6
\$\begingroup\$

This is a SPI Clock. 6.25MHz frequency. I see something like a reflection. I don't think it's from cross talk on my board from another signal, not sure.

Added as requested: This is my system diagram (notice I have a stub). Note that the diagram ends at the "Connector". There is a cable to the SPI Slave beyond the "Connector" but that is not even connected when doing the measurement.

system diagram

What kind of terminations should I try to get rid of this reflection? Should I try to play around with the termination values or try to remove the stub?

oscilloscope trace

\$\endgroup\$
  • 1
    \$\begingroup\$ You might want to tell a bit more about the board. \$\endgroup\$ – pipe Sep 15 '16 at 23:23
  • 2
    \$\begingroup\$ (a) That doesn't look like any reflection problem I've seen before; (b) More details needed of the hardware at each end and the physical SPI link (length, cable type etc); (c) Waveform rise time is quicker than fall time, suggesting unequal drive - odd for normal SPI - again, more hardware & link info needed; (d) Slow down the serial clock (and signal) frequency & review what the waveform looks like then. If that glitch stays, then this suggests it isn't related to frequency and hence not to termination. (e) What's the history of this design - new? Or tested but new modification? Or ...? \$\endgroup\$ – SamGibson Sep 15 '16 at 23:24
  • 2
    \$\begingroup\$ Do you have good supply bypassing on the MAX14611? What do you see upstream of the resistor? The timing is such that it might have to do with when the MAX14611 output changes. Can you get that on the scope on a 2nd channel? Does your board have power and ground planes, or just a few small traces threading their way around? \$\endgroup\$ – Chris Stratton Sep 16 '16 at 15:09
  • 4
    \$\begingroup\$ Given that the glitch is much faster than the signal edges, this is definitely NOT a transmission-line termination issue on this particular wire -- the glitch is being coupled from some other nearby signal. As @ChrisStratton says, possibly an output of the Maxim chip. It might also just be contamination of the measurement (e.g., bad grounding on the scope probe) -- is it causing any actual functional problem? \$\endgroup\$ – Dave Tweed Sep 16 '16 at 16:34
  • 1
    \$\begingroup\$ Measure ground noise. And if exists, improve probe method by removing ground clip and tip to use only 10:1 probe pin and barrel between short test wire added 1/4" apart. If ground noise disappear, so too will glitch. Use 2 ground pins to verify loop inductance has been eliminated using this low inductance probe method. The ground current is due to output capacitance of cable and your layout and driver ground current or stray EM coupling and impedances involved. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Sep 17 '16 at 16:12
2
\$\begingroup\$

I had similar issues in one of my designs, but that issue was related to driving 5V-tolerant input (as it was declared by the datasheet - device itself is 3V3) with 5V TTL signal.

I tried a number of termination circuits, starting from series termination as you have it now on the circuit, ending with series resistor and capacitor. With network of resistor and capacitor I was able to smooth the spike, but it affected edges badly.

During the course of my investigation, I got to know that series termination resistor should be as close to source as possible. In your case it is close to destination. This is the first thing I would change.

Next, "5V tolerance" in the mux's datasheet strains me. Such tolerance, to my understanding, while is great in terms on extending application of the device, goes at the expense of additional internal components which would make device "tolerant" (e.g. diodes, resistors). These components may cause some bad effect (however I can only tell you basing on my case), including kind of changes in impedance and thus signal distortions and spikes.

My second advice, if first will not help up to your requirements, is if you can find native 3V3 I/O mux and use it instead.

\$\endgroup\$
2
\$\begingroup\$

The glitch is too short and the signal too low frequency for this to be a reflection. This really looks like something the 3.3 V to 5 V level translator input is doing to the signal. I don't know for sure what is going on, but I would try two things:

  • Move the resistor close to the source, not the destination. For these kinds of resistors, you should always do this.

  • Lower the resistance, or lose it all together. A 2 inch digital trace on a board with 6 MHz clock shouldn't need a series resistor in the first place.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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