#milkymist IRC log for Tuesday, 2012-02-14

wpwrakonly very roughly. as in overcurrent / not overcurrent00:05
wpwrakyou'd have to add current sensors and an ADC if you really need to know with better granularity00:06
wolfspraulcan or should we add circuitry to make this more precise?00:06
wpwrakseems an unusual feature00:07
wolfspraulyes. ok then, just came to my mind...00:07
wpwrakthe playful control freak in me likes it, of course :)00:07
wolfspraulnah, I can live without00:08
wolfspraulI don't understand the exact use case right now, besides having some fancy current monitor displaying how much each USB socket consumes...00:08
wpwrakyou could dim the USB leds accordingly ... ;-)00:09
wolfspraulindeed - the more power a usb socket draws, the brighter the light :-)00:09
wpwrakmaybe make them flicker when the current if very high, to suggest power is at the point of failing ;-)00:09
wolfspraulbut no, I don't need that still00:09
wolfsprauland I assume adding this would be quite a complication00:10
wpwrakyou'd need a bunch of sensor chips with resistors, plus some sort of adc (or mcu) to gather the data00:10
wolfspraulI just ran into it because I have a bunch of USB devices and hubs on one machine, and I lost overview over who needs how much power00:11
wolfsprauldo I need to power the hub? or not?00:11
wolfspraulhow much does the wifi dongle need? the speaker? the remote keyboard?00:11
wolfsprauland so on00:11
wpwrakyour multimeter is your friend :)00:11
wolfspraulsounds like we surely don't want that00:11
wolfspraulbut I'm basically guessing around, moving usb from hub to host machine, here there00:12
wolfspraulguessing who might need how much...00:12
wolfspraulfor example usb nand storage - probably rather little?00:12
wpwrakyou could make a little passive pcb with usb a upstream, usb a downstream, interrupted VBUS, and two connectors to put a multimeter in the middle00:13
wpwrakthen you could measure what exactly is going on00:13
wolfspraulok then the guessing is easier :-)00:14
wolfspraulbut surely such a small usb current measuring pcb sounds fun00:14
wolfspraulI could take a short extension cable and open some of the wires00:15
wpwrakyes, that works too. a bit finicky, though00:17
wpwrakwhile you're at it, you probably also want to bring out GND. that way, you can replace the PC with a lab power supply00:18
wpwrakor do things like supplying power from the lab power supply but still doing the signaling with the PC00:19
wolfspraulno no :-)00:19
wolfspraulI know when to stop00:19
wolfspraulI was *wondering* about power when I had a little trouble arranging a pile of USB devices00:19
wolfspraulbut simple trial and error solved the problem already00:19
wolfspraulenlightenment doesn't always win00:19
wolfspraulbtw, a small thing maybe for some, but I had to chew on this for a while00:20
wpwraksuch a setup was rather useful for certain experiments with gta02 and hxd8 ;-)00:20
wolfspraulI feel very confident now talking about Milkymist as a "SoC" project, SoC being the category00:20
wolfspraulbefore I was oscillating between CPU and SoC helplessly00:20
wolfspraulCPU is the more generally known term, but it's misleading and the main point of Milkymist doesn't come across00:20
wolfspraulso it's like Linux is a kernel, Apache a web server, Milkymist a SoC - that kind of analogy00:21
wolfspraulof course sebastien was talking about "leading free SoC" all along, nothing new for him...00:21
wpwrakyes, soc is accurate. sounds a tad smallish. but i guess we can't have everything00:23
wolfspraulsmallish?00:23
wpwrakwell, most people would associate "SoC" with relatively small devices00:24
wpwraksmartphones, pads, this kind of thing. already a laptop has "CPU" and "peripherals"00:24
wpwrakof course, that's gradually changing, too00:25
wolfspraulno I think it's SoC everywhere now, even Intel is being talked about making "SoCs" now00:25
wolfspraulanyway I feel confident about this now, only SoC from me now00:26
wpwrakok. so we're like the big guys then :)00:26
wolfspraul(and I just added Milkymist to the "List of SoC suppliers" in Wikipedia :-))00:27
wpwrakphew .. :)00:27
wolfspraulwe're between Maxim and MIPS00:28
