#qi-hardware IRC log for Wednesday, 2012-02-29

kristianpaulha, for the guys of laser and drones,  http://guerrilladrone.feenelcaos.org/02:13
wpwrakaye ;-)02:55
wpwrakas i said, blender is a mystery to me. the gui seems weird. and full of things i don't need.02:56
kristianpaulgui is changing02:56
kristianpaulbut is not weird, is just all screwup :)02:56
wolfspraulwpwrak: ok, here02:57
wolfspraulI shall repeat02:57
wolfspraulreferring to your openscad, cadmium, cgal comparison02:57
wolfspraulI was wondering where in the toolchain/process you see openscal (for example) used02:57
wolfspraulis it just to make animations of the case, for advertisement or educational purposes?02:57
wolfspraullike an animated (dis)assembly guide...02:58
wolfspraulif so, even if Blender (for example) is not 'scripted', so what? can blender be used to describe case parts?02:58
wolfspraulif it's just about animations anyway, blender may not be so off-topic then02:58
wpwraki didn't look at animations. just "static" 3D models.02:58
wolfspraulmain question: what are the openscad files used for?02:59
wpwraki'm not sure openscad can even do animations. blender does, of course.02:59
wpwraki would want to use them for milling03:00
wpwrakthat is, the mesh the cad program generates03:00
wpwraksend it through a toolpath generator, and then off to the mill03:00
wpwrakof course, the toolpath generator is a bit tricky :) i've solved that for 2D. but for 3D, i'd still have to go via heekscad03:01
wolfspraulmilling, hmm03:01
wolfspraulso it makes most sense for parts with curved surfaces03:01
wolfspraulfor example the current m1 case starts with flat acrylic parts03:02
wolfsprauland the only thing happening to those parts is to cut some holes/edges03:02
wolfspraulno need for a mesh, openscad, milling03:02
kristianpaulor plastic extrude, yeah heekscad is pretty good for that too :)03:02
wolfspraulxiangfu: let the fun start. I will go through a real-time test of your --qi image now03:02
wolfspraulm1 the bug minefield, let's see how far I make it ;-) (nobody's fault, I love the thing and our work)03:03
wolfspraulwpwrak: I see you had to hold yourself back replying to Yi's "found 2 bugs" mail ;-)03:03
wolfspraulwe should proudly declare m1 to be the buggiest computing platform ever03:03
wolfspraulI think we can easily declare m1 to hold at least (!) 1000 bugs :-)03:04
wpwraknaw, it's not the buggiest ;-)03:04
wpwrak(m1 case) yes, i'd consider this a purely 2D problem. (unless you change the case :)03:05
wolfspraulwe should offer our customers a special return right - *if* they don't find a bug in the first hour of using the product, they have a right to return for full refund... :-)03:05
wolfspraul(I've actually seen that with some other open hardware products, and I liked the idea)03:05
wolfspraulso for the current m1 case, any openscad endeavor would mostly be for the purpose of visualizing the case, animations, ads, educational, illustrated guide, etc.03:06
wolfspraulwhich means - for that use case - one might also look at Blender, at least compare03:06
wpwrakif you want to make an unboxing movie, then yes, something like blender should be suitable. not sure how well it can be scripted, though03:07
wolfspraulthat's why I was asking about 'what is the end goal' first03:08
wpwrakas far as i know, it's not parametric, so you'd have to use scripting as a workaround. that is, if you value your sanity ;-)03:08
wolfspraulbecause I'm sure in terms of artistic freedom, blender will give you more options03:08
wolfspraulsay you want to project images, lights, gradients, whatever on the sides03:08
wolfspraulyou want to rotate it, you want to make the case appear in a 3D stage setting, and so on03:08
wpwrakyes. what i don't know is if blender is good for cad. i was looking for construction. "part A being 10.5 mm from part B" vs. "drag it into that corner until it looks good" :)03:14
wolfspraulyes I totally understand03:14
wolfspraulif the end goal is to mill non-flat surfaces, makes perfect sense03:14
zenlunatic sometimes my ben shutsoff when i pick it up03:57
zenlunaticnvm forgot to upgrade the default firmware03:59
jow_laptophm there's a libdirectfb binding for Lua now, might be worth porting to the NN16:49
kristianpauland milkymist ;-)16:54
jow_laptopyeah, for visual scripting16:55
jow_laptopthe examples look neat16:58
jow_laptopvery simple16:58
zenlunaticis there an appropriate entity to donate to for ben development? openwrt?18:15
viriczumbi: què hi fas per aquí?18:29
Aylait would be time to update the "about" page of GMenu2X19:06
whitequarkwpwrak: suddenly pstree/19:12
wpwrakwhitequark: mmh ?21:17
whitequarkwpwrak: I (again) stumbled upon your name in psmisc AUTHORS. which (again) has surprised be a bit :)21:33
