Best practice: Consider interconnection as an extension of the test instrument
When tested on a bench, a product generally has the optimal test environment, with short and direct connections to the stimulus and measurement devices. But in production, the interface between a device under test (DUT) and the test instruments will likely include a network of cabling, a signal distribution or switching subsystem, a connector panel, and device-specific adapters. All of these components do nothing to enhance the pristine signals that were—during product development—passed back and forth directly between the instruments and product on the bench. Instead, they contribute opportunities for signal degradation. In essence, all of the cabling, interfacing and switching should now be considered as an extension of the test instrument and must be allowed for when assessing the instrument’s capabilities. This article focus on how path resistance, line capacitance and insertion loss affect test results and what can be done to minimize their impact.
Objectives of a successful automated test system
As a new product concept makes its way through research and development, validating the design is often a manual process completed on a lab bench. Once the product design has been validated, product acceptance requirements are transitioned to the production test engineering team. Ideally, the product will have been designed with testability in mind, but quite often, production testing is an afterthought. The test engineering team then has the responsibility to create an acceptance test procedure that is part of the manufacturing process, and which is efficient enough to support production demand. This procedure is frequently automated to help accelerate test times and reduce errors typically associated with manual interventions by test technicians. Application code running on a PC controls test instruments in a rack that are supplying stimulus to and measuring the response from the product being tested. The challenge of creating an automated test platform is often increased when the platform is required to be versatile enough to test multiple types of products. Product specific test adapters that are used to route system I/O to the device being tested introduce multiple transition points that can corrupt the integrity of test signals.
Journey of a test signal
Figure 2 shows a dual channel function generator, which is connected to multiple points on a device under test (DUT). A small switching subsystem is used to distribute the outputs of the generator. The I/O of the switches and the Instrument is wired to a common receiver panel and then connected in an adapter that is specific to the device being tested. Finally, this adapter is wired to the DUT.
The path that the signal follows introduces multiple attack points on the integrity of the signal. Let’s say in this example; the DUT requires an input pulse with a minimum width and rise/fall time to initiate a sequence of events. In this simple example, a the pulse generated by the function generator first travels through a few feet of cabling to the test interface, a connection point and more cabling through the test fixture, back to the first layer of the switch, back through the interface and fixture to the second layer of the switch and finally out to the DUT. In essence, the 10+ feet of cabling, connectors and switch contact points have become part of the function generator.
The cabling, switch contacts and connectors in the signal path do nothing to improve signal integrity. Instead, they are working against it. The impact on the signal is not always easy to calculate or predict, especially when there are multiple possible paths designed into the system. The result is uncertainty in the integrity of the signal, the resulting measurements and, ultimately, uncertainty in the quality of the product. So while the function generator is likely performing to its stated specifications, once in a test system, its output once through the signal routing system may end up suggesting otherwise.
In such a system, the function generator will be specified for several specific characteristics, such as rise time, bandwidth, source impedance and amplitude accuracy. The signal routing system contributes capacitance, path resistance, insertion loss, reflected power, and various opportunities for impedance mismatching and signal attenuation. If the design of the routing system is not thought through, an attack on the integrity of the signal can happen. That can result in uncertainty in the measurements being made, and more importantly, put the integrity of the product about to be shipped in question.
Taking on this risk is not an attractive option. The goal of any system design becomes making the transmission path as transparent to the signal as possible without simply reverting to a manual setup. So, where do we start in an ATE signal routing design process to ensure that we have a reliable system in place—that is also repeatable, maintainable and provides the lowest cost of ownership over the life of the project? To begin, let’s take a deeper dive into understanding how an interconnection system works against the systems engineer, and then what can be done to limit its impact on signal integrity. We will focus on path resistance, signal attenuation and insertion loss.
- Path resistance: This is the resistance that is added to a signal path due to cabling, connectors and switching. It is fundamental knowledge that a resistor in an electrical circuit provides several useful purposes, such as part of a filter network reducing noise interference or as an attenuator bringing high voltage signals into an acceptable range of a connected device. Resistance in an interconnect system, on the other hand, creates unwanted voltage drops, loss of transmitted signals and heat among a few side effects. Resistance in a signal routing system results in something less than what is expected at the signal’s destination (Figure 4).