wpwrakbtw, i think you missed this one in the backlog:00:31
wpwrakwolfspraul: the weird little corner of the jtag board does actually have a use: it helps you to align the board with the connectors. otherwise, you would have to "feel" for correct alignment. sort of fumblish.00:31
wolfspraulsaw it00:31
wolfspraulgood point :-)00:31
wpwrak(just noticed that when re-seating the jtag board i had plucked out to look at the memory card connector when you asked about ubb)00:32
GitHub88[flickernoise] wpwrak pushed 1 new commit to direct-midi: http://git.io/Mby8Gw00:33
GitHub88[flickernoise/direct-midi] MIDI documentation (still WIP, on-going) - Werner Almesberger00:33
wpwrakwriting documentation helps to understand those use cases ;-)00:34
kristianpaulxiangfu: hello01:42
kristianpaul(replying to yday Hi ;-))01:42
xiangfukristianpaul, Hi. good morning01:46
kristianpaulhey :)01:46
xiangfuwant let you know I finish a simple payload.py for your m1 using mining core01:48
xiangfu: https://github.com/ngzhang/Icarus/blob/master/miner_software/scripts/payload.py01:48
kristianpauloh great01:48
xiangfukristianpaul, also: http://en.qi-hardware.com/wiki/Icarus#Verify_FPGA_core01:48
xiangfuusing linux command. :)01:49
kristianpaullet me try that last one01:49
xiangfuif it not working. try payload.py. I tried the linux commands. somtimes it not working. don't know why. but payload.py always working.01:50
kristianpaulno work.. trying the python script..01:50
xiangfuit should reply in 1 second01:53
kristianpaulerghh no, damn i'll re-sinthesize this and if i can re-enforce timings..01:53
Action: kristianpaul my fault01:53
kristianpauloh wait, xiangfu https://gist.github.com/182245401:58
kristianpaulbut i cant reproduce...01:58
qi-botThe Firmware build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-20120214-0111/01:59
xiangfuyes. you got correct result.01:59
kristianpaulbut just onces...01:59
xiangfuyou mean only work once?01:59
kristianpauldamn01:59
kristianpaulyeap..01:59
kristianpauloh GOD01:59
kristianpaulit took AGES01:59
kristianpaulwait a second i can reproduce it again now with time ;)02:00
kristianpauldamn i tought was timing on my side but this seems to be slow fpgaminer on my side ;)02:00
kristianpaulok may be i can push to a repo now ;)02:00
kristianpauljust in case xiangfu what other hash i can try :) ?02:01
kristianpaulhmm now i took longer..02:04
kristianpaulshould i reset the fpga...02:04
kristianpaulxiangfu: https://gist.github.com/182249802:08
kristianpauldamn dont know what it works and dont ...02:16
Action: kristianpaul dont trust too much the uart serializer..02:16
kristianpaulthere is no reset logic..02:17
kristianpauldamn i needs a mcu..02:18
kristianpaulmay be mico8 :)02:19
xiangfuany  test needs time to find out:)02:20
xiangfu38.164s02:20
kristianpaullol ;)02:24
kristianpaulis too slow..02:24
kristianpauloh well.02:24
xiangfuas least it working :)02:27
wpwrakwhat dollar equivalent does it produce in those 38 seconds ?02:40
Action: kristianpaul will not answer that02:41
xiangfu:D02:41
xiangfuwpwrak, oh. about the projector. I make is wrong. it's not 4mX5m. it about 2.6Mx3M. my home wall is not that big. sorry02:42
wpwrakstill pretty large. and the image looks good ? bright and with contrast ?02:42
xiangfuwpwrak, I can just say it fine. have to with 100% dark.02:43
wpwrakdoes it let you adjust the angle ? or do you have to move the projector to change the size ?02:44
xiangfuhave to move the projector to changed the size.02:48
xiangfuno adjust on angle.02:48
wpwrakmmh, inconvenient02:49
cladamwa(audio lineout plugin detection) added R233 100K @ J2.3; http://downloads.qi-hardware.com/people/adam/m1/tmp/m1r4/Audio_20120208.pdf02:56
cladamwasorry, wrong link above; should be http://downloads.qi-hardware.com/people/adam/m1/tmp/m1r4/Audio_20120214.pdf02:57
cladamwawpwrak, i added R233 100K for detecting LINEout plug-in detection. you could check and view, later we ask joreg. ;-)02:58
cladamwafor audio line-in, current no idea. still thinking... ;-D02:59
wpwrakhmm, might work03:12
wpwrakdon't forget to tell joerg there are no clamping diodes in the fpga :)03:12
wpwrakactually, there are, towards gnd. but it should be disconnected then anyway. yeah, might work03:13
