#milkymist IRC log for Friday, 2011-07-29

GitHub78[autotest-m1] xiangfu pushed 1 new commit to master: https://github.com/milkymist/autotest-m1/commit/b48ae83238b75c65bc40eff129d989cab037dbf400:22
GitHub78[autotest-m1/master] fix warning - Xiangfu Liu00:22
awgood that i flashed 8 pcs successfully by shorter BEN usb cable. ;-)01:40
kristianpaulaw: :-D01:51
wpwrakaw: smells like a signal integrity issue02:25
awwpwrak, yeah....felt like that usb high speed accompanies most transmission lessons we didn't do well during design...I'll also watch if the vga problems is dependent to this after 20 pcs then well-known from data.02:29
kristianpaul(vga) yeah, thats still on the road..02:30
wpwrakwhat's the VGA problem ?02:31
kristianpaulno signal02:32
awDDC is pass, then just no screen shows up.02:32
wpwrakah, repeatably ? or just sometimes ?02:33
awsome boards02:33
awfrom yesterday's data, I was easily to meet a vga no screen. this morning I've tested 5 pcs then vga is okay...surprised. surely the sample base is not bigger. let's accumulate more said like 15 ~ 20pcs successfully. well..just guessed...;-)02:37
wolfspraulaw: our test software absolutely has to do a 100% crc check on all data in nor02:39
wolfspraulI think it's not doing that now, argh02:39
wpwrakaw: and if you re-test a board with good or bad vga, does the result persist ?02:40
wpwrakwolfspraul: that would be really bad :)02:40
awwas still there no matter what i replugged vga cable more times02:40
wpwrakkamikaze, blindfolded and wearing earplugs ;-)02:40
wpwrakaw: okay. did you try to reflash any of those with bad vga ?02:41
kristianpaulno screen shows up = no signal detected by the screen or just blank screen?02:42
awno signal detected, if have signal detected ,my monitor will have a green LED on. ;-)02:44
awwpwrak, i haven't digged into scope signals though. ;-)02:44
wpwrakthat's one or two escalation steps away ;-)02:45
wolfspraulwpwrak: I would think it persists, and that's not that bad I think :-) just a problem with a chip somewhere02:46
wolfspraulon rc2, you have no idea how many chips Adam had to resolder manually02:46
kristianpaulbut if bios load okay it at least should show a blank screen,that can be swiched with Esc key..02:46
wolfspraulI'm worried that we don't have 100% crc checks in the test software, at least that is my understanding02:47
wpwraknot having crc is bad. particularly if there are signal integrity issues.03:07
wolfspraulxiangfu is traveling right now, but...03:07
wpwrakUSB has its own CRC, but if you hit it persistently enough, some errors will get through03:07
wolfspraulin true foss warrior style, he has his m1 in his backback, and is hacking on the road03:08
wpwrak(i've seen this happen in real life)03:08
wolfspraulI hope we get a CRC version of the test software today or tomorrow03:08
wolfspraulhe has it written, testing now03:08
wpwrakyou may also want to consider switching to full-speed, if that's not too difficult.03:08
wolfsprauland that after our last-minute high-speed fix :-)03:09
wolfspraullet's do one by one. first crc software.03:09
wolfspraulthat will improve our visibility03:10
wpwrakyeah. and all the effort to get people to send back their unfixed boards. those may actually be the lucky ones ;-)03:10
wpwrakCRC is always good to have :) and it'll help you characterize the problem03:10
wolfspraulit's too early to point at high-speed now03:11
wolfspraulalso to keep things in perspective, the majority of reflash ops works03:11
wpwrak(too early) yeah, the libusb issue is somewhere in there as well03:12
wolfspraulwe are testing boards off of a manufacturing run03:12
wolfspraulbugs everywhere03:12
wolfspraulbut all we try to achieve with the boards now is 100% pass, and consistency among them. then the rest that will be discovered in the future is hopefully software fixable.03:12
wolfsprauli'm sure you will not try to tell me that there aren't 'a few' cases of if (board_rev==magic_num) do_silly_thing(); in the typical kernel code :-)03:13
