I saw many DAQ cards from large companies that use galvanic isolation in their product.What you think about their design? If we solve earthing problem ,then there is no danger and system is better than when we do not have galvanic isolation.

Hallo Parviz,

why are you so focused on galvanic isolation? Common mode noise in adequately grounded applications is much less than 10V. So, you will almost never need those hundreds of Volts of common mode range of a optocoupler solution.
On the other hand, consequent use of galvanic isolation would also mean, that each channel is to be isolated against every other. But consequence would be the requirement of an extreme number of additional components.

Let's discuss situation with the help of the following simplified schematic:



In the middle of drawing you can see your data aquisition board, consisting of 4...20mA signal buffer, multichannel ADC, MCU and RS-485 interface. On right side you will find the 'host', or whatever it is called. On left side there's drawn one channel of front end sensor electronics, which delivers 4...20mA signal current.
It's important to note, that there are three different grounds, the signal grounds of each section. Galvanic isolation shall help in exact that case, that between two or more of these signal grounds there is some potential difference, called common mode noise or ground noise.

What are the reasons for ground noise?
First consider the following situation: Each section shall be powered by a mains transformer power supply circuit, which makes no connection between yellow-green earth wire and signal ground, according to requirements and rules of 'Safety class II'. Then signal grounds 1, 2 and 3 are nearly identical and signal ground is floating realitive to earth potential. As long as there is no additional earth connection, application will work, without any problems.
Between ground 1 and 2 only rather small signal currents are flowing and voltage drop along ground connection is neglectable. Between ground 2 and 3 there is some high frequency common mode noise, due to typical RS-485 signals. But because RS-485 works with symmetrical signal routing ground noise is neglectable.
Ground noise can be further decreased by introducing a RS-485 cable with shield. Connect both ends of shield to according signal ground (ground 2 and ground 3) and as consequence ground noise between ground 2 and 3 is very low. Use of this cable shield, which is connected to signal ground at both ends, helps also in preventing 50Hz hum being 'injected' into signal ground via interwinding capacitance from mains transformer from making problems at finite common mode impedance of RS-485 transcievers.
It's worth to note, that ground connection between ground1 and 2 does not see what's going on between ground 2 and 3, so it doesn't matter, at all.

Situation even does not change, when one (and only one!) earth connection to signal ground is made. But situation changes, when there are two or more earth connections, and when they are well separated from another. Then, a ground loop can be built, which can result in some increased ground noise. But this 'loop noise' is often much lower than expected. I have never seen more than a few volts. But, I must confess, that applications I worked with were designed obeying proper grounding techniques, like fabrication of ground grid, which looks like an idealized ground plane.
In our drawing, RS-485 can best handle this additional 'loop noise', because symmetrical signal routing is used. But, 'loop noise' between ground 1 and 2 can be troublesome, because signal routing here is not symmetrical. But even when signal is not symmetrical an symmetrical working instrumentation amplifier can help drastically to reduce loop noise. Have a look at following simplified schematic (no protection and filtering circuitry shown!):



Common mode rejection ratio directly depends on matching accuracy of input voltage divider resistors and resistors of final stage. With precision resistors and careful trimming much more than 40dB is achievable. Circuit can handle up to 100V common mode voltage, which is more than enough for very most cases.
Advantage of use of instrumentation amplifier is, that no additional components are neccessary for achieving any isolated circuitry.
Cable from sensor front end section to data aquisition board section should also be shielded, with shield connected to signal ground at at least one end.

Hope it helps,
Kai