High-speed data testing: six top tips for improving reliability

Author : Tim Mitchell, Anritsu

27 June 2016

Figure 1 - Connector issues: the shorter ‘K’ pin prevents potential damage

High-speed data test equipment is at the very forefront of technical advances in transmission systems. Increasingly more innovative ways are being used to raise transmission speeds (to 400 Gbit/s and beyond) and to find better ways of utilising existing bandwidths.

(Click here to view article in digi-issue)

Though the testing grounds for such systems are almost invariably optical, the “donkey work” of the test process will largely be driven by high-speed electrical instruments such as signal quality analysers (SQAs). These will produce the data bit streams, provide the stressing of these streams through the introduction of jitter, and ultimately provide the ability to analyse the returned signal after its journey through the customer’s device under test (DUT).

When service engineers are testing such complex systems for integrity and performance, the standard optical issues encountered (unclean connections, bad splicing etc.) are well known. However, electrical good practice and an understanding of what pitfalls to avoid are also crucial to maximising system performance and the reliability of measurement. From this perspective, many faults and issues can be avoided if some simple practical precautions are taken prior to testing. 

This article identifies six top issues that affect the test and measurement of high-speed data systems, and discusses them in terms of the procedural and hardware improvements that can increase the reliability of these tests.

Grounding

Before supplying power to the SQA (including the DUT circuit) it is important to make sure that each piece of equipment in the system is connected to ground. This may seem obvious, but having equipment at a different potential is a recipe for disaster. Major items in a complex testing system (like the SQA) may well be connected by earth through the mains, but the components in between can be isolated, causing dangerous potential differences between the various components of the system. The user should measure the voltage between the ground of the SQA and any connected equipment (including the evaluation circuit) using a tester in AC mode. If there is no ground connection, the voltage may be different by up to half the specified AC voltage, risking serious damage to all parts of the system, not to mention the potential danger of electric shock.

Electrostatic discharge (ESD)

In recent years the understanding of ESD effects has increased significantly. However, service engineers visiting customer sites to assess robustness will often find issues with anti-static precautions (or indeed a complete lack of them!). 

Figure 2 - Characteristics of commonly used high-speed coaxial connectors

As personnel move around inside a laboratory, they generate static electricity, as do any devices that are not grounded. This might include cables and connectors, the DUT, and office equipment such as chairs, paper and plastic objects. Before the system is turned on, an assessment needs to be made of the risks these items present and any preventative action that might be required. Grounding all the equipment together is the first step in this process, but many more items will be needed to ensure proper protection:

• An ESD floor, regularly cleaned with ESD cleaner

• ESD wrist straps (with one megohm internal resistors to dissipate the static)

• Anti-static shoes and overalls 

• ESD office peripherals such as chairs and bins

• Removal of all loose plastics and paper from the work environment

• A field meter to check static around the test area

• An air ioniser to be used where PCBs and components are exposed. 

Electrostatic damage is one of the biggest causes of failure with SQAs. Both inputs and outputs can be damaged by the high-voltage static discharging through the input and output circuitry. ESD mitigation and the processes associated with this are not just advisable: they are absolutely essential in any test environment.

Connectors

Figure 3 - A modern 70 Gbit/s vector network analyser

Both in the laboratory and in the field, service engineers frequently come across issues with connectors. These faults occur mainly because of dirty or damaged connectors, especially where customers have attempted to connect to the wrong type of connector, causing splaying of the female pin or bending of the male pin. The splaying of the female pin is frequently exhibited when the longer male pin of the lower-speed SMA is mated with a female K connector. Using the correct shorter male pin of the K connector ensures that the pin will be correctly located by the threads first, thereby avoiding damage to the female connector (Fig.1).

Higher up the frequency range, V (1.85 mm) connectors offer coaxial coverage to 65 GHz but the thread is not compatible with K or SMA connectors. For 110 GHz applications 1 mm connectors are available, for which extreme care is necessary during connection. For these devices a torque spanner should be used which must be set to the correct level to avoid damaging the connectors through over-tightening.

Damaged connectors may also cause mismatches and reflections, potentially causing damage to the hardware and reducing system performance. Ultimately, as the engineer increases data and clock rate speeds, connector care - including using the right connector for the right job - becomes imperative (Fig.2).

One of the easiest issues to address is that of dirty connectors. If these are found, they should be cleaned only with isopropyl base solvent. This involves the gentle use of cotton swabs moistened with isopropyl to clean around the centre conductor, but without causing lateral pressure on the pin. Use of an air line can help to remove foreign objects from the connector as well as to dry it off after cleaning. 

Cables

Many of the above connector issues, including the same issues of frequency response and connector care, will be replicated in relation to cabling. However, when a service engineer is carrying out calibration, troubleshooting or visiting customer sites, a high-speed transmission experiment or a device manufacturing set-up can often be let down by the use of old or unsuitable cables.

As a general rule of thumb, cables that are used heavily (with many insertions or removals) should be replaced each year. If they are just left in place without movement, they will survive longer. Engineers should check the specifications for the minimum bend radii and the number of bends that the cable can withstand. Semi-rigid cables can have better frequency response characteristics than more flexible cables, but can survive less bending. In addition, phase stability and the need for matched cable pairs become crucial higher up the frequency spectrum. As a guide, copper-based cables will show a delay of approximately 55 ps per cm. Periodic testing of cables using a vector network analyser (VNA) to check return loss and make attenuation/frequency measurements is advised to weed out damaged items (Fig.3).

Electrical overstress

Figure 4 - Simplified bias-T circuit

One of the most common issues seen when servicing the receive side of bit-error-rate testers is limited or poor functionality from the main data and clock inputs. A major cause of such problems can be the tester not being aware of the highest rated allowable input, and hence causing electrical overstress. It is easy to assume that maximum output of the SQA transmitter is the same as the maximum input, but quite often this is not the case. Although most inputs will have protection diode circuitry, this can only take so much overpowering, so the service engineer has to be aware of the specifications and limitations of the test equipment. 

Other issues with electrical overstress can be encountered when monitoring a signal line with a probe. Under such conditions, it is important to take sufficient precautions against shorts to nearby elements and lines. 

Bias-T usage

Test set-ups often require the ability to alter the DC characteristics of the SQA transmitted electrical signal. To achieve this, some engineers use a bias-T device (Fig.4). However, such units exhibit inductance characteristics, which leads them to generate high surge voltages under certain conditions: specifically when being attached to, or removed from, a connector; at power-on and power-off; when left in an open-circuit configuration; and when shorting.  

These surge voltages will be reflected back into the RF output, potentially over-powering the output and creating another electrical overstress hazard. It is important to always allow this bias voltage to drop - by disconnecting the power supply to it - before disconnecting the signal line. 

Conclusion

By taking some simple precautions, first before setting up a test system and secondly during the actual running of the system, the user can make a fundamental difference to the success of any high-speed data project. Purchasing a high-quality SQA is a hefty investment, so ensuring that such precautions are adhered to is vital. Such procedures can save a great deal of time and money by protecting the test equipment investment as well as the customer’s DUT, and can also set a standard of good practice to be maintained within the lab or test environment. Most importantly, they will enable those conducting such projects to have the maximum confidence in the resulting conclusions.  


Contact Details and Archive...

Print this page | E-mail this page