Placement of decoupling/bypass caps

[czietz]

Time for some electronics basics post. Tbh, this should not come as a surprise to anyone with electronics knowledge, but still I find that actual measurement results help reinforce that knowledge. I often see decoupling/bypass capacitors that are badly placed. (Before people here feel stepped on their toes, let me add that I’m not talking about a specific board I’ve seen in this forum. It’s more of a general observation.)

Anyway, here is the capacitance over frequency of a typical 100 nF decoupling capacitor (SMD, 0805 type, nothing special about it) connected to a ground plane like you would do it on your PCB. It’s roughly 100 nF, as you can see:

cap100nf_r.png

(The effects of the measurement probe were calibrated out beforehand.)

Now, here is the same capacitor, but connected to about 1.5 cm of PCB trace:

cap100nf_pcbtrace_r.png

In this configuration, it stops being a capacitor between 7 – 8 MHz and becomes an inductor instead! (Inductance is shown in orange and on the right-hand y-axis.) This is because the inductance of the PCB trace dominates the overall behavior. Recall that decoupling capacitors are supposed to filter out high-frequency current spikes. This one won’t do so above a few MHz!

While the actual resonance frequency will vary depending on the exact capacitor type and PCB layout, these exemplary measurements show why you should always place your decoupling caps as close as possible to the IC that needs them.

[rubber_jonnie]

That is really interesting, and whilst I am not an electronics expert, it is nice to see this sort of thing and enhance my knowledge.

[exxos]

Totally agree. A lot of people spend time getting capacitors physical sizes down ( which causes other issues) and totally ignore PCB trace lengths which just adds to the lumped sum of ESL/ESR. In some cases it makes what capacitor is used irrelevant as the PCB issues dominat. And of course, you can end up with a oscillator with the L&C values.