I’ve often wondered why some harmonic emissions were unrelated to obvious clock oscillators, system clock frequencies, or even SMPS frequencies. The answer is likely related to the stability of power supplies (both analog and switching) as well as the power distribution networks (PDNs) of the equipment under test (EUT).
During DesignCon 2015, I happened to run into Steven Sandler, an expert in power integrity and president of Picotest. His recent book, Power Integrity – Measuring, Optimizing, and Troubleshooting Power Related Parameters in Electronic Systems, was published in 2014 by McGraw-Hill. It’s an impressive 335 page, full color, reference book jam-packed with practical measurement techniques on all things “power integrity”. This would include characterizing the PDN, accurately measuring impedance, stability, power supply rejection ratio (PSRR), step load response, ripple/noise, edges, and (for the EMI engineer) near field probing of EMI.
THE IMPORTANCE OF LOOP STABILITY
The reason this whole discipline interests me is that issue of power supply stability – both linear regulators and switching – can create harmonics, not necessarily obvious from the usual sources, resulting in radiated and conducted emissions. Poor stability (reduced gain and phase margins) and the associated peaking in the regulator’s noise path often result in a decaying ring on SMPS switches. In Sandler’s words:
“While such decayed responses seem benign, the harmonic content of the signal is quite rich, with the frequency range primarily limited by the edge speed of the impulse and and the duty cycle of the impulse signal. An example of such a ring, and the associated harmonic content, is shown in [Figure 1]. In this example, we can see the operating current’s impact on the ringing frequency and the very rich harmonic content associated with the ringing. The separation between the harmonic spurs is the reputation rate of the impulse. This may seem counterintuitive, but the lower the impulse frequency the closer together the spurs will be. The result is a very large number of noise signals over a large frequency range. These signals then travel through the system looking for resonant response peaks that coincide with one of these harmonic noise spurs. Poor stability is the most common cause of system noise problems.” (Ref: page 104)
CONTENT
Beyond the EMI issues, the book addresses many other topics:
Chapters 1 through 3 introduce basic measurements and measurement philosophy, as well as measurement fundamentals, such as sensitivity, noise floor, dynamic range, averaging, and use of attenuators and preamplifiers. Chapter 3 closes with the various measurement domains, such as frequency, time, gain/phase, and s-parameters.
Chapter 4 includes examples of the range of test instruments used for power integrity measurements from a variety of manufacturers.
Chapter 5 describes the various probes, injectors and interconnects used throughout the example measurements in later chapters. Several of these were custom designs developed and sold by the author. More on this in a bit.
Chapter 6 discusses the distributed system, including noise paths within voltage regulators, control loop stability, PDNs, and how poor stability propagates through the system.
Chapters 8 through 15 discuss specific measurements related to power integrity and include several practical measurement examples and equipment setups. Topics include how to measure:
- Impedance – single-port, two-port, current injection, and impedance adapters
- Stability – control loop basics, gain/phase margin, bode/Nyquist plots, open-loop and closed-loop measurements, small/large signal – including a unique method of non-invasive measurement
- Power supply rejection ratio (PSRR) – in-circuit, out-of-circuit, modulating the input, choosing the measurement domain (VNA, spectrum analyzer, oscilloscope)
- Reverse transfer and crosstalk – series linear regulators, shunt regulators, point of load (POL) regulators, op-amps, modulating the output current, measuring the input current and voltage
- Step load response – generating the transient and measuring the response
- Ripple and noise – selecting the measurement method, choosing and connecting the equipment, averaging and filtering
- Measuring edges – relating bandwidth and rise time, sampling rate and interleaving, interpolation, effects of high frequency losses, the criticality of the probe connection, printed circuit board issues
- Troubleshooting with near field probes – the basics of emissions, near-field probes, probe orientation, measuring instruments, spectrum gating
- High frequency impedance measurement – time domain, calibration, reference plane, setting TDR pulse rise time, interpreting TDR measurements, estimating inductance and capacitance, s-parameter measurements
ONE ISSUE?
One issue I realized early in my reading was that many of the measurement techniques required the use of the probing and instrument products designed and manufactured by Picotest – a potential negative aspect of the book. I contacted Sandler regarding this issue to allow him to address this point. Excerpting from his reply:
“Yes, most measurements require at least one of our products. This could be construed as self-serving, but that wasn’t really the intent. We make these products because nobody else does. Unfortunately, without them, most data isn’t very good or the measurement isn’t possible. As a result the instrument companies refer customers to us and we also end up doing a lot with the semiconductor companies.”
“The only thing that I find shocking is that I had to design and build these myself; nobody made them. If you were to try some of the measurements without our products, you’d see that the results are somewhere between “somewhat off” to “just downright wrong”! That’s what I’m trying to fix. I’m trying to teach that the equipment doesn’t have to be expensive, but the answers need to be correct! I tried to include good examples of what happens if you leave out the coaxial transformer, as one example. You’ll find it in the impedance chapter (the low frequency low impedance data isn’t correct without it) and also in the PSRR chapter (the DC ground loop has significant influence at low frequency).”
“I think it is also pretty surprising that engineers are just supposed to “figure out how to make these measurements” while nobody teaches HOW. That was my main reason for writing the book.”
SUMMARY
In summary, I feel that power integrity issues like loop stability are additional ways of “peeling the onion” on EMI problems. I firmly believe that most EMI issues can be controlled by good circuit design and PC board layout. Properly-designed circuits should rarely need the typical “band aid” fixes like shielding and the addition of ferrite beads at the last minute in the product design process. Sandler has written the definitive book on power integrity and optimizing the design of PDNs and power supply design is a step in the right direction towards controlling EMI and improving product performance. Did I mention “full color”? This is the first McGraw-Hill engineering book to be printed in full color. Highly recommended.
FOR MORE:
Picotest – probes and connectivity products
Picotest also offers training seminars on these measurement techniques
Take a video tour of Steve Sandler’s test lab
Steven Sandler’s profile and published articles on EDN.com
Sample Keysight app-note showing use of Picotest products
Kenneth Wyatt, Sr. EMC Engineer, Wyatt Technical Services, Inc. holds degrees in biology and electronic engineering and has worked as a product development engineer for 10 years for various aerospace firms on projects ranging from DC-to-DC power converters, to RF and microwave systems for shipboard and space platforms. For over 20 years, he worked as a senior EMC engineer for Hewlett-Packard and Agilent Technologies in Colorado Springs where he provided comprehensive EMC design and troubleshooting services. Kenneth is a senior member of the IEEE and a long time member of the EMC Society where he served as their official photographer for 10 years. He may be contacted at ken@emc-seminars.com.
