Horizon – a free and open source electronic PCB design package
horizon-eda.org
horizon-eda.org
Altium is probably the only clean PCB experience at the moment, and even that sits at 10% CPU on my machine with an empty schematic and no PCB open. The more esoteric tools like Mentor and Cadence's tools (and presumably Keysight's offerings although I can't get a student license through my university * ) are powerful but quite difficult to use working alone in the sense that schematic capture is quite slow because you need to (if I'm not doing something wrong) source your own library or make your own - not really competing with Kicad as such.
* The student license options these companies have are absolutely insane. You'd think they don't want people to learn to use their software. Altium's site doesn't work with non-american email addresses, Mentor doesn't include any RF/Thermal/Mixed-Signal simulation addons i.e. the only reason to learn to use it over kicad, Cadence took about 2 weeks to respond and ended up publishing a Photo of my student license with my name on their website if you know where to look (Indexed by number)
Electrical engineers are willing to put up with so much more it seems.
In general, IMHO. EAGLE, a golden-standard proprietary EDA tool, while still being quite a bit ahead of KiCad on the UX front, features no fewer pitfalls and completely unintuitive controls.
I've heard the KiCAD people are working towards this goal. If they pull it off, it'll be a game-changer. I'm definitely rooting for them.
Describing KiCAD’s UI as “quite natural” could only be true with the “in comparison” qualifier, IMO.
* Although not really in the list, last year Cadence was selling OrCAD for something like $430. Hopefully we get more < $500 USD offerings.
In what way? I've never heard of anyone doing "serious" (like sense FPGA routing etc.) In eagle
I suggest trying DesignSpark PCB. It's not open source, and Windows-only (though I suspect it would run perfectly under Wine). But unlike Eagle, Kicad, gEDA, etc. it has a vaguely sane interface (it's still has a weird zoom though) and beginners can actually use it to design real PCBs.
Although, like any other EDA, you really need to check the footprints or make your own.
Cadence, Keysight, etc. provide the licenses to universities, not individual students, but they are pretty easy to get. Sounds like you need to work the channels through a professor.
The Keysight website doesn't actually have a list of universities in its software program so I didn't bother checking whether I could apply.
> You’ll never get those add-ons in a free package, as it takes institutional knowledge and full-time work to develop it, which cost $$.
Given the amount of malware (sorry - I mean license management) the software comes with its their own fault if a team of one student manages to design their own motherboard or similar. Realistically I just wanted to play with some CAD software.
> provide the licenses to universities,
Not for free usually.
Have you tried CST?
https://www.3ds.com/products-services/simulia/products/cst-s...
I feel like they work on the same principle as Adobe: pirate our software when you're learning, so you get used to it, then pay for it/ask your company to pay for it when you're working. I clearly remember one of the top Google recommended searches for "Altium" was "Altium full crack download"...
Recent versions have an "Update PCB from schematic" option which works very well.
Yes unfortunately the develops are not receptive to improving this. They feel that any attempt to make it more intuitive or easy to use is dumbing it down. Disappointing.
Is there work being done on the project that you feel should be prioritized below certain UI changes?
The concept of having split tools for schematic and pcb just doesn’t cut it any more. Nowadays most companies have transitioned or are transitioning to do layout inhouse, because modern products are so integrated that the “old ways”, doing sch inhouse and sending it to a company for layout has become very cumbersome.
IMO one of the key things for any new EDA package is making a wise choice on how to store their files. Being able to script stuff that works on your files is becoming a necessity, and packages like Altium can’t compete with that easily because they use binary files for the most important (sch and pcb). For me at least, storing info as textfiles is a must to consider a switch, because the biggest issue with Altium is how bad it interacts with tools like git.
but kicad at least has an approachable format. so I wrote some crappy tooling to at least generate symbol and layout from a single description. and I could theoretically write a front end and leverage things like Gerber generation.
