Horizon EDA Version 1.0!

[djac]

"You really have no idea about how electronics are designed by professionals  "

Which way is professional and effective always depends on the circumstances, there is no one professional way of working, so please stop proselytising.

[Feynman]

As other people already mentioned, this is by design, inspired by how it's done in the enterprise world. The schematic/netlist is the single source of truth for the BOM. You can't tell someone else like your assembly house or purchasing department to get you a 1k 0603 resistor. You better tell them to get a panasonic ERJ-3EKF1001 and you get exactly what you want.

I have to agree on this one. Every single company with a decent engineering department i worked with or worked for specified full part numbers for every single resistor. In most cases this was done once with some sort of script. In early development stages one can of course work with some preliminary parts that maybe only have a schematic symbol and maybe a value. But at the latest the final production design should have full part numbers for every single (purchasable) BOM item. This is also really handy especially for capacitors: You don't have to look up every time again if your desired capacitance value is really available in this package with that voltage and with this temperature coefficient. If you have full part numbers in your library, you are 100% sure that this part actually exists.

And that's also the preference of most assembly shops. The reason is simple: Most assemblers have no intention to stock some ordinary Panasonic 0603 10k resistor. In fact most assemblers prefer not to stock anything, simply because it's expensive (of course they happily store stuff, if you pay them to). If they have a full part number, they have the full specification. And thus an easier time finding alternatives if there is shortage on the market or in the rare cases where they actually have something "lying around" from a canceled job or whatever.

In the end the assembly shop has no idea if you really don't care about the rated voltage of a capacitor when you just say "100n 0603".

Another advantage: Nowadays, many shops offer some sort of "instant quote" on their websites, where you upload your BOM and stuff and get a quote within seconds. Not going to happen without full part number for every BOM item.

[dunkemhigh]

If you have full part numbers in your library, you are 100% sure that this part actually exists.

Or existed when the device was added to the library.

[djac]

As other people already mentioned, this is by design, inspired by how it's done in the enterprise world. The schematic/netlist is the single source of truth for the BOM. You can't tell someone else like your assembly house or purchasing department to get you a 1k 0603 resistor. You better tell them to get a panasonic ERJ-3EKF1001 and you get exactly what you want.

I have to agree on this one. Every single company with a decent engineering department i worked with or worked for specified full part numbers for every single resistor.
...

And that's also the preference of most assembly shops. The reason is simple: Most assemblers have no intention to stock some ordinary Panasonic 0603 10k resistor. In fact most assemblers prefer not to stock anything, simply because it's expensive (of course they happily store stuff, if you pay them to). If they have a full part number, they have the full specification. And thus an easier time finding alternatives if there is shortage on the market or in the rare cases where they actually have something "lying around" from a canceled job or whatever.
...

Your statement is contradictory in itself, especially if you want to equip your boards externally.

For the assembler, it is advantageous if a resistor is not over-specified and thus has more straight lines of freedom. In many cases, he can use his own stock and does not have to order anything. This saves enormous costs (larger quantities when ordering). Or if he has no stock, he can order from the company that is offering cheap products at the moment.

What must be fixed are the technical key data that must be adhered to for the resistor so that the circuit functions, i.e. in most cases footprint, resistance value, tolerance. Even the exact power rating is unimportant in many cases and results indirectly from the design (e.g. 603).

Whether the resistor is manufactured by Panasonic or Yageo is irrelevant in most cases.

Of course, there are other cases where the manufacturer and the series must be specified, e.g. when it comes to certain high-frequency characteristics or very precise resistors with a low temperature coefficient. This is normally relevant for the rather smallest part of the resistors.

So again, the point is that an EDA tool does not rigidly define the processes, but has the flexibility to adapt to a wide variety of user needs.

[Gribo]

That's the purpose of AVL (Approved vendor list). Usually you specify one PN, and in a different column, specify Any alternates, AVL only, or Do not substitute. And while a 1K Ohm, 1%, 0603 seems to be a bog standard part, sometimes they are not, and the assembler has no way of knowing that. At the end, it is all about communicating your design intent.

