When the first AT32UC3 parts became available on DigiKey, I immediately bought the only model that was available.  It was a 144 pin LQFP format, but I was confident that I could get the part working eventually as long as I could get it mounted on a prototype board.  As it turns out, (and as you will see in a moment) this was terribly shortsighted of me.

Unlike its more simplistic brethren, AT32UC3A requires three different voltage sources (5V from USB, 3.3V for IO, and 1.8V for core).  There are multiple inputs and grounds for each of the lower voltages.  Using jumper wires for these inputs, along with the clock signal results in (I am quite sure) the noisiest possible RF environment.  One that is hardly conducive to a working, high speed circuit.

There are a couple of other problems, such as plugging in the wrong voltage and damaging the chip.  The final problem I ran into is that in order to get all 144 traces into such a small area, I had to run very narrow traces.  That's well and good until one of the header pins pulls loose because there is such a small bit of epoxy holding it in.

It was with great disappointment this evening that I had to declare my first foray into AVR32 a failure.  If you click on the picture below, you'll understand a bit more about it.  The good news, though, is that I recently purchased an AT32UC3B in an TQFP 48 form factor.  I have learned a good deal from my first experience, and I intend to improve dramatically for my second attempt.  To begin with, I will include all power, USB, and oscillator circuitry on the development board-along with pads for headers.

Now, this is all making a dangerous assumption: the AT32 has a built in USB boot loader just like AT90USB!