FreeCAD - is it just me, or bug city?

[janoc]

Add to that: they were built by programmers that don't understand the root problem of making a board. It is still considered as a polygon pusher. give 'em some drawing routines , store vectors and make some and/or/xor operations and off we go. -FAIL-
Everything else is an afterthought. "but it's all connected with according to schematic" . yup , and we all have seen the kind of schematics they produce ... it's amazing the router technology is not merely converting the flylines into copper. then again ..

With due respect, you are wrong in this case. The Kicad author is actually an university prof in electrical engineering. So I would think that he has made quite a few boards before he has decided to start coding his own tool for their design.

And I am not quite sure what are you getting at with the schematics quality - the schematics editor in Kicad is actually pretty decent, certainly comparable or better with Eagle, LTSpice or tools like EasyEDA. That someone designs crap with it is certainly no fault of the tool, though.

[janoc]

It wouldn't be my approach, but it all depends.
If someone is thinking longterm you can better have a proper looking interface, simply because it's much easier to show for potential investors etc.
It makes or brakes a program.
You can have wonderful things, but if people don't use it because it's to buggy or simply doesn't work well, it's still worthles.
Than it's just a great idea idea sitting on a shelf.

It's to bad that most developers (also in electronics) don't realize this concept.

Well, yes, that applies when you know from the start that you are designing a product for sale. Then you are right.

However, tools like Kicad or FreeCAD are not in this category - and not because they don't cost money. They were conceived first and foremost to address some specific need of their authors which other tools didn't at the time. Or simply because the author enjoys doing the work and is doing it for fun. People making tools for themselves don't care about impressing investors or bugs in functionality they don't use. They want something that solves their problem, pretty UI is less important. And people who write code for fun will work on what they consider fun - which is rarely UI.

It becomes a problem only once the tool "outgrows" their authors and becomes more widely popular. So you need to look at it from this angle. Judging these programs from the same point of view as commercial applications that were designed for sale from start (regardless of price) is not really fair, because their objective wasn't the same -  Jean-Pierre Charras (the original author of Kicad) certainly didn't start it 25 years ago with the goal of competing with Altium or Eagle (neither of which even existed back then - there was Protel for DOS).

[Mechatrommer]

If someone is thinking longterm you can better have a proper looking interface, simply because it's much easier to show for potential investors etc.

maybe they dont think longterm. maybe the idea is, once they got it working the way they want (personal use), they open source it, any user want to make it better, they have to code it better, and then compile it on their own, thats the idea. and a one man band SW development doesnt need investors as in multi million factory producing physical products company, even if they want to think longterm. and if they think longterm, you better be ready to pay for it.

It's to bad that most developers (also in electronics) don't realize this concept.

maybe they do, its just too much commitment on other things thats readily more profitable to them.

[Mr. Scram]

I guess you're right.

It wouldn't be my approach, but it all depends.
If someone is thinking longterm you can better have a proper looking interface, simply because it's much easier to show for potential investors etc.
It makes or brakes a program.
You can have wonderful things, but if people don't use it because it's to buggy or simply doesn't work well, it's still worthles.
Than it's just a great idea idea sitting on a shelf.

It's to bad that most developers (also in electronics) don't realize this concept.

Most engineering or otherwise specialist software isn't designed with a great UX. Doing this right takes a lot of time and effort to perfect, and the focus of development often lies elsewhere. Also, due to their niche position, customers generally don't just go elsewhere because of the looks. Standards, costs and reliability are often considered more important. Finally, there might actually be an incentive to make the software slightly cumbersome, rather than easy to use, because you can then also sell training.

So it's probably a combination of priorities and lock-in.

[b_force]

It wouldn't be my approach, but it all depends.
If someone is thinking longterm you can better have a proper looking interface, simply because it's much easier to show for potential investors etc.
It makes or brakes a program.
You can have wonderful things, but if people don't use it because it's to buggy or simply doesn't work well, it's still worthles.
Than it's just a great idea idea sitting on a shelf.

It's to bad that most developers (also in electronics) don't realize this concept.

Well, yes, that applies when you know from the start that you are designing a product for sale. Then you are right.

