2
\$\begingroup\$

I'm looking to configure a Xilinx Zynq-7000 custom board with LVDS receivers according to the following diagram.

enter image description here

In my setup, the 'IOB' on the left represents an LVDS driver from a radar receiver and the IOB on the right represents the Zynq board that receives and processes the data. My concern is with the 50 ohm cable connecting the LVDS transmitter to the LVDS receiver. Standard 50 ohm coaxial cable is not going to work because there are actually many such connections that all need to fit within a single shielded cable that plugs into the Zynq board. I'm having a tough time finding higher gauge wire that specifies a characteristic impedance and I don't feel like putting two lengths of 100 ohm ethernet cable in parallel to make a single connection.

I was considering using MIL-C-27500 22-awg wire because we have a bunch lying around, but I can't seem to find any datasheet that specifies its characteristic impedance. The cable will be up to 3.5 feet long and will be used to transmit 10 MHz SPI data. A quick calculation using 100 ohms and 50 pF (completely made-up number approximating cable + receiver capacitance) gives a rise-time of 5 ns, so BW = 70 MHz. At 70 MHz, the electrical length is 14 feet so my 3.5 foot cable is starting to look like a transmission line (of course, these are extremely ballpark numbers).

Anyways, I don't think I can get away with using regular old 22-gauge wire (please correct me if I'm wrong) given that signal integrity is fairly important in this application. Where should I be looking for higher-gauge, 50 ohm wire for this LVDS application, or is there a better solution?

\$\endgroup\$
5
  • \$\begingroup\$ LVDS is 100 ohm differential. You could use 100 differential twinax cable or flat flex. Check out Samtec \$\endgroup\$
    – EE_socal
    Aug 15, 2018 at 23:20
  • 1
    \$\begingroup\$ How the gauge of wire has anything to do with differential characteristic impedance? Any CAT5 cable ( any twisted pair) should work just fine under these conditions... \$\endgroup\$
    – Ale..chenski
    Aug 15, 2018 at 23:25
  • \$\begingroup\$ @AliChen Sorry, I never claimed that gauge had anything to do with characteristic impedance, only that I can't use standard 50 ohm coax due to its size. I thought CAT5 was 100 ohms typically? Don't I need 50 ohms for LVDS according to the diagram? \$\endgroup\$
    – pr871
    Aug 15, 2018 at 23:50
  • \$\begingroup\$ Have you considered thin coax such as LMR100? \$\endgroup\$
    – Chris K8NVH
    Aug 16, 2018 at 0:16
  • \$\begingroup\$ As EE_socal already said, the LVDS is 100 Ohm differential. A twisted pair inside CAT5/CAT6 cable is a perfect match. 50 + 50 = 100. See also ti.com/lit/an/slyt163/slyt163.pdf \$\endgroup\$
    – Ale..chenski
    Aug 16, 2018 at 1:06

1 Answer 1

Reset to default
1
\$\begingroup\$

I'm having a tough time finding higher gauge wire that specifies a characteristic impedance and I don't feel like putting two lengths of 100 ohm ethernet cable in parallel to make a single connection.

Cat 5 cable is 100 ohms differential, which is exactly what you want.

Your prototypical coaxial solution uses two 50-ohm cables for the two arms of a differential signal, giving 100 ohms differential.

The twisted pair in Cat 5 gives the same differential characteristic impedance as the original solution.

Even if you used some different cable, for example with 85 ohms differential characteristic impedance, you could resolve the mismatch by changing the value of the termination resistor.

Trying to combine multiple cables to change the effective characteristic impedance could only lead to pain and aggravation.

One difference: The Cat 5 cable won't shield the signal like coaxial does, but this is usually not a problem. Also, the Cat 5 won't provide any ground connection between the two systems, so you may need an additional conductor in your overall solution to connect the two grounds. If the link distance is very long you might also need to connect to the cable through a transformer at one or the other end to avoid common mode currents due to ground potential differences between the transmitting and receiving circuits.

\$\endgroup\$
3
  • \$\begingroup\$ As long as 1.1Gbps bidirectional would work, I would just use a high end isolator and not worry about it. \$\endgroup\$
    – GB - AE7OO
    Jan 28, 2020 at 14:22
  • \$\begingroup\$ @GB-AE7OO, I'm not sure what you mean. How would an isolator solve the problem of connecting two boards that need to be, for example, 2 m apart? And are they available with wide enough bandwidths to pass a general digital signal? And are they available at low enough cost to be competitive with just terminating your transmission lines correctly? \$\endgroup\$
    – The Photon
    Jan 28, 2020 at 16:28
  • \$\begingroup\$ The specific chip I am talking about is the ADN4654./55. These are dual high speed TIA/EIA-644-A LVDS isolators. They support, LVDS, CMI(HDMI), Display Port and others. For LVDS they have a .1/2 Gig bandwidth. They also have a common mode immunity of over 25Kv/uS... They have the output for short cable runs of a couple of meters. The cost is where they fail, at about $13 each, qty 1. I don't know how much 2 transceivers, 2 transformers and a larger cable would cost. And at least in my office they will driver over 12 feet... \$\endgroup\$
    – GB - AE7OO
    Jan 28, 2020 at 23:46

Your Answer

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

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