As you can see from the pic below, the spikes sometimes occur at the edge of regular logic signal and due to the considerably large spikes some system components may not work well or shut down.
How to effectively remove these unwanted spikes ?Can anyone provide me some useful materials, articles or suitable way to solve this problem?
Thank you in advance!
overmars,
It's usually best to eliminate the problem at its source. The most likely source of spikes is "ground bounce" or "Vcc/Vdd bounc". This is caused by:
.
1) Poor routing of Vcc/Vdd
2) Inadequate decoupling. For high speed logic families, the decoupling capacitors must have short connections between the supply and ground. Ground and Vcc/Vdd planes work well.
Regards,
Kral
Please clarify "regular logic signal". Is that a signal you see on circuit board full of logic? What type of logic? Is it the signal at the end of a cable? What is the time duration of that overshoot glitch? You could be seeing the results of a poor transmission line.
If it is a transmission line problem, then I suggest exploring this book:
"High-Speed Digital Design - A Handbook of Black Magic" by Howard Johnson and Martin Graham
Thank you all,
I'm sorry, I drew this signal on this pic as a simple example . I mean many times I can see digital signals carrying the spikes which are not clarified in the specific situation.
I want to search for general ways to solve this kind of problem.
e.g. The glitch is pretty large when going through a system which includes differentiators and integrators ,,,
Thanks ..
If you see those glitches all over a logic board, then it may have poor PC board layout, causing transmission line problems. In that case, definitely check out Johnson's book.
It is also possible that your oscilloscope probe is mis-compensated or has a too-long ground lead, causing visible overshoot glitches. However, that's only a measurement problem, it usually doesn't cause the board to malfunction.
Poor grounding and mismatch between the driver and the receiver can cause these overshoot problems. Sometimes a series resistor could help damping out the spikes, too.
Assuming is not a measuring fake problem, the spike presented in spike1.jpg could be removed by an RC or LC filter, where C could be the bus parasitic capacitance and R could be a small value (20 to 50 ohm). Better results are obtained if you'll use ferite chip beads for L. See for example here: **broken link removed**
Buses spikes can be filtered with 4 line arrays like DA1206 series from Steward.
At Digikey you may buy a sample kit K202 which contain a lots of different ferrite and arrays.