Orcad is a total piece of dog crap!

Yes, FreePCB is only layout. But I don't see where that would be an issue. You can pick any schematic program that you would like to use with it.

I have tried to discuss using a "universal" format of some kind that will allow FreePCB to stay in sync with one of the open source schematic programs. But everyone seems pretty happy with the loose interface using netlist files and change log files for back annotation. I am not a developer, so I don't get a vote really. I can only imagine that the spec of such a file would be simple enough that a common spec could be drawn up and any or all of the open source programs could be adopted to it. Sort of like an open source bus for interfacing hardware, like Wishbone. I know that CodeWright will detect that a file it has open has been changed by another program and can load in the new version. So certainly a schematic or layout program could do the same thing with a shared data file.

Rick

In the open source environment there certainly is hope, in for-pay CAD there isn't, IMHO. Vendors usually do not share information so they are all different. Eagle is one of the best re fwd/back annotation because it's fully integrated. But then they really blew it in the schematic editor by not starting with a hierarchical sheet structure. Beats me why not because OrCad had it all along. I guess now it's too late and they are stuck with the old flat sheet structure.

The netlist interface is kind of clumsy. I do that a lot because I don't do layouts, they are always farmed out. At the end of the day you've got to check your netlist piece by piece, line by line, by hand. Doesn't take a whole lot of time but your eyes burn. Got to do it or the sleep will not be so sound until the PCB is tested.

I'm not sure what you envision here ? Netlist files are a nice simple 'common denominator', and contain the

essential information you need to check. Extensions are things like Attributes, and Rules and PCB footprint control from the Schematic, but for this level of SW, Netlist would seem fine ?

Change logs are also OK, as systems that try to automte the Change process decisions too much can also drop the ball.

Smarter would be an incremental Change handler, that found a difference, and then suggested ways to solve it.

- Swap parts, rename parts, Swap Pins, Swap gates, Defer, etc, with the most brutal Del Part and Del connection, left for last.

'Defer' would be an important one, as it allows you to clean the simple stuff first, and often that makes the larger changes also safer.

But that's really just a machine assisted way to manage the change log, or difference report.

Certainly the common operations could be given keywords, and they should work in all 4 directions the same : PCB->PCB SCH->SCH SCH->PCB and PCB->SCH

Because the (old) PCB -> (new) PCB is one flow, such a standard could be promoted and tested by the PCB side.

-jg

Yes, and this is all "batch" mode processing. I am suggesting that a common file format be considered that would allow two applications to share the same data base. This may be more complicated than I realize, but like I said, I can open a file in Codewright and then change it in another tool. The next time Codewright gets focus, it knows that the file has changed and handles it appropriately.

I am not a fan of batch mode processing between layout and schematic capture. Generating a net list to drive layout is not so bad, but the change functionality seems to be very clumsy with nothing remotely like a standard. It seems like everyone who can bang code thinks they need to reinvent the wheel for this sort of stuff.

My analogy was the Wishbone bus for hardware interfaces between different IP modules. Someone gave this a lot of thought (or maybe not...) and now many open source IP projects use it. So the different modules can all communicate correctly. I think a common data base needs to be developed that will allow schematic tools and layout tools to communicate in an analogous manner.

I don't understand this. What is happening in this "flow"?

Rick

Ah, yes, the problem with this idea, is the SCH.

Suppose you add 2 compoents, and swap 2 connections in the PCB.

If the SCH shares the database, you now have a big problem: On what sheet, and where, do the new parts go, and what if the swapped connection span a sheet boundary ?

Those are operator decsions, and rather hard to Automate

It can work, but the larger the variance, the more clumsy batch mode becomes. I have not yet seen a 'multi-pass' flow, but that would seem the smart way to handle larger divergences Suppose you have 30 changes to handle - start with the simple things, and let the operator (who always knows best!) defer some things for the next pass(es), and that should avoid having to re-draw too much - that makes the batch mode more interactive. Rather like the 'next error' in a Compiler Editor.

I have seen systems that share the PARTTYPE info, but split the SCH Decal and PCB Decal. That seems to be the best

What I was trying to describe was a smarter batch mode (as I don't think you can avoid such snapshot flows) Because one of those snapshots is OldPCB-NewPCB, the PCB developers CAN fully develop, and document the change keywords. They could reserve ADD_SHEET, DEL_SHEET and ADD_OSR, DEL_OSR for example, as SCH keywords the PCN parser should skip.

-jg

ple

Don't forget things like interchangeable components or component inputs, which need to smoothly flow between schematic and layout...

Not such an easy problem ?

Best Regards, Dave

Not really. Let schematic capture do it wherever it pleases, where there is the most space or where it looks best. As long as it puts the stuff there so integrity is maintained. The operator can still move things about if it doesn't look cool.

Where it looks best ? How will the SW make that decision ? ;)

Maybe, but that type of editing can take longer than a simple manual add.

I have seen a system that adds a new sheet and adds OSR tags to all new parts. That's a solution to SOME issues, but high on the ugly scale.

That's why I like the 'interactive batch' idea, Much better to have the sch say 'Adding R75, please choose a sheet and location' and then continue. Giving the operator back control, by being able to pick the next action, seems the best. IME SW coded so that 'it thinks it knows best', is seldom right :)

-jg

Heck, even if it does so buy coin-toss, that's better than something that fell through the cracks because the phone rang while the user was in the process.