wpwraknot sure if lineout_det needs tvs, though03:14
cladamwajoerg suggested me that we can also add tvs on lineout_det in #qi-hardware. let me check. ;-)03:21
wpwrakheh "_03:23
cladamwahttp://en.qi-hardware.com/irclogs/qi-hardware_2012-02-08.log.html#t09:3203:24
wpwraknot sure if it's that what he meant03:25
cladamwaso with the same method, I'm thinking using same theory for audio line-in.03:25
cladamwayeah...but it's okay. we can draw idea fristly then ask joerg for help then, how do you think? ;-)03:26
wpwrak(line in) there goes the nice ground :)03:26
wpwrakyup. leave it to him to tear those ideas apart :)03:26
cladamwawpwrak, yes, although current there goes to ground, but I'll also let there to be reached liek 'low' sc level 'normally' (i.e. not plugg-in yet).03:28
cladamwawpwrak, yes, so i'll edit audio sche page again then for help . ;-)03:28
kristianpaulxiangfu: n8, thanks for the script now i have a basis to debug my niner :)03:30
cladamwawpwrak, btw, i did really apply yet joerg's idea on those DC blocking 2 pins header for un-used audio input path, how do you think this?03:31
xiangfukristianpaul, n803:33
wpwrakmmh, not sure which of the many audio changes you mean03:34
cladamwawpwrak, since i don't know if we need to use more route space for TP47~TP51( which were I added ), but original idea from you was to leave solder-pad only.03:35
wpwrakah, MIC2 and friends03:37
cladamwayes, http://lists.milkymist.org/pipermail/devel-milkymist.org/2012-January/002583.html03:38
wpwraki'd just put the caps close to the chip and NC them. no antenna = no RF trouble, right ? :)03:39
cladamwayou wanted to keep those un-used footprint for DC blocking if one can use them for playing them. but joreg suggested put them in ground when no use.03:39
cladamwayes, your idea thoughts not bad and reasonable, but from Joreg's suggestions that he wanted us to reduce audio input noise as possible, so he said that we may use 2 pins headers to short to ground those un-used caps.03:41
wpwraki think that may be overkill03:42
wpwrakof the traces between chip and the NC caps are short, not much can crawl in anyway03:42
cladamwafrom my simple idea is quite same like you and even just put footprint only and NO test points, and DNP them.03:43
wpwrakand i'd NC the caps :)03:43
wpwrakyup03:43
wpwrakNC == DNP :)03:43
cladamwaokay, so let's not add Test points and DNP them. done?03:43
wpwraksounds great to me :)03:44
cladamwaokay, done. ;-) then I edit the plugg-in and detection idea and ask in list for joerg again. ;-)03:44
cladamwabtw, to have detection idea, can you see your rc2 sch now?03:45
wpwrakrc2 ?03:46
cladamwayes03:46
cladamwaalthough we replaced R23/R24 (6.8K ohm) with varistor V27/V28 now,03:46
cladamwacorrect?03:47
wpwrakM1rc2 ? where are the schematics of that ? and what is there about detection ? and where ?03:47
cladamwayou don't have rc2 sch? checking...03:47
cladamwahttp://www.milkymist.org/mmone/old/rc2_schematics.pdf03:49
cladamwaaudio sch in rc2 is almost quite the same as rc3, but I'm trying to describe my idea on line-in detection to you now. ;-)03:50
wpwrakthanks03:50
cladamwasee page 1 for audio03:50
cladamwathere's R23/R24, now M1R4 we replaced them as V27/V28, so....03:51
cladamwaso my plans for line-in solutions:03:51
wpwrakyup. looks like rc3 :)03:52
cladamwa1. use 100K ohm to pull up on DC3.3V then connects to J1.3pin.03:52
cladamwayes03:52
cladamwa2. J1. 5 goes to AUDIO_GND03:53
wpwrakah, 50% grounded:)03:53
cladamwa3. place back those R23/R24 6.8K ohm back like in rc2 but still keep V27/V2803:54
wpwraki knew that would come :)03:54
cladamwa4. connect J1.3 to fpga un-used pin03:55
wpwrakbut yes, seems to be the only choice. unless the codec itself contains anything that can help03:55
cladamwa5. so Normal = Low(0.395V), Plug-in = high(3.3V)03:55
cladamwayeah03:55
cladamwahow do you think this method?03:56
wpwraksounds reasonable03:56
cladamwai need to check fpga's Vin low threathhood value.03:56
cladamwanow checking...03:57
qi-botThe Firmware build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-20120214-0309/03:57
cladamwawpwrak, page 10 of http://www.xilinx.com/support/documentation/data_sheets/ds162.pdf04:01