wpwrakwhitequark: ah ;-)21:35
wpwrakyeah. join early, plant seeds, get credit for life ;-)21:35
wpwrakDocScrutinizer: i have a little signal integrity riddle: looking at J21 and J22 in the upper right corner of  http://downloads.qi-hardware.com/people/adam/m1/tmp/m1r4/FPGA_J21_breakcompatibi22:33
wpwraklet's try this again: http://downloads.qi-hardware.com/people/adam/m1/tmp/m1r4/FPGA_J21_breakcompatibility.pdf22:33
wpwrakso, we have ground on both ends of the connector. is this generally a good idea ? i think it should be, because it provides more direct ground return and also puts a ground close to supply rail, which in turn should make it easier to bypass components on single-sided boards.22:35
wpwrakwhat i don't know is just how desirable this is, compared to having just one pair of ground pins. would the difference be major ? or negligible.22:36
wpwrake.g., in terms of the maximum signal frequency we can use. the connectors are 0.1" headers. not exactly RF, but we should make the best of them :)22:38
DocScrutinizerplease rephrase "which in turn should make it easier to bypass components"22:44
DocScrutinizerimpact on HF: none22:45
DocScrutinizerI'd rather dedicate one GND to 3.3/5V22:46
DocScrutinizerbasically it's waste of wires on flat cable22:46
DocScrutinizerunless you expect to power massive dirty load via 3V3 or 5V22:47
DocScrutinizerwhere separate GND lines might have a very positive effect for noise floor22:47
DocScrutinizerOTOH there are de-facto standard wire counts for flat cable, and rather unusual ones. Can't say if 20 is the former or the latter category22:50
DocScrutinizermaybe just keep 1 or 2 of those 4 GND lines NC with a open thruhole testpad, to use for mods of any kind22:51
wpwrakbefore we had 18, which was a non-standard count. 20 seems to be fine. (we went to 20 a few days ago, just because of this)22:53
DocScrutinizerat very least route PCB in a way so user easily can cut GND connection of 1 or 2 of those GND lines, then solder a wire to solder side of header22:53
DocScrutinizerbtw I'm not really happy with pinout of J2122:54
wpwrakthe headers may be used with cables. or with boards.22:54
wpwrakah, why ?22:54
DocScrutinizerfigure what happens when plugged in 180° rotated22:54
DocScrutinizeryou get 5V to GND and GND to some GPIO 22:54
wpwrakyou learn from your mistakes :)22:55
DocScrutinizerparticularly L21P_1 and L9N_122:56
DocScrutinizerwill get reverse 5V22:56
wpwrakyes ... but it's kinda hard to fix this, no ? unless we require that a circuit always connects all the grounds. they we could cook up some scheme that shorts 5V to GND if reversed22:58
DocScrutinizeralso for RF/EMI it was better to have VDD on 1,3 and GND on 2,4 (or vice versa)22:58
DocScrutinizerbut that defeats your safety gap from 5V to IO22:58
wpwrakalso, J21 has the "pair of supplies and both ends" concept in M1rc2/3. rearranging this would totally break compatibility.22:59
DocScrutinizertbh I'd consider cut away post 3 on J2123:00
wpwrakwe could do this with J22, though. not sure if it would make things better it worse.23:00
DocScrutinizerand recommend a plug in pos 3 of connector23:00
wpwrakfor keying ?23:00
wpwrakyou mean a rubber plug ?23:01
DocScrutinizerblocks any user that tries to insert plug 180° rotated23:02
DocScrutinizerif user thinks he doesn't need that, it's not us to blame for blue smoke23:02
wpwrakwhy pin 3 and not pin 1 or 2 ?23:02
wpwrakpin 3 would be good for J22, though23:03
DocScrutinizeron J22 you could cut pin4 instead23:03
wpwrakyes, 3 or 4.23:03
wpwrakwhy 3 o J21 ? and not 1/2 ?23:03
DocScrutinizerthought J21 has 4 GND, one less is no problem23:04
wpwrakboth currently have 4 x GND23:04
DocScrutinizerthat'S why I suggested 3 on one and 4 on the other23:04
DocScrutinizermaybe 1 or 2 on J2223:05
wpwrakand 3/4 on J21 ? or also 1/2 ?23:05
DocScrutinizerok, my ultimate advice: cut pin1 on J22. Cut pin4 on J2123:06
DocScrutinizerdedicate adjacent GND to the 3V3 resp 5V line23:08
wpwrakgrmbl. not all of those nipples are created equal :-(23:09
DocScrutinizerdeclare J21:3 as 'dirty' and 4 as 'clean'23:09
wpwraki just tried to make such a keyed female connector. but the rubber/plastic key doesn't fit23:10
wpwrakdirty ? clean ? one is connected to other isn't :)23:10
wpwrak(rubber thing) well, it's a sourcing problem :)23:10
DocScrutinizerin RL yes, but in clean theory we have two distinct lines23:11
DocScrutinizera future rev may introduce a fuse and choke and buffer C on M1 for J21:2,323:12
DocScrutinizerNOT for J21:423:12
DocScrutinizerget my idea?23:13
wpwraknot sure. would you leave the pin unconnected on the M1 side ? or just ground it ?23:13
wpwrak(M1 is female)23:14