wolfspraulmaybe more like hundreds? :-)03:13
wpwrakthere may be some. although the usual approach is to hide them via platform-specific callbacks and such :)03:15
wolfspraulyou get my point03:15
wolfspraulI try to make boards with consistent quality03:15
wolfspraulthat's a good basis for software03:15
wolfspraulso first of all I want 100% crc, and ugly blind spot imho03:16
wpwrakfirmware you can't verify is broken :)03:17
kristianpaulah, got it now, crc is on bios not test software..03:18
kristianpauli took that for granted :)03:18
wolfspraulwe overlooked it before the run, my fault03:19
wolfspraulwas not on my radar03:19
wpwrakkristianpaul: famous last words :)03:19
kristianpaul(hacking on the road) yeah, untill the battery dies, but is nice, i usually i think i'm more productive that way03:20
kristianpaulat least i can focus better03:20
wpwrakwolfspraul: if you can simply read the image back, that would also be sufficient for verification03:20
wolfspraulas an excuse I can say that things are improving in so many areas, that one just slipped through ;-)03:20
wolfspraulthe jtag verify feature is too slow03:20
kristianpaulwpwrak: read back take time. and you can get few surprises from that back as well ;)03:21
kristianpaulso later you dount about whatyou write and what you think you read..03:21
wolfspraulxiangfu is on it, we'll have it soon I hope03:21
wolfspraulnot via jtag, but via test software03:21
wolfspraulyes, so the test software is much better, it can do the test over and over03:22
wpwrakkristianpaul: well, if you don't trust the readback, then fix it ;-)03:23
kristianpaul(backpack) and also batteries for mm1 included? will be cool, i got this voltaic thing (http://www.voltaicsystems.com/usb-battery.shtml) for that porpuse, but i havent tried, i need secure voltage input first :-)03:28
kristianpaulwpwrak: uclinux is happy with a inteerupt every 700us right?03:29
kristianpaulactually i think this will be run in userpace, as osgps linux module looks kinda forgotten..03:30
wpwrakkristianpaul: 700 us should be okay, even for a relatively slow CPU, yes03:31
kristianpaulgood !03:32
awwhy do default of Line & Mic volumn set at maximum not middle after boot up? curious.03:45
kristianpauli remenber this values are not saved, if you modified and waiting to recover next boot will not happen03:57
awupdated http://downloads.qi-hardware.com/hardware/milkymist_one/production/rc3/test_results/m1_rc3_test_report.ods04:47
awlunch time. ;-)04:48
Guest81573Im Caso04:58
wolfspraulGuest81573: hey there ;-)04:59
wolfspraulthat's the web client04:59
wolfspraulthis channel is quite active on anything milkymist05:00
wolfspraulthe time now is a little bad because Europe and Americas are mostly sleeping05:00
Guest81573Perfect, I'll keep it05:00
wolfspraulso it's mostly just us Asians here :-)05:00
Guest81573lazy people!05:00
wolfspraulthe backlog is at http://en.qi-hardware.com/mmlogs05:00
wolfspraulit helps if you always have the same nick, liks 'caso' (but I guess that was taken so they kicked you to another one)05:01
wolfspraulyou can change the nick with /nick some_nick05:01
wolfspraultry that05:01
Cristianok, my name is Cristian :)05:02
wolfspraulseems that one is not registered05:02
wolfspraulit is :-)05:02
Guest97831Cristian is taken so... another one05:02
wolfspraulthe way freenode works is that they have registered names05:03
wolfspraulthen you have some seconds to enter your password05:03
wolfspraulotherwise the system kicks you to an auto-name05:03
wolfspraulyou can try names, no problem05:03
wolfspraulthat should be safe05:03
wolfspraulcristian_casorra: seems to work05:05
cristian_casorrahow I can put a pass¿05:05
wolfspraulgoogle for 'register freenode account'05:08
wolfspraulbut with 'cristian_casorra' you should be safe05:08
wolfspraulso no need maybe05:08
wolfspraulcristian_casorra: alright, so I'll get back to you the moment I have the full new units here with me05:16
