#qi-hardware IRC log for Sunday, 2011-10-16

wpwrakfrom the description, it may just be PWM+FET00:00
DocScrutinizerwell, that circuit is nothing else00:02
DocScrutinizerunless you want to spoil your efficiency with a series-R you'll always need a schottky diode (or seconf FET) and a choke00:04
wolfspraulshould I support him with a small run and opening up the design?00:07
wolfspraulI will find out more data points first :-)00:07
wolfspraulwpwrak: rc4 related question - can Adam start with the design verification of the new gate/4.4v reset circuit?00:27
wolfspraulwe have sold over 30 rc3 already, not bad :-)00:28
wolfspraulone open question was the CE0/WE# pullup, my understanding is that you already verified that at least CE0 does no harm?00:29
wolfsprauleven if it doesn't help much, should we still add it?00:29
whitequarkmaybe someone knows: what will happen if I accidentally swap the wires of an LVDS pair? will it explode? will it work?00:32
wolfspraulthe one big open question about rc4 for me is whether we should change the VGA connector to DVI-I00:34
wolfspraulif we do that, we have to change the acrylic side panel, we need to source a DVI-I to VGA adapter cable00:35
wolfspraulthe box printing also needs to be updated then00:38
wolfspraulwe need to decide whether we want dual-link or not00:38
wolfspraulpossible reasons to not want it:00:38
wolfspraul1. difficult to route the wires space-wise00:39
wolfspraul2. even if not difficult, waste of fpga pins for no practical usefulness (that depends on how many free pins are left, and how likely it is that anybody will want to use those wires)00:40
wolfspraulactually it could be a nice way to get wires out, for hacking purposes :-)00:40
wolfspraulso anyway, that's the only big open question00:40
wolfsprauldvi-i now or stay with vga...00:40
zrafaafter hellow world with CMD bit of ubb today I will try to have CLK = 0 and CMD  = 1 via MMC00:47
kristianpaulbtw rc4 will include a non fully populated M1 target for hardware people? i even actually wanted to buy a M1 the other days but price still high ... tought at least for me00:47
zrafano idea how to check with scope yet. I took the scope and it is next to me 00:47
zrafasleeping :)00:47
kristianpauland others i had talked about M1 and hhdl00:47
wolfspraulkristianpaul: what do you want to leave out? describe the product to me that you would like to buy00:48
kristianpaula M1 with no VJ accesories, no VGA and better exp bus (more pincs and dedicated clock lines)00:51
kristianpaulno case either at least with your prices ;)00:51
kristianpaulot make that an option00:51
kristianpauli do wonder if that can reduce price at least up to 200usd?00:52
wolfspraulok, one by one00:53
kristianpaul:)00:53
wolfspraulfor "better exp bus", I asked many times but got very little feedback still00:53
kristianpaul*g*00:53
wolfspraulplease help us by clearly making some suggestions what that 'better exp bus' would be00:53
wolfspraulspecifically00:53
wolfspraulby definition a perfect 'expansion bus' is hard00:53
kristianpaulsure00:53
wolfspraulbecause you don't know what the expansion will be00:53
wolfspraul:-)00:53
wolfspraulbut just saying "better" doesn't help much either :-)00:54
wolfspraulyou are probably one of the guys with the most thinking into the expansion bus by now, next to Sebastien or Werner00:54
kristianpaulbut double the number of pins and adding at least two dedicated pins for clocks sounds better?00:54
wolfspraulspecifically00:54
wolfspraulneed a patch00:54
wolfspraul:-)00:54
wolfspraulplease understand how small our community is00:54
kristianpaulsure :)00:54
wolfspraulyou are holding prime knowledge in your brain :-)00:54
kristianpaulyes00:54
wolfspraulI need an exact schematic diff00:54
wolfspraulwhich pin do you want to go where00:55
wolfsprauland how do we preserve hardware compatibility with the expansion header we have now00:55
wolfspraulso "better expansion bus" - yes, I'm very interested00:55
wolfspraulalways was00:55
wolfspraulbut what is 'better'? :-)00:55
wolfspraulyou tell me00:55
wolfspraulI think we should preserve compatibility with our current header00:55
kristianpaulok, i got your idea 00:55
wolfsprauldid we change the expansion header between rc2 and rc3?00:56
kristianpaulha, easy if i drop VGA i have some "clock dedicated" pins from there, problem solved ;) (kidding)00:56
wolfspraulI thought there was a change somewhere00:57
kristianpauldont remnber i dint read too much about rc2 at that time00:58
wolfspraulok let me check that now, in the PDF schematics...00:59
wolfspraulwhat do you mean with "no vga"?00:59
kristianpaulremove vga related layout and components from the design01:00
wolfsprauldon't understand01:03
wolfspraulwhy that?01:03
wolfspraulwhat's the display output then?01:03
wolfspraulok I checked the history of J21, the current expansion header01:04
wolfspraullooks like it is unmodified between rc2 and rc3 - good01:04
wolfspraulI think we should preserve that J21, or at least mechanical and electrical compatibility with it01:04
wolfspraulif you have ideas how to improve J21, add to it, add a second one, etc. please specify them clearly01:05
kristianpaulok01:05
wolfspraulI absolutely want to roll such improvements into new runs like rc4, but they need to be clear, then others have a chance to react, and if they don't then it will just show up :-)01:05
wolfspraulI think it's already clear there cannot be the perfect expansion header01:06
wolfspraulby definition01:06
wolfspraulexpansion expands into the unknown01:06
kristianpaulsure01:06
wolfsprauland there are too many subtleties one may realize later "oh, it would be nice if one pin would come from this other bank" etc.01:06
wolfspraulone thing we can do is to maintain stability for the expansion header we have now01:07
wolfspraulcompatibility01:07
DocScrutinizerheh01:07
wolfspraulso if anybody makes a board or connector for that, they can carry it forward01:07
wolfspraulkristianpaul: please explain 'no vga', I still don't get it01:08
wolfspraulyou mean an m1 without any display?01:08
kristianpaulyes01:08
kristianpaulor at least thinking about not populate some components in current PCB (DMX, MIDI, Video-in, DAC)01:08
kristianpaulMy first concern is price, features could get done next :)01:09
kristianpaulBut i do understand there is few people, and will consider a patch before i talk about (will be nice if) features again.01:11
wolfspraulnah talk is better than silence01:11
wolfspraulI just want to move from talk to action01:11
wolfsprauland that requires more specifics01:11
wolfspraullook at J21, think about it :-)01:11
wolfspraulas far as I know nobody right now uses this header01:12
wolfspraulwe don't need to maintain compatibility with something that doesn't exist01:12
DocScrutinizeryou know you don't want *me* to do that ;-D01:12
wolfspraulbut it would be nice, whatever J21 improvement we make in rc4, that a future user of the header could still support rc2/rc3 as well, unless he needs a feature that was added only in rc401:13
DocScrutinizeranyway not populating things for price cutdown doesn't make sense on a batch that small01:14
DocScrutinizerwhat's J21? VGA standard plug?01:15
kristianpaulno VGA, just a expastion header01:16
kristianpaullet me show  a pic01:16
DocScrutinizeraah01:16
zrafaDocScrutinizer: how the price could be reduced in some product like MM?01:19
zrafaDocScrutinizer: sorry for the tooo general question01:20
DocScrutinizerI'm not the business dude, but I think fist you have to look closely at where yur money goes right now01:20
kristianpauli see that one thing looks constant for the run is SMT price01:20
wolfspraulI don't think leaving out chips makes much difference01:21
wolfspraulthe best is to keep 1 unified product, and increase the quality, manufacturability (yield) and volume of that solution01:21
wolfspraulif anybody disagrees - go manufacture yourself, all files are open :-) like kristianpaul did with his case, for example01:22
kristianpaul*g*01:22
wolfspraulour product is already crazy cheap01:22
wolfspraulwhich is good01:22
wolfsprauland we will drive it down further01:22
wolfspraulbut just above committing suicide01:22
wolfspraulthere is a reason you will never find a new unusual product like this for such a price01:23
