Half a croissant, on a plate, with a sign in front of it saying '50c'
h a l f b a k e r y
Bone to the bad.

idea: add, search, annotate, link, view, overview, recent, by name, random

meta: news, help, about, links, report a problem

account: browse anonymously, or get an account and write.

user:
pass:
register,


                                   

Shake Mode for PCB Design

(As usual, I can’t find a category for this idea, only loads of irrelevant ones for other ideas)
  (+1)
(+1)
  [vote for,
against]

It is a feature of the human technically-inclined mind that it will produce, in the outside world, artifacts that are aligned neatly, perhaps in grid fashion, or rows and columns, or some similar apparently ‘ordered’ and neat arrangement. Architects do it, UI designers do it, even educated shelf stackers do it.

Look at a PCB for almost any product of the spirit of the age, and you’ll see components arranged all lined up one way, as if to appeal to someone standing horizontally or vertically aside the PCB. The track traces are gridded like streets in a young and immature city far away, the orientations of components are like buildings in a city, the whole thing suffers from architect disease, and for no real benefit other than comprehension at the design stage (which will be taken over by robots anyway). I’m dubious that the PCB is achieving density in packing in the same way that we do with cities anyway.

What I propose, nay, invent, is a ‘shake mode’ in PCB design, where once everything is connected up in the connectivity network, we virtually ‘grab’ the board design and shake it – quite a lot – so that contents may settle. Where they settle is a stochastic lottery, so we’d have to make the shake mode do a bit of a montecarlo (ie, do it a few thousand times) to find an optimum stochastic scatter.

The result would be what looked at first sight like a hodgepodge of components all bunched together, but because they’d retain their network connections, the traces would still maintain the circuit design. The traces could be given different ‘physical’ properties, such as stretchiness, floppiness, flippiness, flippancy, möbius- avoidability, wingly-wangliness, and the random drawing of willies – to fill in areas or provide intensional obstructions for deflection to aid packing density.

Ian Tindale, Nov 04 2018

an autorouter that makes traces as short as possible https://en.wikipedia.org/wiki/TopoR
[mitxela, Nov 04 2018]

Alpha PCB gallery http://alphapcbdesi...pcb-design-gallery/
Mentioned in my anno [notexactly, Nov 08 2018]

[link]






       // no real benefit other than comprehension at the design stage (which will be taken over by robots anyway) //   

       On the contrary. At the design stage, layout is optimized to get connection runs down to minimum lengths, thus getting the parasitic capacitance down and improving timings and waveforms.   

       The orientation of components is also a function of the way pick-and-place machines are presented with tape reels carrying devices for the PCB.
8th of 7, Nov 04 2018
  

       I have had many arguments with co-workers about this. I have always been in favour of packing things as densely as possible and ignoring fixed angles when it comes to routing. But people really don't like it because it "doesn't look neat".   

       It's true that some pick and place machines will only place things at 90 degree rotations, but with the traces there are no limits these days. People who insist that the traces have to be straight with 45 degree bends are being stupid.   

       There are some cool autorouter programs that I haven't used but make nice and fluid circuit boards, where every trace is as short as it can be, see link.
mitxela, Nov 04 2018
  

       I don’t believe a contemporarily designed pick and place system can’t place in any orientation required. What is this, the 1950s?   

       Just shake the design, let it all fall into a bunch at the bottom, then cut the pcb into that shape, it’ll be about a third of the size you started with, all the components are near each other, and the surreptitiously drawn penises will take care of parasitic capacitance and inductive coupling and other such right-hand rule and left-hand rule charge artifacts. Or you could use other shapes, I just assumed it should be penises. Either way, it’d be a far more organic design. In fact, the chips could be elliptical or circular, with a radiative pattern of lands or contacts like a flower, which would ease the ‘shake’ mode even more so.
Ian Tindale, Nov 04 2018
  

       You don't necessarily want every trace to be "as short as possible". For instance, a clock signal may need to propagate from a source to two (or more) recipient devices synchronously (kind of the point of a clock, durrrrr ....) but if one track is very short and the other very long, and the clock frequency (or rather the rise time of the edge) is sufficiently high, then there will be a problem.   

       There are a lot of factors affecting the layout of high-speed multilayer PCBs.   

       // What is this, the 1950s? //   

       If you're in Norfolk, it's the 1850's. If you're further south, yes, it might still be the 1950's where you are.   

       // chips could be elliptical or circular //   

       It's been considered. The actual "chip" inside the package would still be rectangular, as the wafer needs to be divided by scoring and snapping due to the crystal structure. However, a circle has the smallest perimeter compared to its area of any planar shape, so if you want more "edge" for connections it's not optimal. A long, thin rectangular package can have a very high ratio of perimeter to area, but suffers from different lead lengths. A square package is more consistent in track length.   

