2
\$\begingroup\$

This question is not really a problem i am having but a thought that want to get off my mind.

Usually on a reading on a oscilloscope especially on a square wave and sometime triangle waves, there are these spikes or ringing.

See sample images from google below

enter image description here enter image description here enter image description here

I too have sometimes observed this when i am using an oscilloscope but i pay it not mind. From what i read it would seem that this is sometimes caused by bad probing. It got me thinking is this a oscilloscope specific phenomenon? if its caused by probing that would mean that whats actually happening on those lines are very clean signal (maybe not as steep voltage change but there is no ringing)

\$\endgroup\$
7
  • \$\begingroup\$ Probing can also cause ringing. \$\endgroup\$
    – Andy aka
    Jan 16, 2020 at 16:24
  • \$\begingroup\$ You'll want to google "Gibb's phenomenon" or "Gibb's effect". \$\endgroup\$ Jan 16, 2020 at 16:34
  • 1
    \$\begingroup\$ @MarcusMüller Gibb's is a mathematical artifact, I'm not sure it's very helpful in understanding real ringing. You'll never see that kind of oscillation at the edges in a real oscilloscope trace because it would require a transfer function that's not realizable (the ringing would have to begin before the edge so it's not causal). \$\endgroup\$ Jan 16, 2020 at 16:58
  • \$\begingroup\$ Ringing period is roughly 11ns, or 91 MHz. Many X10 attenuator probe tips resonate at this frequency, some include a damping resistor to reduce it, others don't. Are you using a X10 attenuator probe, with its ~6-inch alligator ground clip to make this measurement? If so, then what you see may very well differ from what's really there. As Tony says, additional measurements are needed to discriminate real from fake. \$\endgroup\$
    – glen_geek
    Jan 16, 2020 at 17:20
  • 1
    \$\begingroup\$ @MarcusMüller, Notice that, in the traces that are shown, the overshoot and ringing mainly happens at the end of each transition. (The first image shows just the tiniest blip of an overshoot at the beginnings.) If the phenomenon had anything to do with Gibb, then I would expect to see symmetry between the start and the end of each transition. \$\endgroup\$ Jan 16, 2020 at 22:55

3 Answers 3

6
\$\begingroup\$

That's called ringing, and it's an effect of inductance and capacitance in the circuit (often just parasitic). Picture it this way...the wire from the driver to the receiver has intrinsic inductance, and most digital inputs have [gate] capacitance.

Once the output starts driving the line, current builds up in the (inductive) trace until the capacitance fills up to the same voltage as the driver. There's still a current flowing through that inductance, though, so it continues pumping the capacitor higher until the current stops...now the voltage drives the current back to the driver. That's the first peak, and the next time the current stops is the first valley. Back and forth until resistance dissipates the stored energy.

Of course, the trace is actually a transmission line with distributed inductance AND capacitance, as well as a speed of light limitation, but that's about half a semester worth.

This can affect the signal integrity of the circuit when the ringing is large enough, as well as causing permanent damage to ICs over months or years of repetitive overvoltage/undervoltage.

Of course, the scope probe becomes a part of the circuit when applied, and can change the behaviour of the node.

\$\endgroup\$
2
  • \$\begingroup\$ I do a lot of pcb design for personal uses, what i am getting from what you said that might apply to my future design choices is that, scope aside; since this not a scope only phenomenon, Is to make my signal traces as small as possible and as short as possible this will reduce the ringing on the line am i correct? \$\endgroup\$
    – Jake quin
    Jan 16, 2020 at 17:22
  • \$\begingroup\$ Shorter is better. A 20-50 ohm resistor near the source helps a lot. Thicker is generally better than thin, but check your impedances as well; any wire is a transmission line in practice. Probably most important is to keep current loops small; a solid ground plane helps a lot with that. Remember current always has to return, and the closer the return path is to the signal path, the better off you'll be. \$\endgroup\$ Jan 16, 2020 at 17:48
3
\$\begingroup\$

It is certainly not specific to an oscilloscope measurement. Fast edges relative to transmission line length and any unterminated or poorly terminated line results in (real) ringing at a frequency related to the length of the line.

Rule of thumb is that the unterminated length in FR4 is about 1.5 * period of oscillation (in ns), so your relatively long ~13ns period represents about 20" of trace or somewhat more length in a badly-terminated coax cable.

A bit long for a PCB trace, but not far off a scope probe cable length.

\$\endgroup\$
2
  • \$\begingroup\$ I am having second thoughts of what unterminted in traces mean. From what i know it traces that are not connected to anything right? But that would be very uncommon for a pcb traces(exempt maybe antena traces?). I do a lot of designing pcb and i would like to follow the best design practices. Any information about good design choices even if its a small improvement is very helpful \$\endgroup\$
    – Jake quin
    Jan 16, 2020 at 17:35
  • \$\begingroup\$ @Jakequin That's a huge subject. Just a taste- sometimes we put small series resistors at the driving end of traces (eg. 33 ohms) to keep them from ringing too much. Look up microstrip and stripline. Controlled impedance and matched trace lengths are widely used for high speed digital signals. Eg. USB, HDMI, PCI and so on. resources.altium.com/pcb-design-blog/… \$\endgroup\$ Jan 16, 2020 at 18:12
1
\$\begingroup\$

This is classic Ground wire inductance MEASUREMENT ERROR.

(0.5nH/mm and probe custom low C coax capacitance ~ 60pF/m) (LC resonance > 20MHz)

Solution . Search this site for proper 10:1 scope probe RF measurement methods with tip & ring only or terminated Coax with R divider to 50 Ohms.

  • repeat experiment using above methods until you get textbook quality results with calibration square wave on DSO.
  • you do the search
  • when that is perfect and your experiment rings. Fix ALL the signal paths using coax or twisted pairs .. Or Ignore error and use Scope 20 MHz filter.

I see no one was motivated to appreciate this.

http://www.cbtricks.com/miscellaneous/tech_publications/scope/abcs_of_probes.pdf

What you expect ....

enter image description here enter image description here

But what you actually have that causes a 2nd order overshoot problem.

enter image description here

One way to improve response, either use an other R divider on board or direct to 10:1 probe with tip & ground clip removed. enter image description here

Or more convenient with nearby signal and short ground track. enter image description here

You may accept 3% overshoot within 30ns as a good measurement.

enter image description here

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.