Should open hardware be designed with closed software?

[osmaneralp]

I have an open hardware project that I want to design and make available to the community. The problem is that I don't know how to use any of the open CAD tools, such as KiCad. I know Eagle and Altium Designer. Eagle is not open, but at least it has a free version. I would prefer to use Altium, but of course it is a closed system. Can I use Altium or Eagle for an open hardware project?

--Osman

[Bertho]

In principle, you can use whatever you like and of course it would be very much more preferential to use free software for all steps.

The primary freedom involves that you can study the design and alter it for your own purpose. That means you publish the schematics and the firmware source, including all documentation, under a free license. With these anyone could redo the project/hardware with the same functionality.

You can compare it to a scientific publication. Any person, with skills in the field, should be able to reproduce the results from a scientific publication. Having access to any materials, tools and procedures is an important part of being able to reproduce results and make it accessible to a greater audience. The more parts locked up, the fewer can benefit and the cost rises.

Using a Kicad or gplEDA improves the degree of freedom and makes it more accessible to other users. It enables wider collaboration and ease of use for more people. The same argument goes for compiler, assembler and other tools, like mechanical drawings, for creating the final design. I can only encourage you to learn free tools and tool-chains.

[krivx]

If you can use an open format then the software you use really doesn't matter. Can Altium export to something can be edited with free software? (apart from Gerbers obv)

[ResR]

ExpressPCB? The software is free to download, quite easy to use for my experience. Maybe this is what you look for?

[jancumps]

You should do as you please
You are doing the work, so it's your call.
It would be a never ending story. Would you have to run the oped source CAD on Linux on an open hardware computer?

[Bored@Work]

Can I use Altium or Eagle for an open hardware project?

It is an ongoing debate if using commercial software for open hardware is OK. It doesn't feel right, but many open hardware developers do so. Very common seems the use of Eagle.

[krivx]

You should do as you please
You are doing the work, so it's your call.
It would be a never ending story. Would you have to run the oped source CAD on Linux on an open hardware computer?

This is not so much of a political problem as it is a practical one if you want others to contribute to or use your project. What is the point of releasing design files if they are locked into file formats that are priced well outside the reach of most individuals?

[jancumps]

You should do as you please
You are doing the work, so it's your call.
It would be a never ending story. Would you have to run the oped source CAD on Linux on an open hardware computer?

This is not so much of a political problem as it is a practical one if you want others to contribute to or use your project. What is the point of releasing design files if they are locked into file formats that are priced well outside the reach of most individuals?

Yes.
The only practical way to work together is when you use compatible tools, regardless if it's open or closed hardware.
But if the schematics are published in a readable way (scanned print of a schema created with closed CAD?) and the license is open, it should be ok to call it open.

[8086]

An open design is one that has it's content revealed to all.

You can simply publish PDFs or images of schematics and it can still be open.

The software package is irrelevant. You're still making the information available.

[mikeselectricstuff]

Can I use Altium or Eagle for an open hardware project?

It is an ongoing debate if using commercial software for open hardware is OK. It doesn't feel right, but many open hardware developers do so. Very common seems the use of Eagle.

This is one of those "would be nice" things but it makes no sense to insist on it.
Having to use non-free tools to do PCB layout is no different to having to use a non-free soldering iron to assemble it.
If all the people who bleat on about this got stuck in and helped to make, say, kicad as good as, say, Altium then that would be real progress. Until then it's just hot air and doesn't achieve anything.

[jancumps]

...
If all the people who bleat on about this got stuck in and helped to make, say, kicad as good as, say, Altium then that would be real progress. ...

That can happen. In the software area the open source IDEs for Java are competitive with the commercial IDEs - for capabilities and usability.

[ve7xen]

Of course it is still open regardless of the tools that you use, but I do think it's important for open hardware to use open development tools, otherwise I see little value in their "openness," it's really no different then than publishing copyrighted PDFs of the details. It's up to you of course, but I'd urge you to (especially since you're just starting out) use open tools as much as possible. You will make your work potentially much more useful to others.

Having to use non-free tools to do PCB layout is no different to having to use a non-free soldering iron to assemble it.

