1
\$\begingroup\$

I am trying to design a circuit on (double sided) flex PCB that happens to have CSI-2 signals on it.

I would like to find out what is wrong with my deductions presented below, or what part of differential signaling have I misunderstood.

When I tried to calculate the trace widths and spacing, given the parameters of Flex PCB (extremely thin), It seems difficult or at most impossible to manufacture. On top of that, Polyimide and adhesives have only approximate dielectric constants available. (I am not working with controlled impedance Flex PCB)

But I realized, that FPC cables are used all the time in Raspberry (and almost all, camera modules and interfaces).

like this one

Since they are single-sided, I thought that they get around impedance matching by placing the ground next to the differential traces.

MIPI pinout as example

After a quick check on some online impedance matching calculator, I came to believe that those ground conductors adjacent to pairs of differential pairs have nothing to do with impedance matching at all. (so are those just to reduce crosstalk?)

I do realize that differential pairs theoretically do not need any ground plane — and that it just provides shielding and reducing crosstalk. But I always believed that those are absolutely mission critical aspects and I never really wondered (up until I faced this PCB design) that it never really was the case?

Are my deductions correct?

Should I just delete the ground plane under the differential pairs on my flex PCB and call it a day?

\$\endgroup\$
3
  • \$\begingroup\$ I also find it hard to believe that any impedance matching is achieved using those flat flex cables. You need a reference plane beneath the differential pairs, the weak coupling of the adjacent gnd lines will never be strong enough to provide a reference plane for the diff pairs. \$\endgroup\$
    – Timothy
    Commented May 10, 2023 at 10:55
  • \$\begingroup\$ No, you don't need a reference plane. For example, many cables around you, such as unshielded CAT cables used for Ethernet, just have the wires without a shield, and they do have correct impedance without a reference plane. Many flat flex manufacturers can provide a rough impedance approximation for their cables, so therefore impedance matching is possible. \$\endgroup\$
    – Justme
    Commented May 10, 2023 at 11:32
  • \$\begingroup\$ You can have impedance matching on diff pairs without a reference plane, think Ethernet cables. \$\endgroup\$
    – Voltage Spike
    Commented May 10, 2023 at 16:14

1 Answer 1

5
\$\begingroup\$

CSI-2 interface data pairs require a 100 ohm differential impedance.

The flat flex cable used with Raspberry Pi CSI-2 camera interface has 0.5mm pitch.

A cheap unshielded laminated polyester FPC cable with 0.5mm pitch has about 100 ohm differential impedance in the configuration where a differential data pair wires are between ground wires.

So basically, they don't work around it. The cable pitch is simply chosen so that it has a geometry that has a differential impedance that is very close to the required impedance.

This will vary somewhat due to the differences in geometry, manufacturer and materials, so the matching is not perfect by all means, but it is a surprisingly accurate match.

\$\endgroup\$
1
  • 2
    \$\begingroup\$ "A cheap unshielded laminated polyester FPC cable with 0.5mm pitch has about 100 ohm differential impedance in the configuration where a differential data pair wires are between ground wires." Do you have any link/article/calculation that support that? \$\endgroup\$
    – shlomi
    Commented Aug 26, 2021 at 11:40

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.