wolfsprauland you let me know in advance if you have any questions after reading the web etc.05:16
wolfspraulyou can always pop in here and ask, please don't hesitate05:17
wolfspraulright now it looks like I may have a unit for you in about 2 weeks05:17
wolfspraulI've been saying the same thing 2 weeks ago though :-)05:17
wolfspraulbut we're almost there...05:17
cristian_casorraok perfect, I'll check all the information in the web site, right now I can't because I have to make a protocoll for a dimmer controller... so... this weekend I'll try05:18
wolfspraulcristian_casorra: sounds good, we are here...05:28
aw0x85: just got d2/d3 dimly lit which cant reconfiguration after finished firstly rendering then I powered cycle.07:27
wolfspraulnot good07:27
aw0x85 was finished all functionality test by test tool07:28
wolfspraultry to power cycle again :-)07:28
awnow..it can reconfiguration again. ;-) after maybe 3 minutes. ;-)07:29
awlet me record firstly. :)07:29
xiangfuwolfspraul, Hi07:40
xiangfuimages CRC check in test image. : http://dpaste.com/581313/ now. I needs fix the BIOS check. all other look fine.07:48
wolfspraulxiangfu: do you mean Adam can try using that image?07:50
xiangfuno. let me first fix the BIOS problem. I think there is something wrong in my code.07:51
GitHub61[autotest-m1] xiangfu pushed 1 new commit to master: https://github.com/milkymist/autotest-m1/commit/ac80ffe862d5d0a9a991864c94469cdbe861006b08:23
GitHub61[autotest-m1/master] tests_images, first version not tested - Xiangfu Liu08:23
xiangfuthe commit log have changed to "tests_images, first version hardcode the CRC and length"08:34
xiangfuseems git push -f don't send one message here.08:34
GitHub195[autotest-m1] xiangfu force-pushed master from ac80ffe to 540f5fe: https://github.com/milkymist/autotest-m1/commits/master08:36
GitHub195[autotest-m1/master] tests_images, first version hardcode the CRC and length - Xiangfu Liu08:36
xiangfuoh. it just late08:36
xiangfuaw, just finish the CRC check in test image. new file is here: http://milkymist.org/updates/current/for-rc3/boot.540f5fe.bin08:39
awxiangfu, great.  tks. ;-)08:40
xiangfuaw, you are welcome ;)08:40
awxiangfu, you put it at last one item 'l', seems that i have to select it firstly. ;-)08:43
xiangfuoh. I can put it at first.08:54
aw0x60 with CRC check: http://downloads.qi-hardware.com/hardware/milkymist_one/production/rc3/test_results/60-results08:55
xiangfubut it's ok right? press 'a' or 'l'08:56
xiangfugreat. all passwd08:56
awit must be put at the first item i think. ;-)08:57
awbtw, e.g. when i got midi failed, i have to reload test image again. it's a little not convenience. but it's okay now.08:58
GitHub53[autotest-m1] xiangfu pushed 1 new commit to master: https://github.com/milkymist/autotest-m1/commit/4e5327357e05f617b959107173506aac117048d908:59
GitHub53[autotest-m1/master] move the images test to first - Xiangfu Liu08:59
xiangfuaw, sorry why it failed, you have to reload?08:59
awthat means if i knew a midi failed, i can only use step by step items to test. ;-)08:59
awsince the send/receive is too fast, so that i can't speed up to press a "e" to let it stop. then i have to reload image to use step by step item test.09:01
xiangfuaw,  new one: http://milkymist.org/updates/current/for-rc3/boot.4e53273.bin09:01
xiangfuaw, whenever you press 'e' , it will stop midi test.09:01
awxiangfu, alright, tks.09:02
awxiangfu, yes. but the message/log will become too many. ;-) forget about this currently. :)09:06
wpwrakaw: btw, how do you actually test MIDI and DMX ? do you have some sort of loopback cable or some test device ?10:40
awwpwrak, i made a loopback cable for midi(TX/RX), and bought a dmx cable connected to dmx(TX/RX)10:44
awwolfspraul, qi-hardware wiki now can't enter from here. :)10:45
wpwrakgreat. so it's a real test :)10:45
wolfspraulaw: I am upgrading10:49
awwpwrak, http://downloads.qi-hardware.com/people/adam/m1/pic/rc3_test_example.JPG10:56
