11
\$\begingroup\$

I'm facing a dilemma on how to route correctly in order to minimize noise on a data signal line. In the example bellow there is a 2-layer flexible PCB (which is very thin). The copper layer is 1OZ and the routing is 5mil. Red is Top layer and blue is bottom layer.

The blue wires are connected to LEDS which turn on or off many times a second, max current is 10mA. The red is 3.3V 30Mhz data signal with no error correction or protection in the protocol, just plain data. I have no doubts that there will be errors and this is fine, I just want minimum errors.

What would be the correct way to route the data signal (red) in order to minimize the noise the blue routes generate on it? The examples below just show the principle, I can't really 'go around'.

This is a very limited design in space so I have no room for LVDS. I can only make the routes thinker or double them.

Go 'over' a blue line Multiple paths Go straight Thicker trace (10mil)

Thanks.

\$\endgroup\$
  • 1
    \$\begingroup\$ Is your red signal a low-impedance signal? In that case it's very likely that you won't get any measurable interference; close-coupled interference is really something that applies mostly to high-impedance signals and/or signals with large loop areas. \$\endgroup\$ – user36129 Sep 15 '14 at 9:25
12
\$\begingroup\$
  1. Route straight, without any kinks, therefore minimize inductive coupling. Expanding according to @Majenko's comment: Cross your traces at right angles.

  2. Route as thin as possible to minimize capacitive coupling.

  3. Route as a single trace. Multiple signal connections that only slightly differ in length will cause signal skews and distortion.

\$\endgroup\$
  • 3
    \$\begingroup\$ One of the first rules you learn in digital PCB design is to always cross your signal tracks at right angles to each other. \$\endgroup\$ – Majenko Sep 15 '14 at 9:39
  • \$\begingroup\$ Wouldn't that argue for splitting the traces in a left and right branch, both of the same length ? You'd have the same crossing area, but since the left and right branch cross at different points the distortion picked up from the underlying traces will be different. I admit it won't matter much here since the LED signals are just a few Hz (1E8 meter) \$\endgroup\$ – MSalters Sep 15 '14 at 20:13
  • \$\begingroup\$ @MSalters Splitting the trace doesn't give you any pros. Thus I don't think it's even worth considering. \$\endgroup\$ – Dzarda Sep 16 '14 at 10:27
5
\$\begingroup\$

From the original question

flexible PCB (which is very thin)

Are you familiar with IPC-2223 and have a copy?

Are you aware where on the flexi the bend will occur?

The layout in the 1st will more than likely rule that layout routing out.

As to #2, #3 or #4 these appear at right-angles, so they will be less susceptible to pick-up, BUT their characteristics might not be ideal. Have you determined what the conductor impedance is with respect to your circuit?

You can use SaturnPCB to help determine how thick your trace needs to be, with respect to distance to 0V.

\$\endgroup\$
  • \$\begingroup\$ No bends, it is all straight. \$\endgroup\$ – Gilad Sep 15 '14 at 11:04
  • \$\begingroup\$ the flexi won't flex? not even a bit? \$\endgroup\$ – JonRB Sep 15 '14 at 11:09
  • \$\begingroup\$ length is 300mm and will be bent over a 200mm (diameter) ball so not much bending. \$\endgroup\$ – Gilad Sep 15 '14 at 11:11
  • \$\begingroup\$ Please read IPC-2223, this will definitely rule out the 1st layout. \$\endgroup\$ – JonRB Sep 15 '14 at 11:34

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.