SPI signal showing up on other pins of ATmega1284P microcontroller?

Question

I'm using an Atmel ATmega1284P MCU (40-pin DIP) for a project and noticed some low frequency spikes on the ADC0 pin (pin 40). After some troubleshooting I discovered that the spikes were perfectly synchronized with the SPI communication on pins 6 & 8.

I was able to replicate the issue using the following minimal circuit:

^{simulate this circuit – Schematic created using CircuitLab}

And the minimal code:

#include <SPI.h>

void setup()
{
  SPI.begin();
}
void loop()
{
  SPI.transfer(B00000000);
  delay(61);
}

Monitoring pins 8 & 40 using an oscilloscope I can see that the SCK is correctly jumping to 5V every 61ms, but what is weird is that I see a roughly 400mV spike on the ADC0 line. Here's a screen grab from my scope, the ADC0 probe is in yellow and the SCK in blue:

For reference, here's the pinout for the 1284 DIP:

I'm really not sure what is going on here, but it seems like some sort of grounding issue. I've tried adding decoupling caps between pins 10 & 11, as well as pins 30 & 31, but it made no difference. Any ideas on how to eliminate this spike on the other MCU pins?

Shot of the Vcc trace with 22uF electrolytic decoupling caps near both Vcc & AVcc:

Answer 1

Pin 40 is an input i.e. it is relatively high impedance (compared to an output or a power supply line). That's the first important thing to note.

Next you connect your o-scope probe to it and now you have basically a high impedance pin connected to what could be described as a medium length antenna.

Is it any surprise that something might be seen? You might also find that if you hold the scope probe close by in several areas, fast transients on the SCK line also couple to the probe.

General layout of breadboard could easily cause the spikes to appear (to a lesser extent) on any other pins. If it's on a PCB with a decent earthplane and the scope earth wire is connected to the closet 0V pin you may still see a little spike - this would be due to the scope probe acting as a loop antenna.

Answer 2

Following on from my comment, so the noise is on Vcc, which means it's probably everywhere. A 22uF electrolytic is not a decoupling cap, far too much inductance & ESR. Ceramic disc if thru-hole, or MLCC (multi layer ceramic chip) if SMD, at 100nF (0.1uF) will usually do the job. Sometimes in thru-hole a combo of 100n & 10n works better, depends on the cap type, PCB layout, etc. Put the decoupling cap as close to the Vcc/AVcc & Gnd pins as humanly possible, shortest (thru-hole) leads as possible. Further explanation: http://www.baldengineer.com/a-1uf-decoupling-capacitor-is-too-much.html And as also commented by Ignacio, that app-note is important, especially putting a low-pass LC filter between Vcc & AVcc. Then take another shot & show us the difference :)