awthe midi loopback cable I followed MIDI spec (15meters) to make a long distance (actually I used a 16 meters). ;-)10:57
wpwrakaw: cool setup !10:58
awwpwrak, sorry that I was trying to search qi wiki to find the link I followed though. ;-)10:58
wpwrakyeah, a lot of cable ;-)10:58
wpwrakaw: (link) worked fine, thanks :)10:59
awwpwrak, after our qi wiki restore, I'll find it and post to you. ;-)11:01
awwpwrak, here you are: http://www.midi.org/techspecs/electrispec.php11:03
awupdated : http://downloads.qi-hardware.com/hardware/milkymist_one/production/rc3/test_results/m1_rc3_test_report.ods11:06
aw1. from this morning, there's no d2/d3 dimly lit more immediately after reflashed by using shorter BEN usb cable. ;-)11:07
wpwrak(midi spec) ah, nice. thanks !11:08
aw2. but I got can't reconfiguration while counting rendering times: 0x63 and 0x85.  :(11:09
awwell...time to dinner11:10
awsee later11:10
wolfspraulroh: hi there, how are things?11:11
rohwolfspraul: good.11:13
rohfinished lasering yesterday evening and slept >12hours now11:13
wolfspraulyou mean all logos lasered?11:15
rohon 80 of the 160 lids11:15
rohonce centered per case in the middle. so if you do not like the logo, just put it to the bottom11:16
wolfspraulwhy would I want that? you don't like it?11:16
rohwolfspraul: well.. the customer could want that11:16
rohits also nice without the logo11:17
wpwrakroh: oh ... the grooves are on the outside or the inside ?11:17
wpwrakpheew :)11:17
rohi lasered it mirrored11:17
wpwrakvery good. i just had one of those "oh shit" moment ;-)11:18
wpwrakwith a horror vision of M1s loaded with dirt, etc. ;-)11:19
rohwpwrak: well.. imagine a clean, eben and reflective surface on a dj-booth.. its not only dirt which would stick there.. also all kinds of .. eh.. poweders11:20
wpwrakyeah, M1s could fetch a nice resale value ;-)11:22
rohwill do final qc now and hopefully find a proper package11:30
wolfspraulok, then shipping11:34
wolfspraulhow did we ship last time?11:34
wolfspraulzecke is actually flying over on Sunday (to Beijing), but I wouldn't want to load the customs VAT stuff upon him...11:34
rohi think last time it was dhl11:40
lekernelGenode birthday party in August in Dresden => http://pastebin.com/J9gNurfu11:50
lekernelI will probably go (and have space in the car from Berlin)11:50
rohwell.. too short after the camp12:00
lekernelmwalle, your components are on their way :)12:25
wpwrakroh: insufficient detox time ? ;-)12:29
rohwpwrak: simply plain madness.12:29
rohwpwrak: not because of detox but teardown etc.12:29
rohwpwrak: what most people dont know. there were festivals last weekend and there will be one next weekend just few km from the campsite with a lot of 'personal overlap'12:30
wpwrakaii ... sounds like fun :)12:30
rohso after the camp is after weeks of continous event/party for some12:31
wolfspraullekernel: I understand you want to sell some m1 at the camp?12:31
wolfspraulI think from now on you should always have a few full boxes with you that you can just sell :-) for cash or whatever payment form you like :-)12:31
rohHA.. a notch in the clouds... LESS rain. bbl. need to use that12:32
wolfspraulmaybe there is some paperwork to be considered though, if people want a receipt, VAT, etc. but as a 'small businessman' you may even be exempt from charging vat for a while, don't know. or you just ignore it.12:32
rohignoring is a very bad idea.12:33
wolfspraulsometimes I love China12:34
wolfspraulI will have re-adjustment problems when I get back12:34
wolfspraulall this artificial bloated stuff... :-)12:34
wpwrakwolfspraul: they'll put you in jail for doing what you can get away with in china. wouldn't that be ironic ? :)12:35
wolfspraulnot ironic at all12:36
wolfspraulthere would be far more Chinese in jail in Germany than are in jail in China12:36
wolfspraulthink about that for a while, he he :-)12:36
wpwrakexactly ! :)12:36
wolfspraulI am very aware of this, no worries.12:36
