3
\$\begingroup\$

I am interested in transmitting both digital data and power over a single coaxial cable between two FPGAs. Communication should be at least half-duplex. Communication speed should be on the order of 50-100 Mbps (something like ECL might be required?). Power is coupled in via an inductor biasing network in the same way as some "power over coax" GMSL SERDES pairs use. Here is a block diagram:

enter image description here

I'm looking for suggestions on how to fill in the black boxes, which take bi-directional IO to/from each FPGA and translate it into something capable of driving and receiving from a 50-ohm terminated (both end) coax that is biased with a voltage rail (power over coax; 12V in block diagram is an example anywhere from 5-12V is fine). I would like to use a minimum of parts. Data serialization, error detection, DC balancing, etc can take place inside the FPGA using e.g. PonyLink or something similar. I'm looking for the simplest and smallest translation scheme.

Why do I want to use a single coax? Because the circuit on the Vout side of the block diagram will be carried on a mouse's head for neural recording purposes. Any additional cabling is a mechanical hinderance to the animal. We are currently using a DS90UB95x SERDES pair and a 0.33mm OD microcoax which works well, but is power hungry. We want to reduce power consumption and size because we do not need the large data rates afforded by GMSL devices, which are meant for cameras typically.

Thanks.

\$\endgroup\$
8
  • 1
    \$\begingroup\$ Since the data rate is so low and a cable on a mouse is probably short I would be tempted to directly connect an IO pin to the cable with no active electronics. Perhaps a line driver if you didn't have an IO that could drive 50 ohms. That's as small as it gets. \$\endgroup\$ Commented Jan 15, 2023 at 17:05
  • \$\begingroup\$ What is the environment space in which the wire-attached rodent can wander around? Is the overall bulk of data the same in both directions for a long-term period? 50-100 Mbps is fairly high <-- is that data rate really necessary given that you say this: we do not need the large data rates afforded by GMSL devices? \$\endgroup\$
    – Andy aka
    Commented Jan 15, 2023 at 17:12
  • 2
    \$\begingroup\$ I'd put a capacitor in series with each of theose 50 ohm termination resistors. There's no need to waste DC power heating them up! \$\endgroup\$
    – Neil_UK
    Commented Jan 15, 2023 at 17:24
  • 1
    \$\begingroup\$ I think you should try simulating a single direction scenario (transmit and receive). That shouldn't be too hard to implement then, when that looks good, add the bits to make it bidirectional (also not too troublesome). A simulator will eat this problem and give you a really good basis for making modifications and improvements from the comfort of your PC. A simple scrambling scheme for the data is easily implemented with a few D types and exors. \$\endgroup\$
    – Andy aka
    Commented Jan 15, 2023 at 18:25
  • 1
    \$\begingroup\$ Go for a peak-to-peak data amplitude of about 1 volt and use a comparator for detection (possibly a MAX999). I might consider a balanced transmission and that means inductor power feeds and inductor power-take-offs on both wires. Having said that, it's coax so it shouldn't be a problem single ended. \$\endgroup\$
    – Andy aka
    Commented Jan 15, 2023 at 18:33

0

Your Answer

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

Browse other questions tagged or ask your own question.