# Differential stripline characteristic impedance measurement

How would you measure the characteristic impedance of a differential pair (stripline) if you don't have access to a Vector Network analyzer? I have access to an oscilloscope and a signal generator.

Most of the traces are buried, but I can desolder parts at both ends of the traces to have access to the pads connecting to the traces.

A: The VNA (and/or Spectrum Analyzer) is designed to show impedance over frequency, while the oscilloscope is designed to show voltage over time. These are fundamentally different, yet also the flip-side of the same proverbial coin. So an oscilloscope will not give you the same type of information that a VNA would.

But it can give you enough information to make a partial impedance match, provided the 'scope and function generator are fast enough, and only through trial-and-error.

Desolder parts at the ends. Connect the function generator (with a matching stripline impedance; 50Ω/100Ω whatever), to the input pair. Terminate the pair at the other end with similar resistors. Probe using the oscilloscope at the injection point and start injecting squarewave pulses. If the termination resistors are too high or low, there will be reflections visible. Change the termination resistor values and try again. When the termination resistors match the line impedance, those reflections will disappear. The value of those resistors is the characteristic impedance of the line.

Note, this will not work so well for high frequencies due to parasitic inductance and capacitance, which add two more dimensions to the problem.

• Yeah no. Characteristic impedance is not purely resistive due to losses, that's why we say "impedance" and not "resistance". And iterative guessing without some analysis of the signal and whether it means the current attempt is too high or too low is inefficient enough for exploring a single degree of freedom, for two degrees of freedom it's outright insane. Commented Apr 1, 2022 at 15:31
• You are correct in that impedance is not purely resistive, and the proper tools are the best way to go. But if you only have an oscilloscope and pulse generator, this is the only way to improve matching. Yeah, it's likely going to be ineffective for gigabits... but if you're targeting that kind of speed then you can afford the tens of thousands of dollars for proper tools, right? Commented Apr 1, 2022 at 16:29
• I'm not saying that an iterative method of reducing reflection can't be the best way of making do with the equipment available, I'm saying that somewhat more guidance is needed to explore in a two-DOF search space than blindly "change the resistor values". The next value to use should be informed by some analysis of the observed reflection. Commented Apr 1, 2022 at 16:39
• Well, if somebody's going to use this technique, I would hope they have enough understanding of basic SI principles to know whether to increase or decrease the value of the termination resistor based on the observed waveform (under damped, over damped, ...). Commented Apr 1, 2022 at 16:59
• Thank you all for your responses. I would not mind soldering a termination resistor, but iterating may risk damaging the PCB. To clarify, I was wondering if there was an analytical way, say feeding in a 10MHz sine wave, then a 100MHz sine wave via 50 Ohm coaxes (similar to the asymmetric / single ended driver solution in hparchive.com/seminar_notes/a-210.pdf), measure amplitudes and phase shifts between source, insertion point and termination, and somehow deduce the impedance? Commented Apr 2, 2022 at 20:38

If you have a pulse generator with short rise time (several times shorter than transmission line length) and high speed oscilloscope, you can make time domain reflectometry (TDR) measurement and calculate characteristic impedance. TDR setup is quite simple for single ended transmission lines, but for differential lines it is not so straightforward. Read for example this application note http://hparchive.com/seminar_notes/a-210.pdf .

• Arseny, thank you for your response and the link, I really appreciate it. The traces are about 10cm long, so the corresponding propagation delay is roughly 670ps. My signal generator tops out at 100MHz, rise / fall times in the 1ns range. Commented Apr 2, 2022 at 20:33

Measurements to useful accuracy will be difficult with your equipment. You could make a Jim Williams pulse generator AN-47 (see Appendix D), but producing a balanced output with good accuracy seems difficult.

Do you have a microscope? If so, butcher one board and make a cross-section where you can measure the dimensions (ground plane spacing, track width and spacing), and use the known mathematical approximations. If you have only one board and cannot risk damaging it, you need serious equipment.

You may have enough track for static capacitance measurements to be useful, but I guess the issue will be whether you can measure the coupling capacitance accurately or whether end strays will dominate.