[Lukas]

A bit later than usual, Version 2.0 is out: https://github.com/horizon-eda/horizon/releases/tag/v2.0.0

In addition to a wall of text changelog, there's also a blog post about the most important new features: https://blog.horizon-eda.org/new/2021/05/31/whats-new-2.0.html

[SiliconWizard]

A bit later than usual, Version 2.0 is out: https://github.com/horizon-eda/horizon/releases/tag/v2.0.0

In addition to a wall of text changelog, there's also a blog post about the most important new features: https://blog.horizon-eda.org/new/2021/05/31/whats-new-2.0.html

Impressive work!

[Bob Sava]

A bit later than usual, Version 2.0 is out: https://github.com/horizon-eda/horizon/releases/tag/v2.0.0

In addition to a wall of text changelog, there's also a blog post about the most important new features: https://blog.horizon-eda.org/new/2021/05/31/whats-new-2.0.html

Impressive work!

"Custom values on symbols" - I have to check that out!

[Warhawk]

The 3D model projection feature looks very interesting. Great job. I need to give it a chance the second time.
(KiCad user here).

[Bob Sava]

As other people already mentioned, this is by design, inspired by how it's done in the enterprise world. The schematic/netlist is the single source of truth for the BOM. You can't tell someone else like your assembly house or purchasing department to get you a 1k 0603 resistor. You better tell them to get a panasonic ERJ-3EKF1001 and you get exactly what you want.

I have to agree on this one. Every single company with a decent engineering department i worked with or worked for specified full part numbers for every single resistor.
...

And that's also the preference of most assembly shops. The reason is simple: Most assemblers have no intention to stock some ordinary Panasonic 0603 10k resistor. In fact most assemblers prefer not to stock anything, simply because it's expensive (of course they happily store stuff, if you pay them to). If they have a full part number, they have the full specification. And thus an easier time finding alternatives if there is shortage on the market or in the rare cases where they actually have something "lying around" from a canceled job or whatever.
...

[...]
What must be fixed are the technical key data that must be adhered to for the resistor so that the circuit functions, i.e. in most cases footprint, resistance value, tolerance. Even the exact power rating is unimportant in many cases and results indirectly from the design (e.g. 603).

Whether the resistor is manufactured by Panasonic or Yageo is irrelevant in most cases.

Of course, there are other cases where the manufacturer and the series must be specified, e.g. when it comes to certain high-frequency characteristics or very precise resistors with a low temperature coefficient. This is normally relevant for the rather smallest part of the resistors.

So again, the point is that an EDA tool does not rigidly define the processes, but has the flexibility to adapt to a wide variety of user needs.

It appears, with new pool structure in v2.0, it would be possible to create pools of generic parts with just values and footprints.   If that's still too much, one could create single part for each footprint.  Then hide all parameters while exposing description via 'custom values on symbols' feature, also in v2.0, where one could specify part parameters.  I probably would just stick to generic parts with individual parameters, however.

There are some scripts on github that could be the starting point: https://github.com/RX14/horizon-pool/tree/master/scripts

[SiliconWizard]

I agree with the process being a bit too rigid, that was discussed quite a while ago.
Now I haven't looked at the new component database, so this point may have been addressed. I fully understand the point of enforcing all parts to be fully defined (instead of being generic) for a final design, but as I and several others have said, not all workflows are the same and the way to get there varies quite a bit. I would have favored the possibility of using generic parts with customizable attributes (values, tolerance, footprint, whatever), and just add a check in the BOM generator (or maybe already at the schematic-to-layout step, to ensure that footprints are all correct): each generic part would yield a warning or error, then it would give you the opportunity to change all generic parts in a design for fully-defined parts. I find this process much more suitable for most users. But that's just my opinion. And mabe the new version actually allows this?

As to Horizon's 3D viewer, KiCad should certainly take a hint. It's very good.

[Feynman]

The 3D model projection feature looks very interesting. Great job. I need to give it a chance the second time.
(KiCad user here).

Yeah, looks like someone was tired of exporting a 2D-projection from an external 3D-CAD tool and then importing it back to the PCB tool (like me)

[free_electron]

On the concept of libraries i offer a different view.

