FPC cable for RS-485? 3

[blcpro]

I really wanted to search the posts before asking, but it seems the search engine is down, so please be kind if this has been asked before:
RS-485 bus would usually work best with twisted-pair 120 ohm cable. However, In a new application we are developing, a very flexible cable of this description doesn't exist (for a robotic application with a very small bend radius <1" for the cabling). In the past we have had success using FPC or flexible printed circuit cable for sensors and power and such, but not for a serial bus. Would anyone reccommend such a cable for RS-485 (5V bus)? The distances would be short, less than 8 feet total of bus length, and the bus drivers are slew-rate limited to speeds of about 500kbps. There would be approximateley 4 nodes on the bus, and the cable may be near a DC power cable.
We have not used RS-485 in a design before, so we just want to make sure we make the best choice among available options.

Thanks!

 

[MacGyverS2000]

A continuous 8 foot FPC is not going to be cheap, but spec-wise it's well within the range of what you ask. Using a two-layer FPC you can emulate a twisted pair for 485. Are you sure FPC is what's really needed here?

Dan - Owner

http://www.Hi-TecDesigns.com

 

[blcpro]

Maybe FPC was the wrong name, looking into our purchase records, the cables we usually get from Parlex are actually called FFC, flat-flex cable. They specialize in custom length and termination options, but also sell short-length jumpers in Digikey. These are one layer, as far as I know, and all signal lines just run parallel from end to end and terminate in ZIF ends. It sounds like two-layer would be very expensive and perhaps not as flexible? I have not seen what you described in a standard catalog (emulating twisted-pair, that is).
I am also looking for other types of cables which are highly flexible with small bend radius, which we also use, but are not avaiable in twisted pair (with the same flexible properties). How would a three-conductor flexible cable (with the 3 wires twisted so as to give the cable its flexibility) work for RS485? One conductor for ground, but all three cables appear to be twisted in parallel, so I don't think its really a true twisted-pair. What do you think? I've been told a cable like this has impedance of 60ohms from one cable provider.

 

[blcpro]

In answer to "Are you sure FPC is what's really needed here?", we are currently using FFC cables for similar application very successfully (not for RS485, though). The key is the light weight of the cable and the small bend radius. The light weight allows for smaller motors and less motor current used in the design. The smaller bend radius also means the overall size for the cable-chains can be smaller, and thus lighter and less of a burden on the motors.

 

[MacGyverS2000]

Yep, familiar with Parlex cables... they wanted to practically rape me for using some of their cables with a bit of the cover cut out to attach SMD components to. I wanted to use one of the standard cables sold by DigiKey but with holes. It would have been done by a laser, but I want to say the cost was going to be over $25k for 200 4" long, 4-conductor cables. Nice cables, but not THAT nice.

Two layers is the only way you can get twisted pair in FFC/FPC... so again, are you sure you need twisted pair? What's the noise environment? Any particular reason you can't use standard multi-conductor wire? A thin gauge would be almost as thin as a Parlex FFC, can be ganged for higher currents, and would have no problem meeting the <1" bend radius.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[blcpro]

Havn't used RS485 before, so I havn't really tested the noise immunity with various cable types. I'm just trying to get a start with the best-case scenario, if possible.

The environment is a multi-axis robotic arm. The cabling to the arm is snaked along its length, and will include 24V DC power, logic signals, and the RS485 bus, and possibly some stepper motor power cables as well. These most likely will be all in separate cables, but also likely bundled close together in the same cable chain. The cable chain keeps the cables together and formed properly for repeated bending/flexing as the arm moves. So as for noise, there is the possibility of stepper motor power cables, 24V DC power supplied to motor controllers embedded in the arm (attached to the rs485 bus), and the likelyhood of close proximity to some motors and/or solenoids.

I am still investigating the use of standard multi-conductor wire, similar to what we have used before, but the flex cable is more attractive for the reasons I mentioned in the previous messages.

We currently get long cables from Parlex (about 10 feet) and the cost is minimal, though the cable is standard.

 

[Comcokid]

I've frequently used RS-485 with IDC connectors on ribbon cables in 19" rack systems where the total length was about 30 feet with no problems. There was no real noise sources other than many AC cords plugging into the back of the rack-mounted equipment.

 

[blcpro]

Comcokid, what was your bitrate in that application? I'm hoping for a minimum of 115Kbps, and up to 500Kbps. I can't use that kind of cable in my application, but its similarity to the flex cable is encouraging. Thanks!

 

[felixc]

There is IDC cable that has twisted pairs. Made by 3M I think. At every specified distance the pairs are untwisted to allow to connect regular IDC headers at these places. When you say a low-bending radius, it means that the cables will continuously bend and unbend as the machine works?

 

[blcpro]

felixc,
Yes, the bending is constant, and the cable usally has a rating of number of bending cycles at the specified bending radius. A good flexible cable will have this number in the millions of bend cycles. IDC cable would not normally last this long, but I have seen the cable you describe.

 

[Comcokid]

My application with the IDC cable was done at a previous employer. I looked for the manual I wrote on the equipment, but couldn't find it. I seem to recall that the programmer wanted the various units to communicate quickly - we tested to the maximum baud-rate of the UART successfully and then backed off to a more reasonable baud, I think it was 57.6K (the maximum may have been 115K for the equipment used).

However, from a transmission-line point of view, the impedance between twisted and untwisted wires is little. Twisting allows the net RF field exposed to a balanced line to cancel. The differential driving also supresses any common mode sensitivity. Also, 115K baud is, in today's world, fairly slow.

You can try what was done on my project. Route a cable poorly. Cram the maximum baud you can through it. Verily you have no problem, and then back-off on the baud to provide additional margin on top of that.

 

[blcpro]

I'm attempting to do just that. So far my tests have been limited to using the FFC cable near to a running DC motor, and the results have been acceptable. I have not done a full-scale mockup because the machine doesn't exist yet, so as for routing the cable poorly, the best I can do is leave it as a rats nest pile on my workbench with all the power and motor cables running under it. But I've been pretty pleased that I havn't noticed any noticable noise with a 500Kbps signal. The only noise I have seen is actually coming from the OUTPUT of the RS485 receiver. The actual differential signal on the cable looks very clean, especially in the 250Kbps range and lower, which is great for my app. The slew-rate control of the drivers makes a very nice signal even at 500Kbps, so I'm not as worried as I once was. By the way, thanks for everyone's help so far. I'm feeling much more confident now.

 

[MacGyverS2000]

Hmmm, just thought of this... kinda neat, now that I think about it. Make a single trace FPC that winds back and forth in a serpentine pattern with little cuts in the polyamide layer on the inside of each serpentine curve. Wind two such beasts together, such that the cuts in the polyamide match up and slide into each other (this is a tough one to describe, even ASCII art would be tough to make) wrapping around each other. this would effectively create a dual-layer FPC that simulates a twisted pair, using only single-layer FPCs. It would be thin, more flexible than a dual-layer FPC, and offer a reasonable CMRR.

 _   _   _   _  
/ \_/ \_/ \_/ \_

Copy the above pattern onto a single-layer FPC. Cut small nicks in the polyamide in each indention. Tie one end down with a clamp and start braiding

Dan - Owner

http://www.Hi-TecDesigns.com

 

[blcpro]

Although I had imagined something like that myself, I think the assembly would be tough!

 

[felixc]

A pattern like this could be interesting with a double-sided flex pc board, or could be done by the folks who are doing membrane keyboards.