Homebrew PCBs

Robert Hedan wrote:

> Would it have to be PC-driven?
> 
> I've been thinking a lot about the via pins and working on a basic design;
> possibly using a soldering iron, solder wire feeder and a stamp/anvil to
> 'squish' the solder flat (for multi-layer boards).  The thing with all these
> operations is that they require specific coding:
> 
> - select a diameter for vias on a LCD input screen.
> - holes of matching diameter in the DRILL file will be treated as vias.
> - via process:
>   * drill.
>   * stamp a small circle of the via pad on both sides, to make a depression
> around the hole, or maybe use a wide V pattern.
>   * fill hole with solder.
>   * stamp solder mound with a top flat stamp and bottom anvil (a simple
> solenoid should make a great stamp, maybe spring-loaded to reduce the
> vibration on impact).

   I think this will not work as easily as just thinking it up.  There is a 
reason what I described has just a wire as the above, through, and below 
connection, namely it is very simple and connected and will work easily.  What 
you're describing here will have little if any functional advantage, and the 
notable disadvantage that it'll be unlikely to work in a stable manner across 
time and many holes.  By all means knock yourself out if you want to do it, just 
realize ahead of time it'll probably be 10 times longer to get it working and 10 
times more prone to problems once running than a simple wire system.

   Note that there is little point in reengineering a perfect system even if you 
could.  Boardhouses can still do it better, since they're doing mass quantities. 
  The goal for light duty for home prototyping etc is the simplest, easiest 
method that does what's needed good enough to work.  A pretty, flat solder via 
simply isn't really necessary, and will likely have a huge premium to 
accomplish..  Hand soldered wires work great for 95% of what most people need 
for prototyping.  Automating that so it's fast is the simplest, most effective 
vs the effort way that I've run across.  Rivets etc wouldn't be too bad, but the 
wire only needs bending, and can use the same movement that does X,Y location. 
One stone that kills two birds tends to be more reliable than seperate methods.

> 
> I'm trying to find a flexible way of making vias, without chemicals, that
> would allow me to use multiple layers one day.  I might be content making
> 2-sided prototypes for now, but I'd like to be ready for the future.
> 
> I don't really care to have your code controlling the motor control.
> Ideally, I'd like to be able to control the pulses, and have a black box
> that makes the motors work.  Look at the Gecko system, it is a similar
> two-part system if I read the stuff right: a single controller and
> individual drivers for each motor.

   That is what the dumb mode is for.  And look at what you're comparing it to, 
Gecko is about 10X more expensive than it could be.  Serious overkill for 95% of 
people just wanting to have basic computer controlled motion.  You can even work 
things up to have nearly all of their features with very low cost if you want to 
go to the effort.

> 
> If you incorporate the 'controller' aspect within your driver board, I lose
> all flexibility to add customized tools in the future.  Maybe if you make a
> PIC that we would pass parameters including all g-code data required to make
> both arcs and lines.  This PIC would then communicate with the drivers and
> have protected software just like the drivers.

   I think maybe you are not seeing the idea of dumb mode.  There is no 
difference between having my PIC there simply echoing to drive the drivers and 
not.  Software somewhere else still does the running.  I could stick an entire 
HPGL and G code engine into the PIC, but it would be sort of silly vs putting 
that software on the PC.  I have an HPGL interpreter that handles what Corel 
Draw puts out, and some utility would be gained by putting it in the PIC, but 
it's not that big of a deal to me.  Might even do G code eventually, since a 
single unit that did the G code and also all utility functions needed would be 
nice.  No hurry there either though, there are plenty of good CNC programs out 
there already.

   While it has an intelligent method for doing it's own point to point for 
developing PC software without having to bother doing the low level stepping, no 
one was planning to be there with a gun to make you use it.



   And in general the software that works through the parallel port simply 
doesn't offer direct chopping or similar, the port is too slow.  Extras like 
that have to be done in the driver side and don't affect much of anything back 
upstream.  And easy to turn off, a remote and LCD give access into internal 
settings, that's what they're for.

   My original rough prototype was done in early 2002, and looked at well before 
that and a few times since.  I've already looked at about all of the basic 
issues like this several times or more.

   And for sure no one will worry if you cut out the PIC and only use it for 
drivers.  You could split the board at the drivers and do your own glue and 
control.  Not going to tie the gates together in any fashion on the board 
though, you need seperated gates for braking, chopping, and etc.

Alan