Manufacturers and manufacturer part numbers stored in a library should not be used to pull a final bill of materials. let me explain :

- unique parts : 1 function , 1 footprint , no pin compatible substitutes : you can encode manufacturer and manufacturer part number directly in the database. there is no alternative anyway.
- multisourced parts : still 1 function, 1 footprint but possible variety of manufacturers / part numbers. how do you deal with this in a library ? list all ? list a few ? list the cheapest ? (which can change hour to hour) list the approved ? list the ones in stock ?

The second case is the huge issue with such an approach.
A true parts catalogue (i deliberately do not call it a library. the library is the CAD portion : symbol / footprint ) decouples that.
The CAD portion contains
- footprint
- symbol
- a number of useful parameters so that part selection can be done in the cad system. working voltage, value dielectric for caps, forward current, reverse voltage for diodes etc.
- a suggested manufacturer and manufacturer part number so you can pull a datasheet ( this covers both cases mentioned above : unique and multisourced )
- a textual description for this part
- unique identifier for this 'element'. This unique identifier is "data-less" it contains no identifiable part information. so no numbers like IC-TI-LM741 or RES-1K-0805-1%

The BOM database portion translates the unique identifier to "orderable parts". This translates the one (unique identifier) to many (part numbers) portion of data
i picked 1k 0805 1% in the cad , dropped symbol and footprint. When the bom needs to be generated part 0923540-00 (index-revision) is accessed in the database and it creates a line item that says

<designators from CAD>, <textual description from CAD>, <manufacturers from DB><manufacturer part numbers from db><vendor from db><vendor part number from db>
R214,R215 : Resistor 1K 0805 1% , Vishay CRW0805-102-DQR , Yageo RM8-1K00-P  Digikey  RN-7362, Mouser PQ2154, Avnet RES1K08051p

The pick and place file can do the same
<designator><x>,<y><top/bottom><rotation><textual description from CAD>, <manufacturers from DB><manufacturer part numbers from db><vendor from db><vendor part number from db>
R214 ,2.254,9.521,T,270, Resistor 1K 0805 1% , Vishay CRW0805-102-DQR , Yageo RM8-1K00-P, Digikey  RN-7362, Mouser PQ2154, Avnet RES1K08051p
R215 ,3.745,5.958,B,45, Resistor 1K 0805 1% , Vishay CRW0805-102-DQR , Yageo RM8-1K00-P, Digikey  RN-7362, Mouser PQ2154, Avnet RES1K08051p


The cad library editor needs to talk to the database. whenever a new part is created in the cad library , a new index number is requested form the DB. the CAD creates a new object with this ID, attaches footprint and symbol, textual description and whatever other data is at hand. if you have manufacturer and part number you can enter it and it will get loaded in the DB ( not in the CAd library ! loaded into the DB under the index). The cad tool can simply present a simple UI into the DB to do basic operations of adding the initial load.
The DB portion is accessible through another UI (more detailed) where you can edit the "textual" data.

Whenever a bom or pick place is generated the CAD tool pulls the latest data from the db based on the unique index.

The net result is:
- At design stage you are not "locked in" to manufacturer / part number. if you have it , load it, if not move on.
- at production time you can simply make sure your DB has the latest ordering data based on whatever criteria you need. ( could be price, could be signed contracts, could be whatever is in stock at digikey , could be whatever you have in house.)

This can be stored in a simple single-file db like Firebird or even a tabular format like a CSV file or excel file.
The key is the decoupling of the CAD and ERP/PLM data. The ERP/PLM data is in flux (prices change, product lines go obsolete, get replaced, contract change. not all companies work the same way) Storing this outside the CAD gives maximum flexibility.
Bonus : you can crosslink this DB and integrate it with part search engines like octopart or digikey data pulls through their API's if you want. This is all doable now as this data lives outside of the CAD world. IF the data updates , the CAd pulls it during BOM and P&P generation so You now have a live system with complete flexibility for both engineering and procurement/planning/production

