This is good information. Before I bias the result, I will let a few more respond to the size question, but you have given a pretty good argument for the sizes you recommend.
As to the issue of feeders, I am heavily biased against them for this type of design, and prefer something perhaps a bit of an upgrade to what you were saying with having an area off to the side. I am thinking at a minimum a moving plate that you can place rows of cut tape on (indexed to an edge) and then the plate would move to the next row of components as it switched from one component type to another. I also have a concept that is a bit of an upgrade from that, which would allow for more components to be stored (twice as many) and would not increase the footprint of the overall machine. It would almost be a conveyor style that would bring a row of components into place right under the picking head, and then it would index to the next row once complete with that component type.
There seems to be a popular panel size in use by PCB manufacturers of about 10.5 x 16" , So anything bigger would be a "special".
All of my bigger PCB's would fit 2, 3, 4, 5, or 6 onto a panel.
Note if you are using a stencil, you try to make a "set" of PCBs approximately the size of a sheet of paper i.e. 12" x 10" to make the stencil economical.
But for the machine you are thinking of with paste dispensing, you can do a single PCB at a time (this possibly reduces the niumber of feeders in use too)
A quick measure of what "big" PCB's I have lying around the lab I have 5x6 2.5x4 3x6 5x7 3x4(many) 1.5x5 all of these are revenue PCB's , So I could load 100% of my PCB's over the last 3 decades with a 5 x 7 work area. It should also be borne in mind that any PCB longer than ~8" is likely to be flexed sufficiently to crack MLCC capacitors. This puts an upper reliability limit on PCB size.
The exception to the size rule would be LED illumination strips (so maybe have a provision where you can have the pcb hanging outside the work area by removing some feeders)
So I would recommend an absolute minimum of a 6 x 8 or a more reasonable minimum of 8 x 10.
What will happen in practice is the dimensions will be limited by the feeders, (not the PCB size) (e.g a 8mm feeder might be 1/2" wide, so 20 of them for 10")
From a "number of feeders" perspective, it makes more sense to be able to pickup from the work area , i.e. use double sided tape to stick down 8" lengths of tape with bigger IC's , just next to the PCB. This requires some software support as you can't gaurantee you have stuck the tape down straight. Real PnP nachines do this anyway so they can pick from waffle packs. On the basis of using half of the work area for grid based pickup, I'd make the work area something like 12x10 or 16x10.
You can refill the "grid" really quickly using a hand pickup pencil (made from a 1ml syringe)
It is a whole lot easier to reply to comments when I am not working 12-14 hours a day.
I wanted to elaborate a bit on some of the things that you have brought up in this comment. The alignment of components will be critical to be able to push into the smaller component sizes and pad spacing.
I see this as a two pronged effort. The first effort is to ensure that the electromechanical design can actually provide the accuracy and precission that is required for the task. I have seen a lot of systems that are not designed for this, and they suffer. Even in the 3D printer world, many of these groups only speak of the min step size, but not about the repetability of their process. I think that this causes a lot of confusion and is perhaps borderline false advertising.
The other prong of this effort is to look at how to integrate a camera vision system to help in the alignment of the component.
I have a bit more time now that I can elaborate on my previous comment. One of the reasons that I think that they were having some issues for paste dispensing is that they were using a pneumatic method. While this is common, there are issues of compressability and the fact that you need to control the built up pressure in the system in a way similar to how you would control an inertial device. Added to this, there is a lot of flex in the tubing and such that connects the pump to the syringe. This is greatly influenced by temperature. This means that efforts to calibrate it are going to be difficult. I can also imagine, but I have not actually looked at in depth, that there are differences in viscosity in the solder pastes and that these too can change with respect to temperature.
To combate these, I am thinking of one of two things. The first, if going with the airpump method, there is a need to reduce the amount of line between it and the syringe. I would look at doing that by mounting the pump right to the head of the divice. There are a lot of small pumps that would be suitable for this. The other option is to go to a mechanical system. The mechanical system is going to be the most precise as it does not suffer from the "inertial" effects of the pressure system, nor the variability.
I am wondering, I have not heard too many comments about the build area size that people would be interested in for this type of machine. I would love to hear your thoughts and why you think that it should be that size.
A lot of what you are talking about gets to the heart of the mechanical design which governs accuracy and precission. Calibration also plays a part. If the machine has a constant offset, this will continue to propagate into the component placement. I feel that many of these things can be overcome with good electromechanical design.
I was talking to a company that made a small desktop machine and they said that doing 0603's was about the limit for paste dispensed from a pneumatic driven syringe, and that finer pitch IC's than 0.8mm were also a problem. I for one would be happy to place all the fine pitch IC's and QFN's by hand and then us the machine to place all of the jellybeans. Currently I do the lot with tweezers for prototypes. DFN's are a breeze to do by hand as are TQFP's, the only things I have limited success with is BGA's. If I used a paste mask they would probably work out, the solder on the balls seems to be unleaded which has such lousy flow characteristics.
Many manufacturers of membrane switches use conductive polymers for their circuit traces.. and Ariel's solution would be great for their prototypes. But most of the rest of the industry can't tolerate the higher resistivity of this material for traces.
Low current, low speed, low density.. generally will work fine using conductive polymers.
But even my simpliest designs today often involve USB 2.0 lines (with controlled impedances) or modest levels of current (0.5 amp). All point back to copper.
Regarding conductive epoxy, here's a repeat of a comment I made last year:
Around 1990 there was a promising attempt to use CNC technology to make circuit boards by Ariel Electronics (California). They created a gadget called the Circuit Writer which extruded conductive plastic wires onto a substrate, basically a 2-D plotter with an extrusion head. I actually visited Ariel and saw a Circuit Writer working. I don't think the technology got anywhere, but maybe it was just ahead of its time and with newer 3D extrusions this could be done practically. For more info, Google "ariel electronics circuit writer".
Your comments show much experience. Agreed - conductive polymers are not going to replace solder for every application.
Like so much of life... the devil is in the details.
Cost: I can get both solder and conductive polymers cheaper than what you quoted... and what ever it is today, will be different tomorrow. Often the difference is based on where you are.
Shelf life: similar observations to above (we don't like to keep solder for more than 6 months and never reuse by placing excess back into storage)
Curing: agree.. some curing heat is generally required. But much lower temps than solder. Especially the higher temps of ROHS solders.
Smearing: correct can be easier to be corrected when using solder. But always a bad situation. Even solder balls create problems.
Wicking (surface tension): this depends on the pcb surface finish. OSP (organic surface protectant) type finish, solder doesn't generally wet beyond where you put it. ENIG or HASL variations generally the solder will wet entire surface. And, yes , the surface tension can re-align a component with solder. Some designs and facilities depend on this characteristic of solder. But if often this creates as many problems as it solves (tomb stoning, floating off center because on thermal issues, solder balls floating around in the assembly). Better solution: just put the part where it belongs... and expect it to stay there. And don't bump it until it is cured or reflowed! I have seen articles by "experts" claiming the lack of wetting and automatic re-alignment with many new ROHS solders helps them reduce solder bridging! .. go figure...everyone seems to view this differently.
An item many have ignored: ROHS vs NON_ROHS compliance on components, specifically the surface finish of the connecting points. It is generally not recommended to mix soldering materials and component types. This is not an issue with conductive polymers.