wolfspraulnobody is crazy enough and has the capabilities to do it01:23
wolfspraulfor those that prefer to work with dev boards full of bugs that will never be fixed - go ahead :-)01:23
wolfspraulit's funny that over the years I noticed a dev board pattern01:23
wolfspraulthe people that are saying "look this 19.99 USD subsidized dev board is so much better" are mostly the ones that have never bought a dev board or done anything with one01:24
wolfspraulwhereas people who are regularly buying and using dev boards are *amazed* that M1 comes out at only 499 USD and happily buy one (I have such customers)01:24
wolfspraulinteresting, eh? :-)01:24
wolfspraulso my focus is on making rc4 as good as possible now01:25
wolfspraulfirst the reset ic / nor corruption improvements, then other little things we learnt in rc301:25
kristianpauli donk think they remenber the word subsidized 01:25
wolfspraulyield and manufacturability must go up01:26
wolfspraulthis will help everybody later01:26
kristianpaulDocScrutinizer: http://wikisend.com/download/348752/M1_rc3_labled_u7u19u20.JPG01:26
wolfspraul'better' expansion header - yes please!01:26
wolfspraulupgrade to adv7181c01:26
wolfsprauland so on01:26
wolfspraulmaybe leds for the ports01:26
wolfspraulmaybe upgrade to dvi-i01:26
wolfsprauljust full power improve everywhere ;-)01:27
wolfspraulkristianpaul: please help with specific ideas for better expansion header(s)01:27
kristianpaulOkay01:27
DocScrutinizerkristianpaul: sorry, I wouldn't know how to use that01:28
kristianpaulI do np :)01:28
DocScrutinizer(actually I probably could get it working, by downloading it, storing it to some temp dir, renaming it .JPG -> .jpg, then opening it in file browser 01:29
DocScrutinizer)01:29
kristianpaulahh, i tought you mean about J21 ...01:29
kristianpaulDocScrutinizer: http://kristianpaul.org/~paul/tmp/M1_rc3_labled_u7u19u20.JPG01:31
wolfspraulseems he needs lower-case .jpg01:31
kristianpaulah wait01:31
DocScrutinizernp01:31
DocScrutinizeralready watching it01:31
kristianpaulDocScrutinizer: http://kristianpaul.org/~paul/tmp/M1_rc3_labled_u7u19u20.jpg01:31
DocScrutinizeras long as minme tipe isn't octet stream ;-D01:32
DocScrutinizermime type01:32
DocScrutinizerpost connector01:33
DocScrutinizerk01:33
DocScrutinizerpretty easy to extend01:33
DocScrutinizerif you leave out 2 posts, then the old plug will fit on subpart of new extended J2101:35
DocScrutinizercompatibility: check01:35
DocScrutinizerlayouters will hate me01:36
DocScrutinizerOMFG what's that upper right? USB?01:37
kristianpaulYes01:37
DocScrutinizerand those are meant to last?01:37
DocScrutinizerlooks flimsy01:38
kristianpaulIn the case i bet yes01:38
DocScrutinizerhow about using standard dual USB breakout with post connector and ribbon cable?01:39
DocScrutinizerwell, adds to BOM01:39
DocScrutinizerhaha for the xtal 40°01:40
DocScrutinizer45°01:40
DocScrutinizerquite cool design :-)01:41
DocScrutinizerwhat are those funny little red arrow stickers?01:42
DocScrutinizerand what's that module lower left?01:44
DocScrutinizermust be a PSU01:46
kristianpaulyes01:46
kristianpaulhttp://www.ohwr.org/attachments/365/8b10b.png 01:50
DocScrutinizerone tip from dirty every day experience: make your testpoints vias with the nice golden ring around the hole - WAY easier to place and *keep* a probe on such a via01:54
DocScrutinizerand just in case you're going to permanently contact such a testpad, you can use true thru-hole wiring instead of just soldering a wire on top of the pad (which has a certain tendency to come off when somebody pulls that wire a bit)01:57
DocScrutinizerif you use 1.2mm holes then there are even plugs that could plug in01:58
DocScrutinizerat least I think I've seen such plugs01:59
DocScrutinizerfor J21 it seems possible to extend it to "north" quite a bit. To maybe 38 or 40 pins (minus the 2 you need to cut away to allow old 18pin connector to fit)02:01
DocScrutinizerwolfspraul: toldya you don't want me to look at it ;-D  Whoever is your layouter, will hate me forever for this suggestion02:02
DocScrutinizerkristianpaul: http://www.ohwr.org/attachments/365/8b10b.png puzzles me02:05
kristianpaulDocScrutinizer: http://www.ohwr.org/documents/3602:07
DocScrutinizerkristianpaul: well, nice. But that'S not related to the module lower left side labeled U11 if I can read that02:11
DocScrutinizeris it?02:11
kristianpaulnope, thats a regulator i think02:13
kristianpaulhttp://www.ti.com/lit/gpn/pth04000w 02:14
DocScrutinizerthat'S what I thought, PSU aka regulator02:16
DocScrutinizerI get a bit of headache seeing those connectors with massive force associated and consequently with screws to hold the component against housing, at more than one side of PCB: VGA and DMX02:18
DocScrutinizerscrewing both to the case inevitably introduces mech tension to the PCB, even without somebody jogging the plugs in those receptacles02:20
DocScrutinizeraah wait, VGA has no case-mount02:22
DocScrutinizerthough it probably better had02:23
wpwrakwolfspraul: (WE#/CE0) i've tried WE#. does no harm, correct. may help a little. data is a still a bit foggy.03:53
wpwrakzrafa: (CLK/CMD via MMC) great !03:56
wpwrak(exp bus) extension north is probably hard. must be a busy area.04:16
wpwrakbut yes, the rest of the surface is pretty much covered :)04:16
wpwrakwhat i don't like so much about the header is that there's no peer to let an extension board to form a mechanical "bridge"04:17
zrafawpwrak: my problem is the RON :)04:17
zrafawpwrak: so far, the level of RON I drunk is okey to work yet04:17
zrafawpwrak: but i have the bottle near04:17
zrafaand there is coca yet04:18
wpwrakzrafa: ;-)))04:18
zrafawe will see how long04:18
zrafaI will have energy and patience for this task04:18
wpwraktry increasing the coke to rum ratio. then you'll last longer :))04:19
zrafawpwrak: I have understood some important bits so I think that I did some advance at least04:19
wpwrakkewl !04:19
wolfspraulwe can remove the J3 audio codec expansion header, I think04:25
wolfspraulwpwrak: do you want Adam to help with some design verification?04:25
wolfspraulor should he wait a little longer because it overlaps with your nor tests?04:25
wolfspraulAdam is still busy with rc304:25
wolfspraulbut at some point I want to move forward to rc4 with all resources I have04:26
wpwrakhmm, i have to get him a labsw. he has a lot more boards to play with :)04:26
wolfspraulthat's such a useful board, maybe we just make a small run :-)04:28
wolfsprauland give away to some of our trusted core qi hackers04:28
wpwrak(rc4) i'm currently on the pull-ups. WE# looks safe (as expected) but i'm not sure if it helps or not. my test data shows some divergence - one run says it got ~20% better, the next says 50%.04:28
wolfspraulI want to accelerate04:28
wolfspraulso that's another idea04:28
wolfspraulquick labsw run04:28
wpwrakto make up for the 50%, it has now blown up flickernoise :)04:29
wolfspraulso the next for you is to also try CE0 ?04:29
wolfsprauland the gate and 4.4v reset ic?04:29
wpwrakyup, in that order04:30
wolfspraulI would just make all those mods and once and see whether the corruption is gone for good :-)04:30
wpwrakthe problem is that i still seem to have external factors. maybe i just need more cycles to exorcise them, but maybe not04:31
wolfspraulshould we include leds for ports in rc4?04:31
wolfspraulSebastien is traditionally a bit reluctant about leds04:31
wpwrakah ? how come ?04:31
wolfsprauland I don't know how many free fpga pins we actually have left04:31
wpwraka lot, i think :)04:32
wolfspraulI vaguely remember roh proposing something along those lines many moons ago04:32
wpwraksee page 3 of the schematics04:32
wolfspraulroh suggested 4 leds in the corners, for illumination effects04:32
wpwrakheh :)04:32
wolfspraulI am fine with all this, leds for ports, leds for corners, etc. as long as nobody feels the product is being ruined04:32
wolfsprauland if there are free pins in the fpga, its' probably better to move them somewhere rather than let them die under the bga04:33
wpwraknot sure how constrained the board is with regard to the routing, though04:33
wolfspraulthat's assumign there are no routing problems...04:33
