Oli wrote:... I think it should offer holes for the stepstick / polulu, but also have SMT pads coming off these holes for a new design of board that uses a castellation array on the edges of the board. It would keep the normal holes for compatibility. That way the new board could be easily hand soldered into place and two large castellation pads could be used at each end for proper heat sinking to the motherboard, something the current designs lack (proper heatsinking through the pcb). It is unlikely they will ever need to be removed if they are the current limited variety...
Yup i though of that, i've done a few modules already using castellations - which i really like- no flaky , costly, and space consuming connectors - and you dont pay for them when you dont need the option.
The problem i see with castellations is it only works if the "on-board" version was never populated with parts, unless you have a means of desoldering those parts - in the case of the QFP package with lots of thermal vias underneath to a heavy ground plane, that is very difficult to do without a hot plate and IR/ hot air rework station.
I was looking for a way to support a surface-mount strip header or socket for the pololu's which could double as a castellated SMT land, but the trouble with those is the single row SMT hears use zig-zag style bent leads which waste a lot of board area and don't leave enough room for the IC and parts in between. and they're significantly more expensive than plain 0.1" strip headers.
Considering that the actual use of that option (installing pololus to replace a burned-out motor driver) may actually be quite low to nil, i am ok with having through-hole pads for that.
I may extend the top pad outward and make it rectangular to accept a castellated module later, as you suggest.
i will make the pad hole barrels fairly snug to the pins with minimal pad on the underside to reduce the risk of shorting when a heatsink is used on the back. When installing pololu's, I recommend installing the female sockets on the pololu's and the male pins on the board; When pins are installed along with a heatsink i would recommend putting kapton tape in between, and sanding down the pointy end end of the pins first - then assemble the board to the heatsink, carefully insert the pins, check for shorts with a meter, and solder the pins from the top side. a little tricky but not too much so.
Oli wrote:- I think it might be worth considering including atx power supply sockets. They are used as the power source in most builds and modifying connectors etc is a real pain.
I agree, especially since this allows me to power the board from the standby 5V rail and connect the remote power on-off control to an i/o line and monitor the power good signal. also the remote sense terminals will be properly terminated. The connectors are inexpensive - The only downside is the connector takes a lot of space ( the board will have to get bigger) and is only available as through-hole with very thick / long pins. I wish there were a SMT version. I considered making a custom lead bending jig to make it into a SMT connector but its seems like more trouble than its worth.
Oli wrote:- I am slightly skeptical of the ffc type connectors. I know it would be a tidy solution and in principle a good one but they are not really designed for high currents. I guess we'll see. Costs on custom lengths of this ffc are what I see as another problem. I don't really see a big problem with normal low temperature / silicon cable provided it is routed correctly. I have another solution that I have PM'd you about
. I have to go back and reread your PM on that - The FFC cable is available in standard lengths like 8", 12", 18" etc from Digikey and in stock for a reasonable price, $5-10 typ in small qty. Molex will custom make them for me even in smaller qty (like 100 pcs) for much less than that per piece price, typ less than $2, sometimes less than $1. Depending on the source and pitch, they are rated 1-3A per conductor. i am using 30 conductor , with 4 used for signals and 26 for power, thus 13A should be fine. probably more. Also remember once target temperature is reached the current is cycled on/off at a pretty low duty cycle, so even if a little heating happens in the cable, it shouldn't be significant. What i'm more worried about is the plastic material of the connector body, and how it might degrade or weaken over time, being mounted to the hot plate. that's partly why i like having it mounted backward and the cable sandwiched between it and the carriage so it cant accidentally pop open in use. I agree there isn't any major problem with using regular cable, i just find this option neater and a little more elegant so i thought it would be worth trying in the prototype. Eventually i envision replacing all of the cables with FFC, so i though trying a test case starting with the most challenging section was a good idea.
The connectors are cheap and don't take much space, so if its not popular or doesnt work well, i just wont install them, or leave them out of the production version.
Oli wrote:Imagine having a single board that means that you can buy Barts kit, the ORDuino board for less than $150 and an ATX power supply to have it mostly complete. This would be a real winning formula. Very exciting stuff
Thanks !