wolfspraulI'm still here after all. everything has a price, one way or the other.12:36
wolfspraullekernel: Adam has two interesting observations with boards 0x63 and 0x8513:19
wolfspraulhe is doing full 'rendering cycles' now on each board, where he boots to render, lets it render 30 seconds I believe, and then power cycle. 10 times.13:19
wolfspraulon 0x63, the notes say "1. while counting rendering times, can't configuration more from @ 6 th power on 2. then reflash again successfully 3. can't configuration more from @ 4 th power on"13:20
lekernelsounds like this horrible reset bug is still not gone13:20
wolfspraulon 0x85, they say "1. finish 1st rendering test then power cycle again, then can't reconfiguration(d2/d3 dimly lit); then can reconfiguration after maybe 3 minutes"13:20
lekernelif the rework correctly done on those boards?13:21
wolfspraulhe can later do measurements, first still go through all boards13:21
wolfspraulthe 0x85 is interesting. basically it works again after waiting a few minutes.13:21
lekernelor maybe the particular representatives of the family of pesky diodes used on those boards have higher capacitance/reverse current13:22
wolfspraulit is possible that one could actually demonstrate this on more boards, if you only do enough cycles? maybe something we can improve in the startup/bitstream/rtems boot procedure?13:22
wolfspraulsure that's also possible13:22
wolfspraulthat would be easy to fix :-)13:22
wolfspraulok, just a heads up now13:22
wolfspraulright now, when the smallest unexpected thing comes up, the board will not be put into 'available' state13:23
wolfspraullater a lot of this may disappear under closer investigation or with small hardware fixes13:23
wolfspraulAdam is using a test image with full CRC now, btw13:24
wolfsprauland the shorter USB cable13:24
wolfspraulaw: you saw that lekernel thinks later you should check on 0x63 and 0x85 that the fix2 rework was 100% correctly applied, with all values correct etc. maybe replace the diodes?13:27
awwolfspraul, hi sure. they are of course at the right but maybe have specifications shifted after my soldering. including capacitance/reverse probably, so I'll replace for new part when i back ti see them. thanks for reminds.13:31
wolfspraulgood luck with testing! :-)13:32
wolfspraulkeep telling us anything suspicious you see...13:32
awyes...but hope not more though! ;-)13:33
awjust updated on wiki, alright; i am out now. will continue testing tomorrow, night. :)14:48
mwallelarsc: systemcalls like clone need all registers16:20
larscdo they?16:38
larscthe idea was to put r1-r10 in the clobber list for scall16:43
larscso we don't have to save them16:43
larscso we basically say the state of r1-r10 is undefined after a syscall16:44
larscsince scall is most of the time at the end of an function and r1-r10 is caller saved the restored values of the registers aren't going to be used anyway16:45
mwallelarsc: yes a new process/thread is created by copying all registers16:47
mwalleso the kernel needs them16:48
mwalleand r11- are callee saved and handled by gcc anyways :)16:49
larscthats why we can't trash them in the syscall16:50
larscbut the new process doesn't care about the contents of r1-r10 either16:51
larscor shouldn't16:51
larscthe only problem would be that clone() needs fn and arg16:52
larscafter the syscall16:52
mwallelarsc: why should we need to save them (r11-) ? if a function use them it'll save them16:52
larscyou mean in the kernel?16:54
mwalleany compiled function :)16:55
mwalleinterrupt exceptions only save the caller saved registers16:55
larscuhm, yea, but we need to restore r11- when cloning a process17:07
larscbut i guess to properly support ptrace we need to store and restore them17:21
larscmwalle: another thing: should we add a "soc" node to the dts and put all on-soc components in there and all other outside of it?17:26
mwallelarsc: whats outside?17:42
larscmemory, buttons, leds17:47
larsci'm just not sure if that would work17:47
