06:23 <R0b0t1> DocScrutinizer05, hey

11:10 <enyc> https://neo900.org/git/?p=misc;a=summary

11:10 <enyc> wpwrak pointed me at ages ago...

11:15 <enyc> i'm sure there was a discussion about documents showing what project doing and what stages needed to get through to achievable completion, useful to attract more interest/buyers/etc

13:37 <wpwrak> enyc: yup, this needs input on how the GDC situation will be handled. that opens a number of fairly different paths. so once this is decided, work on that roadmap can proceed. otherwise, it would just end in fairly general and vague terms, i.e., not have more substance than what we already have.

13:44 <enyc> wpwrak: fair enough, do notify me of any drafts if i'm about

13:44 <enyc> wpwrak: actualyl' i'm not sure what the GDC_situation is even

14:27 <wpwrak> enyc: nikolaus (of GDC) announced that we won't have time to work on the neo900 layout (as planned, though much postponed, also in no small part due to the PP freeze) in the next months

14:28 <wpwrak> enyc: this means that we either sit this out (not very attractive) and hope that he'll have time in, say, half a year (uncertain), or that someone else takes care of the layout

14:30 <varu|zZz> this is pcb layout? & have options been identified?

14:31 <wpwrak> enyc: now, this creates a number of variables. e.g., what EDA tools will be used. for it seems that eagle is only suitable for this sort of task with extensions we don't have access to. also, the workflow between joerg (and to a lesser extent me) and that "someone else" would have to be defined. and of course, that "someone else" could be joerg. which seems to be the direction towards which this is heading.

14:33 <wpwrak> varu|zZz: yes, PCB layout. options currently on the table would basically be joerg doing the layout our contracting someone or a company specializing in such work

14:36 <wpwrak> varu|zZz: the choice of person(s) also interacts with the choice of tools. i.e., until now we had (gdc, eagle + private extensions). alas, it seems that (*, eagle), i.e., without those private extensions, is not suitable for complex layout work.

14:36 <varu|zZz> ahh, i see

14:37 <varu|zZz> wpwrak: see PM

14:42 <wpwrak> the choice of tools is a bit delicate. eagle is proprietary but inexpensive and can partially be used gratis. there are a number of "big" proprietary systems that we can expect to be suitable for such layout tasks, but they're expensive, design files could not be usefully shared with the community (if that's a concern - we could of course still publish PDFs and such), and the amount of learning before they can be used properly may be non-ne

14:42 <wpwrak> gligible.

14:42 <wpwrak> there's also kicad, which is open and nice, but i don't know (yet) if it would be suitable for this kind of endeavour.

14:43 <wpwrak> (yet) i guess i'll have a better idea when i've done the layout for anelok, using kicad's new push-router. there, i'm committed to kicad, and if all else fails, i'll just do the layout manually, like i've done in the past.

14:44 <wpwrak> but manual routing wouldn't really be an option for something like neo900. that's also what eliminates plain eagle, which in fact seems to be even less capable than kicad in this regard.

15:02 <DocScrutinizer05> well, eagle has a new autorouter mode that seems to yield almost good results, but takes literally ages (like 2 weeks or more) when used with a grid resolution as needed by Neo900 (0.05mm) and doesn't integrate well with any manual hints or edits

15:04 <DocScrutinizer05> it's quite possible that I'm not experienced enough to set the right parameters to make the autorouter do a decent job

15:07 <wpwrak> i don't think a "batch" autorouter can be an adequate tool for this sort of work. you want something interactive: you make the strategic decisions, the interactive router figures out the pesky little details

15:09 <enyc> are there irc/discussion channels about such tools, you can ask about, find others who do such work, even .... ?

15:09 <wpwrak> also, for "batch", you need to formally specify every single detail. with interactive, if there's a place where you just know a trace shoudn't go, you just don't try. much simpler :)

15:10 <DocScrutinizer05> wpwrak: ((interactive)) yes, exactly

15:12 <DocScrutinizer05> enyc: well, not exactly for eagle. And when you ask in ##pcb or #electronics or sth like that, you hardly find anybody doing such demanding designs with eagle, and the other tools have the downsides wpwrak already listed

15:16 <DocScrutinizer05> so we have the option that I (an EE with very limited PCB layout skills) is going to try to do something with eagle that nobody else done before

15:17 <DocScrutinizer05> though this doesn't apply yet for the proto_v2 where we have rather relaxed restrictions on layout

15:17 <DocScrutinizer05> we will face exactly this problem for proto_v3 then

15:19 <DocScrutinizer05> or we already find some new layouter right away and this might mean we need to switch the toolchain

15:19 <DocScrutinizer05> or I do proto_v2 and we get a warranty from GDC that they will do proto_v3 and beyond

15:20 <DocScrutinizer05> it's a very complex decision to make

15:25 <DocScrutinizer05> just evaluating the eagle router is already a daunting task: you have like 20 parameters (or even more) which have more or less subtle effects on how autorouter works, this means there are between 2^20 and 1000^30 settings to test, and each run takes ages

15:26 <DocScrutinizer05> obviously no feasible approach

15:28 <DocScrutinizer05> an experienced layouter knows his tool and the settings that have proven to be useful, resp which dials to turn to optimize for a particular scenario

15:29 <DocScrutinizer05> I don't

15:29 <DocScrutinizer05> not for this class of layout at least

15:34 <DocScrutinizer05> now should I try to acquire the needed experience (aka learning)? Should I not care about the demanding stuff that's waiting for us in proto_v3 and simply do a proto_v2 "botch" that doesn't solve any of the layout problems we eventually will face? Or should I already try to find a layouter who is experienced enough to do such demanding layout? (hint: it's about the huge 0.4mm FPBGA of OMAP3)

15:37 <DocScrutinizer05> more hints: TI says it needs *at least* a SIX-layer PCB to layout for OMAP3, and this is with in-pad-vias that eagle can't do at all

15:44 <DocScrutinizer05> even more hints: GTA04 afaik uses 8-layer, N900 uses 9-layer or even up to 11 layers (can't recall the exact layer count)

16:02 <DocScrutinizer05> wpwrak: deducing from how unhappy our layouters at OM were when I asked them to change a singe capacitor footprint, and the moaning about "I need to push all those other traces a 1.5mm to the right" they uttered, I gather that PADS also has no proper push router

16:33 <nf`> for layout you guys already have the dimensions of the N900 PCB figured out? Having the outlines with connector positions already mapped is a good start

16:41 <DocScrutinizer05> sure, that's been accomplished in proto_v0/1 and by werner

16:52 <nf`> DocScrutinizer05: excellent

17:10 <wpwrak> DocScrutinizer05: (pads) or maybe they just didn't know how to turn it on :) when they fixed the infamous crystal traces of hxd8 (dash), the new lauout lady moved things around easily enough. not sure whether that was PADS, though.