Re: gEDA-user: Prototyping with SMTs

["Phil Taylor" <phil@xxxxxxxxxxxxxxxxxxxxxxxx>]

Samuel

I spend more time these days drawing pcbs and just demo'ing with the actual
parts.  For some things there is no other way to go: e.g. smt's that cannot be
realistically breadboarded and circuits where your pcb traces become ckt
elements (fast analog work).

When it gets to the point of ordering a $35 breakout board or ordering a $33
custom board with exactly your circuits on it ... there is really nothing
standing in the way of 'breadboarding with pcbs'.

You'll still want to have a selection of .100" breadboad adapters for your
favorite footprints, since there is so much more selection of silicon
(especially discrete transistors) available in smt only.  This way you can
still incrementally test analog parts of a design or certain sub-blocks as
needed.  Of course you do this as much as possible before jumping off on a new
pcb design.  

One other low-ball approach I use occassionally is to mount small smt parts
upside down on .100" radio shack proto board with double stick tape.  I add
.100 pins (beforehand actually) and hand solder poor man's smt breakout leads
from 30ga. kynar wire wrap wire with tweezers and my 1/64" XYtronic iron.  You
can go down to about .025" pitch without getting really crazy.  This is really
a late-night last-ditch technique but it works fine.  put leads between the
smt pads on the part (works esp. well for leadless parts) and the .100" pins,
starting with the .100" pins first, trimming and bending to shape with the
tweezers.  Hit it with a little ball of solder and drag the ball off the part
and back up the lead for a tight little joint.

If you have to do parts with more than 14 pins this way, you'll get really
really bored.

Phil

"Samuel A. Falvo II" <sam.falvo@xxxxxxxxx> wrote:

> I'm old-school.  Whenever and whereever possible, I prefer to
> prototype what I'm working on incrementally.  For example, my Kestrel
> 1 computer is a clear demonstration of this.  It's built on one of
> those cheesy Radio Shack breadboards, runs at 4MHz (although the
> oscillator used in the pictures on the website is 1MHz), and took me
> about two weeks to build.
> 
> I would place the board's power first, then test.  Once that test
> works, then I'd place the CPU, along with a NOP generator so that it
> has some (bogus) instructions to fetch.  Test.  Once that passed, I'd
> place the address decoder -- test.  Then the RAM itself -- test.  Etc.
>  I simply would not advance to the next stage without fully testing
> the first stage.
> 
> The problem with SMTs is that they're very poor at allowing someone to
> incrementally build with them.  Wire-wrap sockets for SMT chips are
> expensive if you can find them, and they are usually quite limited in
> the number of pins they support.  One could perhaps find commercial
> SMT-to-DIP or to PLCC conversion modules, thus allowing for easier
> wire-wrapping, but you're still pretty much limited to 68 pins.  Good
> luck trying to do this with a 144-pin FPGA.
> 
> You could always make your own PCBs that break out all the pins of a
> chip onto a bunch of 40-pin headers, I suppose.  However, even this
> has the problem of being very expensive when you think about it, as
> PCB runs of these kinds of boards is typically done in very small
> quantity.
> 
> What are your approaches to dealing with this problem?  Thanks.
> 
> --
> Samuel A. Falvo II
>