4
\$\begingroup\$

I have a signal source that sends fast pulsing square waves in the MHz range (encoder pulses from a motor) down a long transmission line to a receiver. I have tried many solutions and different types of terminations to mitigate the transmission line effects that showed up as overshoot/ringing on rising/falling edges (that exceed the receiver's absolute maximum input voltages), as well as crosstalk coupling between channels.

The solution that worked the best for this problem turned out to be diode clamps using 4.7V zeners arranged in the schematic below:

schematic

simulate this circuit – Schematic created using CircuitLab

The effect of these diode clamps can be seen in the below oscilloscope screenshot, where the clamped purple trace is juxtaposed with the unclamped yellow trace. As you can see, the rising/falling edges get clamped to a clean corner for the purple trace, which is desirable. clamped unclamped juxtaposition

However, I am worried about the tradeoffs and downsides to doing this. Few things I have in mind:

  • No series current limiting resistor.
    • The encoder (with a line driver) is specced to source a maximum of 20 mA. Using a 4.7/0.02=235≈220 ohm resistor negated the clamping effects completely. Even using a lower 100 ohm resistor erased any clamping effects. I think it is hard to choose a good resistance to use because I cannot identify the other impedances accurately (encoder output impedance, transmission line impedance, input impedance, zener impedance)
    • However, I am thinking this might be fine because of the origin of these voltage spikes are from impedance mismatch reflections, so its not like its drawing current directly from the encoder driver.
  • Ambient current draw. Say one of the channels ends up at logic high 5V, the zener will conduct and sink away current continuously.

Are there any things I'm missing? What do you guys think of this solution?

EDIT1 (additional information requested by @AnalogKid):

  1. The "long transmission line" is a 6ft long unshielded PVC encapsulated cable. There are no twisted pairs and all signals are single ended. However I have estimated the impedance of the transmission line to be around 50-100 Ohms (from measuring inductance/capacitance with LCR meter, then corroborated by trying different parallel termination resistances). I have tried parallel (and AC) termination but ditched it because it didn't get rid of falling edge oscillations, and I didn't like that extra power was dissipated in the parallel path. Series termination cannot be implemented at this time.

  2. The driving device is a Maxon ENX 10 EASY encoder with a built in line driver.

  3. The driver is not source terminated. It is very difficult to source terminate because the mechanical design of that portion is frozen, although could possibly be a last resort.

  4. I am mainly worried about the ringing damaging the receiver (LS7366R) as the oscillations exceed the absolute maximum input voltage rating of the IC. If the energy of the ringing is so small, do you think it is OK to just let it happen? FWIW the system has been working perfectly fine over hours of testing.

  5. I am using the 4" long alligator clips. Although I did use the short ground springs and compared it to the 4" long alligator clips and found the difference to be negligible, so I stuck with the alligator clips due to ease of use. Perhaps I can revisit this.

\$\endgroup\$
5
  • \$\begingroup\$ Please update your question with more information. | What is the makeup of the "long transmission line"? Coax / unshielded twisted pair (UTP) / shielded twisted pair (STP) / lamp cord - ? If it is something with a known characteristic impedance, have you tried terminating the line in a resistive (at 10 MHz) terminator? | What is the part number of the driving device, or what is the circuit if discrete? | Is the driving point source-terminated? \$\endgroup\$
    – AnalogKid
    Oct 24, 2023 at 23:05
  • \$\begingroup\$ Also, all of the energy in the wave - baseband, harmonics, reflections, ringing, whatever - comes from the driver, simply because there is no other source of current in the circuit. The ringing is not dangerous to the driver because the energy in the ringing is so small. | Finally, how long are the ground leads on the scope probes? A 4" ground lead has enough inductance to interact with the circuit. \$\endgroup\$
    – AnalogKid
    Oct 24, 2023 at 23:05
  • \$\begingroup\$ Thanks for your time. I have updated the question with all the relevant details mentioned above. The one thing I wanted to get your opinion on was whether you think one could just leave these oscillations alone (see point 4 above). \$\endgroup\$
    – petemoss0
    Oct 25, 2023 at 0:12
  • 1
    \$\begingroup\$ LSI parts are usually robust, but this one is CMOS, and I did not see anything in the datasheet about input ESD or transient protection. Without it, over- or under-voltage ringing could cause latch-up. Contact LSI and see if they will share a partial schematic showing any input protection. \$\endgroup\$
    – AnalogKid
    Oct 25, 2023 at 3:16
  • \$\begingroup\$ Thanks for the suggestion to contact LSI, not sure I would've done that. Here's their response: "All inputs of LS7366R have clamping diodes to both VDD and VSS. These diodes may get forward biased if an input overshoots VDD by +0.7V or undershoots VSS by -0.7V. In extreme cases forward bias junction currents can lead to latch-up. But the overshoots and undershoots in your case are not strong enough in terms of both amplitude and duration to cause a problem. For additional precaution however, you may add current limiting series resistors (2K) between the driver outputs and the LS7366R inputs." \$\endgroup\$
    – petemoss0
    Oct 25, 2023 at 19:52

3 Answers 3

4
\$\begingroup\$

First, Zener diodes, especially low-voltage ones, have what we call a "soft knee": they conduct noticeably below the breakdown voltage. Together with the input series resistance, they form a non-linear voltage divider, distorting the signal. This current is temperature dependent, meaning that a circuit that is acceptable at room temperature may be unacceptable at high temperatures.

