atmega2560 board

[duskglow]

Some of you might remember the soldering travails I was having... well I think I got it.  Here's the "finished" product.  I soldered on four bypass caps on the four power/gnd pairs on each side of the chip, and added some male headers.  I have some female 26 pin headers on order, I'm going to stick it on a larger grid board and add a 16MHz crystal, route some power and add a reset circuit, add SPI and JTAG headers, and add some headers to export the data lines.

Think this is something that, if I were to actually make this a "real" board, I could do something with semi-commercially?  I don't see any bare-bones 2650 adapter boards, just big old dev boards.

I just hope I didn't somehow fry the chip.  That would suck.

Oh, and in my travels I found this board: http://www.siphec.com/item/ATm2560-UT-TB.html  It's a bit expensive (about $80 from what I hear) but it's insanely high powered for an AVR microcontroller board.

[Kremmen]

A nice looking board. And the soldering work appears quite tidy as well, so no worries there.
A couple of comments though:
Firstly, do think about locating the bypass caps considerably closer physically to the power pins. To be sure this is not a gigahertz FPGA, but nevertheless the long(ish) traces from caps to chip don't do any good. This breakout was supoposed to be only for the 2560? In that case you could drop a couple of vias right next to the pwr pins and put SMD bypass caps on the other side of the board - as is usually done.
Another suggestion would be to add power regulator(s) and perhaps a clock/crystal(s) unless you want to keep this as a pure breakout and nothing else. It could make life easier for the prototyper though if this board could handle its own voltages from a generic DC supply. But of course that adds complexity so just a suggestion.
A third and final suggestion would be a separate programming header, either the 10 pin JTAG or in Atmel's case, a 6 pin PDI (2560 uses PDI,right). Lately i have moved over to the Tag-Connect header which is nice because it is a) tiny and b) is not actually a component, just a trace pattern so it saves board real estate without adding cost. Any of these alternatives would avoid the hassle of mesing with individual pins and every so often getting it wrong...

[digsys]

Definitely agree about the decoupling - WAY too light. Being a single sided PCB, it's going to be tough to add a couple tantalums and
ceramic caps though. If you don't want to go double sided, maybe run the pads under the IC and solder thin wires through before you
solder it. You'd need to dremmel a tiny slot so the IC wouldn't sit up though. Tricky.

[duskglow]

The other side of the board have other traces for an SMD chip of different pitch.  I can't use it.  I bought the board off of ebay, I'm not quite ready to make my own just yet.   But I'm considering designing my own board to let me do that right.

[Kremmen]

Oh sorry, i thought the board was your design too. Well, no problem. Just grab a free version of some commonly used PCB design tool such as KiCAD, Eagle or the one i use, DipTrace. Designing a breakout board like this one is a breeze and a good practice job. Then just hand the gerbers over to say ITead and Bob's your uncle.
Lots of people on this forum and elsewhere who can assist in the practical questions.

[duskglow]

I think I did a pretty good job for what I had available, but yeah, I think that's my next step.

Considering that I couldn't alter the board at all, is the way I did it too bad?

[Kremmen]

This is the challenge of these ready made breakout boards, that you are constrained by the default layout with no possibility to account for the specific needs of a chip. What you have now done will work up to a point, but it is not possible to say offhand where you will start to see problems. The bypass caps are there to supply the switching spikes that can be pretty abrupt. Avoiding multiple simultaneous GPIO state changes as well as configuring the MCU for soft switching will help but not remove the problem.
So your configuration is not an ideal one but whether it is "too bad" will depend entirley on the particular application you try to implement. The only thing you can be reasonably sure of is that you get problems if you try to max the MCU.

[duskglow]

The good news is it's a relatively low freq MCU - 16MHz.   I could probably throttle it down to 8 and not lose too much.

Think I might break out the CAD software soon and do this right.

I'm really only doing this for prototyping - the ultimate goal has always been to make a custom board with what I need.

[Kremmen]

One point worth noting is that clock frequency as such does not directly impact the rate of transition of various signals; the edges will be quite fast anyway. As it is these edges the bypass caps are there to supply, lowering clock frequency may not be an automatic road to success.
But try it and you will see what happens.