mwalleimho the soc is actually the whole dts, memory is outside of the csr bus, thats ok, dunno for the leds/buttons17:52
mwallei would not make it too complicated, imho there is no extra gain17:52
mwallethere is still the fml bridge missing, eg where do we put the code to flush it17:54
mwallelarsc: there is no way to build openwrt out of tree, is it?17:55
larscnot sure, never tried it17:57
larscthe devicetree documentation has an example where the soc has its own node17:58
larsc This node is used to represent a system-on-a-chip (SoC) and must be17:59
larsc  present if the processor is a SoC17:59
larscquoted from the documentation17:59
mwallefrom epapr?18:07
mwalleembededded power application platform requirements18:08
mwallefrom power.org18:08
mwallelinux kernel documentation?18:09
mwalleyeah but keep in mind that this is very power pc specific, eg they search their properties or root nodes by of_find_node_by_type(NULL, 'soc')18:17
larschm, ok18:21
kristianpaulif mm soc dont have a RTC or clock, how to you implemnted that in linux?18:29
larscwe don't18:30
kristianpaulah :)18:30
larscthe clock will be reset each boot18:30
kristianpaulbut busybox have  a ntp client right?18:35
kristianpaulahh, but there is no driver for minimac2...18:35
mwallekristianpaul: not yet :)18:36
mwallemaybe this weekend18:37
mwallebut i still dont have a working initramfs and/or kernel :(18:42
mwalleFreeing unused kernel mem: 100k freed18:42
mwalleinit: memory exhausted18:42
larscno memory?18:44
mwallei dont think so :)18:44
mwallei guess sth is wrong with my initramfs18:47
mwallei have to go18:49
mwallesee you tomorrow18:50
lekernelmh, with rtems cvs head, the flickernoise binary works in QEMU and freezes at boot on the board20:42
lekernelI get this warning from QEMU though: qemu-system-lm32: milkymist_sysctl: timer0: trying to write a value greater than the limit. Clipping.20:42
lekernelmaybe the context switch timer is broken20:43
lekernelno, broken UART driver: http://pastebin.com/Vu2uBV9j20:54
Alarmor can I get 'sys/ioct1.h'?21:18
Alarmi see tomorrow21:31
kristianpaulWhat's are trying to do btw?..21:32
lekernelah, it's just newlib mismatches ...21:58
tuxbrainany good CNC machine to drill pcb that works under linux?22:33
rohwe got a mill... whats your fileformat? gerber? gcode?23:07
kristianpaulwich mill then roh ?23:09
rohi got one here, which will be retrofitted with another high-speed spindle, the current one has 3000rpm. also i can access another mill which already has 2 heads with 20krpm23:10
rohor do you need a buying-recommendation?23:12
kristianpauli think23:12
tuxbrainbuying recomendation, but if you give that for free I will be happy :)23:12
rohtuxbrain: mills are highy application specific.. for pcbs only there are special mills.. but they are quite pricy23:13
kristianpaullike werner one?23:13
tuxbrainI see ones about 6k¬ but only works with windows23:14
rohbut these are dunno sw23:14
rohwe use emc2 (linuxcnc) for milling. which runs on rt linux23:15
tuxbrainyeah roh23:16
GitHub73[flickernoise] sbourdeauducq pushed 2 new commits to master: https://github.com/milkymist/flickernoise/compare/a29b914...25cafb223:17
GitHub73[flickernoise/master] Bump version number - Sebastien Bourdeauducq23:17
GitHub73[flickernoise/master] RTEMS CVS head support - Sebastien Bourdeauducq23:17
tuxbrainI see that beasty https://www.probotix.com/FireBall_Comet_cnc_router23:17
tuxbrainusing EMC223:17
tuxbrainall setup and ready to use for 3,5k$ but dont't know how good it will be for pcb23:19
kristianpauldont look very trusty for PCB tasks..23:21
rohlooks nice, but expensive23:22
rohwe got more mill for less23:22
kristianpaultell us you source :-)!23:22
tuxbrainroh can you sell them to me?23:22
rohwe got this one http://www.alibaba.com/product-free/218784744/Syil_BF20_Milling_Drilling_Machine.html?newId=218784744&pn=1&pt=10&t=5&cids=23:22
kristianpaulor i'm confuse and with got you mena build? :)23:23