However, tools like Kicad or FreeCAD are not in this category - and not because they don't cost money. They were conceived first and foremost to address some specific need of their authors which other tools didn't at the time. Or simply because the author enjoys doing the work and is doing it for fun. People making tools for themselves don't care about impressing investors or bugs in functionality they don't use. They want something that solves their problem, pretty UI is less important. And people who write code for fun will work on what they consider fun - which is rarely UI.

It becomes a problem only once the tool "outgrows" their authors and becomes more widely popular. So you need to look at it from this angle. Judging these programs from the same point of view as commercial applications that were designed for sale from start (regardless of price) is not really fair, because their objective wasn't the same -  Jean-Pierre Charras (the original author of Kicad) certainly didn't start it 25 years ago with the goal of competing with Altium or Eagle (neither of which even existed back then - there was Protel for DOS).

Ok, yes very good point indeed!
I guess I naturally think about making things a bit more user friendly if I share stuff.

To be clear, I am not judging, I am simply making observations.
At this stage you can say that KiCad has arrived on a stage that they (REALLY) need to start working about these things.
I guess my reason for mentioning it, is simply because to make people more aware of it.
I noticed a lot of frustration and heavy discussions around these kind of subjects.


Btw, with all the respect, but the fact that someone is a professor doesn't always mean that know it all.
We are talking more about practical experience in this case.
So what works or doesn't work for people.
For that reason almost every professional program has the ability to completely change the interface to your own taste.

[hermit]

Add to that: they were built by programmers that don't understand the root problem of making a board. It is still considered as a polygon pusher. give 'em some drawing routines , store vectors and make some and/or/xor operations and off we go. -FAIL-
Everything else is an afterthought. "but it's all connected with according to schematic" . yup , and we all have seen the kind of schematics they produce ... it's amazing the router technology is not merely converting the flylines into copper. then again ..

With due respect, you are wrong in this case. The Kicad author is actually an university prof in electrical engineering. So I would think that he has made quite a few boards before he has decided to start coding his own tool for their design.

And I am not quite sure what are you getting at with the schematics quality - the schematics editor in Kicad is actually pretty decent, certainly comparable or better with Eagle, LTSpice or tools like EasyEDA. That someone designs crap with it is certainly no fault of the tool, though.

I wondered about that statement as well.  gEDA is another example of people using the software, writing the software.  In gEDA it really shows in the workflow sometimes. The main problem I had with it was I'm not a code monkey so the learning curve was just way too steep so I switched to Kicad.  Yeah, used Eagle when I had too.

[hermit]

To be clear, I am not judging, I am simply making observations.
At this stage you can say that KiCad has arrived on a stage that they (REALLY) need to start working about these things.

Kicad is in the process of rolling out a new release.  While most of the developers don't visit the forum they do get feedback on what the users want/need from it.  It does what I need but my 'need' is very simple.

[b_force]

To be clear, I am not judging, I am simply making observations.
At this stage you can say that KiCad has arrived on a stage that they (REALLY) need to start working about these things.

Kicad is in the process of rolling out a new release.  While most of the developers don't visit the forum they do get feedback on what the users want/need from it.  It does what I need but my 'need' is very simple.

This is going offtopic (a bit to specific).
I have seen the new release. And I think they should put energy in things that have more priority (like the interface)
Don't get me wrong, it's great what they have done so far, but the lack of certain things is frustrating as hell.

[asmi]

I have seen the new release. And I think they should put energy in things that have more priority (like the interface)
Don't get me wrong, it's great what they have done so far, but the lack of certain things is frustrating as hell.

What is a priority for you might not be so for someone else. For one, I didn't really have a problem with KiCAD GUI per se, my issues were mostly with half-implemented, outright buggy or missing features, which ultimately forced me to fork out on Orcad PCB Editor. BTW Orcad's workflow is very similar to KiCAD's, so time spent using KiCAD was not all lost to me. But before I switched, I've managed to design a somewhat complicated 4-layer board in it, so it is definitely possible even if not simple.

[free_electron]

With due respect, you are wrong in this case. The Kicad author is actually an university prof in electrical engineering. So I would think that he has made quite a few boards before he has decided to start coding his own tool for their design.

With all due respect to the professor : How many of his boards are in daily use in industrial/commercial settings ? 1 ? 2 ? 5 ? 100 ? It is easy to make 1 that only has to work on bench in a perfectly controlled environment and is allowed to goof up.