wolfspraulsure, but first we need to define what we want04:33
wolfspraulanother thing is the need of the metal sheet at the bottom side04:34
wolfspraulwith every layout change and run we loose a chance to remove the need for it, assuming our tests are regression-safe04:35
wpwrak(sheet) do you know why specifically it's needed ?04:35
wolfspraulsort of04:35
wolfspraulI stood right there when the test failed, so I can describe04:36
wolfspraulwe put a thin large metal sheet on the test table04:36
wolfspraulthen on top of that a plastic insulation layer04:36
wolfspraulon top of that we put the assembled M1 (in acrylic case)04:36
wolfspraulwe let M1 run04:36
wolfspraulnow we induce a 6kV discharge into the metal04:36
wolfspraulthe metal sheet04:37
wolfspraulwhat happens is that there are some visual corruptions in the rendering04:37
wolfspraulafter 2 or 3 such shots, m1 just freezes or goes dark04:37
wolfspraulif we turn the m1 upside down, the test passes04:37
wpwrakfunny :)04:38
wolfspraulso it's some field that goes through the insulation plastic and acrylic into the pcb04:38
wolfspraulif the distance is a bit more (such as when m1 is upside down), m1 keeps running04:38
wolfspraulbut if the pcb is only maybe 1 cm or so away from the metal, maybe it corrupts memory04:38
wolfspraulmemory corruption was Sebastien's wild guess from seeing the behavior04:39
wolfspraulthe problem is to identify the 'sink'04:39
wolfspraulwould be time consuming04:39
wpwrakit may be hard to reproduce, too04:39
wolfspraulso we just added this metal sheet between acrylic and pcb, and indeed that fixes the problem04:39
wolfspraulno04:39
wolfsprauldefinitely reproducible04:39
wolfsprauleasy04:39
wolfspraul4kV also shows it04:39
wolfspraulwe tested this many times04:39
wpwrakah, good.04:39
wolfspraulthis 6kV shot creates some field04:40
wolfspraulthe field enters the pcb and corrupts maybe the memory04:40
wolfspraulthe field gets a lot weaker with distance04:40
wpwrakDocScrutinizer51 probably knows a lot about EMI :)04:40
wpwrakyes, all that makes sense04:40
wolfspraulthe way to fix this in the right way would be to find the exact 'sink', where the field/corruption enters04:40
wolfspraulbut since we found the metal sheet workaround, we put that aside04:41
wpwraknot sure how to fix it, though. perhaps just a continuous ground plane at the bottom would do the trick04:41
wolfspraulnow, with every layout change we have a chance to try something04:41
wolfspraulif we are sort of sure we don't introduce regressions, there are almost no costs to that04:41
wolfspraulthere are lots of ground planes already04:41
wolfspraulbut I don't know the exact details of the planes, vias, etc.04:41
wolfspraulif we could remove the neet for the metal sheet, that would be a big improvement in terms of manufacturability and cost04:42
wpwrakthe bottom has many split planes04:42
wpwrakso field goes into the comes out through the other. current flows between.04:42
wpwrakis the metal sheet so expensive ?04:43
wolfspraulis it easy to make a change that might help04:43
wolfspraulor risky?04:43
wolfspraulit's very labor intensive04:43
wolfsprauland it looks bad too04:43
wolfspraulthat metal sheet plus insulation layer alone is probably 30% of roh's work04:43
wolfspraulthe next big thing are the buttons04:44
wolfspraulanother 30%04:44
wolfspraulthe entire rest is 40% of the work04:44
wpwrakwas that testing done after or before we shorted all those chokes ?04:44
wolfspraulall my guess, roh can feedback if I'm wrong04:44
wolfspraulshorted what?04:44
wolfspraulmaybe it would also help if we moved some of the stuff on the bottom side to the top?04:45
wpwrak(buttons) ah, let me show you something :) http://downloads.qi-hardware.com/people/werner/tmp/m1butwood.jpg04:45
wolfspraulalthough there's probably a reason it's there...04:45
wolfspraulyes I saw that pic already :-)04:45
wpwrak(shortened) the Ls that "connected" ground planes.04:45
wpwrak(but left all the spikes to go through the chips, making them rather unhappy)04:45
wolfspraulyou mean L3 and L1904:45
wolfspraultesting was done before that04:46
wpwrakyes, probably. don't remember the numbers.04:46
wolfspraulif Joerg has any ideas what low-risk improvements to try about this ESD problem, great04:46
wpwrakcould be that L3 and L19 fixed that issue as well04:46
wolfspraulthat would be great04:46
wolfspraulbut I won't hold my breadth, this 6kV thing looked like a bomb04:47
wolfspraulcan those 2 small fixes really help? I don't know04:48
wpwrak(buttons) i'm a little concerned about bringing ESD into the box, otherwise i'd cut them from aluminium. that would look nice. (would take hours, though)04:48
wpwrakthese 2 small fixes aren't so small when you consider what they do ;-)04:48
wolfspraulwe have a relatively good feeling saying that the ESD shot created a memory corruption04:49
wolfspraulit was a software problem that eventually made it freeze04:49
wolfspraulthat was our 'feeling' after repeating this a number of times04:49
wpwrakbefore you basically had two antennas (regarding EMI) and a chip in between. now it's two antennas and a wire. and the chip on the side, happily staying out of the mess.04:49
wpwrakcould be. i don't know what things depend on these two chips. i think sebastien said one provides an important clock04:50
wolfspraulif I remember correctly, with 4kV it showed the problem only sometimes, maybe it took 2-3 shots04:50
wolfspraulat 6kV almost every shot brought it down a little more ;-)04:51
wolfsprauland at 8kV there was no hope04:51
wpwrakso if that gets upset, who knows what can happen. we already know that it can freeze the system.04:51
wolfspraulI think the test requires 4kV and 6kV is optional, or something like that04:52
wolfspraulthere are ways to explain oneself out of this, but that was not our goal04:52
wpwrakif it's easy to do, i'd take a shot with the shorted Ls. they address quite specifically this problem domain04:53
wolfspraulanyway I just wanted to mention it again04:53
wolfspraulevery relayout and run is a chance to improve this, as long as we are sure not to introduce regressions04:53
wolfspraulSebastien would have to go back to the lab04:53
wolfsprauleasy - yes, sort of. but still work.04:53
wolfspraulcost small, if any. that guy is very relaxed.04:54
wolfspraulbut I cannot help with this, the lab is in Berlin04:54
wolfspraulhow much does one of these esd guns cost actually?04:54
wolfspraulthat particular test is very low tech04:54
wpwrakmaybe sebastien can pay him a visit ?04:54
wolfspraulso unless there is a shortage of knowledge, it's easy to play with this at home :-)04:54
wolfspraulyes Sebastien could04:55
wolfspraulif he feels this is an important enough thing to work on04:55
wolfspraulmaybe such a gun can be bought for 100 USD or so?04:55
wpwraksearching ... "ESD Awareness Training" hehe ;-)04:56
wolfspraulthen we need to document the test setup a little to not electrocute ourselves, then we can play04:56
wpwrakhmm, nothing under "esd *" from the usual suspects04:56
wpwrakthe "now i have the gun. let's see if i can upset *this*." phase will be expensive, though :)04:57
wolfspraulnot sure04:57
wolfspraulthere are different levels of 'pass'04:58
wolfspraulthe highest level is that it passes without any issues or interruption04:58
wolfsprauleven then you have different strengths of shocks04:58
wolfspraulat some point everything goes down04:58
wolfspraulthe second level is that it keeps running, but with problem04:58
wolfspraulthird level is that it shuts off or reboots04:58
wolfspraulfourth level would be permanent damage, I guess :-)04:58
wolfspraulsomething like taht04:58
wolfspraulmanufacturers have, as with many of those things, a gree leeway in how they intepret all this and how well they actually want to engineer their products04:59
wolfspraulsince everything will fail at some point04:59
wolfspraulgreat leeway04:59
wolfspraulwe could have left out that metal sheet and nobody would have noticed or cared or it most likely would ever have mattered05:00
DocScrutinizermhm, and a 75kV electroshocker can easily take out of service every phonebooth and cigarette machine05:00
wolfspraulso when you have such a gun, you start with 2kV05:00
