Lots of people fuss over the exact bypass capacitor placement. Why? Lots of app notes have fluffy comments like “place as close as possible” when it comes to bypass capacitors. Why?
The fact of the matter is:
With fairly solid and closely coupled power and ground planes, the exact placement of bypass capacitors is not important. Get that: NOT IMPORTANT. Just get them out of the way so they don’t obstruct your routing.
I do not know if you noticed this, but that was actually proven by the measurements in this recent blog post on location of the connection points when doing an impedance versus frequency plot of the power distribution network (PDN). Move the connection points – or move the capacitors – it is the same thing.
- What this plot shows is two different locations of connections when we measure the power distribution network. When you think about it, that is really the same thing as two different bypass capacitor placements relative to the chip.
For frequencies up to 250 MHz these two measurements are identical. In other words, bypass capacitor placement is NOT important.
Fussing over bypass capacitor placements may even make things worse because caps placed close to chips with high routing density will tend to block the routing forcing you to add layers.
This also explains why a simple tool like the PDN Tool is a good mechanism for selecting the right bypass capacitors for your design. The specific placement is not important, but the types and number of capacitors are important.
Important: When you don’t have a good solid power plane area, but rely on (even very fat) traces, this does not work. So just don’t do that and make your life easier 🙂 An example of this is the funky Zedboard we examined a while back.