I disagree. Requiring liquid solder for assembly is quite different than requiring a very specific software package from a specific company (and likely a specific version). Use a Hakko, Metcal, Weller - or hell a homemade reflow oven made by Black and Decker. It's the difference between having a simple, general purpose tool that can easily be improvised, for which many options exist, versus a single purpose, very specific tool from a single company with no potential whatsoever for alternatives. Even in the case of things like device programmers, the programming protocols are (generally) available and it's at least conceivable that open or alternative options could exist (or be made by) for an enterprising user. In the case of EDA suites that's not remotely possible.

[firewalker]

Can I use Altium or Eagle for an open hardware project?

It is an ongoing debate if using commercial software for open hardware is OK. It doesn't feel right, but many open hardware developers do so. Very common seems the use of Eagle.

This is one of those "would be nice" things but it makes no sense to insist on it.
Having to use non-free tools to do PCB layout is no different to having to use a non-free soldering iron to assemble it.
If all the people who bleat on about this got stuck in and helped to make, say, kicad as good as, say, Altium then that would be real progress. Until then it's just hot air and doesn't achieve anything.

The analogy would be with a "non open hardware soldering iron". It is a loop tbh.

As RMS would say "Free as freedom, not as free beer". 

Alexander.

[free_electron]

let me do a thought experiment...

if i can't use tool xyz because it is proprietary and cost money.

Then why would i be allowed to use an AVR processor. That core is also proprietary... only atmel has it so you are forcet to use the atmel cpu. you could whine the design s not 'open' as you are forced to 'buy' the atmel cpu while you have a few pic's laying around that were cheaper or free..

if i made it using a specific pic you could complain : but that pic has a special function that no other cpu in the wolrd has. your design is not 'open enough' ....

so we'd end up with designs consisting only of 1k resistors and 2n2222 transistors. as those are the only parts you can get anywhere. any part that has a single vendor with no alternative would automatically be barred.

so goodbye any kind of design using a chip that has no alternate.... goodbye arduino , goodbye raspberry pi , (both contain sole source parts)

just a though experiment...

[c4757p]

I say drop the buzzword and just say what you can and cannot do with your stuff. Who cares if you can validly call it "open"? Publish your files and say "here's what you can do with them", and if people want to bitch that it's not "open" if you need expensive software to work on it, let 'em.

[David_AVD]

It seems that "open hardware" is often more like "public domain hardware" in a practical sense.

[ve7xen]

let me do a thought experiment...

I would simply say that the requirement to use AVR is part of the final product and not part of the production process. Obviously at some point you're going to need to put down some money for actual hardware, but to the extent possible, I think the reliance on proprietary meta-things like C compilers and EDA suites should be minimized to maximize utility to those attempting to make use or contribute to your project. In most cases it doesn't affect the final product whether you use open or proprietary tools, so open tools should be used if the hardware is to also be 'open'.

I'd go even further and say that to the extent possible, open designs should use openly documented parts.

Of course this is my personal opinion and has nothing to do with whether the design itself is "open" or not, which is a much simpler question to answer.

[Bertho]

just a though experiment...

Although an interesting argument, it is quite a problematic one. Basically all components are proprietary; even a 1k resistor and a 2N2222 transistor. All hardware components that make up a design are hard to manufacture at home. There may be second sources for some, but many components only have a single source.

Even the opencores initiative suffers from the problem that the target CPLD/FPGA chips are proprietary. An ASIC, although feasible with some serious funding, is not free either. The internal architecture, gates, layout and interconnects are all proprietary. You can then, of course, make your own machines and so, but then only to find that the chemicals you need to make the darn chips are not even available for hobbyists; and when you start making your own, then I seriously doubt that the black helicopters will stay away for too long.

Open hardware has a limit, simply financial, to be done with the four freedoms as RMS defined. Software has no tangible substance, hardware has. Even a lesser take on free is not available with hardware. There is, in the current global climate of hardware, no feasible free model that persists through all layers.

I'd really would like to make a completely free processor on par with AVR/PIC/ARM/MIPS/etc. That can be done if the open hardware community can come up with some serious funding in the order of $10^10 and a darn good plan. And then cross your fingers that you will not die in a patent war. If you though mobiles were bad going with patents, hardware is just as bad or even worse because it becomes a global problem, whereas software patents are primarily limited to the US.

