Estimated reading time: 4 minutes

Repeatability, Reproducibility and Rising Frequency: The R3 Predicament

One of the more popular editions of The Pulse in 2011 was the two-part article “Transmission Lines - a Voyage From DC.” After a brief hiatus, The Pulse returns to revisit the topic with a slightly different look at the rising frequency challenges from a general measurement perspective.

Starting again from DC and working through the frequency bands, I'll look at what is realistic to achieve and where economic compromises may need to be made from a practical perspective.

Precision vs. Accuracy?

Before getting in too deep, it is worth remembering that the advent of calculators and computer software sometimes leads technicians and engineers from many disciplines into the false sense that a measurement is more precise than is actually the case.

Probably the first experience of this for “old-timers” was the advent of the digital multimeter. Lots of decimal places gave the illusion of terrific accuracy – but pick up the data sheet and you will discover that perhaps there was a spec of 3% of full scale, ± a handful of millivolts! Yet the 4 or 5 decimal places on the display may have led to the conclusion that there was an order or two of magnitude more accuracy than was truly the case – a trap which is not so easy to fall into when reading off a traditional analog scale.

PCB fabricators have many a similar tale to tell of designers specifying drill tolerances or finished thicknesses to a specification far more precise than the tooling or materials were capable of producing. With education and experience come the knowledge that you have to understand what you are measuring and what the limitations of the measurement equipment and the representative sample you are measuring are, plus the ability of the national standards bodies to provide calibration standards with known accuracies and uncertainties.

Measurements Through the Frequency Spectrum

We return now to the frequency journey topic, and taking voltage, or, at higher frequencies, power transmission measurements. At DC, the capability of precision meters for accurate measurements is quite astounding; the combination of very high input impedance and averaging techniques means it is quite possible to make resistance and voltage measurements with high accuracy and repeatability. For voltage measurement the interconnection between meter and device under test can be quite simple – and the interconnection cost is quite low.

At power line frequencies of 60 or 50 Hz, measurement is still straightforward and simple meter leads still offer a very accurate interconnect to the metrology equipment. As frequency rises up through and above the audio band, then at some point coaxial or transmission line cables become necessary for accurate measurements. An interesting observation here is that as the frequency ramps up, not only do the measurement instruments increase in price, but the interconnect to the measurement sample begins to rise in price, and at a much more rapid rate. The price of the interconnect increases as a proportion of the measurement system more rapidly than the measurement instrument itself.

At frequencies where transmission line characteristics need to be measured, the price of the interconnect moves from dollars to hundreds of dollars, and once in multi-GHz insertion loss territory, then the interconnect costs start to head into the thousands of dollars. At these frequencies, the whole topic of measurement must be viewed as a system, with the measurement instrument, interconnect, and finally, the transmission line under test, including its launch and termination connections, playing vital roles in the quality of measurement.

It is interesting to look at the specifications for vector network analysers (well, it is if you enjoy reading specifications) and try to deduce how accurately you can make a measurement. VNA manufacturers generate pages of documentation detailing the combined effects of the actual test frequency, the types of cables and probes used and the accuracy of the calibration artefacts*, along with the expected insertion loss of the device under test. To answer the question, “How accurately can I measure sample X?” takes a fair amount of analysis – and maybe surprisingly, knowledge of the expected end result before you start. Some vendors even provide software tools for the end-user to input all the variables prior to making a measurement so the end accuracy can be estimated before the measurement is taken.

Figure 1. SDD21 – dB loss per 1000 mm from 0 – 20GHz.

Does all of this complexity mean high-frequency measurements are unnecessary? Certainly not, but it does place a responsibility on the designer who is specifying a high-frequency measurement to ensure that specification is within the measurement capability of their suppliers – and that the specification is not “gold plated” by a non-specialist tightening a spec to an unachievable level “just in case.”

In conclusion, as you encounter the need for higher-frequency measurement in PCB fabrication, take care to understand the capabilities of not only the measurement equipment, but also the interconnecting cables, precision probes, and the design of the test PCB itself. In a future column I’ll be discussing more on test sample design, including R&R, and how in order to achieve the best gage reproducibility and reliability in your high-frequency measurements, you need to work with your supplier and your customer to ensure everyone has a thorough understanding of the measurement equipment and the test piece design requirements.

*At higher frequencies even national standard calibration artefacts can have surprisingly high uncertainties—especially to the engineer who is normally accustomed to low frequency measurements.

This column appeared in the July 2012 issue of The PCB Magazine.