Critcial points :
- the unique ID is immutable. once created it cannot be reused. it is alive ( part exists ) or dead ( part is obsolete) it is never reused. even on a dead part the ID is still 'used' ( to mark it as obsolete, potentially point to substitute)
- the db is a WORM (write once, read many): any alteration of text data creates a revision of a part. Data is never lost. The CAD tool can flag you are using an out-of-date revision and show you the list of newer revisions. it is up to the engineer to approve/reject. in most cases automatic approval is fine (if the parts are vetted to be correct) There can be a flag :
1 = mandatory update ( corporate policy, approved vendor for example , or product line discontinued and substituted or order number changed)
2 = requested approval (may be a power rating change for example the 1/4 watts are obsolete so we swapped to 1/3 watts , same tolerance,value, upward compatible change basically)
3 = need replacing (part obosolete , no footprtin compatible part available )
4 ... and so on.

[nctnico]

A true parts catalogue (i deliberately do not call it a library. the library is the CAD portion : symbol / footprint ) decouples that.

AFAIK this is exactly how the parts library of Horizon Eda has been organised from the start. Just don't get hung up on semantics between library / database; it is the same in this case. However, for some people this turns out to be too rigid.

[free_electron]

nctnico got me intrigued , so i installed it to fool around with it.

at first glance it is very intuitive i could draw schematics and pick parts within the first minute.
rapid fire questions :
- can the keyboard shortcuts be redefined ? saved in an importable/exportable settings file ? that would make it easier to migrate from other tools  . some commands are not intuitive.
- placing a power object requires nets to exist. that requires opening another panel first, defining the netnames. on a port i can drop a port click it and type in the netname i want. (it sys plz_fix when a new port is dropped). power objects should behave the same way. it is counterproductive to first need one window to define the nets, then another window to pick them.
- move operations : always drags wires. make a Drag operator instead. M = move , just what is selected, breaking lines. wiring should not rubberband. D = Drag : rubberbanding of wiring ok
- i managed to draw a diagonal, offgrid wire. how do i edit its endpoints ? when you click a wire it shows a caret at the midpoint. i want carets at the endpoints as well so i can move those. i also want the ability to'break' a wire , or add a bend point.
 -- need handles on begin/end of a section of wire so they can be dragged
 -- need ability to cut wire ( for example hit the / key , then click on a point and it snips a section exactly 1 gridspace long )
 -- need ability to add a bendpoint to a section or at least move the center caret on-grid

pcb editor
-------------
- too much visual clutter. pads with rounded corners show crosshairs, circles and other clutter
- place via terminates routing. when in the middle of routing in want to be able to change layer and go on. so placing a via should shoot the via , then continue on other outside layer ( until you start supporting stacked vias )
- garbage collector : two track segments, running in same direction , with same width need to automerge to 1 segment.

nifty idea's
- auto teardropping on pad entry
- auto teardropping on track width changes


library
-----------
Thats a big mess.

- there is no rhyme or reason on package names. no convention is followed like IPC naming convention that tells you pitch , body size , pincount etc
- many non existing names. there is no such thing as a 5 pin sot23. the correct name is a sot25
- pads, courtyards, centroids , assembly data . all non existing

[dunkemhigh]

there is no such thing as a 5 pin sot23

On the contrary, there are zillions. Perhaps you mean Diodes Inc no longer use the designation, but there are more sources than them.

Hmmm. Seems there are indeed zillions, but some big names are going, or have gone, SOT-25/26. Still, devices are specified as SOT-23-5 so it's a valid designator even if it's not 'official'.

[free_electron]

are going, or have gone, SOT-25/26.

nope. it has always been sot23 25 26. in the heydays it was labeled sot23-5 and -6 but then when they moved to smaller IEC stepped in and said "this needs to be uniform. you want 323 353 36 so make it 23 25 26 . Afaik the -5 -6 were never ratiified as iec standards.
just like it has always been sot323 353 363 sot523 sot553 and sot 563

[dunkemhigh]

One lives and learns.

[Lukas]

- move operations : always drags wires. make a Drag operator instead. M = move , just what is selected, breaking lines. wiring should not rubberband. D = Drag : rubberbanding of wiring ok

Dunno, to I've never found that behaviour useful. For disconnecting net lines from symbols, there's the "disconnect tool".

