View Single Post
  #123   Report Post  
Posted to rec.crafts.metalworking
Jim Wilkins Jim Wilkins is offline
external usenet poster
 
Posts: 3,146
Default Use of primitive tools

On Dec 10, 9:17*am, Joseph Gwinn wrote:
In article
,
*Jim Wilkins wrote:

[snip]







When circuits include a PIC or other programmable device the schematic
isn't enough. Little stubs of circuity stick out of it with no hint of
when or how they are active. Once a member of the original design team
leaves some faults in the product can become impossible to diagnose
and repair even in-house.


Companies don't release unpatented "trade secrets" hidden in the
software.


Devices like these are an extreme example.
http://www.xilinx.com/products/devices.htm
The compiler includes a randomizer so that for instance it won't
repeat a compilation that assigns a critical clock signal to an
excessively long path. This means that the very same schematic may
work perfectly one time its compiled and not the next. I spent quite a
while driving between my CAD room and the radio site, through the snow
and wolves, trying to get every function working.


A subcontractor had exactly this problem in the 1990s. *Drove them nuts
because while the compiler was randomizing paths and pinouts, the
artwork on the prototype board strangely didn't automatically rearrange
itself to follow. *

Completely stopped integration until the engineer broke the undocumented
proprietary CAD file format and figured out how to force the pinouts to
remain stable.

I don't see how randomizing paths and pinouts helps their sales or
protects anybody's intellectual property. *I bet this comes under the
heading of "don't ascribe to malice that which can be adequately
explained by incompetence".

Joe Gwinn-


Auto-routing is something like chess, a pattern-recognition problems
that a human can sometimes do better than a machine. As I understand
it randomizing prevents the tightest routing channels from filling up
and blocking completion the same way each time. The PADS autorouter I
used in the 90's would jam in ways I could easily see how to fix by
moving components, since I could evaluate the effect of lengthening
their interconnections.

I never met another PC board layout person who understood the circuit,
let alone had designed it, so the tool had to be able to fix its
problems automatically.

Xilinx explained their randomizing algorithm in the thermodynamic
terms of metastability, the peaks and valleys of activation energy,
which I followed from knowing chemistry but won't try to explain here.

jsw