Kicad - GUI is Horrific!

[ebastler]

Did the simple test again...
Actions: G hit, mirror, than shifting .. see attached
it gets as this as already posted while after the mirror action the snapping gets into account what results in a mess!!

Your result is certainly different from the one I get. Did you do the mirror operation while still in drag mode, as I described? The mirror and rotate operations work differently within this mode.

Also, which version of Kicad are you using? I tried 6 and 7, they behave the same in this respect.

Edit: Oh, hang on! I tried this again with a different footprint and surrounding schematic, and did indeed run into problems. Seems I was lucky with the footprints I dragged around before, where I did not run into any "collisions" with the surrounding wiring.

Ok, so this does not always work. And if it goes wrong, the results can be quite messy and confusing indeed...

[hpw]

Did the simple test again...
Actions: G hit, mirror, than shifting .. see attached
it gets as this as already posted while after the mirror action the snapping gets into account what results in a mess!!

Your result is certainly different from the one I get. Did you do the mirror operation while still in drag mode, as I described? The mirror and rotate operations work differently within this mode.

Also, which version of Kicad are you using? I tried 6 and 7, they behave the same in this respect.

Edit: Oh, hang on! I tried this again with a different footprint and surrounding schematic, and did indeed run into problems. Seems I was lucky with the footprints I dragged around before, where I did not run into any "collisions" with the surrounding wiring.

Ok, so this does not always work. And if it goes wrong, the results can be quite messy and confusing indeed...

Yes, some success if you drag all, all gets mirrored....

but how on a large/complex schematic/PCB... as impossible.   

Hp

[2N3055]

Did the simple test again...
Actions: G hit, mirror, than shifting .. see attached
it gets as this as already posted while after the mirror action the snapping gets into account what results in a mess!!

Your result is certainly different from the one I get. Did you do the mirror operation while still in drag mode, as I described? The mirror and rotate operations work differently within this mode.

Also, which version of Kicad are you using? I tried 6 and 7, they behave the same in this respect.

Edit: Oh, hang on! I tried this again with a different footprint and surrounding schematic, and did indeed run into problems. Seems I was lucky with the footprints I dragged around before, where I did not run into any "collisions" with the surrounding wiring.

Ok, so this does not always work. And if it goes wrong, the results can be quite messy and confusing indeed...

That was my point..

[propellerhead]

I don't often remake symbols mid-design in the schematic^[1], but where I do that kind of change on the PCB I do tend to disconnect any tracks before the operation. I far prefer a blank space into which I will re-route the tracking than the mess of rubber-banded connections that's just a, well, mess. I think for the schematic, in the example where a pin is moved, I'd just naturally check afterwards that things are as they are meant to be.

1. Thinking back, mid-design symbol changes seem to fall into two camps: small ones (like in the discussion) which I make before connecting stuff because they're tend to be obvious, and big ones where the symbol will be split into two or more smaller parts (or vice verse). In the latter it's just simpler to vape the connections and do them all again.

The in-between case in schematic capture is where I find the greatest value in proper wire-to-pin connections
surviving the symbol edit process:  Where you need to make relatively simple changes to a high-pin-count symbol.
I mean, you always do your best to get the symbol right in the first place and visualize the placement and organization
on the sheet, but as the saying goes, "Everyone has a plan until they get hit in the mouth."  Once you actually start
drawing it it's not at all uncommon for a completely different organization to reveal to you that it's more readable.

So if I'm bringing a bus or two and a handful of discrete (control) signals off of a CPU or big FPGA or something, and
I find, for example, that I now want to draw the address bus going up the drawing instead of down, the change of
direction for that bus ripper is going to have implications for the drawing of those adjacent signals.  So if I open up
the symbol and move that bus 50, 100, 200 thou to make room for routing those other signals, suddenly I've got
32 pins that have changed location.  Let's see, now... do I want those pins to rubberband so I can just move the
ripper and line everything up again, or would I prefer that all of those pins are now either dis- or mis-connected? 
Arriving at the right answer shouldn't take a lot of cycles.

