I have given some thought to this, and I am not too sure that I have a solution that I completely like.
I would love to hear some more thoughts from the readers on this area. I can say that this is one area that may not be as easy as it might first appear. The other professional pieces used in pick and place machines cost much more than the (order of mag or two).
I had not only thought that there would be a need to calibrate the depth, but also the runout. One thought is to have it so that the head would have a spring stop in place of a hard stop. This would mean that as the component comes to the surface, you could even command it to go past theh surface, but a spring in the system would prevent stalling of the motors.
It is interesting to see some of the facts on solder epoxy. I have never used it before. I think that for familiarity it may be best to go forward with standard solder past, but we can keep the other in the mix. I think that many of the items may be interchangable. The main issue that I worry about with conductive epoxy is the fact that solder mask will not give any help in reducing shorts. This will mean that not only will precission despensing be a requirement, but also placement of the epoxy and components as the tolerance stack is now dealing with three different steps. The solder past gives a much larger tolerance on position, volume dispensed, and component placement.
How are these issues controlled with conductive epoxy?
I'm guessing you would plan on using luer lock type nozzles to keep costs down. So some issues here:
The taper fit of luerlock means the actual height might change depending on how hard you push it on, and different diameter nozzles would be different heights too, so some method of calibrating installed height is needed (maybe just drive down to a stage and see where the pressure changed on nozzle suction).
You could use a 16g nozzle to pick up everything from 0603 to a 44pin TQFP, otherwise you need to change nozzles, so need a nozzle rack somewhere, and a means of changing nozzles.
And related to picking up off the feeder, how do you avoid going too high/too low when picking up with the nozzle (too low can flip the part or its neighbour)? What if the nozzle snags on the tape (or more usually snags on the cover tape) ? Is there some sort of spring to avoid damage?
Another important issue is the alignment of parts as they are picked up, typically the parts on the tape will be anywhere with +/-0.5mm of centre and up to 5deg off alignment, so for parts bigger than 0805 one can rely on surface tension to straighten them when reflowed , but 0.65mm pitch could be be misplaced by one pad.
Early PnP machines use two plates that "squeeze" on to the part , simultaneously centreing and aligning in one axis, if these metal plates are gold plated and insulated, you can also check resistance/polarity of the part. Later machines used lasers to check the part was the correct size and right way up. Early machines could only rotate 0deg or 90deg (using a on/off pneumatic cylinder), to get 180deg you needed to "shuffle". Later machines can rotate to any angle. For the hobbyist a 0/90deg may be all that is needed (except you can't do a radial array or arc of LED's.)
Verification: One strip of MLCC caps looks the same as all the rest, how do you mark them? How do you ensure you haven't mixed up two strips when placing on the robot? How can you tell the part hasn't somersaulted during pickup (SOD-323 and SOD-523 are bad for this) ? Should the pickup head read a barcode on the strip? Can you even measure the suction to prove there is a part there?
Wholeheartedly agree with all your points except the last (conductive epoxy).
I have actually used conductive epoxy with large area PV panel bonding. But maybe you've had better experience than me.
I accept that the epoxy will have better dispensing properties (smaller beads, better wetting) than solder paste. However for the hobbyist / small jobber there are some issues.
Cost: a 50g pot of conductive epoxy is ~ $400 vs a 250g pot of solder paste at $50 (both pots would produce roughly the same volume of joints)
Shelf life: Solder past is good for a year or more in the fridge, premix epoxy has to be delivered refrigerated and lasts maybe 3months, two pack epoxy is good for ~ 6months on the shelf, maybe 24hrs after mixing (depends on type), you have to discard what what you don't use.
Curing: You still need to heat cure the epoxy, maybe 100C for an hour, not as critical as solder profiles, but slows down production.
Smearing: If you smear solder paste (e.g. due to bumping a component, it will just ball up, and if you smear across pads it will usually fix itself during reflow, epoxy would just make a hard to fix short circuit). Epoxy pays no heed to the solder mask.
Wicking: You can lay down a long thin paste line, and the solder from all over the pad will suck to the actual area where the pin makes contact. Secondly a slightly misplaced TQFP or fine pitch will suck itself into position due to surface tension.