- Insertion loss and impedance mismatch: The issues we have discussed so far describe the impact system wiring has on the overall integrity of a transmission path. When transmitting signals that have higher frequency components, a key specification of interest is bandwidth. Bandwidth ultimately defines the highest frequency signal that can be passed through a transmission line. In an ATE system, bandwidth is governed in part by the insertion loss at the various connection points in the signal path. Switch contact points and interconnects provide ample opportunity for impedance mismatches that reduce system bandwidth.
Contacts and switches are typically specified for insertion loss at a frequency at which the power received through the switch is half of that generated by the source, which occurs at the -3dB point. If you consider that multiple relays can end up in a signal path, the cumulative effect of insertion loss can be significant, even at lower frequency signals. Figure 11 shows insertion loss plots for two Pickering switch modules, one is comprised of individual SP4T multiplexers, the other includes internal wiring connecting SP4T relays in series to form SP16T multiplexers. Insertion loss is cumulative, and the SP16T module’s effective bandwidth at the -3dB point is reduced accordingly.
Non-sinusoidal waveforms are comprised of multiple harmonics. Even a 140 MHz square wave will be impacted by a switching system that has an overall bandwidth of 500 MHz since the higher-order harmonics that contribute to the sharpness of the edges will be dampened, impacting the fidelity of the square wave as discussed in the previous section.
Therefore, it’s important to ensure that the switching components selected, and the signal path routing system, provide enough headroom to satisfy the test requirements with which you are presented. But also consider that the system you are designing today may be required for other needs later on.
Gigabit Ethernet and other high-speed interfaces
Gigabit Ethernet, USB and other high-speed serial communications interfaces have become prevalent in products in the automotive, semiconductor, defense and aerospace industries. It can be challenging to preserve these signals since they can be low power and ever-higher frequencies. The original USB specification was limited to 12 Mbit/s, while the latest proposal for USB 4 is planning for speeds more than 3,000 times faster: up to 40 Gbit/s. (Figure 14) Ethernet likewise has evolved to operate at higher speeds and with fewer data-carrying lines. Single-pair Ethernet is an emerging technology targeted toward automotive applications that aim to bring the same high-speed data as CAN or other standards with fewer connections and simpler cabling.
These signals can be particularly challenging due to their low power and higher frequencies requiring appropriate cabling to preserve the signal. Their connections also require matching the impedance, performance, and other specifications defined by each technology’s standards organization. USB and Ethernet both use twisted-pair cabling to reduce crosstalk and interference from other signals. Some versions of these standards also require shielding in the cable for the same reason. Adding these features can increase the complexity, size and cost of the cable. As always, these trade-offs must be considered in the context of the larger test application.
Designing custom cables
Once the design goals are understood and acceptable trade-offs determined, the task of designing the cables for manufacturing is typically difficult. Many organizations do not have a dedicated software tool for defining cables, connectors, or their pin outs. This can result in confusion or missing detail in the design, as well as a lack of documentation, which can make it difficult to reproduce or standardize any cables used.
Therefore, especially in the case where there is no dedicated in-house resource available, test teams will work with custom cable manufacturers who can help with the selection of components and by providing documentation on a finished design.
Signal routing for test would be considered to be perfect if the path between the test instrument and the device under test was electrically ‘invisible’. Of course, this is asking the impossible. The serious point here is that test engineers need to take care of the whole measurement channel by considering everything between the instrument and the DUT, including cables, switching subsystems, connectors and mass interconnect products.
The best switch module is rendered virtually useless unless the cabling and the connector types are appropriate for the application. Too small a wire gauge results in increased lead resistance, causing serious voltage drops that impact test results. A long cable may have too much capacitance and signal attenuation. The result can be damage to your switches and even the DUT when discharging the energy in the cable, and it can turn a fast-rising signal into a slow, rolling waveform. Poor connectors may wear out too soon, creating intermittent problems; EMI may be an issue, and there are many other factors to consider.
Selecting a flexible switching platform that can cover a range of application requirements and provide for future expansion provides a solid foundation for a test system design. Pickering’s modular switching portfolio provides maximum flexibility for system designs offering the widest range of products covering DC to light.