So, as much as I like the though experiment, it is more like a fantasy (for now) in my view.

[edit] in the light of the NSA scandal, I'd like to see really open hardware at all levels. A nice thought. Maybe a crowd-sourced project? 

[AlfBaz]

This problem seems to have two issues.

If you use "open" tools to create an "open" design then the open hardware community may languish due to limitations imposed by the open/free tools available. It becomes increasingly difficult to publish multi-layer high speed designs and only very few dedicated people would choose to take on such a task that could be easily accomplished with more efficient yet proprietary tools.

On the other hand, not using the open tools see's development of them languish due to the lack of their use or having them being pushed to their limits, allowing developers to see opportunities to enhance the tools

The other problem to contend with is competition from proprietors of closed tools. The open hardware/software movement has seen steady growth and take up which companies are trying to capitalise on in an effort to promote their closed tools.

In my humble opinion the open community should take every opportunity to advance the open designs available and not limit itself by mandating hobbled and limited tool sets

[EEVblog]

Can I use Altium or Eagle for an open hardware project?

Yes. The majority of open hardware would have been done in the "defacto standard" Eagle. There are quite a few that use Altium as well, such as myself.
Depends entirely on how much of a purest you want to be.

[marshallh]

In my humble opinion the open community should take every opportunity to advance the open designs available and not limit itself by mandating hobbled and limited tool sets

This

[EEVblog]

This is not so much of a political problem as it is a practical one if you want others to contribute to or use your project. What is the point of releasing design files if they are locked into file formats that are priced well outside the reach of most individuals?

That's right, it's up to the individual designer to decide that. If you as the designer really want people to contribute to and expand the design, then you would chose the best tool that facilitates that, be it Eagle, KiCAD, Altium, or whatever. i.e. if you know that all your friends who might want to contribute to the project have access to Altium, then there would be absolutely nothing wrong with choosing Altium, in fact it would be the best choice.
Those who don't have Altium, tough luck, deal with it.
It's the the old problem of not being able to please everyone.
You use KiCAD or some other open CAD software, then you'll still get people who use Eagle (the majority is seems) complain that you didn't use Eagle. You can't win, so just use to tool you want.

[free_electron]

Although an interesting argument, it is quite a problematic one. Basically all components are proprietary; even a 1k resistor and a 2N2222 transistor.

you don't get what i'm saying : there is 100 manufacturers making 1k resistors. if you only have 2k resistors in house you can make 1k by paralleling two 1k's. or putting 10 100 ohms in series.
if you don't have a 2n2222 you can stick in a bc107 or bc547.
if you don't have the atmega328 and only a pic .. you are stuck...
if atmel goes tits-up the entire design is dead. chances of 1k resistors disappearing or 2n2222 transistors disappearing are nil. it can be replaced. there are alternates.
not so for a one of a kind chip... you could port it to a different cpu but what if i use a special feature that no other chip has. that is what i'm getting at.
let;s say i make an open source oscilloscope using a massive altera fpga , analog devices dsp and a maxim ultrafast a/d convertor. i release everything : 8 layer board floorplan , verilog code, dsp code . the whole shebang.

if you are going to whine it was made with altium you can also whine 'why 8 layers,  i can't etch that at home... why altera i want xilinx .. why bga i can't solder that .. why surface mount and not thru hole ? why verilog and not vhdl ? why an analog devices dsp and not a pic ? why this super expensice ceramic chip with heatslug. couldn;t you do this by multiplexing 4 cheapo ones ?

there is  just no pleasing everybody.

an open hardware design should release the following:

- the source files , irrespective of what cad tools they were made with. lots of cad tools can read each other formats.
- a pdf version of the schematic and pcb layers scaled 1:1
- the gerber and nc drill files
- all sourcecode

with that information you can have the board made and build it -as-is-
or you can pull in the gerber , decompile it (not my fault if your cad tool does not have that capability. get a better one) , or even better import my cad files and run with em ( again not my fault if your cad tool cannot read my cad tools files. My cad tool eats almost any cad format out there, mentor, pads, allegro, zuken, eagle , orcad... it eats everything. if kicad can't read my files: go hound the kicad makers to build importers. not my fault your tool is limited. if every tool was identical we would only be using 1 tool.

by releasing the above i have given you all you need. you have the full sources in machine readable format. what more do you want ?

