How can I resolve noise problems?
Submitted By: Erik Malund FAQ Last Modified: 07/10/06
- Noise can show its ugly face in many ways. You may see occasional resets, bad data and who knows what. Beware; however, many "noise" problems I have seen have been software bugs.
- Get rid of each and every ground loop.
- Use a supervisor chip, not the RC reset.
- Put 100nf capacitors directly across all chips.
- Put a 4.7 or 10 uF tantalum directly across the '51.
- Use a 4 layer board with ground and Vcc planes.
- Use differential serial communications (422/485).
- Use a good power supply.
- Put an VDR at raw power input.
- Opto-isolate all non-differential signals.
- Put the thing in a steel box (make sure you use one big enough to take care of the heat).
- Put ferrite beads on all wires at the place where they enter the box.
- Do not drive capacitive loads (like ribbon cables) from 'normal' CMOS outputs, use chips designated 'buffers'
- slow signals can be filtered by RC networks
The easy way to see if a problem is, indeed, noise is to move the circuit to an electrically quiet location and replace all external connections with resistor and LED for output and resistor and switch for input. Noise is particularly a problem in industrial environments and when driving inductive loads.
The cure:
Rules 1 through 4 apply to any design, the remainder apply with diminishing importance to critical designs, ALL apply to designs for noisy environments.
Hint: A 485 transciever transmit or receive a differential signal, no one says you can not use it for signals other than serial transmission and a differential signal is far more 'noise resistant' than a single ended one.
this link http://www.intel.com/design/au...210313.htm has more information
Add Information to this FAQ: If you have additional information or alternative solutions to this question, you may add to this FAQ by clicking here.