[propellerhead]

The workaround for workarounds is to not break shit unnecessarily.

[2N3055]

I don't often remake symbols mid-design in the schematic^[1], but where I do that kind of change on the PCB I do tend to disconnect any tracks before the operation. I far prefer a blank space into which I will re-route the tracking than the mess of rubber-banded connections that's just a, well, mess. I think for the schematic, in the example where a pin is moved, I'd just naturally check afterwards that things are as they are meant to be.

1. Thinking back, mid-design symbol changes seem to fall into two camps: small ones (like in the discussion) which I make before connecting stuff because they're tend to be obvious, and big ones where the symbol will be split into two or more smaller parts (or vice verse). In the latter it's just simpler to vape the connections and do them all again.

The in-between case in schematic capture is where I find the greatest value in proper wire-to-pin connections
surviving the symbol edit process:  Where you need to make relatively simple changes to a high-pin-count symbol.
I mean, you always do your best to get the symbol right in the first place and visualize the placement and organization
on the sheet, but as the saying goes, "Everyone has a plan until they get hit in the mouth."  Once you actually start
drawing it it's not at all uncommon for a completely different organization to reveal to you that it's more readable.

So if I'm bringing a bus or two and a handful of discrete (control) signals off of a CPU or big FPGA or something, and
I find, for example, that I now want to draw the address bus going up the drawing instead of down, the change of
direction for that bus ripper is going to have implications for the drawing of those adjacent signals.  So if I open up
the symbol and move that bus 50, 100, 200 thou to make room for routing those other signals, suddenly I've got
32 pins that have changed location.  Let's see, now... do I want those pins to rubberband so I can just move the
ripper and line everything up again, or would I prefer that all of those pins are now either dis- or mis-connected? 
Arriving at the right answer shouldn't take a lot of cycles.

That is exactly the scenario that I mentioned.

[2N3055]

A workaround for that is to use labels liberally.

Use of net labels is INSTEAD of schematic connections. It has its use where it makes sense, like power distribution or signal busses...
What I have seen by people where "schematic diagram" has no wires at all but only net labels is not a diagram. That is literally like you would write a netlist manually (like we used to do for SPICE)... You can do that in excel, that is taxative list, not a schematic DIAGRAM.

[woody]

A workaround for that is to use labels liberally.

Use of net labels is INSTEAD of schematic connections. It has its use where it makes sense, like power distribution or signal busses...
What I have seen by people where "schematic diagram" has no wires at all but only net labels is not a diagram. That is literally like you would write a netlist manually (like we used to do for SPICE)... You can do that in excel, that is taxative list, not a schematic DIAGRAM.

As a Kicad user I am interested in this discussion and find it (apart from the incidental throat-ripping) quite enlightening. The way I use Kicad (and used Eagle before that) is that I make use of labels extensively. I only use wires in isolated parts of the schematic. These connect to the rest using labels. CPU's and the like only have labels. The same labels I use when designing the firmware and naming pins. That might be like writing a netlist manually but I find that the mixing of labels and actual wires on a diagram gives me a very good overview of what I am trying to achieve when I draw a schematic. Much better that having wired connections all over the place. Detail where I need it and helicopter view where I don't. Mind, I'm not designing the next i9 motherboard, so maybe that is why this works for me.

And AFAIK I never ran into the problems described. If I move, rotate or mirror a component I never use rubberbanding. Just delete connections and rewire. if I change pins around on a symbol in a library I delete it in the schematic, replace and rewire it. I never ended up with unwanted connections.

But it is nice to read how other EDA software tackles problems that I did not know existed 

[2N3055]

A workaround for that is to use labels liberally.