Things change when you have to make 100's of boards that are in mass production at rates of 100.000 to 100.000.000.000 copies per board. I have done PCB layout for harddisks. They make 1 million of these drives per day, and have been for the last 5 years ...  I have done boards that run at 70 GHz+ , i have 8 layer boards with 4 ounce copper per layer that carry 400 ampere. All mass production

Signal integrity , manufacturability and lots of other things come peeping around the corner. The CAD tool needs to know how to handle that.
For a lot of these entry level or homebrew tools such things are an afterthought, or even worse, the people making the tool haven't even heard that such things need to be done or such technologies exist ! How about embedded passives ? capacitive layers inside the board? resistive layers ? copper slugs ? complex design rules to mitigate the formation of anodic filaments ( like limiting how close drill strikes can be ? reliability things like tenting / encroaching , teardropping , silkscreen clip under small parts, soldermask sliver mitigation. ) Copper balancing , aspect ratio control driven by copper thickness. The software needs to be able to handle that and crosscheck that the layouter does not make violations against the rules. Most of these rules are unknown to even experienced layouters.

PCB design is NOT easy. it is not just throwing down some lines and holes. you can not do it with MS paint or autocad or a vector drawing tool. There is a difference between drawing a board and designing a board. Many PCB packages out there are good drawing tools, they are not good design tools. Simply because they behave like a vector drawing tool with some whizzbang in it. There is no 'board problem awareness' in them.

[Mr. Scram]

While I cannot do anything else than to agree with all of that, I doubt that a professor in the field wouldn't know about it at least in a fleeting sense, and probably in a fairly profound sense.

[b_force]

While I cannot do anything else than to agree with all of that, I doubt that a professor in the field wouldn't know about it at least in a fleeting sense, and probably in a fairly profound sense.

I don't wanna put down any professors, but my personal experience is exactly the opposite.
For all of the reasons free_electron is mentioning.

[Mr. Scram]

I don't wanna put down any professors, but my personal experience is exactly the opposite.
For all of the reasons free_electron is mentioning.

Correct me if I'm wrong, but I don't see any reasons listed why a professor wouldn't know about his field in at least a cursory sense.

As a side note: almost 2 billion hard drives or boards in a single revision is an impressive feat. There can't be too many people doing something similar, or we'd all drown in hard drives

[free_electron]

While I cannot do anything else than to agree with all of that, I doubt that a professor in the field wouldn't know about it at least in a fleeting sense, and probably in a fairly profound sense.

judging on the students rolling out of university : when given a board to layout ( let alone make a correct footprint ) : it is a disaster. If a few students are bad . You can blame the students . If the whole shift is bad ... it ain't the students : it's the education , or lack thereof . And that is solely to blame on the teachers. They simply don't know any better.

[free_electron]

I don't wanna put down any professors, but my personal experience is exactly the opposite.
For all of the reasons free_electron is mentioning.

Correct me if I'm wrong, but I don't see any reasons listed why a professor wouldn't know about his field in at least a cursory sense.

As a side note: almost 2 billion hard drives or boards in a single revision is an impressive feat. There can't be too many people doing something similar, or we'd all drown in hard drives

a few years ago the total volume of harddisk made in a year was 650 Million. shared between 2 manufacturers : WD and Seagate . That's 1 million a day, each manufacturer.

When you do boards for such volumes ( and these are complex boards. high speed, diff pairs, impedance controlled, length matching , power distribution , signal integrity. DDR memories , signals at 3 to 4GHz )
every corner is looked at. if you have 0.1% scrap due to a marginal footprint causing solderability errors : you will be tarred and feathered ! These boards have to be manufacturable in bulk, assembled in bulk and they have to work electrically too ! These drives are sold with 5 year warranty. if it turns out they have massive returns after 4 years because a chip is failing due to not enough thermal vias underneath ... there is going to be some fingerpointing at the layout designer.

There is not a sane person that would attempt that with Kicad. You coudl do it but it would involve a massive amount of manual work and manual checking. Simply because the tool is unaware and lacks the features required to do the crosschecks. Yes you can do a DDR4 layout in Kicad. You can also do it with masking tape on vellum. But you will be ripe for the mental bin afterwards. You need software that you can feed the playing rules so that it can verify in realtime that what you are routing is according to the rules.  Attempting such work using a ripup and retry approach is sheer insanity. Tools that are stuck in : user place , calculate , crosscheck , flag cycles so the user can retry are useless. You need  systems that can do this in realtime , as you are routing. Computers are bad at solving the spatial problems. ( that's why there still are no working autoplacers , and autorouters are mainly crap as well ).

