With all the issues of the N4, at least is is most likely going to be assembling PCB's within a few days. The crusty software and fiddly feeders are significant limitations, but it offers more functionality than I had for the 1st 6-9mos of Quad ownership.
Feeders - not so bad. Really peel box is a pain. Although it is always possible to install something like fishing lure. Like "madell" did a lot of years ago on first versions.
If you are going to produce always the same things and there is no need to change reels, everything is OK.
Larger TQFP's with fine pitch - forget about them. Need to align them manually anyway. For limited quantities that's not a great problem. Largest problem is software.
So... Things that are not so simple to solve or "must know":
1. must add a dummy component and skip it at position where bottom left fiducial is located. Otherwise machine isn't going to find fiducials in automatic mode. Relative coordinates goes negative? Not a big problem, just China.
2. pattern compare by size really sucks. Tolerances aren't described, pattern list updated only on reboot. Must be great to avoid component placement on side... if possible to get it work correctly...
3. Impossible to switch vibro on/off and change it parameters on the fly (when machine already runs). In addition in my case must rise vibrofeeder for about 1mm or more, otherwise nozzle doesn't touch component perfectly even if Z = -1.
4. "jointly" vision fails on components with significant height (like HC-49). Not a big problem.
5. how works "large component" mode - nobody knows. For me it doesn't work at all. Always fail.
6. digi-reel size: yes, true. That fails. Some reels are 3mm larger than commonly used ones. In addition they are made from hard plastic (AVX tantalum capacitors for example). Hard to install.
7. new project = must enter all feeder settings again. There is no separate "default" feeder list. Only XY and feed/peel strength are default. Wish to see Z, orientation and vacuum delays too in default list.
8. tray feeder current position must be updated manually on each project restart. There is no "pause" and request to change tray when last slot processed. Impossible to bypass pick failure and proceed with next "pocket", always tries the same.
9. coordinate list isn't relative but absolute instead. In addition they use chinese metric system. No comments.
10. last 3 feeder positions are hard to align because downlooking camera doesn't move so far.
11. if component lies onto uplooking camera glass, strange things begin to happen. May be some additional checking necessary, not so hard to implement in software. BTW - DIP switches on uplooking camera aren't documented, anybody knows what they do?
12. on "skip" components are blown away from nozzle at actual position instead of trash tray location.
There is no any video on youtube where machine runs with hood closed. No reason to do that because something always goes wrong.
Another thing: nobody knows if there are no additional surprises with software like "switching to chinese language". May be it self-destroys after 1000 runs or something similar. Or is designed for use until 2019. Maybe somebody already tried to forward date?
Opinion - one N4 is not enough to be sure. Must have alternative solutions. OpenPNP - unreasonable if hardware is not "open source" with all schematics and binaries.