FFC Cable Interconnect Question

[Shalomie42]

I am working on a design that uses an FFC cable interconnect. We made 2 boards with an FFC going between them to prove out the design. It works fine; however, I opted to use the el cheapo FFC cables & connectors. Working with them in our system is not very robust and I am unhappy with them. Therefore I want to graduate to a more robust solution.

The FFC cable is carrying very low speed LVDS signals (<100MHz). In the design of the PCBs I put all the signals onto the FFC cable in GND-SIG-SIG-GND fashion, so there is a GND between each LVDS pair.

I am looking at this Molex system for the FFC interconnect: https://www.molex.com/en-us/products/part-detail/150221230. There are a few reasons. The FFC cables themselves look way more robust than the el cheapo ones and the connector assemblies at both ends are latched / snap in. Everything is COTS. Price is not an issue.

Here is my question. I have been trying to do a little research as I am not an expert in EMI, PCB design, or return currents. The 15022 FFC cable has what they call a "grounding tape" as part of the stack up. I was thinking to myself that I could get rid of the GND pins between pairs and just use the grounding tape for the return currents. That way they are directly underneath the pairs. It also lets me go with a smaller pin count interconnect. I talked to my hardware guy briefly and he said that the grounding tape is for shielding and should not be used for return currents.

Can I use the grounding tape for return currents (same GND connection at both ends to PCB GND) or would I still have to keep the GND-SIG-SIG-GND pin lineup? Am I only supposed to use the grounding tape as a shield and connect to a SHIELD pin at one end?

El cheapo FFC stuff that we are using now. Not exact part / vendor, but same exact stuff: https://www.amazon.com/Yootop-Flexible-Ribbon-200mmx20mm-7-87x0-0787/dp/B07C4P6S43/ref=sr_1_1?crid=16XBQAJ4G20MQ&dib=eyJ2IjoiMSJ9.0WXUcHWFgjvCbYdHnNnZVR9t92xw48wmUSpGET5cplMCLmMk5tUmYrxNNUYv0f5jQyQI-gN01jiegj3V6iHFV-0mcbb1B161mMGD4Hv6_ciuYeGXS7y2372pNCaKdfUEaw4PSBkCNHjDhL7LpYFU7qj9asYMcM5BCFiRl1faFHH0w7czxxYN1ftWzHMl6NVnoUdTUr-Q8qaQKpuiQiPiXi6VBNcGuE0HE-6Qc2gfJzM.owx1gQupv3FuRihRint6a5wEwrp2rwex33KSvtVjD5A&dib_tag=se&keywords=ffc+cable&qid=1754003556&sprefix=ffc+cabl%2Caps%2C108&sr=8-1.

Additionally using something like a Hirose FH simple FFC latch connector

[thm_w]

Generally you do not want to use shield to carry return currents: https://electronics.stackexchange.com/questions/704788/is-using-the-shield-in-a-cat-cable-for-return-current-a-bad-idea

You want the shield connected on both ends to a chassis/shield/enclosure where possible. Then have the internal signal grounds still as you have them. Sometimes this is not possible.

Bit of info here: https://configurator.nicomatic.com/sites/default/files/library/2020-07/FFC%20Catalog.pdf
Seems standard to either have the shield as the first and last pins, or as an external connection, depending on the connector.

How many positions is the current FFC?

[Shalomie42]

Right now its a 50 pin FFC. I would be saving maybe 10 - 15 pins if I could use the grounding tape. The 50 pins is not an issue, I just thought maybe I could save a few. I should use the grounding tape as a shield. The main goal is not to save pins, but to make a much more robust interconnect.

The way the 15022 connector works is that it has 5 spring loaded prongs that make contact with the shielding tape. You make an assembly out of the FFC cable itself and a few extra components on both ends. That allows the whole thing to snap into the connector on both ends, which makes a robust connection. I don't think I need to use extra pins like some other connectors using the first & last pins as shield as its built into the interconnect for the 15022.

