Probing Two Points
High-Speed Digital Design Online Newsletter: Vol. 5 Issue 12
September 18, 2002
Electronics & Telecoms Manager
Professional Development Centre, University of Oxford
TO: Students of previous CPD courses at Oxford University
Welcome to the CPD newsletter.
Each issue presents a brief technical paper from one of the instructors in our "High-Speed Digital Engineering" series (see listing at the bottom of this page).
The newsletter is a great way to keep up-to-date with the latest trends and techniques in High-Speed Digital Engineering. Through the newsletters, you can read questions and answers from other students that share an interest in your area of expertise.
I hope you enjoy reading, and thank you for previously attending an Oxford CPD engineering course.
Probing Two Points
Christian Carlen writes:
How can I accurately measure the skew between two clock-signals that are physically 20 inches apart? If I try to connect both signals to the same oscilloscope, I have to choose one ground-reference- point. As a result, I have created the problem of a very long ground-connection for at least one of the two probes. This has me concerned about ringing and delays.
Furthermore, the delay along the ground plane would be somewhat like (20 in.)*(180 ps/in.)=3.6 ns. I am concerned this delay will affect my measurement.
How about this: Suppose I build a magnetic-field probing loop with one probe by connecting its ground-wire to its probe tip. I can use this structure to probe the first clock signal (without having to touch ground). Using this first probe I can try to catch the rising edge of the first clock signal and use it to trigger the second probe that is directly connected to the second clock-signal. The accuracy of this approach is not great.
Another idea I had was to bring the two clock signals with two coax-cables physically together to a single point where I can attach both probes with short ground-wires. If I make both coaxial cables the same length, the loading and the added delay shouldn't affect the skew-measurement.
How do I resolve this problem?
The answer is simple: Just ground each probe near its respective point of measurement.
The reason that works is a little complicated: the natural inductance of each probe cable acts like a kind of primitive balun, which is what makes the circuit work.
A balun is a type of transformer with two windings: a signal winding and a ground winding. The signal and ground wires go through the core together, entering on the left side and coming out the right (in the ASCII-text picture the wavy symbols in the middle represent windings, not resistors).
The signal and ground wires on the left side of the balun attach to your circuit. These wires then pass through the core of the balun and appear again as signal and ground connections on the right side. Because the signal wire passes directly through the core from left to right, you can pass a DC signal through the balun.
MAGNETIC CORE signal left -----/\/\/\----- signal right ground left -----/\/\/\----- ground right
If you place equal but opposite currents on the (+) signal and (-) ground wires (as you would when looking at a differential signal), the two signals induce opposite magnetizations in the core. As a result the balun has no effect on differential signals--they pass right through unscathed.
The balun does affect common-mode signals, because common-mode signal and ground currents both pass through the core in the same direction (left- to-right). The full inductance of the balun impedes the flow of common-mode signals passing from left to right.
Your probe cable acts like a kind of primitive balun. It doesn't have a magnetic core, but it does have two wires (signal and ground) that proceed from left to right through a long length of cable.
Any differential-mode signals (i.e., signals present between the signal and ground terminals of the probe at the left end of the cable) pass through the circuit undisturbed.
Any common-mode signals passing from left to right through the probe (with the same direction of current on both wires) are impeded by the full inductance of the circuit. That inductance is defined by the length and configuration of the probe shield, and the return path flowing through the scope ground connection to the earth and back to your circuit. In typical configurations with 1- meter probes the common-mode inductance of your probe amounts to a couple of microhenries. At a frequency of 100 MHz, an inductance of 1 microhenry has an impedance of approximately 600 ohms.
The common-mode (balun) inductance of your probe impedes the flow of common-mode currents on the shield of the probe. This allows the probe ground to touch ground points in your circuit having high- frequency potentials slightly different from the ground potential on your scope without drawing large amounts of current through the probe shield.
When you use two probes, the common-mode (balun) inductance allows you to probe two points, using a different local ground reference at each point, without drawing excessive currents on the probe shield wires. The differential signal on each probe (difference between signal and ground at the probe tip) comes through, while common-mode signals (related to differences in the ground potentials at the two local ground reference points) have little influence on the measured signal.