wolfspraulcan test your household stuff a little :-)05:00
wolfspraulI bet [75kV]05:00
wolfspraulDocScrutinizer: did you read up on the test above?05:01
wolfspraulany ideas what we might improve in the board?05:01
DocScrutinizerimprove for what? which test?05:01
wolfspraulphew :-)05:02
wpwrak(test your household stuff a little) yeah, that;s what i meant with the expensive part ;-)05:02
wolfsprauljust above, a few lines05:02
wolfspraulwpwrak: yes but you start with 1kv, 2kv05:02
wolfsprauland then you slowly go up05:02
wolfsprauland as you go up, you will see when stuff starts failing05:02
wolfspraulI doubt the very first thing you see is a permanent damage05:02
DocScrutinizerESD test? weardll, dunno the bo05:02
DocScrutinizergrr05:02
wolfspraulDocScrutinizer: I describe it again05:02
DocScrutinizerESD test? well, dunno the board05:03
wolfspraulput metal sheet on table, put plastic insulation on top05:03
wolfspraulthen Milkymist One, assembled in acrylic case, on top of that05:03
wolfspraulshoot 6kV into metal on table05:03
wolfspraulwhat we observed looked like a memory corruption05:03
wolfspraulm1 kept running, but with strange visuals05:03
wolfspraulwhen we turned the m1 upside down, it was stable05:03
DocScrutinizernot exactly a surprise05:03
wolfspraulin upside down, the pcb is about 3 cm further away from the metal that gets the shock05:04
wolfspraulwhen we put a metal insulation layer below the pcb, it was stable also with the pcb on the bottom05:04
wolfspraulso the order then was05:04
DocScrutinizerjep, or like 8 times further, I guess05:04
wolfspraul1. metal on table05:04
wolfspraul2. plastic insulation05:04
wolfspraul3. acrylic05:04
wolfspraul4. metal sheet inside case05:04
wolfspraul5. pcb05:04
wolfspraulthe pcb was < 1 cm away from the metal that got the shock05:05
wolfspraulthat was stable05:05
wolfspraulbut when we remove #4, we get what looks like memory corruption05:05
DocScrutinizeryup05:05
wolfspraulfor rc3, we kept the metal insulation layer inside the case, but it was a lot of hard work05:05
wolfspraulso we are thinking what long shots we can try in rc4, something with low risk of bad side-effects, to improve this05:06
wpwrakDocScrutinizer: that was also before we shorted those chokes/beads connecting certain ground areas. so i wouldn't be surprised if this alone could change the odds of the ESD problem still existing.05:06
DocScrutinizerwell, improve the metal sheet design to reduce work with it05:06
wpwrakmake a metal case ;-)05:06
wolfspraul"improve the metal sheet design" ?05:07
wolfspraulit's a lot of work no matter which way you turn it05:07
wolfsprauland only needed because the pcb itself is not designed to keep the esd impact away05:07
wolfspraulwhich is most likely possible in some way05:07
wpwrakwhy is it so much work ? that's what i don't quite get05:07
wolfspraulbecause it is05:08
wolfspraulmanual labor05:08
DocScrutinizerA) 0.6cm vs 3.6cm may well make the difference between direct spark impact to some chip and the discharge going to the cable or desk's steel frame in the top-down case05:08
wolfspraulnot just running a software loop a couple times more :-)05:08
wolfspraulmetal needs to be sourced, cut05:08
wolfspraulholes need to be drilled05:08
wolfspraulinsulation layer needs to be glued on it05:09
wolfspraulthat thing along was 30% of the workload for the entire rc3 cases05:09
wolfspraulalone05:09
wolfsprauland it's completely unnecessary05:09
DocScrutinizerno, it's not05:09
DocScrutinizera metal shielding against ESD is mandatory quite frequently05:10
wolfspraulyou think there is no way to get rid of it?05:10
DocScrutinizerprobably not05:10
wolfspraulmy guess is that it's a kludge only needed because we combine a lust of meeting standards with lack of resources :-)05:12
DocScrutinizerESD tests are assuming discharge to a GND level metal or other conductor "shield" of any kind. No PCB / circuit can be built in a way such that ESD in middle of the board won't directly impact on any arbitrary chip 05:12
wolfspraulno05:12
wolfspraulnot middle of the board05:12
wolfspraulI described the test setup already twice now :-)05:13
DocScrutinizerisn't the source metal plate in parallel to the PCB?05:13
wolfspraulyes05:13
DocScrutinizerso discharge can hit middle of PCB05:13
wolfspraulthe spark was applied to the side of the metal plate05:13
DocScrutinizerin real life even more likely to happen05:14
DocScrutinizerthat's irrelevant05:14
DocScrutinizeryou could replace the metal sheet inside by a wire going 5 times zickzack across the plastic case bottom05:15
DocScrutinizerfaraday cage05:15
DocScrutinizeror more like lightning protector05:15
DocScrutinizerget 5mm alu sticky tape on endless reel05:16
wolfspraulok I am pretty sure we could make layout changes to the pcb that would let us pass this test05:16
DocScrutinizeryou can't, basically05:17
wolfsprauladd another layer to the pcb to just move the metal to that layer05:17
DocScrutinizer?05:18
wolfspraulno?05:19
wolfspraulmetal layer is metal layer05:19
DocScrutinizerthe problem is your chips aren't *under* that layer05:19
wolfspraulyes they are above05:20
wolfsprauljust like with the metal sheet we added now05:20
DocScrutinizeryou don't need a metal sheet, you just need a wire at same location05:21
wolfspraulalright we don't get many practical ideas out now :-)05:22
DocScrutinizerconnect that wire to central GND point (usually minus of power supply, sometimes EE decides to have GND of input as central GND point)05:23
wolfspraulmaybe someone buys a gun, then we can more easily play with it05:23
wolfsprauland maybe we can find the area of the board where it enters, and a simple metallic tape at that location may be enough?05:23
wolfspraulI don't know05:23
DocScrutinizerrun a gnd trace all along the sides of your PCB05:23
wolfspraulI'm sure there are better ways to improve it than what we have in rc305:23
wolfspraul[gnd trace along sides] ok, we can ask around and see whether someone thinks this is a bad idea05:24
DocScrutinizerthe point where it enters is irrelevant if you catch the discharge spark halfway at the location where your metal sheet is now05:25
wolfspraulwe have a lot of space on the bottom. actually I'm wondering why we can't move all components to the top.05:25
wolfspraulspark?05:25
wolfspraulthe spark enters the metal sheet far away, say 30cm or so away from the M105:26
DocScrutinizerwolfspraul: what is your take on what actually happens to the electrons after they entered the source metal plate ?05:26
wolfspraulmy definition of 'spark' may be a little simplistic, I thought it's just the thing you can see...05:26
wolfspraulsure the metal conducts05:26
wolfsprauland then there's the plastic insulation on top of it05:27
DocScrutinizerit will go to GND one way or another05:28
DocScrutinizerone way was directly to some metal of the desk it sits on05:28
DocScrutinizeranother was to cable (e.g vga, or power, or usb whatever) straight thru the cable's isolation05:29
DocScrutinizeryet another was straight thru the layers 2) and 3) missing 4) metal sheet into a chip on 5) PCB05:29
wpwrakwolfspraul: (sheet) ah, you built it from scratch. i see. i though you got it already pre-fabbed, from one of your famous corner shops :)05:30
DocScrutinizerif it doesn't do a real discharge, then it induces capacitive interference into the most protruding parts of PCB, even thru the layers 2 and 305:31
DocScrutinizer...unless you got either a metal sheet or a few wires there to form a faraday cage05:31
wpwrakperhaps replacing the M1's bottom acrylic with metal would help. a tad messy to produce in low volume, though.05:31
wpwrakmaybe kristianpaul can find a cheap place with a CNC mill :)05:31
wolfspraulno no05:32
wolfspraulthose are separate solutions05:32
wolfspraulof course we can build a faraday cage05:32
wolfspraulwe can build a metal case05:32
wolfspraulpeople use copper spray on the inside of plastic parts05:32
wolfsprauland so on05:32
wolfspraulbut that is not what I am after right now05:32
wolfspraulI am trying to design the pcb better05:32
DocScrutinizerprobably two stripes of metals sticky tape across the diagonals of the arcrylic wil suffice if you connect them to GND05:32