       Circular packages would have a much poorer packing fraction. Components such as electrolytic capacitors, and inductors, are cylindrical because the physics of their operation and the methods of manufacturing require that geometry, not because designers prefer it.
8th of 7, Nov 04 2018
  

       Well then, such a pair of traces would be tagged with such a constraint, so the design would have to bear that in mind. Not just bear, perhaps lion and tiger it too.   

       The thing is, nothing in nature is laid out like a PCB, and I think this is an indication of our maturity as makers. If you’ve ever taken apart a living thing, everything’s packed in there together, in 3D, in optimum orientations, and where it should be (given evolution, sometimes daft circuitous routes are evident, such as in the brain; in the eye; in the genitals – but even so, they’re not in an equidistant grid!).
Ian Tindale, Nov 04 2018
  

       // nothing in nature is laid out like a PCB, and I think this is an indication of our maturity as makers. //   

       We agree. Nature is rubbish at "design".   

       // If you’ve ever taken apart a living thing, //   

       Numerous times.   

       // everything’s packed in there together, in 3D, in optimum orientations, //   

       ... nearly impossible to access and diagnose, no replaceable parts, long lead times for modifications, systems often exhibit repetetive faults that are never satisfactorily addressed, no technical specifications available, no user or maintenance manuals, every unit a one-off build - no consistency.   

       Just rubbish, utter crap. Why do you tolerate it ? Some organic functionality can be useful in its proper place, but building complete systems that way is arrant foolishness. No wonder your species has made such a total bollocks of things.
8th of 7, Nov 04 2018
  

       Yes - length matching, differential pairs, etc all goes into the design rules, along with minimum trace thickness, min spacing, via size and so on.   

       I think the real reason people are opposed to organic-looking PCBs is simply that they are harder to design.
mitxela, Nov 04 2018
  

       Not harder to design; harder to get to work well.   

       Form follows function.
8th of 7, Nov 04 2018
  

       More has to be known about the quantum EM field as a holistic.   

       For this to work to nature's prescription, all components have to be individually working so they will shake to lowest dynamic state, burn out or half burn out in the 'as is' working field set.
wjt, Nov 07 2018
  

       I like this idea. There was a bit hoo-ha about evolutionary algorithms for circuit design a little while ago. Build a few thousand virtual circuits and cull the worst. Multiply and vary the best and rinse & repeat for a bunch of generations. It doesn't seem to have moved much beyond research. But that was about the circuit, not the physical spacing. Looking at a lot of PCBs I often get the impression they're not spatially optimized. I wonder if there's some sort of PCB "all-time classics" I could look at, there must be some real works of art out there beyond the consumer garbage I've seen.   

       //Nature is rubbish at "design"//   

       When I hear of computer code that pulls neat tricks like the multiple overlapping reading frames in some viruses... Or a machine that works as long as an alligator in the everglades.
bs0u0155, Nov 08 2018
  

       // they're not spatially optimized. //   

       No, because they're electrically optimized, which is the dominant design pressure. With most machines, delivering primary functionality overrides aesthetics every time.   

       // an alligator in the everglades //   

       ... or a Great White shark ?   

       <Hooper>   

       "Mr. Vaughn, what we are dealing with here is a perfect engine - uh, an eating machine. It's really a miracle of evolution. All this machine does is swim and eat and make little sharks, and that's all !"   

       </Hooper>
8th of 7, Nov 08 2018
  

       // I wonder if there's some sort of PCB "all-time classics" I could look at, there must be some real works of art out there beyond the consumer garbage I've seen. //   

       The closest thing I found in a couple of quick Google searches was a gallery showing what a company is capable of in PCB design: [link].   

       I think it would be worth starting a thread on the EEVblog forum for that. If you don't, I probably will (eventually).
notexactly, Nov 08 2018
  

       Baby shark, doo doo doo doo doo doo
Baby shark, doo doo doo doo doo doo
Baby shark, doo doo doo doo doo doo
Baby shark!
Ian Tindale, Nov 08 2018
  

       Can you hear that whirring noise, [IT] ?   

       It's the sound of the greatest lyricists in history spinning in their graves ...
8th of 7, Nov 08 2018
  

       // it’ll be about a third of the size you started with//   

       A bold claim since a PCB is usually no bigger than it needs to be.   

       Clearly the strategy would be to use a PCB at least twice the required size and cluster components in one corner apart from a few that randomly occupy the other space.   

       Or you could put a fraction of the components on the board and construct the rest of the circuit in free space as if the board started growing electronic mold.
bigsleep, Nov 10 2018
  
      
[annotate]
  


 

back: main index

business  computer  culture  fashion  food  halfbakery  home  other  product  public  science  sport  vehicle