| ID: | 64 |
| From: | ge...@igor2.repo.hu |
| Date: | Sun, 4 Sep 2016 17:58:52 +0200 (CEST) |
| Subject: | Re: [pcb-rnd] pcb issues: grid |
| in-reply-to: | 62 from Chris Smith <sp...@icloud.com> |
Hi Chris, On Sun, 4 Sep 2016, Chris Smith wrote: > Hi Igor2, > >> On 4 Sep 2016, at 05:28, gedau@igor2.repo.hu wrote: >> >> Finished the first implemention. Tested them with save-load round trip, looking at preferences/sizes and bottom text (didn't inspect rotue styles). >> >> Chris, please test it on your system to see if round trips are the same as mines. If you have mainline from git in a Linux virtual box, could you please try whether it loads all 3 formats properly? >> >> (Note the new format combo box in the Save As dialog box. Default is the old centimil for now, until we can confirm the human readable format doesn't break) > > Mainline git reads all three formats properly. Thanks for the tests! >>> A. a natural unit .pcb; this would be the default .pcb long term. It would save using %06$mS; as far as I can tell, this would work with any mainline starting at least from 2011. This uses unit suffixes and "random" units. >> >> This should be very similar to mainline git files, except the excess zeros removed. Human readable, no rounding error in round trip. > > I honestly don?t know where you?re going with this one. I have several lines in a .pcb which have mixed units, i.e. offsets in nm and thickness in mils. IMHO, that?s not readable, it?s just confusing. > > I?m of the opinion that the ?natural? units should be whatever the drawing units are for the .pcb, with enough precision to represent nanometers. If I use footprints in other units (mils) then I understand that these are ?foreign? units and will be converted to the natural units on import. What matters to me is consistency ? I don?t mind so much about rounding errors when converting foreign units, rather that the result is exactly the same over save/reload cycles. If you don't do mixed unit boards, this feature is totally useless or even confusing. Unfortunately most of my boards are mixed-unit: almost every part I use are imperial, while all drill bits, board dimensions/outline and much of the DRC are metric. I also want to read my .pcb files with a text editor sometimes. In a perfect world each Coord would have an attached unit info, so the software would remember how it happened - if you configured your drill holes in mm but placed your pins on a 100 mil grid, pcb-rnd would track those holes in mm and the coords in mil. But it's unfortunately not like that, and it'd be a huge edit (and potentially some more runtime overhead). Instead we can let the code guess on save. I think the original intention behind %mS was this. The actual implementation, inherited from mainline, has some heuristics to find the best fit unit, but I have to admit it doesn't work as I'd like it to work. Instead of finding the "roundest", it mainly considers magnituede. I have plans to change this, and assign weigths (preferences) to units. Then if I have a 0.8 mm hole, it will print that as 0.8 mm because "that looks better than 800 um or 31.5 mil, but when I have a 10 mil wide trace it would be printed as 10 mil because it looks better than 0.254 mm. Of course all this is optional. I totally understand your need for consistent units. It wouldn't help me much with my mixed unit boards, tho: if the above heuristics get it right only 70% of the time, I am already much closer than having it wrong 50% of the time. Nevertheless I will probably do a few things to make this as painless as possible: - I will probably introduce a 4th format that prints all values in mm - there will be a configuration setting about what the preferred/default "save as" format should be; anyone can have his own user config that sets his preferred format as default - since .pcb files can have embedded config items, it's then also possible that a given design insists on using a specific (sub)format - so if multiple users work ont he same design with different preferred settings their clients won't VCS-fight, because a plain 'save' will always use the same format <snip> >> - value 0 is intentionally unsuffixed > > > Why is this, out of interest? I am not sure, it's how I inherited the code. I think it's somewhat reasonable in the "most natural unit" case: any unit is as natural as any other unit for value 0. I am not sure I want to keep it like this, as it may be confusing not to have units there but have units everywhere else. Regards, Igor2
Reply subtree:
64 Re: [pcb-rnd] pcb issues: grid from ge...@igor2.repo.hu