wolfspraulor 'as good as possible'05:32
wpwrakthat would be a partial metal case - just change the bottom plate, keep the rest of the concept05:32
wolfspraulbecause from the feedback we got from the testing guy, he thinks this can definitely be fixed inside the pcb, provided enough resources are thrown at that05:33
wpwrakif you want a quick fix, see if it's already fixed :)05:33
wolfsprauljust think about any modern smartphone, how thin they are, how much stuff is inside - how do they pass those tests?05:33
wolfspraulour problem if anything are our minuscule resources05:33
wpwrakDocScrutinizer: do you know how much such ESD guns / ESD simulators cost ?05:33
wolfspraulwe cannot just throw 100k USD research money at this to make it 'go away'05:33
DocScrutinizeryou can't design PCB so much better that no bad effects will occur to chips directly, no matter if they are on bottom or top of PCB05:33
wolfspraulwell I dout we can say our pcb right now is 'perfect'05:34
DocScrutinizerwpwrak: no, but I guess they are rather cheap05:34
wolfspraulespecially when you haven't looked at it in detail yet (not saying you should, just pointing out a hole in your logic)05:34
wolfspraulmy initial point was innocent05:34
DocScrutinizerit's basically just a capacitor that gets charged to a defined voltage05:34
wolfspraulI wanted to remind myself and others that every new run is a chance to get rid of the need of the metal sheet05:35
wolfspraulwhich would make roh happy :-)05:35
wpwrakdo you actually need the isolation on the metal sheet ?05:35
wolfspraulthat's to avoid that the metal bends and shorts something on the bottom side of the pcb05:35
DocScrutinizerwolfspraul: we for sure can improve a few things on PCB, but that won't solve the problem, just reduce it05:36
wolfspraulespecially after someone unassembles and reassembles the case05:36
wpwrakyes .. but is that likely to happen ?05:36
wolfspraulreduce it may be enough to pass the test05:36
wolfspraulno it's not likely to happen05:36
wpwrakokay, good05:36
wolfspraulwe could have left out the entire metal sheet and just bend the rules like 99% of other commercial companies would have :-)05:36
wolfsprauland 100% of Chinese05:36
wolfspraulmaybe that's why they are so successful :-)05:36
wolfspraulthe right thing to do is to remember this thing with every run, and re-test once in a while, and maybe find a cheap fix that lifts us above the testing criteria05:37
wolfspraulthen everything is 'perfect'05:37
wolfsprauland the metal sheet can be removed05:37
wolfspraulit's a waste of money and resources and distraction anyway05:38
DocScrutinizeranother suggestion: place some posts on gndplane that protrude way further than the sensitive parts next to them05:38
wolfspraulcan you translate it down to my dumbness level?05:38
wolfspraulfirst idea - run a gnd trace around the outline of the board05:39
wolfspraulright?05:39
wolfspraulbottom side I guess? or bottom and top?05:39
DocScrutinizerright05:39
wpwrakif you want to remove it and are worried about being able to sleep at night, i'd just test M1rc3. two major offenders have already been fixed there, and a search for more hasn't turned up anything else that looks suspicious05:39
DocScrutinizerbottom and top, and connect to real GND only in one spot, which is the central GND spot05:40
DocScrutinizeryup, there've been those incredibly stupid chokes in GND05:41
wpwrakDocScrutinizer: GND around the PCB ... how does this mix with connectors that have metal at the edge ?05:41
wolfspraulno worries I can sleep great05:41
DocScrutinizerthose are like building a parabolic dish for ESD05:41
wolfspraulI think one reason for the closedness in the hardware industry is that they really depend on hiding a lot of stuff :-)05:41
wpwrak(-:C05:41
wpwrakwolfspraul: you've heard it ;)05:42
wolfspraulwe have this sadistic mix of wanting to comply perfectly, even with idiotic laws and regulations that nobody follows, combined with a total lack of resources to follow05:42
wolfspraulthat combination leads to being stuck05:42
wolfspraulan ousider might call it stupid05:42
wolfspraulall Chinese will call it stupid05:42
wolfspraul:-)05:42
wpwraki think they quickest way to get rid of the metal sheet is to just test again. the problem may very well have retreated beyond safe limits by now05:42
wpwrakall the rest is good design advice, of course05:42
wolfspraulthis is actually a quality level that makes sense, I'm just saying in general we do have this mix sometimes05:43
wolfspraul"hey, I have found this law here, look, we have to do this and that"05:43
DocScrutinizerwpwrak: mix?05:43
wolfspraulthen I ask around the industry, and nobody even knows the laws, and everybody is busy couting their profits and paying taxes and bribes05:43
wolfsprauljust kidding05:43
wolfspraulback to making M1 the perfect ESD machine...05:43
wpwrakDocScrutinizer: if there's a connector with metal at the edge that doesn't go to "common ground". how would that fit with the "ground around the pcb" rule ?05:44
wpwrakmaybe the answer is "the connector should go to common ground" :)05:44
DocScrutinizererrr05:44
wolfspraulhmm05:44
wolfspraulwhat does this translate to exactly on m1 then05:44
wolfspraulmaybe we even have that already?05:45
wolfsprauland what about those "posts on gndplane that protrude further than the sensitive parts"05:45
wpwrak(posts) lightning rods :)05:45
DocScrutinizereither you have a separate shielding which goes to spark catcher, or you got just one GND (like on cinch) then it goes to analog GND and NOT spark catcher05:46
wpwrak(have that already) yes, maybe. now that the split ground planes have grown together again, we may indeed have that already05:46
DocScrutinizer(lightning rod) yeah, or simply electrolytic capacitors for example05:47
DocScrutinizerthe aluminium case is on - usually?05:47
DocScrutinizeron minus*05:47
wolfspraulno aluminum case05:47
wolfspraulacrylic case05:47
DocScrutinizerelectrolyt capacitors have no acrylic case ;-D05:48
wolfspraulah sorry, lost the context05:48
DocScrutinizerwolfspraul: the whole point is to have the most protruding conducting parts be GND level05:48
wolfspraulok05:49
wolfspraulbut I don't know what that translates to in the m1 pcb :-)05:49
wolfsprauland unfortunately the layout files are in Altium format still05:49
wpwrakthat would actually be the feet05:49
wolfspraulso hard to quickly converge around some specific small ideas now05:49
wpwrakyes, kicad when ? ;-)05:50
DocScrutinizerlooking on an arbitrary PCB next to me I see most protruding parts being xtal, capacitors, a huge choke witj siferrite core05:50
wolfspraulwe can move to kicad, I'm super interested in that, but... the problem is once we do that we loose the ability to work with our layout house05:51
wpwrakDocScrutinizer: this is what the bottom looks like http://downloads.qi-hardware.com/people/werner/m1/tmp/back.png05:52
wpwrakah, the layout is outsourced ?05:52
DocScrutinizererrr05:52
DocScrutinizera photo of a real PCB would help better05:52
wpwrakwolfspraul: one thing that may help with interference hardening is to move long traces traces to an inside layer05:55
wpwrake.g., the SDRAM address lines (those with the R7x and R8x resistors)05:56
wpwrakon the other side of the FPGA, there are also what's probably NOR lines with a similar pattern05:57
wpwrakinteresting. the M1 schematics don't mention what value the series resistors of the SDRAM have05:58
wolfspraulwpwrak: layout outsourced sounds so draconian05:58
wolfspraulI think we have moved way past that05:58
wolfspraulthe M1 electrical pcb files are continuously in Altium format, and yes, we have hired outside consultants to help us with those files, layout, review, give advice, etc.05:59
wolfspraulthere are 2 layout personnel that by now are relatively familiar with the board's layout05:59
wolfspraulif we move to kicad, we loose the ability to get their feedback06:00
wolfspraulso I'm hesitating still06:00
wolfspraulcan we import gerber files into kicad to transition the layout?06:01
wolfspraulthe gerbers are probably available for download somewhere, since we put all files online and they have been sent to the pcb and/or smt fabs06:02
wpwrakyes, we have the gerbers. not sure it they're a nice way to get started, though06:03