I'd really would like to make a completely free processor on par with AVR/PIC/ARM/MIPS/etc.

There is already one in the list you give : MIPS is a free core. anyone is allowed to make a MIPS processor without paying royalties. you can only call it a MIPS if it passes the MIPS compliance test suite though. the testing costs money. but there is no licencing  fee to use.

There are plenty of cores that have been made open. 8051 for example is open. intel specifically released it years ago ( that's why  years ago there was a sudden boom in companies making processors based on the 8051. there were no more royalties to be paid to the instruction set )


Now, the sword also cuts the other way.

Mandating a project is done in kicad or geda in order to release it as 'open' will turn a lot of eagle users away. It will also turn a lot of other tool users away.
Telling someone he can't use state of the art when doing a design and blocking him to use pen and paper will have a very simply reaction : f-you.

It could be even worse : instigate a rule like "you can only make open sauce stuff if you are using open sauce software on open sauce os's": boom you just killed of 90% of the open hardware makers. your whole ecosystem will come crashing down.

i am willing to give you my design for free. asking me to learn a different os, toolchain and imposing all kinds of other restrictions that are a burden to me is simply counterproductive for me. I won't bother, so you won't get my designs.

[Bertho]

you don't get what i'm saying

Oh but I do get what you are saying. The premise of the thought experiment can be interpreted to be "there needs to be a second source for any component" or "hardware needs to be free too".

I just took the more extreme approach in the thought experiment. Please remember, hardware will never be free as in beer because you need to cover the expense of the physical world.

an open hardware design should release the following:
- the source files , irrespective of what cad tools they were made with. lots of cad tools can read each other formats.; - a pdf version of the schematic and pcb layers scaled 1:1; - the gerber and nc drill files; - all sourcecode

And that was exactly my point in the first reply in this thread...

My answer to your thought experiment is not representative of what I think is possible or necessary in the real world. You invited borderline (extreme) thoughts with your thought experiment.

There is already one in the list you give : MIPS is a free core. anyone is allowed to make a MIPS processor without paying royalties. you can only call it a MIPS if it passes the MIPS compliance test suite though. the testing costs money. but there is no licencing  fee to use.

We can agree that there are lots of free designs for cores and whatnotelse already available. However, your thought experiment invites, or maybe even mandates, to think the path all the way through. The fact that I cannot make a chip at home does not mean that the actual physical execution of the design should be proprietary. That is why I said, lets build a fab that produces open hardware in an open way. Wishful thinking, maybe, but would be really nice.

[amyk]

There is already one in the list you give : MIPS is a free core. anyone is allowed to make a MIPS processor without paying royalties. you can only call it a MIPS if it passes the MIPS compliance test suite though. the testing costs money. but there is no licencing  fee to use.

MIPS isn't completely free - some instructions are patented.

SPARC on the other hand is, but it's a "big" architecture that no one has really used in anything outside of huge mainframes and the like.

[osmaneralp]

This is not so much of a political problem as it is a practical one if you want others to contribute to or use your project...

This is a good point that I had not considered. My original thoughts were that I would complete the project mainly for my own benefit, and on the side I would make it open for anyone else to copy. I had not thought that other might actually help with the project. If I want others to contribute, I can see the importance of using an easily available free tool the tool that anyone can download and use. Unfortunately, the free EDA tools are not necessarily the easiest to use or the most widely used. If a free tool has very few users, the pool of contributors would be very small. A "closed" tool that has thousands of users might be a better choice than an open tool that has very few users.

If I am doing the project on my own and just publishing the results, it sounds like the prevailing opinion is that I can use whatever I want as long as someone else can duplicate the results with a different tool. This makes perfect sense to me, and this is the route I will probably take.

Many thanks for the advice. It is an interesting discussion.

--Osman

[westfw]

CAD packages aside, I get an uneasy feeling about "open source hardware" projects that can't actually be built on an individual basis.   BeagleBoard, for instance, may be fully "open source", but ... *I* can't build one for a reasonable cost.  I might get a certain amount of "warm an fuzzy feeling" from the sense that perhaps I could have it built FOR ME, if it were important to my business, and should the supply dry up, but it would be a significantly expensive proposition.

