PROTEL Autotrax - known issues
Ever spoken to someone about a software package they use every
day, and been surprised to find that they never, ever, have any
problems with it? Never lose a file, never get an unintended result?
Sometimes it is just a matter of knowing what NOT to do!
You must size strings in multiples of 12 thou,
if you don't they will display as if they WERE multiples of 12 thou,
until you come to manufacture, where they may change.
The hole sizes in all the protel PCB libraries are not really
correct. They are often odd values like "23 thou". I believe
that what was intended was that a designer could layout their
board, and AT THE END decide that they wanted, for instance, to
use no IC sockets, so 28 thou was the best size, or IC sockets,
so 32 thou was the best size - and then they could do a global
hole size edit of just hole sizes matching "23 thou" to these
new, real, sizes. Actually, a far better technique is to make
them all 32 thou and be done with it, as often this last step
get reliably missed in the rush to get prototypes made!
Check all your hole sizes.
You should not select autoroute "pre-route" while you have SMT
pads on the board - it will often add little PTH via pads beside
the SMT ("stringers") even if they don't need them, and even if
they short to neighbouring pads.
In Autotrax, SMT pads will only be produced without a hole if the
hole size is set to 0. Be careful that you specify a hole size of 0 for SMT pads
Octagonal pads are rotated by 22.5 degrees in gerber plots.
Most australian manufacturers then rotate them back again for you.
I elect not to use octagonal pads for this reason.
This issue may be duplicated in other versions of Protel software
Sometimes when you rotate components with
rectangular SMT pads on PCBs -
or perhaps when you rotate block selected areas, I'm not sure -
you will get the X and Y values for the pads swopped.
That is, everything will rotate by 90 degrees, but if the longest
dimension of the pad was horizontal it will still be horizontal.
This is usually immediately obvious. It is not random - it
happens when you do something unusual first (I can't quite
remember what - perhaps it happens when the pads are not part
of a component, but are "exploded" or primitives - email me if you
know what causes this!)
The "netlist:length" command in PCB does not seem to yield a useful
measurement
The PCB design rule check is pretty reliable - but reading the
report takes some experience - it tells you that there is a
fault, but often twice, and often in a way you are not
expecting. (For instance if you reverse the ground and
VCC connections on an IC, you may get hundreds of reported
errors on a large PCB, when there is only, really, two.)
DRC sometimes shows an error for strings in copper when there
is not one - be careful though - some manufacturers somehow
screw up translation of copper stings - I have had shorts.
I tend to leave a bit of extra room around copper strings now.
If you have two pads with the same pin number in a component
(which happens with some tactile switches) then PCB DRC
will throw up a lot of errors, - especially if you then
use the tactile switch as a jumper. The only solution is
to show the four pins on the schematic - or make a PCB library
component with two pins and two soldered unnumbered pads.
When producing a multilayer board, Vias cannot be connected to the
Ground plane or VCC plane. The solution is to use individual pads as vias,
as any pad can be individually edited to connect to the ground plane or VCC plane
as required. However, be very careful with any translation of the final multilayer
PCB file to other formats - while Protel V2.8 generally handles the translation
correctly, many other Protel versions lose the plane connection information
saved in an Autotrax file, although the outside layers of the PCB will seem
to be correct. Have any translation done by someone who knows what they
are doing!
The Autorouter included in Autotrax is not really very good.
You are much better off manually placing tracks. I find the
autorouter most useful as a yardstick test after I finish placement,
before I begin manual routing. (Remember to do this on a COPY
of your work!) If the autorouter can achieve 95% routing,
I know the board will be a very easy manual route job. If the
autorouter can achieve 85%-90% routing, I know the board will be a
moderate manual route job. Also, using the autorouter to process
just preroutes cleans up the "rats nest" that is used to guide placement
during manual routing.
Issues for Schematic
If you overlap a line and a pin in schematic (instead of joining
them exactly end to end) then a connection will not be made.
Helpfully, however, POST will show the line segment as an error.
You get used to this - and it is actually part of the way the
package operates - its a side-effect, not a bug as such, - the
package has no real way of determining your intentions if you
do not explicitly connect point-to-point - you may have intended
to cross over the pin and connect to another line, for instance
(seems ridiculous, but it is possible).
Sometimes you can manage to save highlighted tracks in Schematic.
Not really a bug - but it does surprise you when it happens.
If you make very large multipart components in schematic, you
will chew through the available memory very quickly. One such
component that is tempting is an mutlipart DB25 RS232 connector,
where you can place the individual pins one at a time, and the
program automatically knows the signal name. A few of these and
you will run out of memory. The standard connector pins that
schematic provides need the pin name entered every time, but
they use virtually no memory.
Issues for Schematic Libmake
Libmake is a dodgy program, it tends to crash.
Be very methodical when editing schematic libraries,
make regular backups. Open a library, edit or create a part,
back out, save and quit. Then do the next part, -
backup between if you are really cautious. Libmake seems
to work fine if you follow this pattern. But when you do a
whole session, editing a lot of parts without saving, then it
often seems that Libmake crashes just before you finish :-(
Issues for Plot
This program will seem to run more slowly if its window is not "maximized".
Also, because it expects to print directly to the printer hardware, some new
types of printers will not operate. Generally, HPLSR300 is a good type of
printer format that will certainly still work with just about all HP printers (and
others - the brother HL1240 for instance). Otherwise you can either track
down an early protel windows demo version and use that to print your
autotrax artwork (a lot of effort, really), or use postscript and ghostview
(for very good results, but an investment in installation) or use Matthais
Hartl's PCX drivers - link below.
I think that covers it. I've laid out hundreds (no exaggeration)
of boards in Autotrax 1.61. I think I would know of any other bugs
by now - but then again, maybe there are some things I just don't do,
so I just haven't discovered.
Email if you find something we should add.
Links to other Autotrax material
Can you suggest something we can add to this page?
I'd appreciate feedback by email to
...
It would also be useful to have a page or manual describing hints, and
how to use Autotrax - we haven't had time to do this - does
anyone have the material, or is it online somewhere already?