Dear Medhi, your problem is not only magnetic field but your circuit's current driver is probably not strong enough to drive the current over that distance (resistance in cable also). If you really want reliable data at low current use an I2C chip on each side.  You get these in 3V or 5V, while the I2C is easy to command and address through a microprocessor (presuming this is what is generating your data). 

With I2C you may go up to 100m, see example on this link

http://www.esacademy.com/en/library/technical-articles-and-documents/miscellaneous/i2c-bus/frequently-asked-questions/i2c-faq.html

Regards

Gerro

Dear Gerro Prinsloo , Thank you very much for your answer. you know that we designed the new device as shown above to lessen the cost as much possible. We bought 100 m copper cables is less than 20$. we have a large network. I wonder if you know about the cost of I2C because I found it with the high cost on the internet and according to your answer we need more I2C for each 100m.

thank you

Mehdi

sounds like you have something similar to 4 conductor 24AWG wire which has .0842 Ohms/meter resistance ~17 Ohms for a 200 meter loop. Max current @ 3V in this loop is 176mA with no load. If you need to power your end-point devices over the wire, you'll need to consider the ~17 Ohms in series with that load (e.g. a load drawing just 50mA will drop .85V on the line leaving only 2.15V to power the load and provide any return signal which will also experience the resistance losses).

There are many options that can provide communication over your existing wire but you'll need to provide more details in order to decide which path to take. What is the desired data rate? Does the communication need to be two-way? How many end-point devices are on the network? What is the network topology? Can you use wireless for the data communications while using the cable to provide power to the end-point devices? What kind of processing power is available at each end-point? What is the physical environment like? Industrial, indoor, electrically noisy, etc.?

My first suggestion would be to go wireless with the data and use all four conductors for power to minimize voltage drop to your end-point devices. Hope RF makes a transceiver module that will operate down to 2.2VDC ~25mA in transmit mode and transmit data over several different frequency bands that you can specify when ordering and have a range of around 200 meters in open space. They are available from many hobby electronics providers for about $7 each. Look for the RFM12B.

I2C is a single-ended, capacitance limited system designed for modest distances and is not an apt choice.  Use balanced transmission such as RS485 or LVDS for high speed if at all possible.  Assign your conductors as follows: if going clockwise you have conductors A, B, C, and D, place ground on A, data+ on B, power on C, and data- on D.  This ensures equal noise coupling from power into both data lines so that the common mode rejection of the data receivers can ignore it.  For fast edges or high data rates be certain to use proper, balanced termination of the data lines so that reflections are not an issue.

Dear Ron Reiserer and Orin Laney

I mad a simple sketch for my network for more clarification, 

Please see the attached PDF file. 

Your comments is highly appreciated 

Mehdi 

To make this 4 conductor flat style cable work, you will need to place balanced data on the two center conductors and power on the outside conductors.  The hot side power conductor must have a high frequency bypass cap right at the connector on both ends and the ground conductor termination should also be right at the connectors.  Use a proper connector that cannot be inserted reversed.  Notwithstanding that the conductors are not themselves twisted, include some rotation in the cable lay (3 to 4 times per meter) to help maintain balance of ambient noise pick up.

Since your power is apparently coming from a switching supply such that the ripple is the problem, this must be addressed directly.  Either use linear regulation at the sending end, or insert a differential mode choke, which is nothing more than a common mode choke with a winding reversed.  But, it will have to be a large one of the sort intended for power line filtering and not some tiny surface mount part.  Best practice is actually to use two in series, one in standard common mode configuration and the other differential, because neither mode is helpful.  Common mode radiates noise and diff mode couples into your data.  The usual HF bypass cap directly across the choke output should be included.

I have concerns regarding bandwidth over the distance quoted and mention termination again, but the above  will at least keep power noise out of the data.