Consider the original BeagleBone, or the "Chumby Hacker Board."  Both were open source, but as far as I can tell, they might as well not have been...

an open hardware design should release the following:
- the source files , irrespective of what cad tools they were made with. lots of cad tools can read each other formats.; - a pdf version of the schematic and pcb layers scaled 1:1; - the gerber and nc drill files; - all sourcecode

It's interesting that many of the OSHW designed released using free CAD packages (EAGLE, gEDA, etc) do NOT also release gerbers/etc.  If you want them, you're expected to generate them yourself...

[free_electron]

That is odd indeed.
Gerber is the lowest common denominator. What good is a design if i cant modify it. Lets say it is made in eagle , but not the free version.  I dont haveeagle and will not buy it. How do i get the gerber to make the board ?

If i have gerber i can at least have a board made. If i have gerber i can modify it or import it in whatever cad tool i use.

The more i look at all that 'open' stuff the more hurdles i see ..
Can't do this, can't do that. If it's got restrictions it can't truly be open....

[amyk]

Someone asked about converting PDF to gerber in one of the other boards here a short while ago, and got it done, so that's another avenue for "opening".

[mikeselectricstuff]

Here's another aspect  - if you know a tool well, you'll want to use it. You may decide at the start, or somewhere along the way to open-source the design. Being in a non-free or non-mainstream format may be slightly less convenient to people, but at least the essential parts are there - PDF of schematic, PDF and/or gerbers of PCB. Most projects aren't so complex that someone who wants to improve it would have to spend a lot of time re-entering it.
But at least the design is out there.

Any prejudice against non-open tools may discourage anyone who isn't interested in learning new tools from making their design open, so the bottom line is that there will be fewer open designs out there, which is worse for everyone.
 
IMO the current OSH 'charter' is already a little too prescriptive.
My feeling is that it is entirely reasonable for someone to decide to to say "here's a design, feel free to modify it & improve it, but if all you want to do is a straight copy then I'm not going to help you do that by giving you produciton-ready files. You need to put some effort in yourself" 

[westfw]

Lets say it is made in eagle , but not the free version.  I dont have eagle and will not buy it. How do i get the gerber to make the board ?

Well, if it were in EAGLE, you would download the free version and use it to generate gerbers.  EAGLE's freeware version seems to be relatively screw-free - you can view, print, or run the CAM processor on design files that exceed the restrictions of the license.  You can probably even edit them slightly; the main "real" restriction is that you can not move a component to a position that is outside the size limits.  For example, I had no trouble outputting gerbers for the Google ADK2012 board (an 'extended' Arduino Due; 80x127mm, 4 layers.)

If i have gerber i can modify it or import it in whatever cad tool i use.

Really?  I can't imagine that an imported gerber is very useful for anything beyond the most basic fixes.  Do the high-priced packages really manage to reconstruct packages, nets, and components, or a schematic, from what is essentially a group of shapes and line segments scattered across several layers of PCB?

[tom66]

I release schematics for my projects, but I -rarely- release gerbers.
My reasoning is, schematics: great for repair, understanding or change to the product.
Gerbers: great to completely copy the product without putting any effort in. You can't edit a gerber. You may as well start from scratch with any more than basic changes to an already compact PCB, too. (Most of my designs use every square inch.)

I don't know though, perhaps this view is less of open source view and more of open design. I want my products to be easy to repair, maintain and learn from, but conversely I don't want someone to start making carbon copies without putting -any- effort into at least engineering it differently or improving it in some way.

I also release schematics as PDFs. GPL-EDA has some awkwardness with symbol footprints, and a lot of my library is custom. Without the toolchain and symbols being set up properly, there is no point in having the *.sch.

[mikeselectricstuff]

For the majority of designs, which are not especially complex, it isn't a huge effort to re-create a PCB in the package of your choice - the bulk of the work for a PCB is placement, and then secondly  figuring out the routing paths. The actual routing after that is pretty trivial, so to create a new version, maybe 95% of the work is already done.
It helps if your PCB SW can load a gerber, as you can use this as a template as a guide to place components over, but even without this it's not a huge  effort to re-create something from a drawing.
 

