On 9/20/2014 11:01 PM, Jeff Liebermann wrote:
On Sat, 20 Sep 2014 18:25:05 -0700, John Larkin
wrote:

I found one of my layouts from 1985:
http://802.11junk.com/jeffl/PCB-Layout/
It's a light pen interface card for the IBM PD as a 16 bit ISA card. I
did a lousy job and am not very proud of it. However, it does show
what was typical of 1970's PCB layout technology. If anyone wants
details or more drawings, please say something as all of this is going
into the trash in a few daze.

Ooh, curved traces. I was taught to never do that, on the theory that
the tape would eventually creep in the corners.

That does happen if one stretches the tape when laying a trace. It's
especially bad with narrow traces. Traces will move, especially if
the layout is left in the sun. I used a rubber roller from my wet
photography kit, to flatten the traces and make sure they're properly
stuck to the mylar. (Incidentally, note that I used acetate instead
of mylar in the above layout. Not a good idea and I forgot why I did
it).

For RF, rounded corners are a problem due to impedance bumps.

From the source you cite:
"If you use a radius greater than three times the line width, you will
have a transmission line that is almost indistinguishable in impedance
characteristics from a straight section."

So where is the problem?


Sharp
corners are equally bad due to reflection problems. The compromise is
a chamfered corner (mitered bend):
http://www.microwaves101.com/microwave-encyclopedia/480-mitered-bends
which unfortunately also makes a tolerable fuse at the bend.

My understanding the reflection idea is also a myth but rather the real
issue is the impedance change due to the added capacitance of the
corner, which is also supported by your reference. Impedance changes
will also cause reflections, but the signal does not reflect from the
corner itself like a light beam.

--

Rick