wolfspraulfor example the planned change from vga to dvi-i connector06:03
wpwrake.g., you'd have difficulties making things "snap"06:03
wolfspraulespecially if we go dual-link, we have a lot of new wires06:03
wolfspraulthat would be done by the same layout people as before, using the Altium files06:03
wolfspraulAnson Lin and Vera Cheng06:04
wolfspraultrying to convert them to KiCad makes no sense, that will cost us more than not asking them to help at all06:05
wpwrak;-)06:05
wpwrakit's not something you do in a weekend anyway06:05
wolfspraulyes we just have to be realistic here and choose the best path forward for the product06:05
wpwrakyou also need to update your components libs, footprints, etc.06:05
wolfsprauleverything stays 100% open anyway06:05
wolfspraulthe files06:05
wolfsprauleven if they are in a bad format that makes our style of collaboration difficult (read: that will be replaced sooner or later, especially if we are successful)06:06
wolfspraulso...06:06
wolfsprauldo we have some actionable rc4 esd ideas now?06:06
wolfspraulI'm still not 100% clear06:06
wolfspraulI can point Adam to this chat who can in turn point Anson and Vera to it06:07
wolfspraulmost likely nothing will happen then :-)06:07
wpwraka smoother ramp-up would be a Ya done in kicad. build a larger kicad-savvy group in qi-hw. then M2 wouldn't be so hard either.06:07
wolfspraulyes but we are talking about rc4 now and I want to upgrade it with all the nice things we learnt in rc306:07
wolfspraulM1 rc406:07
wolfspraulthe esd issue just came back to my mind06:07
wolfspraulsince I think if we are lucky we can get rid of the metal sheet without much effort06:08
wpwrakone thing that bothers me with kicad is that there's this major change to all the file formats hanging in the air. it's really have that past us already. that thing's been in the air for something like 1-2 years by now06:08
wolfspraulwith some luck...06:08
wpwrakDocScrutinizer: if you just put isolated metal tape on the PCB over some critical traces, without making an effort of grounding it, it should also help to keep the field away, no ?06:09
DocScrutinizera lil bit, depends on how much it's coupling capacitively to GND vs to trace06:10
DocScrutinizerworst case thing get worse ;-P06:11
wpwrakwolfspraul: here's your ESD gun: http://www.edn.com/archives/1996/091296/graph/19di1fg1.htm06:13
wpwrakmore details: http://www.edn.com/archives/1996/091296/19di1.htm#fig106:13
wpwrakDocScrutinizer: yeah, the area should be considerably larger than the trace. otherwise you'd indeed just have added one more trouble amplifier ;-)06:15
wpwrakseems that ESD guns aren't that cheap. http://www.testequip.com/sale/details/HAE0002.html   USD 700006:19
wpwrakthere should be simpler ones, though. hard to google.06:20
wolfspraulDocScrutinizer: thanks a lot for your feedback!06:27
wolfspraulwe have been a little weak and unsystematic in the past in translating your feedback into actual improvements in the runs, but I believe a few things have made it, and we may become more systematic over time ;-)06:28
wpwrakhmm. there are interesting "related" products: http://www.zapper.com/html/ultra_small_blood_zapper.html06:28
wpwrak"blood electrifier" damn. i need that :)06:29
wpwrakyeah, seems that the ESD simulators you can actually find a price for are all quite expensive06:31
wpwrakprobably > 50% going into insurance :)06:32
wolfspraulwhy that expensive?06:36
wolfspraulwith my China brain I'd give it 100 USD06:36
wolfspraul(small) electrocution risk included, of course06:36
wolfspraulI shall add it to my sourcing list, next time I'm near a market... :-)06:38
wpwrakyou should visit the market and see what you find.06:38
wpwrakexactly ;-)06:38
wolfspraulbut, and I am not kidding you. if I would buy one in China I would only operate it wearing some insulating glove that I am sure is sourced and ideally made from a foreign brand06:40
wolfspraulI wouldn't value my own life at 10 USD or so, fair Chinese market value06:41
wpwrak(gloves) ;-)))06:41
wolfsprauljust judging from all the most basic things that are falling apart in my household06:41
wolfspraulamazing how many things one can get wrong06:41
wolfspraulreally06:41
wpwrakhow many electrical fires did you have to put out so far ?:)06:42
wolfspraulso yes, seriously. only with a glove I can trust, sourced from a foreign brand. I am not kidding.06:42
wolfspraulwell06:42
wolfspraulthey know each other06:42
wpwrakso the glove will be more expensive than the ESD critter ;-)06:42
wolfspraulso when I turn on the mini oven and microwave together, the fuse jumps out06:42
wpwrakthat's a good start :)06:42
wolfspraulbecause the cabling and pretty much everything is typically so crappy, you work with small fuses just in case06:43
wolfspraulhopefully the fuses are made in Japan06:43
wolfspraulI am sure one can get that wrong too06:43
wolfspraulfine but I would insist on such a glove06:43
wolfspraulusing a made in China ESD gun06:43
wpwrakso not 100% adapted to chinese standards yet ;-)06:44
wolfspraulI take enough risks even that way, it could explode etc.06:44
wolfspraulbest would be to hire a Chinese proxy worker06:44
wolfspraulremote hands, literally06:44
wolfspraul:-)06:44
wpwrak;-))06:44
DocScrutinizerwolfspraul: I fail to imagine a glove tested for 20kV isolation (20 as you need a safety margin, you dunno if that 8kV tester really works at 8kV - which is btw the most expensive thing of such a device: regulation of the high voltage to an exact value)13:22
Action: DocScrutinizer idly ponders to suggest revamping a 20ct piezo gas lighter13:23
wolfspraulDocScrutinizer: yes, thanks for the notice :-)13:24
DocScrutinizerprobably better meets the specs than a chinese ESD gun ;-D13:24
wolfspraulmaybe there really are no Chinese ESD guns, and that's why the foreign ones are still expensive13:25
wolfsprauleven the Chinese may be unable to sustain the personnel losses from using their own ESD guns :-)13:25
DocScrutinizerfair assumption13:25
wolfspraulI wouldn't be surprised the least if that were actually the case13:26
wolfspraulI have made stranger realizations in China...13:26
DocScrutinizeralso there's simply no market for such a device 13:26
wolfspraulbecause nobody cares about ESD? :-)13:26
wolfspraulhe he13:26
wolfspraulwe become the cynicist club here...13:26
wolfspraulit's not that bad - I will keep my eyes open13:27
DocScrutinizeras chinese don't care at all about ESD tests, and export won't be interested for know reasons13:27
wolfspraulactually in Taiwan I've heard taking those shots is not that bad13:27
wolfspraulthey use it as some kind of mutprobe13:27
wolfspraulwhen you get a strong esd discharge at home, how strong is that compared to a 4kV shot from a gun? don't know...13:28
wolfspraulof course if something is wrong with the gun, all bets are off, I guess :-)13:29
DocScrutinizerprobably quite similar, as those 4kV got picked to simulate exactly that13:29
DocScrutinizerthere's even the HBM, human body model. taletelling, no?13:30
DocScrutinizerhttp://en.wikipedia.org/wiki/Electrostatic_discharge#Simulation_and_testing_for_electronic_devices   http://en.wikipedia.org/wiki/Human_body_model  >>In both JS-001-2010 and MIL-STD-883G the charged human body is modeled by a 100 pF capacitor and a 1500 ohm discharging resistance. During testing, the fully charged capacitor is discharged through the resistor connected in series to the device under test.<< 100pF isn't really lethal 13:33
DocScrutinizereven at 8kV13:33
DocScrutinizerthe PSU will be ridiculously weak as that's cheaper, so no danger to get more power from that side either13:34
DocScrutinizerHBM and MM frequently used to test ESD13:34
DocScrutinizerhttp://en.wikipedia.org/wiki/Machine_model13:35
DocScrutinizerduh, scratch that last link13:35
DocScrutinizer>>Guidelines and requirements are given for [...] types and points of discharge on the "victim" product...<<13:39
DocScrutinizerall this just to get a set of test(-conditions) that most closely simulates your everyday inhouse synthetic carpet ESD event when grabbing the device13:40
wolfspraulyes13:41
