Suspect noise on comms wires?

[cupoftea]

Hi,
Just today seen a big test jig with unshielded ribbon cable comms wires and it was working intermittently. I suspect noisy comms. There was an approx 15way ribbon cable (approx 30cm long) between 2 boards. Can you confirm that the comms signals in this need to be differnetial type?, and not the unipolar (TTL) type that you commonly see between components that are near to each other on the same board? So, eg, the ribbon cable should contain eg CAN bus signals, or RS485?, rather than TTL?

The test jig was for a switching circuit for three phase mains.
..............................-----------------...............................

Also, presumably on a test jig, which will not be made in volume, you should do super-low-bit-error-rate comms?.....eg, the most noise immune comms protocol that exists? With the highest magnitude signals (bits), so that SNR is maximised?
Also, common mode chokes should be used at each end of the cable to reduce bit error rate?
Also, shielded cable will reduce bit error rate?
And if its ribbon cable then it must have GND interleaved with all signals? (to reduce BER)?

BTW, what is the comms protocol with the lowest BER?

Presumably slower comms will always be less noisy than faster comms?.......and if possible, comms should be slowed down so much that the comms lines can be RC filtered?

 

[betwixt]

I've sent TTL over much longer cables than that without errors but you do have to be careful how you do it. Differential is always preferable because it is inherently (almost) immune to common mode interference. Remember that TTL inputs are not intended to terminate TTL outputs so there will be an impedance mismatch that can cause reflections and ground bounce, particularly over longer cable lengths. If the signals are always driven, you can add a terminating resistor at the receiving end to absorb most of the unwanted signals.

https://uk.farnell.com/c/passive-components/resistors-fixed-value/fixed-network-resistors?resistance=330ohm&network-circuit-type=dual-terminator

Brian.

 

[cupoftea]

Thanks, what sort of comms protocols woudl you send over a ribbon cable? (RS232, RS485 , SPI, I2C, CAN, MODbus, Manchester coding, TTL?)

 

[ZASto]

RS232, RS485

All those listed above are hardware layers.
Protocols are your software implementation of transfering data to/from point A/B.

 

[betwixt]

Some of those are not protocols but you can use almost anything if the speed isn't too much of a issue.

You have to understand that the high frequency emissions from a cable are primarily caused by the fast rising and falling edges of the signals carried along the wires. They are an essential and integral part of the signals, if you deliberately remove (filter) them so the rising and falling edges are more gradual or rounded, you also restrict the speed and usability of the signal itself. You can get 'slew rate limited' drivers that deliberately restrict the current that can flow at their outputs so the edges are less steep but they also have a lower top speed for the same reason.

One of the ways to reduce the effect is to use differential signals and keep the two anti-phase wires as close as possible to each other, typically twisted together like in CAT5 and CAT6 network cable. The theory behind that is if the wires carry identical signals but one inverted with respect to the other, the magnetic field around them is largely cancelled. The differential receiver re-inverts one of the signals then adds them so any common mode interference picked up is also cancelled. You can use differential for almost all digital and analog signals but of course it doubles the wiring per signal.

Brian.

 

[cupoftea]

ways to reduce the effect is to use differential signals and keep the two anti-phase wires as close as possible to each other, typically twisted together like in CAT5 and CAT6 network cable. The theory behind that is if the wires carry identical signals but one inverted with respect to the other, the magnetic field around them is largely cancelled. The differential receiver re-inverts

Thanks, thats the best explanation ive heard of that, after many hours of internet seeking.
And i take it that the ground of the two diff signals gets taken with the pair?...so it has to be threee wires really?

 

[ZASto]

Thanks, thats the best explanation ive heard of that, after many hours of internet seeking.
And i take it that the ground of the two diff signals gets taken with the pair?...so it has to be threee wires really?

No, not needed.
Consider LAN cable. It has 4 pairs -> 8 wires. No ground.

 

[betwixt]

Actually, the ground does have to be there but by balancing the data there is effectively no ground current. Any difference in ground potential from one end to the other through the resistance of the ground wire becomes almost irrelevant. Most digital receivers sense anything above a few hundred mV between the wires before noting its polarity. Ones and Zero are sent as one polarity or the other.

LAN cables are slightly different because they are transformer coupled at both ends. The original question was about ribbon cable so I assume it also links grounds for other purposes.

Brian.

 

[cupoftea]

Actually, the ground does have to be there but by balancing the data there is effectively no ground current.

Thanks, even though these's effectively no ground current, i assume that due to high frequency coupling, the ground wire has to be closely runnin alongside the two diff pair wires?.....kind of like a twisted triplet of wires? (and this triplet presumably being shielded for noise mitigation).........or if in a ribbon cable, then the ground should be in the middle of the two diff wires?
Though i assume using ribbon cable is generally bad practice for diff pair comms?

 

[KlausST]

Hi,

Whether GND is needed or not depends on the curcuit and signals..

With a RS485 receiver IC you usually need GND, otherwise the input will float with respect to receiver_GND and thus one can not guarantee input_common mode limits.

With opto_coupled signaling you don't need GND. (Mind the characteristic impedance mismatch)

With transformer coupled signals you don't need GND. But it needs DC_free signaling, otherwise the transformer core saturates. Here comes Manchaster coding into play. It is no interface, no necessarily hardware, no protocol. It's just an additional coding layer to get rid of DC when transmitting data.

what sort of comms protocols woudl you send over a ribbon cable? (RS232, RS485 , SPI, I2C, CAN, MODbus, Manchester coding, TTL?)

This is a mixture of some different things:
* RS232, RS485, TTL .... describes the voltage levels
* SPI (and UART) ... describes an interface, which signals there are, the job of the signals, how data are sent (but both are able to work with RS232, RS485, TTL)
* I2C and MODbus ... is an interface specification, with voltage levels, signal description, timing and protocol
* Manchester just forms one serial data stream into another serial datastream. It's main benefit is, that the datastream becoms DC free and thus enables the use of transformers for electrical isolation.

Klaus

 

[betwixt]

Ribbon cable certainly isn't optimal for differential signals but if you look at the individual wire connections, it is best to keep the differential carrying ones as close as possible to each other. So a better layout would be:
signal1+
signal1-
ground
signal2+
signal2-
ground

and so on.

Looking at it mathematically, if we call the wanted signal 'W' and EMC signals 'U':
At the transmitter you produce W+ and W-
The receiver sees W+, W- and U.
Now, re-inverting W- and adding it to W+ gives 2W but as U is common to both wires U added to -U gives zero.

Brian.

 

[KlausST]

Hi,

Ribbon cable:
I made good experience with GND | S+ | S- | GND over non twisted standard ribbon cable. Several MBaud over a couple of meters via RS485.
I also replaced one GND with VCC (as power supply of slaves). Capacitor between VCC and GND close to the connectors.
This way the VCC signal has the same AC impedance as the usual GND.

If I remember right, then I got a differential signal impedance of about 150 Ohms (not sure about the value, and it surely depends on exact ribbon cable type) ... then I got almost undistorted signals at 1 MBaud.

There also are shielded flat ribbon cables. A bit expensive, though.

Klaus