Terminating Ribbon Cable

[Noway2]

I have two questions both regarding the design I have been working on, on different topics, so here is the first of two posts.

This question is in relation to the board where the pressure transducer and other analog signals will be monitored. I am planning on connecting this board to the main computer by way of a ribbon cable. The cable, typical 100 Ohm cable could potentially be between 3 feet and 6 feet in length, though I am trying for less. I am debating wether or not to use a smaller cable with a serial link or to use a larger cable and a parallel bus. If I go with the serial link, I can use a differential driver, such as an LVDS that is configured for 100 Ohm cable and design the PCB traces from the transmitter and to the receiver accordingly. The transeivers are about $4-$8 for a quad device package.

My question is, if I go with the parallel bus route, I then need to terminate the cable. What would be the best method of termination? Would AC termination with a 100R and a small capacitor work or is there a better method.

I should mention that I am starting to lean towards the serial approach and run a 3 or 4 wire serial link (x2 for differential), but I haven't ruled out the parallel approach yet.

By the way, I have also tried googling this topic every way I can think of and about all I can come up with are SCSI interface products for a PC.

 

[logbook]

The key thing to mention here is data rate. Pressure transducers don't normally get read very often as they only change relatively slowly. 6 feet of cable is nothing if you are using a 100kHz clock rate for example. You will not have to worry about the termination of the transmission line because the cable is "electrically short".

6 feet = 2 metres
speed of light is 3E8 m/s
2m of cable takes 7ns.
Allow for the dielectric constant and it is now say 15ns.

If your edge speed are in the 1µs region then you will not see any reflection effects of the transmission line.

 

[itsmoked]

I would concur totally with logbook.

You could use straight 0-5V with a "reasonable" data rate no termination.

Or for "vast utility" full duplex 485.

Why not 2 shielded, twisted pairs, in a single cable -> power and signal. Then your remote board is fully remote-able to thousands of feet.

Ribbon cable... parallel ports... So dated. Going for the retro look?

 

[Noway2]

Yes, I am going for the retro look . As a matter of fact I think I might even design in a lava lamp .

My biggest reason for going with ribbon cable is that it is readilly available off the shelf, can be used with locking connectors, is relateively inexepensive, and comes in various widths to make wrong point terminations difficult. Our manufacturing defect rate here is attrocious, with the biggest majority being wiring errors. I have been on an airplane far to many times because some idiot didn't tighten a screw and it wasn't caught in test.

Regarding logbooks comment about edge speeds, most of the logic in the device is driven by an FPGA or a buffer with rise times on the order of a few nanoseconds. I didn't terminate most of the signals on the board, just the read and write strobes. I supposed I could follow the same idea here too as I should still have enough time for the address and data lines to settle before the strobe.

 

[MikeHalloran]

I probably haven't discovered _all_ the ways to screw up mass terminated ribbon cable, but I'll assert that there are a few. If you're going to make up the cables in your shop, invest in a tester.




Mike Halloran
Pembroke Pines, FL, USA

 

[Warpspeed]

Ribbon cables are actually a pretty good medium, provided you use the proper tooling, and provided the ribbons are protected inside equipment, (where they cannot be brutalized).

I once had an application that required two sets of unrelated RGB video signals (plus syncs) going to a pair of high resolution large screen computer monitors. These traveled together down a single six foot ribbon cable. Needless to say this would be rather sensitive to reflections, HF response, and crosstalk. There were NO visible problems and it worked just fine. I could see video components up to 350 Mhz on my spectrum analyzer and the pictures looked crisp.

The trick is to terminate both ends of the line in 100 ohm impedances, and ground each alternate conductor in the ribbon. I know it sounds horrible, but it worked fine, and in this instance it was particularly convenient for the production guys, and lower cost than coax.

Don't underestimate the humble ribbon cable.

 

[itsmoked]

Be aware that there is also twisted ribbon. Especially since you were concerned about noise.

Besides it's pretty.

 

[Noway2]

My goal was to get the ribbon cable assemblies purchased from a vendor becuase of concerns about the assemblers screwing them up, though thoughts of a tester had also crossed my mind as a prudent measure.

My plan was to have the PCB traces from the driver to ribbon be designed for 100 ohms (or at least very close to it) which should match the ribbon cable and keep reflections minimized. Besides, if it doesn't work I can just put a 10x scope probe on it, can't I?

My last thought was to use a differential driver and receiver as opposed to a parallel bus for noise immunity and requirement of a lot less signals, at the expense of a couple of driver chips. The signals or commands to and from the microprocessor would be converted to and from serial by the programmable logic devices which would also tack on a CRC check value.

