At 09:56 AM 5/4/01 -0400, Mamdouh Wahab wrote:
>I use Protel 99 SE on Windows 98 , a couple of times
>when I generate the netlist from the Schematics, some
>connections drop out, they are on the schematics but
>they don't appear on the netlist. Is there a way over that
>without having to review the whole netlist.
How do you know that the connections are "on the schematics"?
Simply because something appears to the eye as a wire does not mean that a
connection has actually been made.
OrCAD, or Tango, used to not connect pins which had the wire drawn too far
into the pin. It was essential that the wire *end* on the pin. Protel will
not ordinarily do this because auto-junction will add a junction if you
cross a pin. But if you turn autojunction off or if you delete what looks
like an unnecessary junction, you will be eliminating the connection even
though it appears to be there.
Ordinarily, also, an unconnected pin will generate an error when one runs
ERC. But one may decide to ignore unconnected pin warnings, particularly if
there are many of them, or one may have the matrix set so that, for
example, unconnected outputs are not flagged. *Every* unconnected pin
should generate a warning. When you are drawing the schematic and you want
to leave an explicit pin unconnected, drop a No-ERC Directive on the pin.
That way the warning will be suppressed.
Only in a dire emergency would I release a net list to PCB when it has
errors or warnings. A properly-drawn schematic should have neither.
Sometimes the cause of a warning can be pretty hard to find, but net
listing is unreliable in the presence of warnings, plus there are plenty of
conditions which can cause errors even if there are no warnings.
One of the common ones is to use two variations of the same net name.
Suppose on one page there is a net called "netl." Then on another page one
names what is intended to be the same net "net1." Do you see the
difference? I think I'll leave this one as a little puzzle to brighten up
our days.... If you know the answer, or figure it out, don't post it until
tomorrow, if you post it at all.
Anyway, *any* difference in net names will probably cause a problem. So
what we can do is to run a list of nets (not the same thing as a net
list!!! -- seems today is puzzle Friday!) and look down it. This won't help
you if the first letters of the net are different such that the net is
listed far away from the first one, but most errors it will catch.
If you want to be *entirely* certain, the only way is to have someone
manually make a net list from a printed copy of the schematic. Then this
list can be compared with the computer-generated list, using the Protel
netlist comparison tool, and any differences can be verified. Why not just
mark off the net list against the schematic? For the same reason that
scientists prefer double-blind tests if they are practical. What you expect
can influence what you see.
Using Sheet Symbols/Port connections is probably safest, since it forces
you to be explicit on all levels, and if a net is misnamed on one sheet it
will either create a warning or error if its name does not match what is on
the sheet symbol, and from then on the name does not matter.
If I were to use Netlabels and Ports Global, or Ports Only Global, I'd
probably write a utility to sort the net names from a list of nets, sorting
on last character first, then next-to-last, etc., i.e., "batx2" would be
sorted after "datx2" (probably immediately after), making it easy to find
*initial* letter errors.
Note a single-pin net warnings is often an indication that a net naming
error exists; this catches many connection errors; but it can happen that
two sections of a net are isolated by an error, and each section has more
than one pin. If all the pins are passive, for example, every net will have
more than one pin and no warning will be generated.
However, if one is thorough with nailing down ERC errors and warnings, and
one double-checks the net names, and, as another wrote, one works on grid,
preferably 0.1 inch grid, the incidence of hard errors can become low
enough that it is more economical to go ahead an make a prototype, which
will catch nearly all remaining errors. Yes, it's an expensive way to find
an error, but it is also expensive to sit there and verify a net list.
In another thirty years, perhaps when we try to name a net "bat2," the
'puter will say to us, "Is this what you want? You have another net named
"dat2." On the other hand, who knows how much of this job we will still be
doing....
"Computer, make me a device which will turn down the heat when I start to
sweat unless I have a fever. And I don't want to wear anything."
"Yes, sir, I can get most of it off the shelf for TC50 and the sensor
interface will cost TC80 to design and build through mak-yr-own.com. Shall
I go ahead?"
"Show me the plans first."
"Here's a display of the system diagram. The complete plans will take about
three hours to show you. Shall I go ahead?"
"Never mind, I think I'll just tell you when I want the heat turned down...."
[EMAIL PROTECTED]
Abdulrahman Lomax
P.O. Box 690
El Verano, CA 95433
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
*
* To join or leave this list visit:
* http://www.techservinc.com/protelusers/subscrib.html
* - or email -
* mailto:[EMAIL PROTECTED]?body=leave%20proteledaforum
*
* Contact the list manager:
* mailto:[EMAIL PROTECTED]
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