rohand added emc2 ourself23:23
rohit has a parport interface box23:23
kristianpaultuxbrain: are you in the same steps of gadyetfactory?23:23
rohheh http://www.bf20.com/index.html23:24
rohwe got cnc spindles and steppers on all axis23:24
roh_big_ steppers. direct drive. no belt. and still handwheels23:24
tuxbrainkristianpaul: from who?23:26
rohbought it at http://www.monotek.de23:27
kristianpaultuxbrain: i mean manufacture own PCB and products etc...23:27
kristianpaullike dangerous protoypes, gatgetfactory...23:27
rohhttp://www.monotek.de/xt/product_info.php?info=p124_Bohr-Fraesmaschine-Opti-BF-20-Vario.html this one + cnc23:29
tuxbrainkristianpaul: I not pretend to sell the result, but we want to avoid the etching solution to prototipe boards23:31
kristianpauldo you have samples of PCB drilled with those CNC roh ?23:31
kristianpaultuxbrain: yeah,one step less is always good (etching)23:32
rohkristianpaul: not really. we wanted to add another faster spindle for pcbs23:32
roh20krpm. the higher the spindle speed the faster you can route23:32
tuxbrainmore than speed I'm concerning on acuracy, I dont mind it the solution takes 3-4h or more in end a 10x10cm board meanwile is acurate23:34
rohwell.. we wanted metal parts and pcbs are a addon23:34
tuxbrainjust the opposite to me :)23:34
kristianpaulbit price i guess23:34
tuxbrainyes and also only works under linux23:35
tuxbrainsorry windows23:35
tuxbrainand the soft inside the machine is a mistery23:35
kristianpauli think will be hardware some DIY/floss style CNC for making PCBs acuurately btu worth the source :-)23:36
kristianpaultuxbrain: may be try asking in #cam ?23:36
wpwrakpcb engraving is a little tricky. you need some system that maintains the exact altitude. not sure how well they do with the modern thin boards.23:36
kristianpaulwpwrak: your cnc is capable of that?23:37
wpwrakmy mill can drill holes and cut through the PCB. i made a few attempts at engraving, but they didn't go so well.23:37
wpwraki have some ides that might yield better precision, but i don't know yet whether they will work23:38
wpwrakthe idea is to scan the surface first, then use the altitude profile to compensate for board warping and such23:38
wpwrakbut again, that's probably only good for 1.6 mm boards. with 0.8 mm, it may have to cut too deep into the board to make sure it always removes all the copper everywhere23:39
wpwrakalso, the finer the traces, the thinner the bits and thin bits are more expensive and break more easily23:40
wpwrakso cost goes up with O(n^2)23:40
wpwraki made my peace with the chemical process :)23:41
kristianpaulsome chemicals and a good laser printer and transfer process could do wonders i guess :)23:42
wpwraktuxbrain: since you're planning to move into new premises anyway, make sure you have a balcony for the etching. then it's not so bad. you can store the acid outside so the fumes don't cause trouble. with HCl+H2O2, etching is quick and you have no liquid waste to worry about23:42
tuxbrainups this is MM channel :P I think I was on qi-hardware on23:42
wpwrakheh ;-) let's continue over there then :)23:43
rohwe got a automatic depth regulator for the 2nd spindle23:48
rohits a vertical sled and a shaft around the spindle which hold the right cutting depth precisely by sliding over the copper surface23:48
wpwrakwhat happens if the board fills with stuff after a bit of milling ?23:49
rohwpwrak: doesnt happen. there is air blown through that shaft23:49
rohwhich pushes stuff away from it via small holes to the side.23:50
rohyou config the depth by rotating the ptfe (teflon) shaft and then lock it in position by using a small worm screw23:50
wpwrakroh: looks nice. needs a relatively potent machine, though, with all the extra parts23:54
wpwrakroh: of course, yours has plenty of power :)23:55
rohtrue. no problem milling real steel23:55
rohwe even have a coolant pump23:55
--- Sat Jul 30 201100:00

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