Itsmoked, I surely hope that I can avoid the twisted and or shielded pair ribbon connections. Given the choice I would much rather put a big paper weight... er ferrite choke around the ribbon . I really took the motto that if it doesn't work get a bigger hammer to heart!

 

[Warpspeed]

There are two different ways to tackle the noise problem.

One way is to use single ended drivers/receivers and ground every alternate ribbon conductor. If there is a proper ground plane layer at each end of the cable, this is effectively just a continuation of the solid ground plane. It will very effectively bond everything together eliminating any differential mode noise. As you rightly suggest, a big nasty ferrite will reduce any high frequency common mode problems should they arise. That would be my preferred choice for a relatively short ribbon cable run, perhaps inside a single equipment rack, or single piece of equipment.

Where a long external cable run is required, and the equipment at each end has a different and possibly varying noisy ground potential, the problems are rather different. In that case a fully differential system would be much better. Twisted pair with balanced signals, and maybe even pulse transformers or opto isolators would be more appropriate. Just treat the ground as a hopeless case, and forget about even trying to effectively bond things together.

I guess the final choice depends upon the faith you have in your common ground.

 

[Noway2]

In this application, everything is a common ground. It is an automotive type system, being a diesel engine system, running on 12v / 24v batteries with sensors and everything all tied back to the engine skid.

Your comments made me think of something that I hadn't considered before. I could run a serial port using the common ground approach and most likely things will be just fine. This board is also the power supply board and while that may sound odd, it is really the ideal place for it. One of the functions of the controller is to determine if the batteries are present or not. Since the presence of a charger can mask a missing battery, I must periodically isolate the battery from the charger and take a voltage reading to see if I have the battery there. Since I have to route both batteries and chargers into this board it becomes a natural place for the power supply, which is a POL converter (another strike against the idea of complete isolation).

The ground difference between the two boards should be minute as I am planning on running a large enough wire between the boards for power and ground to more than source the current without significant IR loss.

My concerns about noise in the project are pretty much a "its something I need to be aware of and plan accordingly" but I believe through proper design practices it can be dealt with.

Again, I do appreciate your taking the time to comment as I now have a bit of insight that I didn't see before that could make the application just as effective, but slightly less expensive.

 

[Warpspeed]

Don't underestimate the noise problems with an engine starting battery and a battery charger !

If the charger uses SCRs, there can be some fairly fast switching transients generated which can sometimes be difficult to eliminate. And a large engine under full cranking amps can cause massive dc drops in ground return lines.

I would be tempted to fit opto isolators to all the serial data lines right at the main circuit board, and make the data lines current loops independent of any grounds.

As you suggest, the usual method of battery testing/detection is to lower the charger voltage to say 11.5v in a 12v system, and see if the voltage actually falls that low. That can be done periodically without disturbing any other equipment connected to the battery.

 

[Noway2]

You have a point about the potential noise from an engine starting and it is a dangerous trap to under estimate them. Perhpas differential is the way to go for the comm lines between the boards, then. The incremental cost is fairly insignificant relative to the cost of the system. In either case it is better safe than sorry, especially since noise issues like to rear their ugly heads only after a number of units have been installed.

Fortunately, I won't have the starting current going through my equipment as that is direct from the battery. I will however have the solenoids for the starter solenoid, water, fuel, etc. I will be powered from two sources that are diode steared both on the supply and ground lines, feeding into a regulator that has a decent amount of filtering on it.

I have no real idea of whats in the chargers. I beliieve that they are switchers and probably use either SCRs or IGBTs for output drive. Never took one apart, though.

Other than the snesors to which I supply the reference potential, everything else is isolated from me. Hopefully the power supply system is capable of supressing the input transients.

 

[Warpspeed]

Your battery charger may be a switcher, but I have never seen one like that. Usually for engine starting applications they are fairly crude, just a mains transformer, a couple of SCRs with a choke/capacitor filter. There is no real requirement for low noise output or fast dynamic response, size and weight are not usually issues either. But cost is.

The battery charger may otherwise be reasonably sophisticated and have a microprocessor controlling it, but the power handling part is usually very basic using SCRs. The output noise is usually specified in RMS, but some rather nasty high amplitude low duty cycle SCR voltage switching spikes can be present, so beware.

The main difficulty with engine starting is that the ground at the battery, and the ground at the engine block will typically change by many volts during initial full load cranking amps. Diode isolation sounds like a very wise precaution.