I don't understand why we don't have vendor supplied footprints in a tool-neutral textual format yet. I know people have tried - who are the contenders?
If they’ve improved it I might give it a chance
I gave this a spin, and it's so refreshingly the opposite of that.
My main barrier to using it for my next project is this - am I going to get 100 hours into a project then hit a brick wall and have to port back over to KiCAD. If I had some confidence around this I think it'd give it a shot.
A gallery on the website of projects completed with it would probably do the trick.
Seriously pleased this project is a thing. In 2020 there isn't really an excuse to be using software with such a poor UI/UX as KiCAD
That said, I am very happy to see new entrants, like Horizon. I watched a presentation by the author in the 2018 FOSDEM and it looked like a great idea even back then, I'm going to try it for my next project to see how it has progressed.
Wayne Stambaugh, the KiCAD project lead, was in that presentation, and they're very much aware of the problems KiCAD has. They're working in that direction, as KiCAD 5 no longer requires you to export/import the netlist, I believe (I'm not sure because I reflexively always do it anyway).
One of the first questions they had towards the Horizon developer was "why make a new tool instead of contributing to KiCAD?", and the answer, IIRC, was "It was easier to make my own", which sounds reasonable given the amount of innovation that went into Horizon and the constraints a project as widespread as KiCAD would have, regarding backwards compatibility etc.
An easy starting point is better import/export of layouts. Designing an antenna in my RF-specific CAD suite is great, but when I actually have to build the thing, all I get is DXFs. Converting DXF to gerbers is a mess in KiCAD. Same goes if I want to reverse the pipe and simulate my layout.
Staying in the OSS realm, QUCS (open source ADS equivalent) has RF circuit simulation, just needs some help implementing auto-layout. This would make it trivial to e.g. optimize a simple transmission line circuit, and have it fabbed. That'd be huge for opening up simple microwave circuits to the DIY/"hacker" community.
The Microwave Office guys knew this 25 years ago, which is why it’s so much better than ADS. It is all tightly integrated and object oriented.
Starting from scratch also allowed me to trivially implement features that KiCad is missing, such as proper panelisation: https://blog.horizon-eda.org/progress/2020/04/05/progress-20...
Why should I learn Horizon vs Kicad first, or vise versa?
- library management that makes sense (though is a bit more complex than KiCad's)
- clean and modern UI
- fast-paced development (if there's a suggestion, that makes sense to me, I'll implement it)
pro KiCad:
- large library of parts
- has been around for a long time, so for every problem you may encounter, there's likely a forum post or so that describes a solution/workaround
- larger community
It's still a pretty clunky workflow but it gets the job done. I really wish it was possible to have a point and click interface in the GUI for downloading individual new parts akin to a Linux package manager, but this is probably difficult to manage since there is no centralized distribution point for parts that are outside of the official libraries.
Glib::Error: No GL implementation is available.
Couldn't find much to help me out figure out how to fix that issue, I assume it's some missing dependency but... where do I go from here?
Thanks for your work, as it currently stands I can't use it but I appreciate it exists nonetheless (I'm on a Mac).
Similarly, gtkmm is so much cleaner than Qt as a modern C++ library with none of the preprocessor / codegen quirks. Unfortunately it limits the platform reach, because GTK isn't really polished anywhere except Linux.
I raise a toast to the obsession with clean code and clean design. Sometimes "works on my machine" is the counter to "worse is better." At the same time, it's pretty inconvenient for me to use the software. Oh well.
Differential pair routing
Pin swapping... ie a differential pair where the pins can be reversed
Pin agnostic routing, ie a MCU or FPGA where it doesn't matter which pin is used
Incremental part placement, unlike kicad where all the footprints get dumped on the board in a big mess at the start
These are things that bother me most in kicad (it does have diff pair routing)
Thanks - looking forward to taking it for a spin
Yes, uses KiCad's push&shove router
> Pin swapping... ie a differential pair where the pins can be reversed
Not really, but swapping P/N in a diffpair is as easy as selecting nets in the schematic and using the "swap nets" tools.