The good software combines the spatial solving skills of the human with the calculation speed of the computer to perform the crosschecks. And these crosschecks can be electrical, mechanical ,thermal ,signal integrity , manufacturability, solderability , current load capacity and many many other things. These things can easily be cast into rulesets that can be crunched by all that horsepower in the workstation. The cad program can show, in realtime, what all options there are. The human can then pick the spatially best one.

In theory , if all is electrically connected it should work... in theory ... and that is what professors tend to excel at.

[b_force]

I don't wanna put down any professors, but my personal experience is exactly the opposite.
For all of the reasons free_electron is mentioning.

Correct me if I'm wrong, but I don't see any reasons listed why a professor wouldn't know about his field in at least a cursory sense.

Oh, well, there is an extremely simple explanation for that.

Because a professor is teaching classrooms and is involved in other fields as well.
Therefore he can't spent the time and energy that's needed to get to know all of these things.

In theory , if all is electrically connected it should work... in theory ... and that is what professors tend to excel at.

This is also exactly my experience.
Absolutely nothing wrong with that, not at all even!
But it's just a different expertise.

A baker bakes bread and a butcher chops meat.
A good professor is smart, but just doesn't always have the practical experience.
free_electrons example is a little much, but I have seen exactly the same experience first hand multiple times when having to deal with interns and professors.

[janoc]

To be clear, I am not judging, I am simply making observations.
At this stage you can say that KiCad has arrived on a stage that they (REALLY) need to start working about these things.
I guess my reason for mentioning it, is simply because to make people more aware of it.
I noticed a lot of frustration and heavy discussions around these kind of subjects.

What gives you the idea they aren't working on it? However, do keep in mind the hobby/non-profit/pet-project aspect of it. It still applies, even though it is not a one man project anymore. There is no deep pocketed company behind it. People were up in arms with excitement that CERN is behind it but CERN people (which were 1 or 2 part-time folks, I believe) are working first and foremost on things that CERN projects needed. So they added stuff like the push-and-shove routing.

And re lot of frustration and heavy discussion - that's the issue. Lot of discussion but nobody actually contributes to the project (whether by code or money) and expects someone else to fix things. As if there was "someone else" to do it. There isn't, it is as simple as that.

Then there is also a lot of griping because Kicad/FreeCAD/whatever doesn't work in the same way people are used from other tools. Or lacks a specific feature. So it is immediately declared useless, piece of crap, good for nothing toy. Etc. As if someone was forcing them to use the software.

Criticism is good but it needs to be constructive. Declaring something unusable only because it doesn't work the same way as Altium/Eagle/whatever and demanding it to be changed, as is often the case, is not really helpful. Granted, some developers may not be the easiest people to work with, but that is not really an excuse. Being able to compromise and to deal with difficult personalities is also part of engineering.

Btw, with all the respect, but the fact that someone is a professor doesn't always mean that know it all.
We are talking more about practical experience in this case.
So what works or doesn't work for people.

I certainly wasn't implying that. But when someone works in the field he did (I believe he is retired now), he certainly wasn't only a theoretician. And the folks from CERN who use Kicad to design and build boards for their particle detectors also don't know what they are doing? CERN is hardly so cash-strapped to not be able to afford something like Altium, so I guess there may be some other reasons why they are using (and supporting) Kicad .

For that reason almost every professional program has the ability to completely change the interface to your own taste.

I guess we are not living in the same world then. Or Microsoft, Autodesk,  Unity3d, Unreal, Blender, etc. (just to name a few things I use on a daily basis) didn't get the memo. At best you can customize colors, screen background and hotkeys (sometimes not even those). And that's that.

Oh and I guess Apple is not professional at all by this standard - the vendors mentioned above are mostly lazy to add customization options, on the other hand Apple actively removes them or discourages their use for the sake of the control of the 'experience'.

Or my Emacs, which is notorious for its customization possibilities, is the poster child for usability. 

Sorry but this is a silly argument.

