Last summer Atmel released exciting news.  They had gone into production with their new 32 bit series of processor, the AT32 (alternately called AVR 32).  Unfortunately the debugging tools are a bit too expensive for me, and creating a circuit containing the processor, RAM, and flash memory would take much expertise that I have not got.  Further, the chance of getting the whole thing running without the complete debugging environment is extraordinarily slim. 

Lucky for me, Atmel released a microcontroller version of the chip.  This version is much more accessible to fiddlers like myself, although it only started becoming available for purchase in the last couple of months.  In the meantime, I slathered at the mere thought of getting my grubby mitts on one of these chips.  Well, I finally did.  The only one available for order at the time was a 144 pin TQFP.  Since I've never worked with this chip, my first task was to create an adapter board that I could use to get the chip running in a test environment.

Consider that this chip is only about 3/4" on each side.  That's 36 pins in 3/4".  12 pins every quarter inch.  It took me over a week just to create a circuit board to break out the 144 pins to standard headers.  The results are in, and the following demonstration shows why photo etching is so much better than the often touted toner method.

As all my designs, this one started in Eagle.  Each time I would modify the traces to maximize clearance, and they never seemed to get better after I rendered the drawing to a TIF file.  As it turns out, the problem was in part that the rendering was happing at only 600DPI.  For parts this small, 1200DPI is much more appropriate.  After printing the mask at this resolution,  the traces looked very clean.  If you look closely you will notice that I rounded the traces.  They seemed to flow better when rounded, and I also like the organic appearance.

After exposing the board and soaking it in developer for a few minutes, I was able to breathe a bit easier.  A cursory examination revealed that the photo mask had been quite clean, and the traces looked sharp after exposure.  Now, these traces are almost impossible to see clearly with the naked eye and my concern was not for the intermittent product, but the finished one.  I dunked the board into the warm, bubbly tank of acid and used the extra time to tidy my workspace.

Unfortunately I tried to throw away the chip I was building an adapter for.  While not paying attention, the chip crashed to the floor.  It doesn't appear to have cracked, but some pins bent.  I paid for that later.  After a short eternity, the board was done etching (ok, 20 minutes).  It looked great.

After rinsing the acid off the board, I did a happy dance.  In my wildest dreams I did not expect that the board would be so clean.  After coercing congratulatory words from Aimee, I cleaned the etch resist and soldered the chip onto the board.Due to the pins bent during my cleaning spree (which is odd, because I don't often tidy anything up) it took quite a while to line the board up.  Despite excellent advice from a friend, I still have not built a reflow oven.  All 144 pins were hand soldered by yours truly.  I have not tested the work thoroughly yet, but a quick review with my handy eye loupe indicates that the project was successful.  More updates to come, hopefully after getting a program to run on the chip!