DocScrutinizerso, given a piezo spark unit from a lighter feels much the same like everyday ESD events, and also discharge gap length indicates voltage is ~in right ballpark, you probably are just fine to actually revamp a lighter for first reproducible tests13:43
DocScrutinizeradditional benefit: proven small risk of auto-electrocution13:44
DocScrutinizer:-)13:44
DocScrutinizerwolfspraul: could you point me to a set of photos of the bottom side of an assembled M1 and the inside of the bottom case? Latter one preferably with and *without* the metal sheet13:51
DocScrutinizerooh, already getting late over there :-D13:54
wolfspraulDocScrutinizer: ok, some pictures that clarify the overall mechanical construction are here http://en.qi-hardware.com/wiki/Milkymist_One_accessories#packing_instructions14:04
wolfspraulwithout components14:05
wolfspraulhttp://en.qi-hardware.com/wiki/File:Milkymist_one_rc3_bare_pcb_bottom.jpg14:05
wolfspraulhttp://en.qi-hardware.com/wiki/File:Milkymist_one_rc3_bare_pcb_top.jpg14:05
DocScrutinizerwolfspraul: thanks :-D14:05
DocScrutinizervery nice, the case bottom is way simpler to think about alternatives to your metal sheet than I thought14:06
wolfspraulwith components I only have some unfocused shots http://en.qi-hardware.com/wiki/File:M1_rc3_hw_fix2_bottom.JPG14:07
DocScrutinizerinstead of using the metal sheet I suggest using 5mm wide endless aluminium sticky tape and attach two stripes across the diagonals14:07
wolfspraul(that wire is not there anymore now)14:07
wolfspraulhttp://en.qi-hardware.com/wiki/File:M1_rc3_hw_fix2_top.JPG14:08
wolfspraulyou mean like a big 'X' across the entire bottom?14:08
wolfspraulthat certainly would look pretty cool too :-)14:09
wolfspraulbut will it work?14:09
DocScrutinizerwill work way better than nothing14:09
wolfspraulthe ends of the 'X' should be connected to the 4 spacers?14:09
DocScrutinizeryes14:09
wolfspraulwell, definitely an interesting new idea14:09
DocScrutinizerestimated work time: ~60s per unit14:10
wolfspraulwhy will that work as well as a metal sheet covering the whole surface?14:10
DocScrutinizerfaraday cage principle14:10
wolfspraulmy naive thinking would be that a lot of pcb ground is not covered by another layer of metal then, no?14:10
wolfspraulI mean "pcb space" (ground is ambiguous here)14:11
wolfspraulah ok [faraday]14:11
wolfsprauldefinitely worth a try14:11
DocScrutinizerwolfspraul: possibly you should contact each of the 2 stripes only to one spacer, to avoid GND loops14:13
wolfspraulshould they connect to each other in the middle?14:15
DocScrutinizerwill get separated by the glue, and that's OK'ish14:16
wolfspraulwhat's better? for the aluminum to make contact in the middle of the 'X', or to be isolated?14:16
DocScrutinizerit's non-critical but better they don't contact - again for GND-loops14:16
wolfspraulok. and only one end connect to the spacer14:17
DocScrutinizerso the "isolation" provided by the glue was already planned in14:17
wolfspraulmaybe make it just stop right before the spacer on the other side?14:17
DocScrutinizeryes, only one end to spacer14:17
DocScrutinizeryes14:17
wolfspraulI mean make it a little shorter to not reach all the way to the spacer on the other side14:17
wolfspraulok14:17
wolfspraulgot it14:18
wolfspraul:-)14:18
DocScrutinizerof course you may go fancy and create nice patterns with a (then uncut) tape forming e.g "M1" or "JR" ;-D14:19
wolfspraulbut it needs to connect to a spacer?14:20
wolfspraul1 enough?14:20
DocScrutinizerthen run one stripe from pattern to one spacer14:20
DocScrutinizeryes14:20
wolfspraulinteresting idea14:25
wolfspraulnow we only need an ESD gun somewhere then we can easily test :-)14:25
DocScrutinizeractually splicing alias contacting cut tape ends may be a bit difficult, so better not cut it at all when doing that pattern fancy14:25
wolfspraulnah, just X will do14:25
DocScrutinizeryep, think so14:26
wolfspraulif that worked, it would be a big improvement over the insulated metal sheet14:26
wolfspraulok thanks a lot!14:26
DocScrutinizeryw14:26
wolfspraullike I said we still have other old feedback from you in the pipeline14:26
wolfspraulremove this that14:26
wolfsprauldon't know where the list is right now and waht is done and not done14:26
wolfspraulwill study :-)14:26
wolfspraulah yes, there's some here http://en.qi-hardware.com/wiki/Milkymist_One_RC3_Known_Issues14:27
DocScrutinizer:nod:14:29
DocScrutinizerthe chanlog should have it all14:30
wpwrakthey go even up to 40 kV :) http://www.professionalequipment.com/guide-to-electrician-safety-gloves/articles/14:34
wpwraknot too pricy either. class 2 for 20 kV peak: http://www.lonestarsafety.com/index.aspx?Command=GroupInfo&GroupID=1157614:36
wpwrakclass 4 costs a little more: http://www.highvoltagesupplies.com/store/product/class-4-insulating-gloves-1115.cfm14:37
wpwrakbut glove + chinese ESD tester would still be an order of magnitude cheaper than a proper ESD tester :)14:37
wolfspraulwpwrak: I realized something thinking about labsw & the led driver board - both are power supplies :-)14:37
wolfspraulright?14:38
wolfspraulI know it's a pretty crude way to start thinking, but that was my realization...14:38
qi-botThe build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-10152011-1004/14:41
wpwrakthe spacers have some fancy grounding ;-) into adjacent split grounds. i wonder if this is really such a good idea ...14:44
wpwrak(labsw) there's nothing power supply in there. all it does it switch ;-)14:45
wpwraks/does it/does is/14:45
wpwrakand it switches slowly14:45
Action: wpwrak realized that i''ll have trouble sleeping when the power-cycling tests are done. i'll miss the clicking of the relays.14:46
wpwraks/realized/realizes/14:46
wpwrakwolfspraul: ah yes, regarding doing the modifications all together or one at a time, there's that problem of just driving the corruption out of the detection range vs. solving its cause14:55
wpwrakif each change makes it a little less likely, i may very well have a run of 10'000 good cycles, but it's still there14:56
wpwrakexponential distribution is a tricky beast. e.g., if corruptions happen on average every 3350 cycles, then i'd have a 5% probability of not seeing any in a test of 10000 cycles. but any user would also have a probability of 5% of seeing it within only 170 cycles. it may not take very long to reach this point.15:00
wpwrakand there's a 1% chance of hitting it within only 35 cycles. and so on.15:02
wpwrakDocScrutinizer: schematics page 8, left edge, J12 through J15 are the feet. how do you like that ? the connect to four corners of the main ground area16:34
DocScrutinizerwpwrak: URL of schematics please16:41
wpwrakhttp://milkymist.org/mmone/rc3_schematics.pdf16:45
apinkI edited the wiki on the "Format Data Partition" page.  The edit is at the end of the ubifs startup section.  I added the line linking /etc/init.d/start to /etc/rc.d/S99start.17:35
apinkI am a newbie to linux and to openwrt.  Please feel free to edit my edit.  I made a best guess at what should take place but may be very wrong.17:37
jow_laptop"/etc/init.d/start enable" should create the symlink accordingly17:39
apinknewTopic:  Could someone point me in the right direction regarding suspend on the ben?  I would like to get more time out of the nanonote so that I can use it at work all day without recharging.17:40
jow_laptopthis is usually cleaner as it also treats the START= variable correctly17:40
apinkjow_laptop:  Thanks.  I will change the wiki.17:41
zearare people responsible for ben liballegro port in here?18:27
whitequarkwpwrak: hi19:28
whitequarkI've tested the board--looks like the boost and logic parts of it work, but now there's another problem19:29
whitequarkthere is a backlight power converter in the LCD screen, it is powered by 12V 300mA19:29
whitequarkwe have a 12V converter module (sealed and so on)19:30
whitequarkrated 2A19:30
whitequarkon the power on, the power converter does "bzzzZzZzZ(khrzzkhrzzkhr){10}zz.", and, during that, the voltage at converter output is ~6.5V19:31
whitequarkafter it silences out (it turns off at overcurrent), the voltage is ~7V19:31
whitequarkthe lamp emits only a very dim light during the bzzzz'ing19:31
whitequarkwhat can be wrong?19:32
whitequarkmaybe you have an idea or two :)19:32