> Pin agnostic routing, ie a MCU or FPGA where it doesn't matter which pin is used
Not really, but you can draw the connections in the board editor and backannotate them into the schematic: https://horizon-eda.readthedocs.io/en/latest/backannotation....
> Incremental part placement, unlike kicad where all the footprints get dumped on the board in a big mess at the start
Yes, there's list showing all currently unplaced packages: https://imgur.com/ZrvRw46
How tightly integrated are the schematic editor and layout editor? Are they two separate programs like KiCad or can you eg. delete a component from the schematic and have it automatically removed from the PCB (but without nuking ALL non-schematic components from the PCB).
Definitely gonna give this one a try.
1. Does the schematic editor support multi-page schematics? Kicad's hierarchical-block approach always feels needlessly clunky to me.
2. Does the PCB editor support planes and copper pours?
3. Does the PCB editor have anything to help estimate a line/diffpair's characteristic impedance?
2. yes, including global or per-net rules for thermal relief, spacing etc. It warns you if nets have not been recalculated on Fabrication output. Adding them is simply: draw a polygon, right-click and "Add Plane", there you have to select any names net.
3. there is length tuning built in, but quite frankly I never had to use it, so I don't know if that is what you're looking for
re 3: I was more thinking about the plane spacing. Like say I'm making a board where I want to carry a high-speed clock signal. If it's single-ended, chances are the transmitter has a 50-ohm output impedance, and the receiver has a 50-ohm input impedance.To get the cleanest signal, I want my trace to have a 50-ohm _characteristic_ impedance (which is a function of the trace-width, distance to the closest plane layer, and PCB substrate's chemical properties).
Trying to model it accurately is a huge pain in the ass, and probably not worth attempting. But it could be nice to have a rule-checker that says something like "for your PCB dielectric constant of 4.7 with 10mil spacing to the nearest plane, your trace should be 18mil wide to get the desired impedance".
There are a bunch of these calculators out there - but it would be super cool to do some rudimentary constraints as DRC!
1. Opening up your first project is not straightforward. Instead of just opening with a blank project and letting you add a resistor, the app asks a lot of rather irritating and irrelevant questions (who cares who the "author" is at creation time, and couldn't it guess anyways?) that could be better decided later. Worst offense: the "create" button is grayed. As in: nope, you won't be able to use me until you answer all the useless trivia.
2. Horizon seems to cling to the "main window launches sub-apps" paradigm (much like kicad). This is a rather confusing UX workflow. At the very least, it clutters up your desktop. Why can't all sub-app live under one roof ? I mean even gimp has finally given up on the broken multiple top window idea ages ago.
3. Couldn't find any standard old-timey thru-hole parts in the default "pool": 2N2222, 2N3904 standard, LEDs. Not fashionable these days for serious work, but still very much in demand by the hobby crowd.
4. To place a part, it looks like you have to select the *exact* component you need right away. Why? . Why can't I just place a generic NPN in the diagram and then refine what part it'll actually be later on ?
5. The part browser : why so many search boxes ? I want a Google-style single search box that's smart enough to match what I'm saying to a given part. Finding a part shouldn't be like entering an SQL statement. Fuzzy matcher please.
6. Right-click on a component: a mile long menu, no "part properties" entry. When I finally locate the "change symbol" entry (in a sub-menu), it sends you to a subtly different UI than the part browser.
7. No navigation buttons in sight. Where's "Frame All", "Frame component"
8. Components are hard (near impossible) to just grab-and-move
[edit]: forgot to mention installation. On ubuntu, this was not exactly straightforward. Choice between installing a binary through flatpak (non-standard thing I don't have) or building from scratch (there went 30mn of my day). Almost made me open the windows laptop that's specifically reserved for running win-only crapware.