DocScrutinizerfor same reason I'd layout board in a way so user can cut up path from J21:3 to GND23:14
DocScrutinizerdang s/21/22/23:14
DocScrutinizerJ22:1 cut // 2 GND_3V3 // 3 3V3_dirty // 4 3V3 clean23:17
DocScrutinizerJ21:1,2 5V // 3 GND_5V // 4 cut23:19
wpwrakwhat would you use 3V3_dirty for ?23:19
DocScrutinizer3V3_clean aka 3V3_digital, same domain as all the IO and GND on 19,2023:20
DocScrutinizer3V3_dirty for dunno motors etc23:20
DocScrutinizer3V3_clean for digital23:20
DocScrutinizerthat's all mere definition for now23:21
DocScrutinizerthough user already should follow these guidelines on his extension board23:22
DocScrutinizerwhen he doesn't need 3V3_clean for the digital part of his stepper motor control, leave it NC23:22
DocScrutinizerwill make a nice separator on flat cable23:23
DocScrutinizerless crosstalk to IO_L1P_A25_123:24
wpwrakhmm. with the current circuit, expansion boards would have to provide their own cleaning.23:24
wpwrakwhich would seem appropriate punishment for trying to hook up motors :)23:24
DocScrutinizerwhatever, you got my suggestion23:25
wpwrakthe clean/dirty rails scare me :)23:25
DocScrutinizercall it raw and digital then23:26
DocScrutinizererr, other way round, of course23:27
wpwrakwe already have "raw". but that's something else.23:27
DocScrutinizers/raw/high power23:27
wpwraki'd just say "don't connect motors, tesla coils, or tachyon colliders" :)23:27
wpwraklet's not over-engineer this :)23:28
DocScrutinizerI'd just say "never short the two 3V3 at your extension board. Use 4 for digital and 3 for other supply purposes"23:28
wpwrakif we really wanted to supply tons of power, we could add 5V/GND header right at the DC input. do as they do in PCs.23:29
DocScrutinizerwho's over-engineering this now23:29
wpwraki'm not saying i want to add this ;-)23:29
wpwrakthat was conditional irrealis (or whatever you call this in english) :)23:30
DocScrutinizerI'm just suggesting one or two sentences in user manual, on recommended use of those lines23:30
DocScrutinizerno engineering at all involved23:30
DocScrutinizeror all just engineering and no real changes23:30
wpwrakmy concern is that we want neither a massive power drawn nor anything sparky in that area23:31
DocScrutinizerso you should add a thrid sentence for that23:31
wpwrakwe probably have 500 mA absolute maximum of 3.3 V and 5 V combined23:31
DocScrutinizerit's not the absolute current but rather the edges23:32
DocScrutinizerslew rate23:32
wpwrakbut i like the keying :)23:32
wpwrak(edges) yup. but then, they're most likely to exist on digital anyway23:33
DocScrutinizerand if you drive a 1W LED from the 3V3, I strongly suggest doing this via J21:3 only, NOT via J21:423:33
wpwrakso with that logic, you'd prefer to connect digital to the scrubber :)23:33
DocScrutinizerI don't feel like discussing this right now23:33
DocScrutinizeruse it or bin it23:34
wpwraki'll file it23:34
wpwraksince we don't have a scrubber at present, there's nothing to be gained by adding design rules that are irrelevant in practice23:35
wpwrakbetter if people who do nasty things design their own scrubber23:35
wpwrakit'll also be closer to the consumer, thus we'll have less EMI from the nasty stuff23:36
wpwrakbtw, are you on the milkymist mailing list ?23:37
wolfspraulgood morning everybody23:44
wolfspraulnice to see Joerg - hope you are well!23:44
DocScrutinizerwell, I kinda grew old during the last 4 years23:45
DocScrutinizerworking 10h/day leaves me wasted23:45
DocScrutinizerotherwise fine23:46
wolfspraulwe have to find a jungbrunnen for you :-)23:46
wpwrakhmm. you started in december. that's 3 months. yet you feel as if 4 years had passed. so you're accelerated by a factor of 16 :)23:46
DocScrutinizernah, that deficit in available power predated the start of job at STE23:47
DocScrutinizerand actually job starts to relax a bit, since quantum to learn per day diminished from a month's equivalent to something more manageable23:49
DocScrutinizerthe hardest part though is going to sleep err 50min ago ;-)23:51
DocScrutinizerto show up at work 8 o'clock in the morning (sth I haven't managed to accomplish a single time yet)23:52
DocScrutinizernext week I *have* to23:52
DocScrutinizer3 day seminar on code qulaity, held by ENEA23:53
DocScrutinizeryesterday I got the shocking info they want to keep me for at least another 4 months23:54
DocScrutinizerand expect to extend that to end of 201223:54
wpwrakmy condolences ! ;-)23:55
DocScrutinizerso I have the exciting perspective to handle another few tickets like "one out of 3000 devices shows 100% reproducible data corruption of downloaded file. Please check driver of the interface used for download!"23:59
--- Thu Mar 1 201200:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!