I have a board (B1) that communicates with another equipment (E1) over RS-485. The RS485 interface circuit consists of MAXIM MAX485. AT89S52 is used to implement a protocol to communicate with E1. All this is 'working fine'.

There are two occassional problems noticed:

1. If I make 10 B1 boards, 5 of them work without any hitch.
Remaining are not able to establish communication with E1.
Cable length between E1 and B1 is about 1.5 metres.
Problem remains independent of terminating resistance (120E) (enabled/disabled) across RS485 D+ and D- pins

Then if I replace MAX485 from newly purchased lot, B1 starts working. Some times even a new IC may not work. I may have to replace one or two more , before I discover 'working' IC.

But when I connect this B1 to a standard RS232/485 converter and connect to PC, B1 is able to send messages. (I have written a E1 simulator software on PC) - so as such there is no problem with MAX485 also. Then I connect B1 back to E1. Communication still does not take palce.

Cable between B1 and E1 has three wires - D+,D-,Gnd. Behaviour is same ,whether I connect this 3rd Gnd wire or not.

What could be the problem?

I can't doubt E1 because it is supposed to be a proven equipment, in production.

2. These boards are mounted in a metal enclosure.
3rd pin (Earth) of 230V cable is connected to E point on PCB and transformer body. E point is not connected to power supply Gnd. RS485 port is on a DB9 socket mounted on the metal body.
Occassionally I get a 'mild Shock' when I touch this DB9. (Not every time), that too only when E1 is connected. If I disconnect E1 then this 'mild shock' is not observed. When I touch E1 itself (E1 has a metal casing) it does not give any 'mild shock'

Any idea what could be the reason?

Voltage between Earth and Neutral is of the order of tens of volts. E1 and B1 are both powered from a Spike buster which is connected to a single AC socket on the wall. Is the problem related to earthing not being of good quality?

Thanks in advance for your thoughts on this.
KG