>Few comments, please ignore if you've heard it/them already.

Tim,

Good Summary, I would only add a few points and comments...

0.  First spend some time working on your layout before you start.  This can
save hours of time eliminating sections of the board that are unreasonably
thick with wires.

0.1.  Decide on your pin numbering scheme before you start/send the board
out.  Most hobbyists/engineering prototypes go by component.pin number.
Most professional wire wrappers require netlists that go by board X-Y
coordinates.  It doesn't matter which you use, just make sure your wirelist
is correct for it (ie don't try to change coordinate systems in your head).

You may want to invest in the little paper numbering guides that slip over
the back of the wire wrap socket pins if you are new at this game (it can be
confusing reversing yourself and the pins the first few times you have to do
it).

>1.  I'd invest in a "proper" wire-wrap stripper such as those made by CK in
>Germany (mine for 30 AWG = 0.25mm is part number 3756).  The built-in
>stripper in the 3-in-one tool gets irksome when doing a lot of wrapping.
>Proper tool also makes it easy to get the length of stripped wire absolutely
>right.

I would agree with this statement if you are going to do a lot of wiring and
for a long period of time.  I agree that the moderately priced tools (the
squeeze powered/cheap electric wrappers) are really not worth the money
because of how easily they will break.

If it's a one off project that doesn't have too many pins, buy the Radio
Scrap 3 in 1.

>2.  When wrapping a daisy chain connection such as a data bus or clock line,
>don't go a>b, b>c, c>d etc.  Go a>b, c>d, e>f etc THEN b>c, d>e etc.  Reason
>is that you may want to change the connections, and if it's early in the
>daisy chain, you have to unwrap the whole lot.  Sometimes an unwrapped wire
>can be re-wrapped, but it won't provide a RELIABLE connection.  This
>matters, since a good wire-wrap connection is very good, certainly better
>than single-sided PCB.  One guy here one time wrote a program to sort a
>netlist to give this coherent two-level wrap.  When used with our
>long-obsolete PCB layout program to get a good placement, meant that boards
>were wrapped a lot faster than the 30 wraps/hour quoted in this list by
>someone who once did it for a living.  I suspect reason is that most of the
>time is taken identifying unambiguously the two pins to be wrapped, and/or
>checking the work with a continuity tester - actually doing the stripping
>and wrapping is a matter of seconds.

Good advice (and something that most people don't do, even when they know
they should).

>3.  If you're going to do any significant amount of wrapping (or for
>sustained periods), I'd advise shelling out the megabucks for a powered
>cut-strip-wrap tool.  The heads on these things cost over 200 UKpound, but
>they make life very easy.  (The heads are easily damaged by inexperienced
>users.)  Main problem with a powered tool is that it works by tension, so
>the wrapped board has a tendency to come out of the assembly frame with a
>pronounced curvature!  There are ways round this, which I won't bore you
>with, especially as any engineer or technician can work out their own.

Same comments to 1. above.

>4.  Just because a design works on wire-wrap, doesn't mean it will work on
>PCB!  Since wire-wrap is essentially point-to-point, and the connections are
>often at almost random orientations (and certainly with no sharp corners
>except at pins), the high-frequency characteristics are totally different to
>those of a PCB.  A PCB may suffer much more crosstalk and interference, and
>there may be signal skews that were absent on the wire wrap board.
>Wire-wrapping is perfectly capable of functioning correctly for years, so
>only change to PCB if quantity or other considerations such as space dictate.

And vica-versa.  I've seen a number of projects that don't work as wire
wrap, but we've gone ahead and embedded them into cards and they've just
worked fine (typically high frequency and the quality of the signals can be
observed to degrade).  Also note that it's just about impossible to do any
type of equal length nets using wire wraping.

I disagree with the comment that a a PCB may suffer more crosstalk and
interference; we have lots of boards here in various logic families running
as fast as 500 MHz without any worry of x-talk.  Now, the boards all have
internal ground/power planes and high speed traces are placed in between
them, but that's the price you pay.

In my department, we design high speed functional test equipment, having to
do something twice is not an option, so we've come up with design rules for
different situations.

>5.  Surface mount and wire-wrap CAN go together.  It is most unfortunate
>that Vero stopped making a wonderful system that allowed you to put together
>what amounted to custom wire-wrap surface mount footprints for PLCC
>packages.  However, there are still lots of companies out there making
>megabucks selling transition connectors/sockets/adaptors.

Been there, done that.  It's a pain.

If you are going to do that, make sure you put down the special sockets
first - I always find I have a lot of fiddling before everything lies flat
and correctly laid out.

