zero loss and infinite isolation—and does not exist yet. What does exist is a method of accomplishing a realistic and effective system design using the available switching technology. It is of utmost importance to know what pitfalls can happen to properly avoid them. Applying analog theories to the digital signal could provide the solution.

Switch Bandwidth vs Digital Signal Rise Time

Switching-system bandwidth is one of the most violated and misunderstood parameters in testing high-speed digital signals. What bandwidth must the switching system have to successfully route the digital signal without adding or subtracting to the original signal? This is clarified for most applications by approximating the rise time and/or fall time of the digital signal with this formula:

356/Rt = F

Where: Rt = the rise time of the digital signal in nanoseconds from 10% to 90% of the signal amplitude.

F = Frequency in MHz.

Simply, the frequency elements comprising the rising or falling edges of a digital signal are higher than you might initially think.

If you know the frequency range of the system and want to know the maximum rise time that the switching system will pass without major misshaping, apply the following approximation:

356/F = Rt

For example, if the frequency range of your switching system is 150 MHz, you can handle digital signals with rise or fall times of approximately 2.4 ns. If the rise time is less than this, a rounding of the signal edges occurs.

Characteristic Impedance and High-Speed Digital Signals

Characteristic impedance is another important parameter for high-speed digital signals. Characteristic impedance may seem to apply only to analog type signals; however, the rise or fall times of digital signals demand attention to this parameter. Without it, false triggering and unreliable system performance result.

The majority of malfunctions are caused by reflections of the signal back to the source due to impedance mismatching. This is quantified in the analog world as voltage-standing-wave-ratio (VSWR) or return loss. Typical impedances for digital signals are either 50 W or 75 W .

Reflection of the original signal back to the source depends upon the degree of the mismatch, the source of the mismatch, and the electrical length of the signal path (Figure 1). Mismatches can occur from any number of sources such as the source impedance, the signal path (switches, cabling, and connectors), or the load impedance.

A VXI-ATE switching system has impedance imperfections. This causes some mismatches, so remember to minimize these when planning your ATE switching scheme. To optimize testing results, use good connectors and cabling between instruments. A switch with poor VSWR (1.5:1) results in significant reflections on a digital signal.

Reflections Caused by Impedance Mismatches

An impedance mismatch causes reflections that create peaks on the digital signal. Overshoot and undershoot caused by a mismatch could render the test invalid (Figure 2). The ATE switching system must minimize these impedance mismatches when routing digital signals to the measurement equipment so that accurate measurements of the real signal can be made.

Minimizing Problems

Impedance mismatches and bandwidth problems in ATE switching systems may be resolved in several ways. To find the best solution, first determine what tests to make and what items must be connected to make the necessary measurements. This possibly includes the switching paths from the signal sources and the signal measurement equipment.

For example, take an application that requires testing 64 UUTs, each with a single input and a single output. In this example, the input to the UUT will be supplied with data from an ECL bit-stream generator. ECL technology is often used to generate digital video formats and can be in excess of 350 Mb/s with data rise or fall times of 0.35 ns.

To successfully connect the bit-stream generator to each of the UUTs without compromising the original signal, a switching tree is best suited. A switching system with one input and 64 outputs is best sub-divided into smaller sections, such as 1 × 8s. The 1 × 8 has higher bandwidth and lower impedance mismatch than that of a larger switch.

The term switching tree is used because the configuration is schematically similar to that of a family tree. As viewed in Figure 3, the bit-stream generator is fed to the single port of the 1 × 8, denoted as SW1.

The outputs of SW1 connect to eight additional 1 × 8 switching sections via coaxial cable or some other matching transmission medium. The output of these sections (SW2 to SW9) connects the signal to the different UUTs.

Although the signal must pass through two different switching sections and an additional length of cabling between the switching sections, impedance matching is better, and its bandwidth is preserved due to the smaller sections the signal must pass through. If this were one large switch, the lower bandwidth and impedance mismatch would render the switch useless for fast rise-time signals.

Crosstalk Nightmare

The term crosstalk expresses the level of extraneous signal that is leaking into the signal of interest, usually from an adjacent crosspoint in a switching array, poorly shielded cabling, ground loops, or a combination of all these (Figure 4). The leakage of signals from cables that are adjacent to the signal of interest could be a major cause of crosstalk, yet is one of the easiest to avoid.

Good quality cabling is advised throughout the ATE setup. Double-shielded cabling does not cost much more than a single-shielded cable, so why not use the better shielded cable type?

Leakage from cable to cable can cause false triggering of the instruments making the measurements, or be evident when viewing signals on a scope. Due to the limited panel space of VXI, coaxial connector types SMA, SMB, or SMC are used on high-performance VXIbus switching modules.

Modules with multipin I/O connectors generally are not acceptable due to the unshielded nature and poor impedance matching, both causing problems for the VXI-ATE. Many times, the density (size) of the switch designed into a VXI module depends upon the number of connectors that can be located on the front panel and not the actual switching inside.

Plan for Future Needs

Due to the ever-increasing data speeds and faster rise times the bandwidth demands of tomorrow’s devices will be higher. When investigating the types of switching available in the VXIbus marketplace, keep future requirements in mind.

There are many different cost levels in VXI switching modules, each with its own features and performance specifications. Thoroughly study your current switching needs and identify possible vendors that can provide solutions to meet your needs. Next, check on the next higher-performance-level equipment to compare the pricing difference. It may be easy to justify the additional cost. If additional performance is required in the future, you will be ready.

About the Author

Norton W. Alderson is the vice president of marketing at Universal Switching. Formerly a senior staff engineer with Matrix Systems, Mr. Alderson has more than 17 years design experience in switching equipment for the test and measurement industry, and holds a B.S.E.E degree. Universal Switching, 7145 Woodley Ave., Van Nuys, CA 91406, (818) 785-0200, email: [email protected].

Copyright 1997 Nelson Publishing Inc.

July 1997

Comments

To join the conversation, and become an exclusive member of Electronic Design, create an account today!