Better ugly than any missed back-writes :-)

Plenty of times. For example the SW in autopilots has undoubtedly saved many lives. There always comes a point where a pilot panics. SW can malfunction but it does not panic.

Agreed in general, but have you never seen "kernel panic"?

Yes, but that is a kludge, here are are discussing the proper way to handle revisions :) Anyone can kludge, that is the problem. Keep that as a 'last resort'....

I was talking about CAE SW, preempting the operator. Most common cause of this disease, is the SW is NOT coded by someone who actually uses it, and features are often chosen by committee. Sometimes the poor decisons made, are simply stunning.

-jg

Yes, I've seen SW do stupid stuff and usually that was due to inadequate design reviews or insufficient field testing.

Just read some CAD NG's for the various programs. Almost every day not assuming operator error costs dearly. Today's case: Someone upgraded to Eagle 5.0 but then decided 4.16 was better. Unfortunately he had touched his custom library with 5.0 and saved, probably realizing the millisecond he hit "save" what he'd done. No backup. So I suggested a trick to get it back but since I did not upgrade I don't know whether it'll save his day.

Now had Eagle stored umpteen backup files like it does for schematics and layouts this might have been prevented. Unless he saved one too many times, that is.

Eh? Is this some new behavior for 5.0? IIRC every time I save my lbrs it does the same .l#n backups as for sch and pcb files.

Mind you, for not having backups, he deserves what he gets I'm afraid. I keep my EAGLE docs in my RCS.

Thanks, I'll let him know. Mine doesn't do that but it may just be a setting.

I don't follow this. If you "manually" add the parts to the layout, don't you have to add them to the schematic as well??? Why would you want to do it in that order, layout first then add it to the schematic? I guess this might be useful for things that aren't normally shown on schematics, like mounting holes or parts that are added as convenient like bypass caps. I have been trained by the layout guys to include mounting holes and any other board feature in the schematic as a part so that there is no concern with the feature not making it onto the board. But often this is done at layout and then added to the schematic. Bypass caps are often added or subtracted as layout permits, so that can be done in the reverse order at times.

Still, I don't see a difficulty. I don't expect the schematic program to "draw" my schematic for me. I just expect it to let me know that there are parts added that I have to include on the drawing. Maybe an extra page with any new parts *is* the right way to go. Rather than trying to "draw" a full schematic, each pin on each part will have a short net with the signal name attached. In Orcad, this provides connectivity. Then the added components can be moved to the correct page and connected as appropriate?

Even so, how is this any different than the batch mode? The real problem is that you are trying to add parts in layout when you should add them in the schematic and then work them into the layout. I do this all the time in FreePCB.

What is "ugly" about it? The added parts are there for you to do what you want with them. Do you feel that the tool is the boss??? When you draw a schematic and import the netlist into the layout program, do you live with whatever layout the layout program gives you by default??? No, you use that as a starting point and move things around to suit yourself.

I don't get what is better/different about batch than sharing a common data base. When a part is added by layout and it has no schematic info, clearly this is the same as back annotating an added part and the software can handle it in a similar manner. The difference is that you can have both tools open and add part in either and have that reflected in the other interactively.

I think you are projecting onto this idea your concerns about other tools you have seen. Try opening your mind to the possibility that this can work well and then figure out *how* to make it work well.

Rick

You seem to have answered your own question. BOTH operation methods should be supported.

The operator shoud be free to choose either order.

I was meaning ugly in the visual sense, and also a little clumsy. I thought this thread was about a smarter way to work ?

I prefer taking what works, and improving where it is clumsy. Modern PCs are much faster, so multi-pass is not the issue it was 5-10 years ago.

Quite often SCHs are created by someone else. Do you WANT parts appearing and vanishing from your PCB, or does the SCH guy want parts removed / added, as you change a footprint ?

Sometimes you might do a 'design-stub', where you try something but later decide to undo that. You don't really want the SCH 'alive' during that process. Or, you'll find a stack of New Sheets added, and a couple of parts on each, that you then have to clean up later... The SCH is 100% electrically correct, but visually unusable.

Or suppose you finish the PCB, and go to clean the SCH - ou decide to delete and add on the correct pages, the decoupling caps, and now have a nice neat drawing. OOps, the 'live' PCB unrouted the traces, when the cap was removed.

Now, you could, of course, defer all this with an 'off button', but then you need another pathway to re-sync, so are back to supporting the batch, or my suggested 'smarter batch'.

It is also common for a PCB pgm to need to be fed from more than one brand of SCH front end.

There are reasons a fully shared database is uncommon :) (I have never seen one)

It IS common (and smarter) to share the PartType info (PinName,GateSwap, attributes), but the Symbols are separate, and so are the design / databases.

-jg

I save the entire state of the operating system and all applications (very easy when you work inside a VMWare virtual machine) every time I take a break and right before doing anything risky like upgrading. Any saving that the application does is just extra insurance. With today's large hard disks and fast processors, there is no excuse for not being able to recover from any hardware or software failure without losing moree than an hour or two of work.

Come to think of it, I believe I never lost more than 15 minutes of real work and I had numerous hard drives perform that spectacular final screech-clunk. Since some of those were on laptops I ended up losing more time because I had to find and drive to the next contract repair shop.