Modbus & ticker

[greenbridge]

The thing is that the unit is used in potentially "noisy" environment. I hope to fix the problem by replacing the current PIC with the one with dedicated data eeprom (like PIC16F17xx).

[Ttelmah]

That massively 'reduces' the chance of problems, but still leaves a small
part of the issue there.
Reduces, since there will no longer be code needed to write to the
program memory. Not having this means the program memory should
not change. Hurrah.
However 'part still there', since the current problem seems to imply that
the chip is getting it's RAM contents corrupted. This may well result in
slightly 'wrong' things happening in the code.
Now, I have PIC's in some very 'noisy' environments, and with really careful
PCB designs, they give no issues at all. Steps:
1) Well smoothed initial supply.
2) Good ground plane under the chip.
3) Protection on every input and output. Every signal coming onto the board
has input resistors and clamp diodes, or opto coupling. Same for every
output signal.
4) Careful layout of signal paths. Anything carrying significant power or
potentially carrying noisy signals routes well away from PIC signals, and
also the ground paths for such signals do not cross the board where
the processor components actually are.
5) Supply layout isolation. The parts feeding the ADC etc., have their own
extra smoothing.
6) Decoupling, decoupling, decoupling. Some things like SD cards for
example, need very good low ESR capacitors close to the device, and
values much larger than the 0.1uF.
7) Shielding. If there is a lot of radiated EMF, then a grounded metal casing
may well be needed. Good old Faraday...

[temtronic]

Mr. T's 'seven steps to success' does WORK !
In my former job, back in '77 I did design/build of Optical Emission Spectrometer computer systems.The rack containing an LSI-11 system and my 32 channel ADC subsystem were INSIDE the spectrometer. EMI, RF ? Think of 20KV and 200-350ma for 1-2 minutes..........
I did all '7' and then some and could reliably get 15 bits of a 16 bit ADC, almost all of the time. It's all about details and doing everything possible to eliminate any way EMI/RF/'noise' can get into your system.

Years later...one rather nasty, repeatable 'event' always occurred between 1 and 2 AM on a Sunday night, putting 'noise' on my remote energy control system (used Bell telephone wiring....). After 2 months of head scratching it was traced to bad Bell wires causing 'crosstalk'. Had them reroute my wires in another cable span..... problem solved....

Sometimes you have to think 'outside the box' and try things that 'this will never work' and magically the problem is solved.

Jay

[greenbridge]

Thanks a bunch

[Killink]