[Mechatrommer]

this is way off-topic already.... my suggestion to OP is to have a look at the FreeCAD code, and see if anything he can contribute to the community

[free_electron]

What gives you the idea they aren't working on it?

The fact that it is not in there from the beginning means they started off wrong ...

We have first discovered fire, then the wheel , then the combustion engine.
When designing a new piece of software is version 1.0 going to let me design at 'fire' level , or at 'combustion' level ?

The same goes for making a new PCB tool. You first need to see what is required of a MODERN PCB tool before making yet another vector drawing program that you can wrench board design on to.
Modern PCB tools go far beyond 'drawing' .bill of materials, variant management , supply line , integration with other tools ( like mechanical cad ). Linking realtime between schematic and PCB.

The fact that kicad has no integrated libraries, and you still need to edit files before you can go from schematic to PCB means the authors do not understand the problem. They treat the schematic capture tool as a graphical tool to make a netlist. A second tool then reads the netlist and lets you wire packages accordingly. The fact that footprint assignment is done in the PCB tool , and not in schematic means they don't understand multiple issues. 

If i select a part in the schematic , this needs to be something that can be purchased. So it has a fully qualified part number , which in includes the pcb footprint. Simply putting down 7805 in the schematic is useless. who makes it ? what package ( can be anything from to220 , to92, sot89 , sot223 to any kind of other package )

Ultimately the PCB design system needs to deliver an accurate bill of material that matches schematic and layout. The pick and place data needs to be there.
The output is a complete set of machine readable data that allows the board house to make the board, and the assembly house to assemble the board.

Efficient PCB design tools set things up in such a way that you only need to enter the minimum amount of data required and they derive all data required from that.
If you need to muck with intermediate files or do things by hand it means the PCB tool is lacking functionality. And all PCB tools lack functionality. We live in a complex world.

But, it does not do any good to make another PCB tools that uses methodology of 40 years ago. That is just an academic experiment in making yet another tool that does things we could do 40 years ago.

[hermit]

Reading a lot of what you wrote I have to wonder when the last time you used Kicad was?  Either that or you didn't take the time to learn the software.  I'm an over educated hobbyist that helps a friend with his startup from time to time.  I'd never pretend to tell someone that depends on this type of software for a living what they should be using, but let's be fair about what we say about the software here.  Some of what  you describe is not the software I've been using of late.

You CAN add footprints in the capture software.  Either as a part edit or as a drop down that brings up all parts.  I don't use BOM but users on the Kicad forum aren't generally thrilled with it so they have written and shared scripts and customizations.  From my perspective I'm at least a tad better off with it than I am with Eagle.  My first was ORCAD pre Windows.   I liked that a lot.  Hit r and a resistor shows up.  Hit c and you get  cap.   

And just to stay on topic.  Someone just wrote a FreeCad plugin for Kicad to do 3-D part modeling. Seeing that some of the more prolific 3-D model contributers are using FreeCad, it is at least good enough for that. 

[free_electron]

1) Reading a lot of what you wrote I have to wonder when the last time you used Kicad was? 
-snip-
2) You CAN add footprints in the capture software.  Either as a part edit or as a drop down that brings up all parts.
-snip-
3) they have written and shared scripts and customizations. 

1 : doesn't matter. it should have been in there form the beginning. If i buy a car it should be complete. Having a car where the wheels and motor will come later is useless.
2 : wrong approach. There is no need to 'assign footprints'  . There needs to be a library where parts are defined. Once you are in schematic capture phase the part should already be defined.
3 : the fact that someone else needs to write that means the original developers did not do a good job.

And that is what i have been saying : the original developers do not understand the problems of real world PCB design. They started off with the idea:  let's make another vector drawing tool that can be used to generate copper structures. What they have done is made a fancy rubylith editor. Orcad 3.1 under DOS was more advanced that Kicad is today. It had integrated libraries, forward/backward annotation , pinswap and many other features that are not in base KiCad. you need add-ons and plug-ins to get there. Not good.

[Mr. Scram]

I respect your professional opinion and experience a lot, but I think you might be a bit off the mark here. It seems to be a horses for courses thing. There is no universal outline to which all PCB design packages should adhere. Not everyone is interested in designed complex n-layer boards with all sorts of high speed signals buzzing about. I could complain that my neighbour's sail boat is utterly inadequate for international container shipping, and I wouldn't be incorrect, but I don't think it is a sensible complaint either.

