Altium Rules

[yobleduwop]

Has anyone sucessfully put in enough rules into the altium autoplacer and auto router to get a desirable results? im sure its possible to get it to make a decent board,
Thanks Joel

[Gladmarr]

I used Altium at work for the last year-and-a-half.  Their autoplacer is total junk!  It usually places the parts in some random array pattern regardless of your board shape.  Their autorouter usually makes several tracks out of one net, and even places several nets on top of each other.  Between that and Orcad's autorouter, I'd pick Orcad any day!  I know this doesn't help you at all, but I saw my chance to get in a dig on Altium and I took it!   

[Paul Perry (Frostwave)]

I find it difficult to conceive of a stompbox that would benefit from auto placement.
Considering the relatively small number of parts, and the constraints set by input & output.
Think to yourself: when the auto placement is operating, exactly WHAT is being optimised. How likely is it to be what YOU want optimised?

[Gladmarr]

...that's pretty much the crux of the biscuit right there.  How does the program know what your priorities are when the autoplacer starts up.  The autorouter idea makes more sense, because it will (ideally) optomize your routes for the part placement you have picked.  Unfortunately, in Altium, that barely works as well as it says it should.  I wasn't using the newest Altium Designer, by the way, the company I was with was too cheap to get the updates (and for $5000 a seat, I can sorta understand), so I can't speak to any improvements they may have made since then.

[MKB]

Backwhe I was designing boards full time, we never used autoplacement as we couldn't get good results (using OrCAD Capture and Layout Plus), it was much faster to simply place the parts with a ratsnest.  And we only used autorouting on one board, a main board for a piece of test equipment that had a bunch of 144 pin flatpack IC's.  Standard hand routing was much faster in nearly all cases.  This was in 15 years of PCB designing.

If you are designing a board with a small amount of very high pin count IC's, maybe autorouting and autoplacement makes sense.  However, for stompboxes, the setup of the net attributes alone might take longer than just hand routing the thing.

[The Tone God]

Awhile back I tried to write a set of rules for autorouter to route basic analog circuits. The rule set kept getting bigger and bigger through every test iteration I made. I finally gave up. So many exceptions and priority shuffling. I now believe that manual routing is not only better it is the only way to route analog boards. The rules set to make an autorouter competent would be enormous.

Sorry people, there is no short cut so stop with the autorouters already.

Andrew

[R.G.]

Auto-magic place and route has been the dream for a long time.

The reality is that it's only usable for digital stuff. Period.

There is no program which will place and route even simple analog effects acceptably or in an acceptable time.

I have experimented with this for years on some very compentent EDA systems, including some that are not publicly available. Same thing every time. The results either never come out or are just stupid at a glance.

I even tried taking boards I'd already done by hand and just putting the parts placement into the systems. Then I'd kick off an autoroute session, thinking that the autorouter could certainly do the routing if I had already done it by hand. It never worked, even when I let a rather large computer system grind on a simple FX layout for a long, long time.

Maybe some day. Not today, or tomorrow. Give it ten years then try again.

[Paul Perry (Frostwave)]

If they do get it working in 10 years time, my bet is that it will be via a genetic algorithm approach.
As the mathematicians would say, it's a "NP complete" problem.

[The Tone God]

I personally do not see even in ten years routers getting up to speed. The problem is with the rule/script nomenclature. It is too limited. It has to be seriously expanded maybe even adding some type of heuristic language to gain the ability to teach the router properly. I don't see any of that happening as analog continues to fall by the way side there is less interest in that type of feature especially considering the cost it would take to impliment. I think autorouters are destined to be intended for quick/cheap/lazy digital and nothing more.

Andrew

[Paul Perry (Frostwave)]

Maybe so, but note that original & completely unexpected antennas have emerged from genetic algorithms. And an antenna consists ENTIRELY of layout artifacts, form our perspective! So, maybe...... I'm going to try to hang around for 10 more years just in case

[WGTP]

Wow, cool stuff.  I'll deal with the layout, but someone come up with a Simple Tube Sounding Distortion using GA (might end up with a ZVEX Nano, but more simple).   

[Paul Perry (Frostwave)]

The problem will be putting together the rules for the prgran to judge which of any two possible circuits sounds most "tube like". A thousand times a second.
It isn't generally appreciated that OUR task (stompboxing) is FAR more complex in its aims, than merely getting a golf buggy on Mars, for example.

[WGTP]

I know that's right.  It's another one of those deals that seems sort of simple, but it is just because you haven't a clue what is going on.  Then when you find out a little it does seem like Rocket Surgery.   

[R.G.]

See my comments on autorouting in another thread.