1
\$\begingroup\$

This application note seems to suggest that for high-speed signals clamping diodes can be used to deal with overshoots instead of termination:

enter image description here

To manage signal integrity issues and protect the input pin, follow the guidelines in this document [...]

Guideline: Use Internal PCI Clamp Diode on the Pin

The PCI clamp diode can sufficiently clamp voltage overshoot to within the DC and AC input voltage specifications when the bank supply voltage (VCCIO) is 2.5 V or 3.0 V.

I realize this will prevent the overshoot voltage from damaging the device, but I assume that "to manage signal integrity issues" means it is supposed to prevent signal reflections and ringing as well. Does it? AFAIK clamping diodes are very widespread in modern ICs, yet external termination resistors are common practice. What benefits does proper termination have, compared to this clamping approach?

\$\endgroup\$
7
  • 1
    \$\begingroup\$ I wonder if the clamping diode guideline is meant only as a "protect the input pin" thing, and not a "manage signal integrity issues" thing. The way it's worded would make it seem like not all of the "guidelines" are for both purposes. \$\endgroup\$
    – Hearth
    Commented May 10, 2017 at 14:54
  • \$\begingroup\$ @Felthry maybe, that's why I'm asking. The claming diode should absorb some of the overshoot, so it will reduce the reflection somewhat, and I'm wondering how much is that "somewhat". \$\endgroup\$ Commented May 10, 2017 at 14:59
  • \$\begingroup\$ @Felthry Also, Table 1 in than AN says "No action required" for the case above (in contrast with "Apply series termination" for other cases), which I tend to read as "Termination not needed". \$\endgroup\$ Commented May 10, 2017 at 15:03
  • \$\begingroup\$ There are many cases where "proper termination" is impossible. Generally trace impedance is somewhere between 25 and 90 Ohms. You cannot just put a 25-90 Ohm resistor on an input because the required DC current will over-tax the driver. Also, you sometimes have complex branching topologies or bi-directional signals. So other strategies have been devised. The clamp will DEFINITELY improve signal integrity by limiting the amplitude of the overshoot. Termination resistors are usually for lower amplitude differential signals, or RF or analog video. \$\endgroup\$
    – user57037
    Commented May 10, 2017 at 15:27
  • 1
    \$\begingroup\$ Oh, yes, you can do that, of course. Should be standard procedure on any signal with a fast edge. You can use 0 Ohms to start, and adjust as needed. I have found that where applicable, the internal drive strength is a better way to control the edge rate than the external resistor, but it is best to keep all options open. But series termination does not guarantee no overshoot. \$\endgroup\$
    – user57037
    Commented May 10, 2017 at 17:48

1 Answer 1

1
\$\begingroup\$

When advanced low Vdd CMOS uses RdsOn of 25 Ohms at 3.3V with 0.2 to 2 ns rise times, designers often are unaware of need to have controlled impedance tracks and results in ringing at half wavelength of track length.

  • low power Schottky diodes with small capacitance are sometimes used to provide a dynamic load impedance of around this driver impedance range to at least provide some matched load to the overshoot.

  • when the transmission line is mostly inductive from poor coupling to gnd tracks or gnd plane near same gap as track width, and with CMOS input Z can be x pF and 1GOhm, ringing and UNDERSHOOT can cause glitches.

analogy

The clamp diodes are like a stiffer rubber end stop to a spring (inductive track) with a weak shock absorber in zero gravity going over a pot hole and the car still bottoms out and then jumps up but with less bouncing shock waves. A properly tuned shock absorber is like a properly tuned transmission line to give controlled step response with minimal overshoot/undershoot and ringing.

No overshoot is a theoretical perfect transmission line matched to between source and terminator which is as I said above can be 25 ohms -/+50% with temp and 10% supply variation. so often terminated with 50 Ohms. with tracks matching the load. But skin effects and dynamic dielectric constant reduction >1Ghz causes some imperfections .

A diode is a poor man's choice for lack of impedance controlled on track layout and design,pick any; coplanar, stripline, microstrip! for 50 Ohm single ended and 100 Ohm differential.

\$\endgroup\$
1
  • \$\begingroup\$ Thanks, but I'm not sure I understand your analogy. A car bouncing off the road in zero gravity will necessarily jump up, since there's no force to push it back to the road. \$\endgroup\$ Commented May 10, 2017 at 17:14

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.