Use of net labels is INSTEAD of schematic connections. It has its use where it makes sense, like power distribution or signal busses...
What I have seen by people where "schematic diagram" has no wires at all but only net labels is not a diagram. That is literally like you would write a netlist manually (like we used to do for SPICE)... You can do that in excel, that is taxative list, not a schematic DIAGRAM.

As a Kicad user I am interested in this discussion and find it (apart from the incidental throat-ripping) quite enlightening. The way I use Kicad (and used Eagle before that) is that I make use of labels extensively. I only use wires in isolated parts of the schematic. These connect to the rest using labels. CPU's and the like only have labels. The same labels I use when designing the firmware and naming pins. That might be like writing a netlist manually but I find that the mixing of labels and actual wires on a diagram gives me a very good overview of what I am trying to achieve when I draw a schematic. Much better that having wired connections all over the place. Detail where I need it and helicopter view where I don't. Mind, I'm not designing the next i9 motherboard, so maybe that is why this works for me.

And AFAIK I never ran into the problems described. If I move, rotate or mirror a component I never use rubberbanding. Just delete connections and rewire. if I change pins around on a symbol in a library I delete it in the schematic, replace and rewire it. I never ended up with unwanted connections.

But is nice to read how other EDA software tackles problems that I did not knew existed 

Good discussion is good. Both sides learn....

Analog electronics is drawn with wires. Buses and digital with buses and Net labels. Power with net labels and some drawn parts..
It is always mixed. But schematic diagram should at least appear to graphical diagram.  So you can follow the signal down the wire....

If you have a pin that is connected to 22 other pins with only net labels scattered across A3 diagram full of chips... On screen you can use highlight net... If printed, it is useless.... hence wire going to all of them..... Also, wth time you develop intuition looking at diagrams.. You can recognize patterns.. There are many reasons why it was invented and is standardized as is......

[propellerhead]

Use of net labels is INSTEAD of schematic connections. It has its use where it makes sense, like power distribution or signal busses...
What I have seen by people where "schematic diagram" has no wires at all but only net labels is not a diagram. That is literally like you would write a netlist manually (like we used to do for SPICE)... You can do that in excel, that is taxative list, not a schematic DIAGRAM.

I agree.  It's useful sometimes, but it's very annoying to have to chase labels around a diagram and search
where the wire is that this one connects to.  As I think I may have mentioned earlier, I've seen extreme cases
in which what should have been a few pages of proper schematics turned instead into a binder full of 11"x17"
sheets, each of which had a handful of gates in which the inputs and outputs were labeled.  It was completely
accurate, and also completely unreadable and useless.  And people do get carried away even on small MCU
projects and refuse to draw a dozen lines because they habitually label the signals.

[propellerhead]

And AFAIK I never ran into the problems described. If I move, rotate or mirror a component I never use rubberbanding. Just delete connections and rewire. if I change pins around on a symbol in a library I delete it in the schematic, replace and rewire it. I never ended up with unwanted connections.

But it is nice to read how other EDA software tackles problems that I did not know existed 

You've been conditioned to accept substandard tools.  The "A"s in CAD and EDA stand for "assist[ance]" and
"automation".  Software that assists and automates would not make you do all that deleting, replacing, and
rewiring.  Instead it maintains the connections and all you need to do is a little visual tidying-up rather than
redoing work you've already done.

Demand^H^H^H^H^H^HExpect better.

[PlainName]

The "A"s in CAD and EDA stand for "assist[ance]"

Actually "aided". Essentially the same meaning, but helps your argument to not have obvious faults in it

Software that assists and automates would not make you do all that deleting, replacing, and
rewiring.

I question this. Sure, the ideal is that the tool does everything for you, but even if it just acted as a simple drawing package it would be CAD since it aids in drawing. There is an expectation that some minimum level of automation would be appropriate for an electronics design tool, but that level can be a relatively low bar. I would suggest that assuming 'CAD' means 'all singing, all dancing' is seeing things that won't necessarily be there.