So you are saying I should connect the shield pin at both ends to CHASSIS and not a single end? I put provisions on one of the boards to be able to do that but not the other. I had watched Rick Hartley's famous grounding video. I could do both ends. The way I was handling it was to through hole plate the edge mounting holes in the PCB and those are connected to metal standoffs to the CHASSIS. I then put 0 ohm resistors in series between SHIELD and the mounting hole. This would allow you to solder resistors to make the connection.

The way the interconnect works is that we have a cabinet. One board is inside the cabinet and then we drill a hole in the cabinet and poke the FFC out of it. Then on the other side of the cabinet is mounted the second board. Distance between boards is <12". If we connect the SHIELD of both boards to CHASSIS, they will be connected to the same CHASSIS on both ends.

[mtwieg]

Here is my question. I have been trying to do a little research as I am not an expert in EMI, PCB design, or return currents. The 15022 FFC cable has what they call a "grounding tape" as part of the stack up. I was thinking to myself that I could get rid of the GND pins between pairs and just use the grounding tape for the return currents. That way they are directly underneath the pairs. It also lets me go with a smaller pin count interconnect. I talked to my hardware guy briefly and he said that the grounding tape is for shielding and should not be used for return currents.

Whether the shield "should" be used for return currents is an odd question if you intend to use LVDS. LVDS should be fully balanced, therefore each pair acts as its own return path.

The grounding tape definitely affects the differential impedance, and provides some level of shielding. If it's acting as a return current path, that's not necessarily bad in itself, but it begs the question of why a return current exists in the first place...

Can I use the grounding tape for return currents (same GND connection at both ends to PCB GND) or would I still have to keep the GND-SIG-SIG-GND pin lineup? Am I only supposed to use the grounding tape as a shield and connect to a SHIELD pin at one end?

The pinout (Sp-Sn-Sp-Sn vs G-Sp-Sn-G) will affect crosstalk significantly, and also the impedance to some extent. I would expect Molex to provide some data on this, but I have yet to find anything on their website. They just state a differential impedance of 100ohm +/-10%. Nothing about crosstalk or attenuation, or what pinout is used for characterization. Can't really make any suggestions without more data...

[Shalomie42]

The pinout (Sp-Sn-Sp-Sn vs G-Sp-Sn-G) will affect crosstalk significantly

Pardon my ignorance as I am not a PCB or LVDS expert. What is the ideal ? Should I forego putting the ground pins in between the pairs ?

The current lineup is

GND-SIG1P-SIG1N-GND-SIG2P-SIG2N-GND, continuing for all the pairs

[thm_w]

Maybe you could get away with the additional cross-talk, but its likely design dependant. eg the length of the cable, signal speed, etc. I'm not sure if there is an easy way to measure this and compare. You could cut the middle ground pins on both ends I guess, that would give some idea.

So you are saying I should connect the shield pin at both ends to CHASSIS and not a single end? I put provisions on one of the boards to be able to do that but not the other. I had watched Rick Hartley's famous grounding video. I could do both ends. The way I was handling it was to through hole plate the edge mounting holes in the PCB and those are connected to metal standoffs to the CHASSIS. I then put 0 ohm resistors in series between SHIELD and the mounting hole. This would allow you to solder resistors to make the connection.

Watch the correct Hartley video:

[mtwieg]

The pinout (Sp-Sn-Sp-Sn vs G-Sp-Sn-G) will affect crosstalk significantly

Pardon my ignorance as I am not a PCB or LVDS expert. What is the ideal ? Should I forego putting the ground pins in between the pairs ?

This is a decision I wouldn't make without signal integrity data from the vendor (in fact I wouldn't even consider using a cable without that data). I'm betting molex has it, but doesn't make it easy to find. I would just send them an inquiry, they typically respond pretty quick.

The effect of crosstalk is difficult to estimate without knowing more about the signals and the protocol being used. Clocks (or anything edge-sensitive) are what you should be most concerned about. Note that having a low baud rate doesn't necessarily save you from signal integrity issues. The rise/fall times are what really matter.