The reason you care about common-mode currents on the shield is that these currents can induce noise into the measured signal.
what you should do
First, working with each probe independently, connect the probe signal pin to its own ground. Now WITHOUT TOUCHING YOUR SYSTEM just wave the self- grounded probe around near your board. You should see no signal on the scope. If you see anything, it's because you have left too large an exposed area between the probe signal pin and its ground connection, and that loop is picking up local magnetic fields. Use a shorter ground.
Second, while keeping the probe signal pin connected to its own ground, touch the probe ground onto the digital logic ground of your system near the point of measurement. If you see any new noise, it's because your system is pumping current through the ground shield of the probe. To fix this problem you must connect a shorter (better) ground between your system and the scope, to encourage the ground currents to flow through your new, shorter connection (instead of through the ground shield of the probe). In extreme cases, people sometimes pass the probe cable through a giant ferrite bead to reduce this source of noise. This procedure increases the common-mode (balun) inductance of your probe.
Third, while keeping both probe signal pins connected to their own respective grounds, touch both probe grounds simultaneous onto the digital logic ground of your system, placing the probe grounds near their respective points of measurement. This should cause little (if any) new noise.
These three tests measure the noise floor of your probing arrangement. If the residual common-mode currents on your probe shields are going to cause any difficulties, you'll see it during these tests.
If you see an acceptably low noise floor during the three tests you may then proceed to probe the actual signals on your board with the confidence of knowing that any wiggles you see are *actually present* in the clock signals, and not the result of noise is being picked up by the probe configuration.
The delay of each probe depends only on the propagation delay of the differential component of the measured signal. The differential component is the difference between the local signal and the local ground reference. The fact that the ground reference point for measurement one and the ground reference for measurement two are different is immaterial.
If the three tests reveal a noise floor too large to permit a clear view of your signal, you may want to try using differential probes. A differential probe used with the (+) wire on the signal and the (-) wire on the local ground draws very little current on its shield, because there is no direct low-impedance connection between the board and the probe shield. A pair of differential probes can therefore be used to measure signals at two different points that have wildly divergent ground potentials.
Whatever type of probe you use, after you measure the skew, swap the probes re-measure. Then average the two measurements. This procedure nulls out any systematic differences in the probes or vertical amplifier channels of your oscilloscope.
Dr. Howard Johnson
Call for Papers
Dr. Johnson will chair a technical session (theme 10, #38, "Signal Integrity Analysis and Design") at the IEEE International Symposium on EMC, to be held May 11-16, 2003 in Istanbul, Turkey.
His session will explore regions of common interest between the subjects of EMC and signal integrity.
Suggested topics include the control of ringing, crosstalk, ground bounce, power supply noise, signal rise time, stray returning signal currents, split ground planes, noise isolation barriers, and the propagation of extremely high-frequency waveforms.
The four instructors in Oxford's High-Speed Digital Engineering series are:
High-Speed Digital Design: Howard Johnson, PhD, is the author of High-Speed Digital Design: A Handbook of Black Magic, chief technical editor of standards for Fast Ethernet and Gigabit Ethernet, Signal Integrity columnist for EDN magazine, and longtime contributor to the Oxford Signal Integrity curriculum. Through his consulting and teaching activities he interacts with over 1000 engineers annually (www.sigcon.com).
High-Frequency Measurements: Doug Smith, MSEE Cal Tech, is the author of High-Frequency Measurements and Noise in Electronic Systems. His early career at Bell Laboratories led him to eventually become a member of the IEEE EMC Board of Directors and a prolific writer (www.emcesd.com).
Signal Integrity, Right by Design: (new for 2002) Ed Sayre, PhD, began his career helping to invent the method-of-moments electromagnetic simulation technique and teaching electronics at the University level. He now serves as the lead consultant at North East Systems Associates, one of the largest independent consulting firms specializing in the design of high-speed digital systems (www.nesa.com).
PCB Design for Real-World EMI Control: (new for 2003) Bruce Archambeault, PhD, is the author of PCB Design for Real-World EMI Control and also the EMI/EMC Computational Modeling Handbook. Bruce works at IBM where he acts as an internal consultant to numerous EMC projects. He is a past Associate Editor for the IEEE Transactions on Electromagnetic Compatibility and a Board of Directors member of the Applied Computational Electromagnetics Society.