>6.  A previous comment said don't waste two colours just on Vcc and Gnd -
>omitting to mention that most prototyping boards provide a Vcc/Gnd grid that
>you link the sockets to with Z links (so called because they are more or
>less that shape - you drop them over power pins and solder in place before
>starting wrapping).  It's also a good idea to solder a decoupler between the
>power pins at this stage - adding after wrapping is not easy, unless you
>aren't bothered about getting the capacitor closely connected to the IC
>pins, in which case why bother at all?  Z links should not be used to
>connect pins such as chip selects and output enables to power - you may want
>to change a permanent enable to controlled enable!

We usually buy sockets with the decoupling caps built it.  More expensive,
but less fiddly wiring.  Using the "Z Links" often require a special raw
card (which ironically enough, we have a ton of because IBM used to have a
standard for this and all prototyping cards were bought with this standard).
Once we run out of the cards, we'll probably abandon wire wrapping all
together.

>7.  It's easy to implement a "branching star" connection scheme in
>wire-wrap, and this may be very useful for analogue signals and to minimise
>clock skew.  Just watch out for the wrapping strategy as per comment 2
>should there be any possibility you might want to modify the net.

I disagree that this is an advantage with wire wrapping.  When I say "equal
Length nets", we typically spec them to 0.1" (100 thou).  With wirewrapping,
you'll find that some wires will be extra long on the board; because of
this, we've had to go to very small co-axes.

If you need precise equal-length nets, the best carrier and wiring solution
we've found is micro-wire boards, which are very expensive and take a long
time to get manufactured.

>8.  An undocumented wire-wrap, or worse, a documented wire-wrap that's been
>modified in unrecorded ways, is an absolute nightmare to reverse-engineer
>for the actual netlist.  It's the sort of mistake lots of people make - once.

The upsetting thing is, people being people make this mistake a lot more
than once.

9.  A wire wrapped board is much more susceptible to physical damage than an
embedded board.  Pins can get easily bent (shorting themselves against each
other) and wires can get pulled.  As well, if the stripper isn't properly
calibrated, you may find that the wires inside the insulation can be nicked,
only to break later (especially when other wires are being EC'd).


About two years ago, we (Celestica) went through a process of figuring out
the costs and advantages of different methods of board assembly.  Wire
wrapping was the worst method by far; the opportunity for defects was the
highest, the time required to build a board was the longest, the signal
integrity was hardest to ensure and the actual cost was the highest.

The reason for the exercise was that because of our accounting procedures,
you had to plan capital to get embedded boards made, wire wrapping was
considered an "expense" item (ie it could be done without going outside, or
if you did go outside, it could be done on an expense basis without a lot of
paper work and prior planning).


The design used for the comparison was 50 Dips, from 8 pin to 40 pin with a
bit of analog and some connectors.

The exercise was designed for Test Engineering, where we do a lot of
electrical designs that are used within the companay and not shipped to
customers.  As I alluded to above, we have some designs with 400+ active
components, specialized connectors, and up to 500 MHz signals running on a
variety of logic families.

We also do RF work up to 18 GHz (although this uses specialized equipment
and isn't really appropriate in this exercise).

The assembly methods compared were:

Wire Wrapping
Embedded Card with the Design Engineer using Cadence and Cooper-Chyan Routers
Embedded Card with an external company doing board layout
Micro-Wire

Micro-Wire, even though it is very expensive and boards can take up to 10
weeks to get built provided the best signal integrity.

Embedded Cards, laid out by an external company was the cheapest, but tended
to have more design problems.  But, if the design of the board was farmed
out by the engineer (ie to a student), the development process will require
less engineerign support.

The Design Engineer doing the layout was the best all round, but is
impractical in some situations (such as the engineer having to work on
different aspects of the project, prepare for the cards coming in).

For Embedded cards, we have fast proto shops that can provide up to 12 layer
boards in less than 48 hours from electronic transmission of the gerbers.
Costs are somewhat proportional to board size/complexity, but lower than
wire-wrapping in just about all cases except for very simple boards (less
than ten components with only IDC connectors).

Even for hobbyists, I would argue that the cut-off point for wire-wrapping
to an embedded card from somebody like AP Circuits is two or three chips.

Over the last two years, for all my hobby work, I've gone from doing
everything from wirewrap to prototyping on a "vector" board (the ones with
the spring contacts) and then having boards built (typically by AP Circuits)
if I want to keep the design.


Actually, for PICs, I should point out that the SIMMSticks are great
prototyping tools (I used one for one of the projects in my book) and can
reduce your prototype build drastically.  Mixing them with point to point
and wire-wrapping can be a very effective PIC prototyping method.


>That wraps up my 2p worth.  Sorry folks, couldn't resist.

Sorry Tim for the long missive on wire wrap versus other forms,

myke

"Aliens are really just the same as us, only their molecules are different"
- Will Robinson