What to do to protect RS485 ICs?

From what you have said, I am not sure you have a common reference between the two systems. As I understand it youy only have the 2 wire A & B connecting the systems. If that is the case it is likely that one is floating against the other and sooner or later you are going to violate the +12/-7V spec of teh RS485. Try joining the two grounds together through a ~100R 2W resistor (Vsquared/R=144/100 <2W). If you can't run the wire, at least try to connect the commons of the RS485 drivers to earth/ground (athough this can be tens of volts difference from one location to another).

If you have got the commons connected, you should also consider the possiblilty that the electrical noise is affecting the processor itself and not just the RS485 connection. You will need to come up with some technique to prove that it is only the communications or not.

I don't see on your schematic the terminating 150 Ohm resistors. Without 2 of these resistors on both ends the schematic will not work well and will be much more vulnerable to HV spikes. See RS485 terminating resistors - you have only polaryzing resistors but you need also terminating ones!

@antedeluvian:
A 24 VDC is regulated to 5 VDC in the main board by 7805, and the same 24 VDC is going to slave board and regulated there to 5 VDC, there.
So the grounds are connected.
Some feedbacks I am getting:
1) The slave card freezes(after some days). This happens when I use 100nF/50V as capacitor.
2) The digital display lights off, then displays correctly, I see a similarity with SerB disconnection. I mean similar thing happens, when the SerB pin is not connected. This happens when no capacitor connected. In this kind of case, the 74LBC184 is damaged.
3) The slave card works slow(after some time). The information arrives 2-3 seconds late generally(not always). This happens when I use 50nF/450V.

An information: I can clearly measure at least 100VAC at an unconnected cable(both ends are not connected). This is a low power voltage. And this always stays on the cable. I think I must think a solution about this. And this is not a single case, I see this at many cases.
More details: (When no capacitors used)If the 220VAC links at the flexible cable(0.8cm) are open, the slave is working. When the 220VAC links' fuses are closed, the problems started. We have prevented this with capacitors up to a point.

@luben111: My schematic is not totally same with the bmp image I have sent in the beginning. If you can see that schematic, the differences are, the 1k resistors(termination resistors) are changed as 100R and added one capacitor of 10nF/400V. And the biasing resistors are changed as 680R. The differential ICs are 75LBC184.

Thanks a lot for your helps and I am looking forward to some more .

Actually, when I say, at least 100 VAC can be measured, the range is much higher. We have seen 200-300 VAC at some places.
Any idea to overcome this?

how you power the slave modules? If the distance is big and you power each slave individually (locally) you need galvanic isulation of RS485

I strongly reccommend that you try connecting your 0V to earth-ground.

@antedeluvian:
The 220VAC links at the flexible cables are isolated from the main power by transformer.
So connecting the 0V to Earth-Ground may not be effective. So instead, should I connect
0VDC to the isolated 0 of transformer? And how should I connect, directly or over a resistor?

I am not really sure of your set up, but the isolated 0 of the transformer is floating like everything thing else. If your power supply is isolated, then there should be no problem in fixing the potential by connecting the 0V to Earth-Ground.

@antedeluvian:
Actually, the transformer is out of the controller and the isolated 220 VAC links are going to the same place with slave card with the same flexible cable, and the 220 VAC links passes over some contacts and comes back near master, 2 times more goes to slave and comes back. If we change the circuit a little, for e.g. if we switch the isolated 220 VAC with isolated 0 VAC, I mean if the 0 VAC goes to the shaft instead of 220 VAC links, the voltage we measure in the unconnected links of the flexible much less, such as 40 or 50 Volts.
I think your point will be very helpful to me, I want to ask one more thing. If I want to connect the supply 0VDC to isolated 0 VAC, should I use a resistor? Same question for earth-ground too. I will try it also.

@luben111:
The galvanic isolation is an expensive solution at this point. I should find a solution on my circuit at the moment.

Thanks a lot.

I must stress that I have NOT followed all your description (my time is limited) so my advice is first to move with caution. I would start out trying to connect from your 0V to earth-ground with a 100R resistor (use as high wattage as you have). Then measure if there is a voltage drop across it. If my understanding is correct there should be very little and you can short across it.

Thank you very much. I will try it tomorrow and write the result tomorrow evening.

I'm afraid that if you don't use galvanic isulated RS485 on big distances your schematic will work only if both ends of the line have same potential. Even if the schematic works today tomorrow some big electrical load could be switched on near your point and big currents will start to flow.

So for big distances you have only two options:
1. Use galvanic isulated RS485, each RS485 point is individually powered locally
2. Use non galvanic isulated RS485 and power all modules from a single point (other modules will receive the power through the RS485 cable).

@antedeluvian:
I have tried today the connection from the 0VDC supply to isolated 0VAC by the means of a 100R / 2 Watts resistor.
It seems to make an improvement. Now I will implement it to the place that we have problem.
I didnt try the earth-ground connection yet, in some installments, there may not exist earth-ground correctly.

@luben111:
While saying the same supply, you mean the regulated 5VDC? Because the 24 VDC is same but regulated seperately at
the communication points. By the way, the communication length is maximum 50 meters if two points of communication
exists. If it is four, then it is max 100 meters.

Thanks a lot.

You Can use Voltage Suppression diode protection SM712 across the lines and gnd....................
https://www.bourns.com/data/global/pdfs/CDSOT23-SM712.pdf

>>While saying the same supply, you mean the regulated 5VDC? Because the 24 VDC is same but regulated seperately at
the communication points. By the way, the communication length is maximum 50 meters if two points of communication
exists. If it is four, then it is max 100 meters.

You need to avoid contacting the remote units device "logical GND" with the real GND. You can power all remote modules from a single place but if your remote units have local coupling to the real GND then you need galvanic isulation. Simple test which tells you are you on the right path - if you disconnect completely the remote unit from the RS485 cable and if you measure the resistance between the remote unit "logical GND" and the real GND you should see resistance > 10 MOhm. If your sensors in the remote units are coupled to the real GND in all cases you need galvanic isulation.