Post AX66D82KGKfrEHoQbY by gsuberland@chaos.social
(DIR) More posts by gsuberland@chaos.social
(DIR) Post #AX63MGPuArjJxqVPtY by niconiconi@mk.absturztau.be
2023-06-26T19:59:00.674Z
1 likes, 0 repeats
The biggest secret of PCB schematic footprints is that the standard library is almost useless if you want to take the classic "readability-first, signal-flow centric" approach - every single footprint needs manual editing. It's just that in the computer age with 1000+ pins, people simply stopped caring anymore, even in analog designs this style is becoming a lost art... Forcing everything to fit inside A4 paper certainly doesn't help. #electronics
(DIR) Post #AX63MH1ptq4TrUFi1A by ignaloidas@not.acu.lt
2023-06-26T22:07:59.775Z
0 likes, 0 repeats
@niconiconi@mk.absturztau.be tbh my extremist view is that schematics should just be done as code, of some sort. I've seen a commercial project trying to do that in a closed beta, I hope to maybe start an open source version of that one day.
(DIR) Post #AX64TknFjW909mYtO4 by niconiconi@mk.absturztau.be
2023-06-26T22:13:58.302Z
0 likes, 0 repeats
@ignaloidas@not.acu.lt It's appropriate for processors with 1000 pins, but I don't think it's going to work for anyalog designs which are designed and interpreted almost entirely visually. The current practice of using net labels everywhere due to digital influence already made anyalog schematics difficult enough to understand...
(DIR) Post #AX64TlVD65J2M77zu4 by ignaloidas@not.acu.lt
2023-06-26T22:20:33.149Z
0 likes, 0 repeats
@niconiconi@mk.absturztau.be Yeah, full analog wouldn't really be possible, but I think stuff like high speed digital would still be fairly doable. The blog post where I found out about the commercial program (and have forgotten it's name since...) used it to try out multiple MCU's for a sensor design - the sensor portion and the CPU portion talked through some set of signals - that got abstracted into an interface, and for testing out, exposed onto a connector on two boards - one for the sensor, another for the MCU. When the MCU was decided upon, the interface was just left to live on board in traces and a single board was routed to production. Which I think very neatly shows the potential that could be achieved with it. E.g. you can abstract whole communication channels, like SPI, and give each signal a special net class - you could still do that with high speed digital.
(DIR) Post #AX64kmipoc4D3scH6u by ignaloidas@not.acu.lt
2023-06-26T22:23:38.315Z
0 likes, 0 repeats
@niconiconi@mk.absturztau.be found the comercial stuff https://www.jitx.com/
(DIR) Post #AX66D64NZj8184l2Wm by gsuberland@chaos.social
2023-06-26T20:04:28Z
0 likes, 0 repeats
@niconiconi first thing I do on a new project is set the schematic page layout to A0. idk why we even bother with ISO paper sizes being the default for EDA software these days.
(DIR) Post #AX66D7Ht2zWkuG5LDk by niconiconi@mk.absturztau.be
2023-06-26T20:09:18.240Z
0 likes, 0 repeats
@gsuberland@chaos.social To be fair, tremendously oversized monitors are only a recent trend. Historically unable to see the whole drawings was the usual compliant. Office printer limitation is also a reason... I use A4 so I can read them on the workbench, I try to follow the "netlist style between pages/modules, classic style within a page" compromise...
(DIR) Post #AX66D82KGKfrEHoQbY by gsuberland@chaos.social
2023-06-26T20:15:36Z
0 likes, 0 repeats
@niconiconi yeah fair, I can see the printed version being useful for debugging. I use a touch screen on an arm at my bench (or rather did before I moved, and will do once it's set up again) but that's not a universal solution.
(DIR) Post #AX66D8dY1wRr5jE9ce by niconiconi@mk.absturztau.be
2023-06-26T20:16:34.942Z
1 likes, 0 repeats
@gsuberland@chaos.social I've long planned to use a similar solution, you just reminded me to actually order and install it... :woozy_baa:
(DIR) Post #AX672tNtz1fGqTaTFw by niconiconi@mk.absturztau.be
2023-06-26T22:43:19.032Z
0 likes, 0 repeats
@ignaloidas@not.acu.lt It's great for development but what about readability (which was my original rant) My opinion is that you can't solve the problem of bad schematics with bad code...I think old-school circuit schematics is a visual form of literate programming (artistic programming?). To be successful, the location of every pin in a component is custom placed at the best location that visually fits together, and when necessary, splitting one chip into multiple symbols, and even grouping subsystems into abstract macro-blocks in hierarchical schematics. The end result is that you can immediately see the entire architecture of the system.The reason that this traditional method lost favor is that (1) doing the same for large digital chips takes too much time and effort. So schematic has degraded from a high-level map to somewhere between a wiring diagram and code for machine to read (with a ton of GOTOs due to net labels). (2) reconfigurable pins make it impossible to group logic functions, at least not before the whole PCB design is finished and frozen. (3) Vendors no longer provide comprehensive technical documentation to users, so easy understandably is just not important. In the 1980s reading Theory of Operation was one of the best ways to learn electronics, even in world-class designs you basically have everything publicly documented.Doing it programmatically can probably solve the problem of bus boilerplate and dynamic pin allocation. But I don't have much confidence of readability of schematics in the wild...
(DIR) Post #AX672u0te2rAnPpc2K by ignaloidas@not.acu.lt
2023-06-26T22:49:18.278Z
0 likes, 0 repeats
@niconiconi@mk.absturztau.be hmm, some good points. But I think that maybe part of the problem is that in general, laying stuff out on a diagram is harder than laying stuff out in code. I don't know how many times have I looked at the documentation diagrams and found them utterly incomprehensible, but opening up the code, even if it wasn't that well written, it was still a lot easier to understand. I do think that a good diagram can beat just making representing stuff as code, but I also think any code can beat a poorly drawn diagram.
(DIR) Post #AX67tQ2ZoWrJBVevhY by JensHannemann@mastodon.online
2023-06-26T22:46:47Z
0 likes, 0 repeats
@ignaloidas @niconiconi Isn't that what CircuiTikZ is all about?https://github.com/circuitikz/circuitikz
(DIR) Post #AX67tQtOe96rpKN720 by niconiconi@mk.absturztau.be
2023-06-26T22:49:49.217Z
0 likes, 0 repeats
@JensHannemann@mastodon.online @ignaloidas@not.acu.lt Not really, it's a drawing and typesetting tool to generate rendered images, it's not a design tool. Nobody designs and understands circuits using CircuiTikZ, in the same way that nobody learns Euclidean geometry in PostScript or MetaPost.
(DIR) Post #AX67tRXoDtR5qfHO1Q by JensHannemann@mastodon.online
2023-06-26T22:56:47Z
0 likes, 0 repeats
@niconiconi @ignaloidas I know that, but that wasn't @ignaloidas's "extremist" premise. If you want schematics as code, I would think a SPICE netlist or some CircuitTikZ code should suffice. With CircuitTikZ, you at least get an awesome diagram. Neither of these are design tools, indeed.
(DIR) Post #AX67tS9jwrmFkJ1g92 by ignaloidas@not.acu.lt
2023-06-26T22:58:47.374Z
0 likes, 0 repeats
@JensHannemann@mastodon.online @niconiconi@mk.absturztau.be well schematics are a design tool, a very crucial one at that. Almost every PCB today starts as one, and there's few reasons to spend much time on them after you finished it.