Microcontroller Circuits With Good EMC Characteristics

Are you talking about the susceptibility of the Pic to interferrence... or the noise emitted/generated by the pic?

I use Z86*** chips and these are horribly susceptible to negative going spikes on the 5v rail causing resets... (I inherited the chip family...but I daresay all chips have their weak spots)

Del

Thanks for your comment.

EMC legislation regulates both aspects. Therefore both are equally important.

Schorschimi

EMC legislation regulates both aspects. yes
Therefore both are equally important.. No !

Susceptibility test results depend to a great extent on how you define 'failure'.
Taking the example of a chemical pump... If you define 'Failure' as a pump running when it shouldn't then maybe a system reset could be a failure.
If you define a failure as 'pump running continuously' then system reset isn't a failure. If you define failure as 'Unit catches fire' then EMC susceptibility isn't a problem at all!

The other big problem is such a device being part of a permanently wired installation in , say a commercial laundry, should probably be tested as a system...this of course never happens...

Emission testing is another kettle of fish entirely...

Del

Is this a commercial product you are designing which will require complying with local electrical and EMC standards (an energy saving device for a wave soldering machine perhaps) or is this a practical or homework exercise?

Is it advisable to reduce the clock frequency? How far?

What does the datasheet for your microcontroller say regarding this. Check the suppliers website for application documents (many are very helpful) regarding clock frequency, suggested board layout and component selection.

Is it recommended to reduce the slope of the signals by capacitors or other means?

How fast are your signals switching? Based on the information provided, turning on a DC load once every 3 seconds is not going to be an issue.

Any other recommendations?

Do you have diodes across the coils of your solenoids (I am assuming they are DC). Nothing like an unprotected solenoid coil to wreck havoc on a microcontroller.

Have you looked on the web, you should really as there are many great guides for circuit design and layout to reduce the effects of EMI. There are also many great guides on explaining what EMI actually is and its effects.

I would suggest you invest in a random waveform generator. Use it to power the part and to input various signals to various pins. First, characterize exactly what corrupts the part's operation in a variety of circumstances. Then go about determining the kinds of events that create the kinds of waveforms that you know disrupt the processor's operation. To go about it in the manner you have suggested would only create a set of circumstances that may not cover all the possible means of ingress nor would it necessarily produce general information that could be applied in a meaningful way to situations that were not represented by your test.

As a PCB designer, I have these suggestions:

Provide a single ceramic capacitor (0.1uf below 8 MHz, .01uf above 8 MHz) on each of the PIC +5V supply leads to ground immediate and not to exceed about .200" from each PIC pinout. I would also include at least one 10uf 10V tantalum cap. near each +5V pinout (there may only be one) on the PIC to ground.

The PCB should be double sided and 2-oz cu (I would only consider 1-oz only if production quantities exceed perhaps 1000 or more and to save some $$. I would route the circuitry intelligently and flood both sides of the PCB with cu. Unless there is some odd about your ground/common reference, suggest that you bring ground to the PCB and plated-thru mounting points. Not knowing just what the SYSTEM is about, I have no opinion as to possible ground loop problems otherwise.

As mentioned earlier, make sure you include reversed biased diodes across each of the relay coils, as near as possible to the relay, SMD or thru-hole. I typically use a 1N4001 (1 KV PIV) or either class. The same holds true on your solenoid outputs.

With these, you should be fine.

I tend to wonder about your mentioning about "20 resistors". Are these lamps actually LED's? If so, consider using SIP (single inline package resistors), saves space for one.

The power supply you cite moves me to say this: In design, I select one where the max. load is about 50-60% of the p.s. rated. Saves for early failure. No sense in pushing it. I too tend to opt for one with PFC. They're only a few $$ more and satisfies all of the market.

Modifying and filtering the PIC is fine. But, consider that the real power switched is around the lamps and solenoids. The lamps are even worse, as their cold (starting) resistance is about 10% of the working resistance. Hence, all noise is 10fold from them.

Replace them with LEDs.

Transient slope control is a very effective way to cut down on the EMC noise, no matter how that is measured. Since logic switches in nanoSec, relays in many milliSecs, slowing the drivers switching speed by some magnitudes affects the operation not at all. But it reduces noise proportionally.