[free_electron]

Oh yes, you can decompile gerbers.... I have done it many times.
Load gerbers anddrill file in camtastic (part of altium). Check alignment. Verify layer order and naming. Hit the 'extract netlist' button. Camtastic takes a few seconds to analyse all copper shapes and assign nets to every element that is interconnected. It looks at the drill coordinates and pads to find the vias and part holes.
Once that is done you click export to pcb and voila. An altium pcbdoc opens with the board now consisting of regulare pads, vias and traces and including a netlist. So you can start pushing stuff around.
When done -> export back to gerber.

Now, of course the thing cant detect components as gerber does not contain information of what is grouped together across layers. But, this is where the human comes in.
You create a small local library by selecting pads and traces. This may take a few hours but is not hard to do. Lets say a board contains 085 1206 some tqfps some so and some connectors.
You really only need to extract one of each , toss em in a llnrary and then perform a sequential place. It goes very quickly. Once done you use the filters to select free smd pads and all crap on the silkscreen and delete it.
There you go. A full board with real parts and traces, decompiled from foreign gerbers.
Now you can draw the schematic from it and use the project compiler to verify that your reverse engineered schematic matches the board. You can use all the tools like the crossprobing to ease the work. It is still time intensive but it does work.

[tnt]

To me the software package used doesn't matter. It's a tool.
As long as it is possible to buy it or get the format documentation (i.e. not some internal tool that's not distributed anywhere), then it would be fine.

Yes, if you don't have it, you won't be able to use the files ... well same thing if you don't have a computer. And to produce the board physically you'll also need to pay.

Including PDF & Gerber version would of course be nice, but shouldn't be a requirement. And for example in my case the PDF and gerber are often _not_ in the git repositor, they might just be on the website or something. This is simply because I really don't like to include autogenerated files (i.e. files derived entirely from others) in the git repo ...

[Alana]

In my opinion if you really want to use someone else open or public domain project all you really need is schematics in PDF and firmware. I had to re-do most of public domain projects i used into my fav CAD simply because its damn outdated [remembers win3.11] but i know it well and prefer over Eagle or Kicad or even Altium. So if i can, anyone can. Simple as pie.

[Tonkabot]

I started with Eagle, then went to PCAD and it was a huge improvement in productivity an quality.  After a couple years with PCAD I tried to go do a board in Eagle and it was like pulling teeth - I had learned all the nice ways to do stuff and Eagle was so much extra work to do simple drag and drop type of stuff in PCAD.

I gave up and went with what I knew then.   Now of course we have migrated to Altium, and I see that as a huge improvement over PCAD, at least once I got over the learning curve.   I only have so many brain cells and I am going to use the best tool I know how to use, it isn't feasible to keep expertise in multiple PCB tools - at least for me.

Distribute the schematic in PDF and Gerbers if you want people making their own boards.

Maybe someone could come up with a interchange format between different tools, such as the .STEP files in the 3d CAD world.   
But none of the proprietary vendors would need or want to do that.

[homebrew]

In my opinion yes of course!

But I restrict my projects to the free version of Eagle. Thus everyone can alter the design easily. I tried my luck with gEDA but in my opinion the degree of reliance of that suite is still quite low. Even the most simple PSU design with rectifier cap and regulator nearly went bust because I didn't realize that gPCB swaped the pinout of the diodes I used in gSchem.

And that's a 5 year old problem being still present: http://archives.seul.org/geda/user/Aug-2007/msg00299.html

What a horror show ...  bummer!

Therefore, just stick to what you are used to ... thats fine!

Cheers,
Peter

[johnnyfp]

I like to tweak open source projects be it software or hardware, and find that most projects have a different way to present there projects. Be it pdfs, gerber, eagle files or whatever, and I always end up coverting it to something that I know and love.
I think it the nature of this "open"ness. Not everyone will know how to use tool x y or z but they will know there own tool. So they will tend to lead to using that.

I think as long as there is a means for a viewer to read and interperate your project, then its considered open.

[resistor]

I don't think the tool used to create it is as important as the format used to distribute it.  An open source design available in a format that can only be opened by expensive software isn't useful to many people.  Publishing gerbers and PDFs helps some, but both of those are (mostly) read-only formats, so they're less useful than having the original in a truly editable format.