Differential signaling over a long(ish) wire

homosapien

I am thinking about having some Prop. projects communicate over wire (~200ft.) amongst themselves. I have read that differential signaling is one of the better methods to do this for noise immunity. The layman's explanation of why differential signaling has good noise immunity is that noise will effect both conductors equally, so the difference (differential) in voltage between conductors will remain the same and the data is preserved.

My questions are:

1) Why doesn't using two conductors, one ground (0v reference), one data from output of Prop. (either 5v or 0v in respect to ground) running between the two Prop. circuits qualify as a differential signaling system? Is it due to the fact that the data line is terminated at the high impedance Prop. pin while the ground line is not, so the lines are not affected by noise the same amount?

2) Can one implement a differential signaling system between Props. by using 3 conductors, one ground and two signal wires, (e.g. a 1/0 on the signal wires being a high, a 0/1 being a low) ? Or is there still the same issue of using the ground as reference for the 2 signal wires?

I am just wondering if I really need to use a RS-485 chip such as the ST485BN to get effective noise immunity...

-N

kwinn

Your guess for #1 is basically correct. Any noise electrical noise induces an equal noise current on both wires. The ground wire has a low impedance to ground while the signal line is higher impedance. This means the current induced by noise on the ground line produces a much lower voltage than the same current on the signal line.

In a differential line both signals are of equal impedance and the noise current produces an equal voltage on both signal lines. Since the data is the difference in level between the two signals that effectively cancels out the noise component.

You can use two signal wires and a ground shield (this is what RS485 does), however the shield should only be grounded at one point. You should use an RS485 or a CANbus driver chip for this Two transistors with pullup resistors would also work but then you would need to ensure one is off when the other one is on.

An alternative to both RS232 and RS485 is a 20mA current loop connection. It has much better noise immunity than RS232 (close to what RS485 has) but can not be easily used on a bus system.

Mark_T

A differential (or "balanced") signal has a constant average voltage (the voltage on the two wires average out to half the supply voltage typically), so that you don't have a radiating antenna.

Two wires where one is ground and one is signal averages out to a signal of half the amplitude, so it acts as a radiating antenna - the signal has been encoded as part common-mode and part differential mode. The common-mode component is the problematical one.

The differential line encodes the signal purely as differential mode, so the antenna effect isn't coupled to your signal.

Because common-mode pickup can be quite large in long cables you typically have to use transformers to couple the signals at the ends of the wire (ethernet often does this).

Also a simple point, but important - a 5V differential signal has 5V of noise immunity, a 5V signal referenced to ground has a best-case noise immunity of 2.5V.

homosapien

Thanks for the replies guys.

Thinking about it, I was pretty sure the different terminations of the ground/reference wire and signal wires were the reason it would be impossible to do true differential signaling without additional hardware. I wanted to make sure it could not be achieved through some programming/wiring 'trick'.

I think I will pursue the ST485BN chip route, as at ~$1, it seems like a pretty reasonable way to add differential signaling advantages with minimal wiring and code overhead.

jmg

homosapien wrote: »

I think I will pursue the ST485BN chip route, as at ~$1, it seems like a pretty reasonable way to add differential signaling advantages with minimal wiring and code overhead.

Buy the slowest device that will meet your speed specs, as that also reduces RFI.

I notice that (finally) LVDS is entering the signaling market, with M-LVDS.

["Multipoint LVDS transceivers (low voltage differential signaling driver and receiver pairs)"]

["Common-mode range: –1 V to +3.4 V, allowing communication with 2 V of ground noise"]

and 100MBd and 200MBd models, in Full/Half duplex (ie RS422/RS485 like )

http://www.analog.com/en/interface/lvds/adn4692e/products/product.html