1
\$\begingroup\$

It's frequently communicated that when working with signals in the Ghz range, routing is extremely important to signal integrity. Pretty much any inclusion of unnecessary stubs is out of the question, traces must be length matched to some fraction of the wavelength, vias backdrilled, etc.

That being said, I've been tasked with an unconventional goal of creating a high speed (1.5Ghz) signal interface along a large circular ring (e.g. a slip ring) of a maximum 3 inches in circumference. The signal will be introduced at one random point in the ring and picked up at a fixed location at another place in the ring.

Specifically, I'm not quite too sure what the implications are / if it is possible to operate at that level of speed given the self induced jitter from the phase shift. Assuming the following, are there any opinions out there on if this is possible or not? If so, what additional mitigation measures can be taken in this particular scenario?

  • Annular ring impedance matching to termination
  • Gold on gold contact to introduce signal
  • EQ/DI/OS SC IC (ReDriver) immediately after for insertion loss
  • Clocked Retimer eventually after to hopefully save the signal
\$\endgroup\$
  • \$\begingroup\$ Various optical data transfer systems are in use on rotating systems, it might be an option if you can multiplex all your signals onto a single signal. \$\endgroup\$ – KalleMP Jun 9 '18 at 8:25
  • \$\begingroup\$ Thanks for the callout Kalle. I've looked at FORJs in particular for this application but the economics don't line up. I'll keep my eye out though. \$\endgroup\$ – Smithers Jun 9 '18 at 8:29
1
\$\begingroup\$

I've done a similar system that may help you understand some of the problems. The payload data rate was 53.46 Mbps and the "ring" was about 600 mm in diameter: -

enter image description here

The important thing about the ring is that it was "balanced" and terminated. The receiver made no contact but sat a couple of mm from the ring. The ring rotated at several thousand revs per minute. Having a balanced ring means low EMI to other systems (also important).

The really clever thing about the design is that the rotating ring's coils were wound in such a way that the static receiver coil never saw a signal inversion or a cancellation point.

If you have serious alignment problems this will not work because it is intended to only generate near-field transmissions and the magnetic loop receiver could not be misaligned more than +/- 2 mm axially and radially. This was because the rotating transmit coils were spaced 6 mm apart. A bigger spacing gives more axial leniency but it then starts to be an emitter.

With a 3" diameter you would want to try and scale things down as frequency rises. Anyway, just a few thoughts.

So it boils down to what basic data rate you are using and how good you are with electronics in the GHz range. My design was limited to 54 Mbps because we used conventional fast CMOS and no modulation scheme.

\$\endgroup\$
  • \$\begingroup\$ The more I have worked with this problem the more I've come to appreciate this answer. Thank you for the excellent insight. \$\endgroup\$ – Smithers Jul 4 '18 at 2:25
1
\$\begingroup\$

Perhaps a graphic to really show what you mean would help, however, what I get from your description is something that will not work and cannot just be fixed by throwing some impedance matching or gold plating after the ring.

This is not just a matter of reflections or a weaker signal, but a 7-8 cm is a significant fraction of a wavelength (to even more than an entire wavelength, depending on the material properties you are dealing with). What you could get is a kind of bad and unpredictable rat-ring coupler, where part of the wave travels one way around the right, and part of it travels the other way. As they could be of different lengths, one will have a different phase shift than the other, and as a result you might actually get destructive interference. The only way I can imagine this working is if you cut your ring and make the connection points power divider/combiners, and terminating the ends at the cut.

\$\endgroup\$
  • \$\begingroup\$ This sounds like a solid shot at a path forward. The addition of termination at the long stubs removes the worry I had about resonance and reflection. I believe this will increase insertion loss quite a bit, but the redriver should help there. Thanks for the comments. \$\endgroup\$ – Smithers Jun 9 '18 at 9:00

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.