whitequarkoh, it's rated 500mA, and 2A is the maximal peak current19:36
wpwrakshould still be enough. during the test, is is loaded ?20:40
whitequarkhm21:04
whitequarkthe lamp is connected, and the 12V converter is attached to HV converter before I turn everything on21:05
whitequarkI guess it is, then21:05
wpwrakodd then. well, check the current ...21:12
wpwrak(input current first)21:13
whitequarkhmmm21:34
whitequarkit now "works" in a different pattern21:34
whitequarkit emits a quiet buzz, which is sometimes interrupted by a "scratch" (sounds quite like a spark, but there are no visible sparks, no heat and no smell)21:35
whitequarkthe lamps glow very dim (but they definitely do)21:35
whitequarkand the current is 170mA21:35
whitequark@5V21:36
whitequarkthe output voltage is 5.5V, and the input is 4.63 (down from 4.92 normally) V21:40
whitequarkI guess it's because I am powering it from a different (USB) power source21:41
whitequarkI've tried to attach a spare lamp I had lying somewhere21:45
whitequarkit actually lights up for a second intermittenly (i.e. it is on 99% of the time, but it flickers), and it turns off soon21:45
wpwrakhmm, odd. dunno. maybe a bad module ?21:49
whitequarkit worked in the notebook21:50
whitequarkor do you mean the 12V boost one?21:50
whitequarkthe noise is emitted by HV coil. neither it nor the controller build up any heat21:52
whitequarkeven if it is the inverter, they cost near to nothing21:55
whitequarkhttp://www.pchub.com/uph/laptop/43-25222-2183/YEC-YNV-C02-LCD-Inverter.html21:55
wpwrak"CAUTION  HIGH VOLTAGE" hmm :-)22:03
whitequarkaccording to leaked FCC test results, it's 1.5mA@1.4kV22:09
whitequarkand the input current is 33mA22:09
whitequarkhm.22:09
whitequarkwait, what?22:09
whitequarkdoes it have some kind of perpetuum mobile inside? :D22:10
whitequarkI've powered it off a SLA accumulator with the same result22:12
whitequarkit's quite discharged, through22:12
wpwrakso what does that have to do with your 12 V converter ? it this a third converter in your system ?22:13
whitequarkfourth, actually22:21
wpwraki somehow wonder if it wouldn't have been easier to just get some cheap tablet and remove the case ;-)22:22
whitequarkI have a board (yes, that's mine too) which charges Li-Poly accumulators and converts their whatever voltage to 5V22:23
whitequarkthen, there's a buck which converts that 5V to 3.3V (actually, 3.47V due to only E24 resistors available) to power the logic of LCD22:25
whitequarkthere is a boost 5V->12V which I've already described22:25
whitequarkand 12V->HV which powers the backlight22:25
whitequarkas the motherboard, a StarterKit board is used... remember I've told you about the board with an LDO which gets so hot that PCB turns black?22:26
whitequarkthat one.22:26
whitequarkso, no switching converters there :D22:26
wpwraksounds massively complex :)22:27
whitequarknot any more complex than a random netbook22:27
whitequarkmy one has IIRC five converters22:27
whitequarknot counting the one in power supply22:27
wpwrakyes. but a netbook can probably do a bit more, no ? :)22:28
whitequarkyes, taking a cheap tablet and removing the case is MUCH easier22:28
whitequarkbut it is also much less fun22:28
wpwrakheh :)22:28
whitequarkthe resulting "thing" will be rougly equivalent by functionality to any ARM netbook22:28
whitequarkit will be five times heavier and less convenient, of course, but on the other hand you'll have all the schematics and datasheets and docs and whatever22:29
whitequarkquite in the spirit of OSHW22:29
whitequark(ahem)22:30
whitequarkI'll happily publish eveything under WTFPL, but I doubt that some brave one exists which will try to replicate this design22:30
wolfspraulI was thinking about building a portable (luggable) Milkymist One case :-)22:31
whitequarkalso, it is quite dependent on the scrap parts I (and my friend who wants that) had lying around22:31
wolfspraulwe can make a competition whose portable computer is heavier and bulkier...22:31
whitequarklol22:31
wolfsprauldo you know this? http://chrisfenton.com/diy-laptop-v1/22:32
wolfspraulhttp://chrisfenton.com/diy-laptop-v2/22:32
whitequarkthe 1st one is super simple22:33
whitequarkduh, the second is too22:34
wolfspraulyes, sure22:35
whitequarkcurrent "motherboard" we use runs Linux on some (I don't quite recall which) high-end Cortex-A22:36
wolfsprauldo you have some pictures?22:36
whitequarkyeah, quite a lot of them. I'm planning to write a small article on making the board22:37
whitequarkI don't currently have any for the whole device, sorry22:37
whitequarkforgot to take them22:37
whitequarkwait a sec, I'll upload them22:38
whitequarkI think that the final device will use a SODIMM module with CPU and RAM, and every other thing will be placed on a "hostboard" I'll make: power converters, smart battery system with two Li-Poly cans, LVDS adapter, maybe some more circutry22:40
whitequarkI'd post a link to the module, but the site of the shop is currently undergoing some maintenance (it's 2:40 here in Russia) and 503's22:41
whitequarkanyway, here it is (in the case the site will be up timetimes): http://www.terraelectronica.ru/catalog_info.php?ID=1177&CODE=677003&Name=TAM-3517&Razdel=%CE%F2%EB%E0%E4%EE%F7%ED%FB%E5%20%E8%20%EE%F6%E5%ED%EE%F7%ED%FB%E5%20%EF%EB%E0%F2%FB%20%E8%20%ED%E0%E1%EE%F0%FB%20%E4%EB%FF%20%F1%E5%EC%E5%E9%F1%F2%E2%E0%20Sitara%20(Cortex-A8%20%E8%20%20ARM9)&TableName=class_19_2_41_5_622:42
whitequarkahh, it is this one: http://www.technexion.com/index.php/arm/ti-sitara-am3517/tam-351722:43
whitequarkwolfspraul: http://imgur.com/a/2V0vl22:47
whitequarkany comments? :)22:56
wolfspraulone sec22:59
wolfspraulok nice! I need to learn more about your project for better overview...23:00
whitequarknot sure if everyone on this channel wants to hear about it23:02
whitequarkI'll PM you23:02
wolfspraulwhy not?23:03
whitequarkah okay23:04
wolfspraulplease keep the channel posted :-)23:04
whitequarkso, the ultimate goal of this project is a notebook made from as many perfectly known parts  as it is possible23:04
whitequark*not notebook, but a tablet23:04
whitequarkthe notebook was a year ago :)23:04
whitequarkthe core of it will be a high-end ARM like Cortex-A23:05
whitequarkwith a lot of memory and Linux on-board23:05
wolfspraulI will follow more closely23:05
wolfspraulgotta run now, bbiab23:05
whitequarkhmm23:06
whitequarkshould I wait for him or just leave that in the logs?..23:06
whitequarkwpwrak: how would you design something netbook/tablet-like without a lot of switching power supplies?23:20
whitequarkyou have just Li-Poly voltage which varies between 2.8V and 4.2V per cell, and you need: Vdd=3.3V, Vbus=5V, Vcore=1.8V, Vbacklight=12V (or whatever), maybe more23:20
whitequarkand a li-po charger is a step-down converter too23:21
whitequarkso, five at least23:22
kristianpaulha, after all i discover a mprintf implementation inside gnsssdr.ru osgpa port to solve the unsigned issue fot the 16bits data23:37
kristianpaulso it can print signed :)23:37
kristianpaulwpwrak: yes i do have friends with CNC and better prices than when M1 case23:38
Action: kristianpaul still waiting invitation for altium live so can get the viewer23:39
kristianpauloh,  OsmocomGMR annonuce :)23:41
kristianpaulwpwrak: is okay to ignore that is still having 0.38% of write fails to the namuru core i ported to M1?23:43
kristianpaulI'm about to implemnt a "ramdon" write test and see how it goes, but i'm little concern that i still no getting zero errors report.23:43
wpwrakthat's register writes ? failures sound like trouble ...23:44
kristianpaulcontrol and status register.. i added some general porpuses there to run some basic test initially to measure troghput and later to verify writes23:45
kristianpauland actually that may explain why some code dint worked in rtems but do i the m1 bios23:46
wpwrakhmm. i think you want to debug your registers before relying on them. otherwise, you'll always suspect a register problem when something goes wrong.23:48
kristianpaulk23:48
kristianpauli'll try this as the torture test http://www.codeproject.com/KB/recipes/SimpleRNG.aspx :)23:52
wpwraknice and simple :)23:55
--- Mon Oct 17 201100:00

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