Secondly, they have a large capacitance, which will slow down fast signals.

A better solution is to use PIN diodes in place of D1 and D2. If you want to clamp to 0 V and 5 V, then connect the far end of the diodes to 0.7 V and 4.3 V (instead of 0 V and 5 V).

schematic

simulate this circuit – Schematic created using CircuitLab

\$\endgroup\$
9
  • 1
    \$\begingroup\$ thanks for your help and information. I'm a bit too far in the design process to add this many components, as well as building a 0.7 V and 4.3 V source, but great information otherwise. side unrelated note I see you are the author of connectorbook - it was a helpful resource during design, thank you for that! \$\endgroup\$
    – petemoss0
    Oct 25, 2023 at 0:18
  • \$\begingroup\$ " it was a helpful resource". So glad to hear! \$\endgroup\$ Oct 25, 2023 at 11:57
  • 1
    \$\begingroup\$ You got it. That circuit clamps pulses below 0 and above 5 V into the capacitors. It would not clamp a DC of, say, 5.2 V, because it would charge the capacitor to a higher voltage. R1 just limits the current in the diodes. A lower value provides better clamping of longer pulses. A higher value reduces the supply current. \$\endgroup\$ Oct 25, 2023 at 16:15
  • 1
    \$\begingroup\$ I'd say best solution for receiving RS422 signals properly is an actual RS422 receiver, not a bunch of discretes for clamping. \$\endgroup\$
    – Justme
    Oct 25, 2023 at 21:20
  • 1
    \$\begingroup\$ OP still hasn't. I had to dig that out from encoder data sheet. \$\endgroup\$
    – Justme
    Oct 26, 2023 at 14:35
2
\$\begingroup\$

Your encoder has a RS422 differential output driver.

It's not intended to be directly connected to a logic IC.

Your problem is most easily solved by having a proper RS422 receiver as a physical interface, and properly terminating the differential wires with transmission line impedance, and if possible, replacing the random cabling with something suitable for differential signals, such as CAT5 cable with twisted pairs.

The output of RS422 receiver is a clean logic signal you can connect to the counter IC.

So basically, you are trying to fix a problem that will not need fixing if done the right way.

\$\endgroup\$
2
  • \$\begingroup\$ Completely agree with you, but logistical limitations prevent me from doing an entire overhaul of the system, so rather than dealing with the problem directly, I can only mitigate the symptoms. \$\endgroup\$
    – petemoss0
    Oct 25, 2023 at 21:29
  • \$\begingroup\$ As a workaround, make and add-on PCB with RS422 receiver before the counter to deal with the problem indirectly. \$\endgroup\$
    – Justme
    Oct 25, 2023 at 21:32
1
\$\begingroup\$

I have accepted @DavideAndrea's answer to this question because it provides a general alternative solution to the problem posed in the title.

However, in my specific case, the signal receiver is a LS7366R, so I reached out to the designer of the chip, LSI/CSI Computer Systems, thanks to the suggestion of @AnalogKid. The following is the exchange verbatim:

Q: "We are using your "LS7366R" in a prototype. Due to transmission line effects, we are seeing ringing on rising/falling edges that exceed the absolute maximum ratings of "Vss - 0.3 to VDD + 0.3". This ringing ranged from Vss - 1 to VDD + 1 and is very fast (settles within 500 ns).

Solutions that mitigate this ringing all have undesirable tradeoffs, so we are wondering if the chip had some sort of under/overvoltage input protection, or if there is any other helpful information that can be provided, such as the failure modes we should expect if these absolute maximum ratings are exceeded. It should also be noted that the system has been working perfectly under hours of test."

A: "All inputs of LS7366R have clamping diodes to both VDD and VSS. These diodes may get forward biased if an input overshoots VDD by +0.7V or undershoots VSS by -0.7V. In extreme cases forward bias junction currents can lead to latch-up. But the overshoots and undershoots in your case are not strong enough in terms of both amplitude and duration to cause a problem. For additional precaution however, you may add current limiting series resistors (2K) between the driver outputs and the LS7366R inputs."

\$\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.