- Navigation Buttons: I proposed a solution here: https://github.com/horizon-eda/horizon/issues/119#issuecomme... This is part of the effort to add icons and toolbars to things that right now are reachable behind the spacebar menu
- Many TH (as well as audio/synthesizer-related) parts can be found in the work branch of my fork of the offical pool: https://github.com/atoav/horizon-pool/tree/work The offical library/pool only includes parts that have been checked by four eyes, so many parts are still open as Pull requests: https://github.com/horizon-eda/horizon-pool/pulls
As for the other points:
- on 1: Starting projects: for me it is only entering a project title and selecting a path, which is reasonable unless you are the type who wants to decide things like these in the end on saving.
- on 2: While I also like single window concepts better (with panes like Blender), I think having schematic, board and pool/library editors in seperate windows is kinda reasonable.
- on 4: You actually can. You are not forced to use existing parts if you use "Place component"
- on 6: Given how the whole thing works, such a thing like "Part properties" doesn't really make sense. In Horizon everything you place on the schematic is either a generic unit (E.g. "Two-terminal resistor") where you can enter a value or a existing part mapped to this generic unit (e.g. a "ERJ-6ENF2003"). This is (on purpose) different from how other EDA-tools do it, but in practise it works quite well. The point on the different assign window is something I'd like to have changed as well : )
- on 8: if you want to move something you have to use the move tool which you invoke by pressing the m key. You can also move things with the arrow keys.
Generally the UX paradigm of horizon is very much built around single key shortcuts or key sequences. And if you don't know the keys just press the spacebar and search for the command and it will tell you right there.
As for now the experience for power users is very fluid, but when it comes to installation, onboarding etc. the whole thing could use a lot of improvement (and are gladly beeing worked on).
Any plans for Mac downloads? And also, does its component model separate out schematic symbols and pin assignment (like DesignSpark PCB and I assume others) or are they inseparable (like KicadPCB)?
This is a good summery (with a nice graphic of the structure of an example part): https://horizon-eda.readthedocs.io/en/latest/pool-why.html
As far as I can tell, the development models also differ significantly: The LibrePCB devs seem to be focused on doing things the right way the first time such as having a proper operation-based undo/redo stack whereas horizon just stores a copy of the whole schematic/board for undo/redo. They also seem to focus more on simplicity for beginners rather than more advanced features such as length tuning and advanced design rules.
Putting development effort where I like enabled me do to the board for my master thesis https://github.com/carrotIndustries/x-band-tx/ (4 Layers, differential pairs, length-matched traces) in Horizon EDA only after about 2 years after the initial beginning of development.
We also talked about ways to increase collaboration between the two tools. A few ways I could see is:
* Router: Extracting the KiCad router into a separate repository / library so it can be used and developed by the different projects.
* Gerber Compatible Stroke fonts: Would be nice to have one font specification and implementation to be used for PCBs. LibrePCB started the FontoBene (https://github.com/fontobene) project to do just this, but currently LibrePCB is the only user of it.
* Regular exchange of ideas and domain specific problems / challenges and solutions. This already happened a few times at c3 and FOSDEM.
1. Does Horizon support an equivalent to the 'Variant Manager?' This is something that will allow variance of build outputs (different component populations, different schematic text, etc).
2. In the docs it seems to mention submitting parts to the default pool. If people submit parts to the pool are there style guides somewhere? (line widths, US vs EU resistor symbols, etc)
3. Looking at a random SMD footprint (SOT-363) it looks like the paste window is equal to the pads. Is there a way to set paste mask expansion like is done in Altium? (Automatically offset paste window from pad by some amount)
No, but there's been some discussion on it: https://github.com/horizon-eda/horizon/pull/347#issuecomment...
> 2. In the docs it seems to mention submitting parts to the default pool. If people submit parts to the pool are there style guides somewhere?
There's https://github.com/horizon-eda/horizon-pool-convention, but that's still work in progress.