Whether something has been intended for something from the start doesn't seem to be too relevant. What something was yesterday doesn't matter, but how it pans out today does. There are numerous examples of something evolving over the years, growing from an awkward tool for another job, into a nice and effective tool. I can complain about Windows not supporting NTFS back when I used it in the 90's, but it sure does now. Dismissing it because it didn't back then seems unreasonable. Whether the community added something or the developer seems mostly semantic. It's a matter of different organisational structures.

As a final thought: whether there should be a library or predefined parts doesn't seem to be an universal truth. It looks a lot more like a matter of opinion. Whether one approach is better or worse than the other, sure, but that too is up for debate.

[hermit]

As a final thought: whether there should be a library or predefined parts doesn't seem to be an universal truth. It looks a lot more like a matter of opinion. Whether one approach is better or worse than the other, sure, but that too is up for debate.

https://www.eevblog.com/forum/projects/favorite-schematic-pcb-software-with-large-database-for-components/msg1318873/#msg1318873

[janoc]

What gives you the idea they aren't working on it?

The fact that it is not in there from the beginning means they started off wrong ...

We have first discovered fire, then the wheel , then the combustion engine.
When designing a new piece of software is version 1.0 going to let me design at 'fire' level , or at 'combustion' level ?

The same goes for making a new PCB tool. You first need to see what is required of a MODERN PCB tool before making yet another vector drawing program that you can wrench board design on to.
Modern PCB tools go far beyond 'drawing' .bill of materials, variant management , supply line , integration with other tools ( like mechanical cad ). Linking realtime between schematic and PCB.
...

I respect your experience in the field, but did you actually bother to read the replies I wrote to others?

Kicad is over 20+ years old program, with the first (unpublished) version from 1992! DOS version of Protel was the standard back in the day when the development was started, SMD were a newfangled novelty and boards with more than 2 layers were almost unheard of (at least in the "market" Kicad was targeting).

So what are you talking about "modern PCB" tools or proof of starting off wrong? Hindsight is wonderful, isn't it? I guess Altium or Cadence have also started with everything done right to today's standards from version 1.0, correct?

Also, how many of the professional grade tools actually do have the features you are describing? Such as the real-time linking between schematics & board. Most that I have seen require you to annotate the schematic/board and then push the changes to the other tool (or "pull" them from there, it's the same). Not exactly what I would understand as real-time. (and, FYI, Kicad is capable of doing both forward and backward annotation - it used to be a bit awkward with having to re-import the netlist before, recent versions have it fairly seamless on a button press).

What is so difficult on accepting that this is an old program, with some design decisions that are more than a decade or two old and that, unfortunately, take time to change? Especially given the resources that the Kicad team actually has. Why people keep assuming that it is some sort of newfangled tool released a year ago by some kiddies that have never seen a transistor in real life? That would be the only case where such analysis would make somewhat sense.

That doesn't mean that there aren't things to improve in it - I believe you have yourself written a fairly extensive post on things that need to be fixed for it to become usable for more advanced/professional work some time ago and I pretty much agree with what has been written there. That's a fair criticism and constructive suggestion for improvement.

But bashing a tool only because it is not designed up to "modern standards" (whatever those are) or because the author(s) didn't have the benefit of 20 years of hindsight when designing it is a bit unfair, IMO.

[Bassman59]

and you still need to edit files before you can go from schematic to PCB

Not true.

The fact that footprint assignment is done in the PCB tool , and not in schematic means they don't understand multiple issues. 

Except, no. You can -- and many users do exactly this -- create what we call "atomic parts," that is, parts in the symbol libraries which include the necessary footprint.

If i select a part in the schematic , this needs to be something that can be purchased. So it has a fully qualified part number , which in includes the pcb footprint. Simply putting down 7805 in the schematic is useless. who makes it ? what package ( can be anything from to220 , to92, sot89 , sot223 to any kind of other package )

Give the part a name that is unique and representative of the part you can buy, and add a simple Part Number field to the symbol and everything you want is there. And, again, the part is in your library, so every time you place DRV134PA on your board, you get the right footprint for your layout, and then you can generate a BOM that includes your Part Number field.

The functionality I describe has been in Kicad for quite some time, like three or four years.

The user group discusses this workflow.