Thank you so much Piotr, I will definetly read this.
There is so many things to understand, I feel like my head will explose. Thank you for helping me
Yes it is. Designing for EMC is one of the many aspects that is inherent in any PCB that is designed for a commercial product, and as a student getting ready for “commercial job” you have to make it a part of your design.
Yes, I guess I already told you a few times you’re biting off bigger pieces then you can chew and swallow at a single time.
Code is also involved in EMC.
There are probably many rules how software should be written to be more EMC robust. As an example you can assume that bit in flash is more robust for electromagnetic disturbance then a bit in RAM.
So you can ask yourself a question: If I copied some data to RAM then should I assume that during 15 years my device is expected to continuously work can I assume that data is 100% stable or may be I should once a day/once a hour/once a minute to rewrite it.
As the last line of defense you can assume watchdog reset. EMC standards specify what the device behavior should be when exposed to electromagnetic fields. How you stand-up after watch-dog reset can decide if device pass or not pass EMC tests.
I recommend you use a single GND. This is simpler and a lot of time it is better than having multiple GNDs that you have to carefully connect and have to be careful with the signals going from one place to a other. A single GND is easier to do right.
Is it correct that you use a 2 layer board? If so, make a GND copper pour on both sides. Try to route as many tracks on the top layer, so that the bottom pour is connected very well. Where there are gaps, because a track goes in between, make sure you connect it to the top GND with vias.
Signals always need a return path. The return current always takes the path of lowest impedance. For higher frequencies, this path is closest to the track. Since you want the return path either:
- For differential pairs: In the other wire of the differential pair.
- For all other signals: In GND.
You have to make sure there is always a continuous GND connection near each wire. That includes a GND via near every signal via.
Keep in mind: The frequency of a signal is not the clock rate. The frequency depends on the slop of the edges, the steeper the slop the higher the frequency. You can easily have a 160 MHz frequency on a µC that uses a 8 MHz XTAL.
There is a lot of details i left out to not make it more complicated. I think it is already complicated enough for you.
My 3 cents about 2 layer boards.
Since I read those articles (from links I mentioned a moment ago) my 2 layer PCB strategy was to route 100% signal and power tracks on top with whole bottom being the continuous GND. I sometimes had to use some 0Rs to not break my GND continuity.
An example of such designed PCB I showed here:
Recently I started to use QFN microcontroller cases and as VCC routing as at the picture above became not possible I decided to use 4 layer PCBs as it is the only way to still have continuous GND layer.
I meant, this is not something that was mentioned in the work I have to do in my intership.
It’s not like I have the choice… I have to work on that even though a smaller project would have been better for beginner like me
Yes. And ok I will follow your advice for now and use a single GND.
From what i understand and what @paulvdh wrote, the project is too complex for your current skill level. You have to expect that the PCB may not work.
Getting help from someone with more experience, to at least look at your board, would be very helpful. If not, produce only a few prototypes and test them.
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.