> 3. Is there a way to set paste mask expansion like is done in Altium?
Yes, you can set it globally in the board.
When producing a placement file for a PCB, can component part numbers be exported ? Neither Cadence, nor KiCAD stores components in their PCB files, only footprints, so part numbers cannot be easily accessed.
Is there any "data first" EDA package? Something more "docbook" and less "Microsoft Word"?
I'll have to find a simple project to test this out on though, because as you say, It needs to be a huge improvement on KiCad to warrant a switch.
Horizon misses nothing (for me) now and has a beautifully efficent UX compared to Eagle and KiCAD. Their library concept is also well thought through, so I don't have the feeling I am wasting my time when I am contributing parts.
The downsides are:
- you still have to create many parts yourself (but parts, packages, etc are json and generating them can help a lot)
- horizon supports only newer OpenGL versions, so if your machine is old it might not run
Other than that I'd take it over all the others every day (and I do).
Horizon docs have this explanation: https://horizon-eda.readthedocs.io/en/latest/why-another-eda...
There are some things about it which drive me nuts and which could easily be done better in an open tool:
- The UI is extremely slow. Clicking on a component often causes the entire app to pause for 1-10 seconds while it loads all of the part parameters and redraws panels.
- It doesn't handle large libraries gracefully. We have something around 30000 parts and the library compiler crashes more often than it succeeds. The Altium sales folks have attempted to get us to switch to vaults but can't explain to us what the data store is, what the concurrency model is, why it's not further lock-in, and why it should be more performant.
- Better geometry importers and exporters. Getting fancy RF geometries in and out is a huge pain. For one thing, Altium's DXF parser renders at a fixed resolution. So if you have some very small feature it will often get turned into a line or a dot. And even if not, it'll be coarse.
- Never do anything in the UI thread that isn't strictly responding to user input.
There are some things it's mediocre and it would be somewhat difficult to implement better in an open tool:
- Really good understanding of mechanical mates. Basically Solidworks-grade constraints for putting things together.
- Tight interconnectedness between schematic and layout rules, and a really powerful and expressive language for writing rules. Altium's rules language is simultaneously constraining and often lacks coverage for obvious cases. As an example, there's no way to do same-net gap analysis. You can make unmanufacturable spaces between two traces of the same net all day long and the design rule checker would never care.
Some stuff Altium pretends to do but doesn't, and it would be fiendishly difficult to do a good job in a FOSS tool:
- Physics modeling. I want to know how much coupling I'll have between nets. I want to see discontinuities in transmission lines. I want to know what will radiate and absorb and how. Basically I want to be able to have my CAD tool also be Ansys HFSS with varying degrees of fidelity.
- Real schematic hierarchy. The schematic compiler is chock full of bugs that it's dangerous to rely on hierarchy and also dangerous not to. I want to draw schematics with the same type of composition and repetition that I use to write software.
Then there's stuff that Altium doesn't even pretend to do and someone totally should write an EDA tool that does.
- Useful diff of schematics and layouts. I want to be better able to tell where I might have made an unintended change.
- Real time collaboration. Google Sheets but schematics and layouts. I want to be able to drop in on my layout artists' work and be able to point things out or share in the burden.
Lately in my office there have been lots of new people coming from places that use Cadence and they're slowly trying to get us to switch over. I haven't figured out what about it is supposed to be superior.
I was about to start learning KiCad, but I think I'll explore this first.
The maintainer seemed to be open to the idea and noted it down for later. So maybe it will be there in the next release : )
Things are going to be different and if that's what's making you quit the software I have to imagine you wouldn't even be using it anyway and you just thought it looked cool despite not being an actual EDA user.
I myself would probably not assume that because there is no explicit button it cannot be done ( I'd rather try the usual thing like spacebar drag, middle mousebutton, etc).
However I understand that this might depend on your background and your willingness to learn the damn thing, which is why lowering the entry bar is worth it imo.