Weird optimization idea

Post your circuits and ideas here
Post Reply
Message
Author
tinker1
Posts: 126
Joined: 19 Jul 2010, 09:33

Weird optimization idea

#1 Post by tinker1 » 31 Jul 2010, 11:59

Weird idea. Probably not useful, but here it goes:

Imagine a quad part, such as an LM324 quad op-amp. You can substitute any of the four amps for another. In a given circuit then, there are 24(=4! ) permutations. Half of these mirror the other if you don't worry about the orientation and power lines.

My idea would be to make 12 fake components with the pin numbers on the package in non-standard places. With each of these, replace the actual component in the schematic, and have PCB Layout autoplace and autoroute a board. Record the total trace length and note which is shortest. Put the real component in the schematic, and move the amps around to match the test that gave the shortest traces.

Now you could probably do it faster by just thinking about it. But this might work better on days when your brain is in a fog. If those days don't happen to you, congratulations.

Alex
Technical Support
Posts: 3897
Joined: 13 Jun 2010, 23:43

Re: Weird optimization idea

#2 Post by Alex » 02 Aug 2010, 00:03

Do you mean using 2+1 power parts in component instead of 4+1? I don't understand how it will optimize schematic or layout. Please explain.

tinker1
Posts: 126
Joined: 19 Jul 2010, 09:33

Re: Weird optimization idea

#3 Post by tinker1 » 02 Aug 2010, 02:24

Alex wrote:Do you mean using 2+1 power parts in component instead of 4+1? I don't understand how it will optimize schematic or layout. Please explain.
In a quad part, you have parts A, B, C, and D. Suppose you draw your schematic using a particular part in a spot. There may be a more efficient layout if you had used A, C, B, D or B,A, D, C , etc in the respective spots. Not a big savings, but it could be worthwhile.

So a method of trying out lets say the A, C, B, D for an LM324 quad op amp, would be to have an artificial pattern with the pin positions numbered this way:

14 13 12 11 5 6 7
1 2 3 4 10 9 8

instead of the normal

14 13 12 11 10 9 8
1 2 3 4 5 6 7

So if that first pattern gave shorter traces in autoroute, then you could swap parts B and C on the schematic, and then use the real pattern to make the real board.

My idea may well be more work than it's worth.

JackBak
Posts: 11
Joined: 29 Jun 2010, 06:28

Re: Weird optimization idea

#4 Post by JackBak » 02 Aug 2010, 04:11

One could further extend tinker1's idea to unused I/O pins on a microprocessor, especially in a QFN or LQFP type package. Of course that would effect the firmware but the PCB layout is usually well ahead of any firmly committed firmware design, so a simple change to a define file (.h in C) would be all it would take -- proper notification to the designer from DipTrace must of course be given.

User avatar
Jojje
Posts: 6
Joined: 14 Jun 2010, 08:31
Location: Sweden
Contact:

Re: Weird optimization idea

#5 Post by Jojje » 02 Aug 2010, 04:56

I think it's "pin swaping", or Automatic gateswap optimization or ?

You can swap pins in the same "part" or between "parts".


http://newsgroups.derkeiler.com/Archive ... 00046.html

http://www.kxcad.net/electronic_Xilinx_ ... g_pins.htm

Or google for "swap pins cad" or something like that.
Jag har regelbunden avföring varje dag klockan sju,
men jag vaknar aldrig fören klockan nio.

electronics-engineer
Posts: 70
Joined: 16 Jul 2011, 12:54

Re: Weird optimization idea

#6 Post by electronics-engineer » 06 Aug 2011, 04:05

I figured out a better way to do this. See http://www.diptrace.com/forum/viewtopic.php?f=4&t=1780 for details.

Post Reply