[propellerhead]

The "A"s in CAD and EDA stand for "assist[ance]"

Actually "aided". Essentially the same meaning, but helps your argument to not have obvious faults in it

Over the years "aided" and "assisted" have been used interchangeably, so it's not really a distinction worth pointing out.

Software that assists and automates would not make you do all that deleting, replacing, and
rewiring.

I question this. Sure, the ideal is that the tool does everything for you, but even if it just acted as a simple drawing package it would be CAD since it aids in drawing. There is an expectation that some minimum level of automation would be appropriate for an electronics design tool, but that level can be a relatively low bar. I would suggest that assuming 'CAD' means 'all singing, all dancing' is seeing things that won't necessarily be there.

Every CAD system I've ever used has saved me that unnecessary effort, so it's not a big ask. 
I simply refuse to use schematic capture that doesn't maintain connectivity and behave that way.

[ebastler]

Over the years "aided" and "assisted" have been used interchangeably, so it's not really a distinction worth pointing out.

Yes, I am sure they have been used interchangeably. It just has not happened all that often.

[propellerhead]

Over the years "aided" and "assisted" have been used interchangeably, so it's not really a distinction worth pointing out.

Yes, I am sure they have been used interchangeably. It just has not happened all that often.

[ebastler]

Pot, kettle, black?

[propellerhead]

Yup!  Gotta keep a sense of humour about these things!

[TimFox]

In "CAT scan" for a medical x-ray tomographic scanner, the "A" stands for "axial".
The popularity of "CAD" has misled people to think it stands for "assisted".

[propellerhead]

And NMR got renamed MRI because the science-ignorant public freaked out over "nuclear"...

[TimFox]

NMRI was re-named MRI.
NMR is still the correct term for its use in chemistry labs for spectroscopy.
Personally, I prefer the German term "Kernspintomographie" =  nuclear spin tomography.

[propellerhead]

Those wacky krauts have a word for everything!

[SiliconWizard]

How did things go from KiCad to MRI?

Anyhow, nuclear bad, AI good!
Don't you want some AI in KiCad?

[thinkfat]

Those wacky krauts have a word for everything!

Us krauts, we like to be precise, and it shows.

[woody]

You've been conditioned to accept substandard tools.

Demand^H^H^H^H^H^HExpect better.

I was going to react to the 'demand' this morning, but overnight it changed 

I am not gonna demand anything from Kicad. It is a tool. It has its shortcomings. I learn about these, work with them and find myself left with a very usable piece of software that gets my job done. That gets even better every year. For a good price, and, more important, without the despicable ransomware business model some alternatives I paid a lot of money for in the past have embraced.

Kicad? I like it. And anyway, wasn't it a poor carpenter that blamed their tools? 

[propellerhead]

Sure, people like to say that, and you might even agree... until you have to put down 0402 passives
and all you have is a pair of fencing pliers and a blowtorch.

Honestly, I really do sympathize with anyone trying to make a choice in a world of SaaS and other
new forms of ugliness, and KiCAD (and other open-source packages, I'm sure, but am not aware of)
offers a badly-needed escape hatch from that world.  But that doesn't make (all of it) it good.
I'm the first to admit that since I actually used it so little, I have no idea how brilliant the layout
and other portions of the package at that end of the design chain may be.  In fact, one might make
the argument that since no package is (or can be) perfect, if you're going to make compromises,
the least desirable place to make them is in schematic capture, because people (like me) who
are offended, alienated, and driven off by the first step in the design flow will never have a chance
to find out how good the rest is.

If it were me, then, in need of a new package (remember, I'm happy with what I've got but would
like to run a more popular and modern - and preferably open-source - package as well), I'd probably
go for a bootleg/cracked version of something older, perhaps Altium.  I've never felt the drive to run
the "latest and greatest"; that early-adopter bleeding-edge stuff has always been a trap.