A while back, I proudly wrote about my newest foray into the realm of AT90USB.  The past few days, I've been working on some code for the board.  Not much, just some simple changes to the stock Atmel CDC demo.  At first I just put the Atmel demo on the chip and plugged it into my computer.  Perfect.  A USB serial port.  Last night, I plugged it into a Linux system.  It pretty much worked.  Tonight, I decided to try getting my two computers to actually talk to the device.

Not so much luck.  I stripped out everything except a blurb which is supposed to print a string when you press a button.  It seemed to work under Windows, but behaved erratically under Linux.  The I realized, it was behaving badly under Windows as well.

The string of text just kept displaying, whether I touched the button or not.  Sometimes if I tried to transmit text, communications would come to a screeching halt and the serial port would freeze.

Finally something tickled the back of my brain.  As I watched text roll across the screen I picked the board up.  The text stopped.  I moved my finger toward the button and the text started before I could touch the button.  It stopped when I moved away.  At first I considered the possibility that the device was mocking me.

It was then that I realized my problem.  The board has no ground plane, no shielding, not even large grounded patches around the perimeter.  At 16Mhz, interference takes on strange characteristics.  After fidgeting with tinfoil (hey, it works for people running from aliens), I decided to go for plan B.  The best fit I had around the house was a piece of unused aluminum flashing.  I glued it to the board, coaxed some solder onto it (apparently aluminum isn't real friendly with solder), and grounded the plane to the top of the board.

Worked like a charm!