On Mon, May 23, 2011 at 7:43 PM, William "Chops" Westfield wrote: > Don't use a UBW32 ? Seriously; that's going to be a major factor > limiting how much you can shrink the board. Mmm. Could be that the > new ChipKit boards change your equation... > The UBW32 is pretty cool and I'm comfortable using it. The layout is nice and I think it's smaller overall than the ChipKit boards. > > There are a lot of components on there, and I don't think you'll be > able to shrink the board very much just by better routing and > placement. Protection circuits are annoyingly large :-( > You're right. I rearranged the board and switched some components to surfac= e mount. Check it out: Unrouted: http://postimage.org/image/22skvifk4/full/ Routed: http://postimage.org/image/22sj7z32c/full/ I got the board down to 4.9x5.6 inches. Not bad and not overly crowded. > In general, your passives are further apart than they need to be. A > set of signal-protecting diodes, resistors, and caps (as you have on > the left edge) can generally fit on a 0.1 inch grid unless there are > other reasons (power dissipation, isolation) for keeping them far apart. > Edited them. Switched to surface mount and pushed them a little closer together. > Consider quad opto-isolators instead of one-per-package devices. > I looked them up, but I found the price and simplicity of 4N25s attractive. I've used them before and I'm comfortable designing with them. > Consider resistor arrays (SIPs) where appropriate (current limiting > for LEDs, especially.) > I've looked at the arrays as you've suggested. I didn't find that it would save much space compared to the 1206 resistors. There are only 8 LEDs on there anyway. I can't quite figure out what your TTL devices are, but it looks like > you have one 20pin DIP (eg U8) driving four LEDs, which is pretty > awful, density-wise. Use higher efficiency LEDs and the PIC32 can > probably drive the LED directly (4mA). I don't know if you can get > optos for low drives or not (Hmm. How much CAN the pic32 drive on IO > pins? I see a 25mA absolute max, but not a "recommended" value...) > You could stick in something like a TLC5940 LED driver and drive 16 > leds from a 28pin chip (with internal current limiting, so you get rid > of the resistors as well.) > The TTL devices are 74 series buffers. I'm keeping the pins of the UC as unloaded as possible. The buffers drive the LEDs, and eventually the MOSFETS, etc. The other thing to think about is that $65 is NOT a lot of money. > About one day's worth of labor at minimum wage, or less than a > billable hour for a research project. (of course, it's REAL money of > the sort that escapes the university instead of going around in > circles making everyone look good, but still...) A TH design that > costs an extra hundred bucks that is robust and can be put together > easily may be preferred over a crowded SMT board that was cheaper but > more difficult to assemble) > You're right. This IS for a university, but specifically, I'm working for a professor researching in biology. Most of their funding goes toward lab tools related to biology. I quoted her a price of a few hundred dollars for everything, including time and labour, and it seems I'm still well within that budget. The project was my suggestion to her to improve certain lab procedures and so on. Also, I'm going to be using PCBGeek for making these boards: http://www.pcbgeek.com/. Their prices seem reasonable, and I get a few boards out of it. I'm only making 3 units, but if I squeeze it down enough, I might get a spare board. --=20 http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .