Hi Mike,
precisely. With limited time, I had to focus on development and support. Getting the complete process (paste printing, populating, soldering) stable is anything but trivial and therefore time-consuming.
* If solder paste printing goes wrong on a sample, I wipe it off and try again. For series production, this is not exactly desirable. With proper tooling, it will get better, but a decent printer (with guided squeegee) will cost you another 2k$. Still, it will be a process to learn and optimize.
* Populating (PnP). Getting components is not too hard, the re-reeling services of the major distributors are good and not overly expensive. Getting to know the quirks of the machine and the SW that comes with it (the SW that came with my machine worked just so-so) was what eventually made me give up. I can only tell about my machine, but the sub-10k China machines all have roughly the same concept:
- The block feeders are a PITA. They limit height and force you to place the aluminum caps (and other higher parts) by hand. The springs used to keep the tape down introduce so much friction that the feeder holes in plastic tapes will occasionally be ripped apart by the feeder mechanism.
- The cheap cameras (composite video, a relic from the 80s) and the image processing with OpenCV are also rather working so-so.
* Soldering will require a proper oven. The T-960 could be a good choice, but will cost you another 1k$. Vapor phase is better, but more expensive. Finding the right profile between unsoldered joints and burned boards will cost you even more time. I have the T-962, which is unsuitable for lead-free.
A PCBA house knows how to do all this.
With a PCBA, you have the extra advantage that defect devices are on them, while inhouse they are on you. And yield is never 100%, see my remarks about the stable process. If you manage 80% in the first phase, I would say you are already pretty good.
I cannot tell about the situation in the US. Over here, there are plenty of PCBA services around. Heck, I could print "Made in Germany" on my devices
In terms of cashflow, it does not make too much difference whether you pay the distributors up front or the PCBA. In terms of turnaround time, DIY is likely to be much faster.
In my opinion, you have three choices:
* The Chinese way: Chinese paste printer (i.e. framed stencils), Chinese PnP (like CharmHigh ets.), Chinese oven. This is what I tried before I gave up. About 10k$ to get started.
* Used devices. Read the posts by rx8pilot. This is what he did, if I recall correctly, it took him like half a year to get it all running. Maybe 20k$...30k$ to get started.
* Assembly service.
Get some offers, do the math (including yield), and set an appropriate price for your own time.