-- need handles on begin/end of a section of wire so they can be dragged

Clicking on the ends of a wire to select the junction it's attached to.

-- need ability to add a bendpoint to a section or at least move the center caret on-grid

If a net line isn't horizontal or vertical, dragging it will bend it. In all cases, there's the "bend net line" tool.

-- need ability to cut wire ( for example hit the / key , then click on a point and it snips a section exactly 1 gridspace long )

You're the first one to bring this up, can you describe some cases where this would come in handy?

- too much visual clutter. pads with rounded corners show crosshairs, circles and other clutter

Turning off outlines will get rid of these.

- place via terminates routing. when in the middle of routing in want to be able to change layer and go on. so placing a via should shoot the via , then continue on other outside layer ( until you start supporting stacked vias )

That's exactly how it works. To get the pcb editor to switch layers automatically, you'll have to set up a layer pair rule.

- pads, courtyards, centroids , assembly data . all non existing

Huh? Pads, courtyard and assembly outlines are there. Are you referring to the pick&place centroids? So far everyone has been happy with the pick&place center being on the package's origin.

[free_electron]

Dunno, to I've never found that behaviour useful. For disconnecting net lines from symbols, there's the "disconnect tool".

the problem with rubberbanding is that it vomits diagonal lines everywhere ! and that takes oodles of time to clean up. select a group of elements. move out of the way so i can add things inbetween. if the move is anythng else than pure horizontal or vertical things become a mess. that's why i want explicit move operators and drag operators.

Clicking on the ends of a wire to select the junction it's attached to.

doesn't work on diagonal lines.

You're the first one to bring this up, can you describe some cases where this would come in handy?

to move a section of schematic out of the way to insert another block. i should be able to slice individual wires , or draw a 'slicing line' anything crossed by the slicer gets broken. ( also need in the pcb tool by the way !)

Turning off outlines will get rid of these.

make it off by default.

That's exactly how it works. To get the pcb editor to switch layers automatically, you'll have to set up a layer pair rule.

why is top -bottom not the default pair ? that pair should be there when a blank project is created.

Huh? Pads, courtyard and assembly outlines are there. Are you referring to the pick&place centroids? So far everyone has been happy with the pick&place center being on the package's origin.

The package origin is not always the pick origin. for layout purposes ( like to be able to place parts on a grid ) you may want to set the origin at pin 1. the pick and place point could be somewhere else ( like the gravity point . you don;t want parts to wobble when the machine is moving. manufacturers specify where the pick point is for 'difficult parts like connectors. some even have special caps that need to be removed after placing. (pinheaders)

in the schematic symbol editor : how do you change the grid size ? like on a  transistor : the arrow is drawn off grid. When i try to manipulate that it always want to snap to the pin grid. how do i get a grid that is , let's say 1/10 th of the pin grid ? note that i do want a grid ! just one that is smaller than the pin grid so i can make nice symbols.

Arcs also have a weird behavior. normally you set center , then diameter , then begin and end angle. this is like click on one point, click another point then drag to create the curve. Not always easy...

[Lukas]

the problem with rubberbanding is that it vomits diagonal lines everywhere

That's been somewhat improved in Version 2.0: https://blog.horizon-eda.org/new/2021/05/31/whats-new-2.0.html#fewer-crooked-net-lines

doesn't work on diagonal lines.

Is the diagonal line between two pins? If that's the case its start and end points are determined with the pins. In that case, just drag the line to bend it.

that pair should be there when a blank project is created.

Good suggestion, just implemented it: https://github.com/horizon-eda/horizon/commit/7c54653cc70ea0e65991a819ceeb3b90fb68bccb

how do i get a grid that is , let's say 1/10 th of the pin grid

Hold down the fine grid modifier key (usually alt, can be changed in the preferences). Should only be required when drawing symbols though.

normally you set center , then diameter , then begin and end angle.

Seems reasonable, should land in a couple of days.

[Lukas]

It's been at while, but now we're at version 2.2: https://blog.horizon-eda.org/progress/2022/01/09/progress-2021-09-12.html

The banner feature this time is support for hierarchical schematics.