wpwrakyeah, easy :)04:03
cladamwaso sound a 0.395V(low) is less than VIL @LVTTL and LVCMOS33. ;-)04:05
cladamwaso should be okay, right?04:05
cladamwathe column on VIL (V, max)04:06
wpwrakyup04:06
cladamwawait, why did you say 50% grounded ? is it problem?04:07
cladamwayou meant one of (J1.5 and J1.3) to audio ground @ 50 %, is not okay?04:08
wpwrakyou're undoing part of the paranoia grounding you added on joerg's advice. but i don't know if it matters04:08
wpwrakprobably not04:08
wpwrakwe're not exactly doing hi-fi recordomg amyway04:09
cladamwaoah..yes. need to undo...04:09
cladamwayeah..so let me edit those idea above in audio sch page then ask for detection in list again. hehe...okay?04:10
cladamwayeah...we 're not doing hi-fi device. should be okay? or don't know?04:11
wpwrakour main use of audio in is currently called "bass", "mid", or "treb". and it's off the documented range by some 15 dB, maybe more. so ... :)04:15
cladamwaaha...i didn't see s/w setting too much, yeah, if like only three catagories, phew~ yes, overkill ... :)04:17
cladamwaokay...later I'll send idea in list again. hope joreg don't think we're crazy though. ;-)04:18
wpwrak(3 bands) that shouldn't stop us from doing better, of course. but that's where we're at for now :)04:19
cladamwa;-D04:20
lekernelsometimes there are happy surprises with ISE... the 13.4 version now correctly infers that wide block RAM with byte-level WE's10:23
whitequarkthey did it accidentally ;)10:29
Fallenouahah10:32
kristianpaulok lets move to 13.4 then..10:46
kristianpaulwtf whty i'm getting arsicgt console instead of webpack !10:51
kristianpaularcsight*10:51
kristianpaulhere we go again... *sigh*10:52
kristianpauli give for now, even with other webwrosers it dint work..10:54
GitHub90[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/0c214b484ec4b342e6338fc6b66d655b3c06f06a12:18
GitHub90[migen/master] Use double quotes for all strings - Sebastien Bourdeauducq12:18
GitHub103[milkymist-ng] sbourdeauducq pushed 1 new commit to master: http://git.io/1g9JBg12:20
GitHub103[milkymist-ng/master] Use double quotes for all strings - Sebastien Bourdeauducq12:20
GitHub148[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/46b1f74e98dcdf179c74845faafda5fca7bee57213:06
GitHub148[migen/master] bus/asmibus/hub: forward data and tag_call - Sebastien Bourdeauducq13:06
GitHub22[flickernoise] wpwrak pushed 7 new commits to direct-midi: http://git.io/8B-erw14:20
GitHub22[flickernoise/direct-midi] stimuli: refuse to bind the same control twice - Werner Almesberger14:20
GitHub22[flickernoise/direct-midi] compiler: set SF_ASSIGNED for control variables to prevent -Wundefined warning - Werner Almesberger14:20
GitHub22[flickernoise/direct-midi] compiler: allow multiple stimuli to set the same variable - Werner Almesberger14:20
GitHub71[milkymist-ng] sbourdeauducq pushed 1 new commit to master: http://git.io/JVgA3w14:48
GitHub71[milkymist-ng/master] README - Sebastien Bourdeauducq14:48
wpwrakwhoa. done ? :)14:54
GitHub189[milkymist-ng] sbourdeauducq pushed 1 new commit to master: http://git.io/WMOw_w14:58
GitHub189[milkymist-ng/master] s6ddrphy: prepare quilt - Sebastien Bourdeauducq14:58
lekernelno, not yet14:59
lekernelbut I have a very precise idea of what needs to be done. now it's just coding and debugging ...14:59
lekernelphew, active low resets... I never got it15:02
Fallenoulekernel: are store words working correctly for you using migen & milkymist-ng ?15:02
Fallenouon sram15:03
lekernelit's something competely useless in a FPGA and, in ASICs, optimizing the reset polarity should be done by the tools, not the coder15:03
lekernelFallenou: yes15:03
Fallenouok15:03
lekernelmirko: you're at in-berlin?15:48
mirkolekernel: kinda.. hosting stuff there..15:55
GitHub76[flickernoise] wpwrak pushed 1 new commit to direct-midi: http://git.io/s9hqng18:51
GitHub76[flickernoise/direct-midi] More work on the MIDI documentation - Werner Almesberger18:51
larsc_hm, what's the result of expressions like 'assign a = ~a' and more interestingly 'assign a = b | ~a'22:38
wpwraklocal overheating ?23:01
larsc_my generic flow control logic model resolves to 'ack_o = ack_i & ~stb_o; stb_o = stb_i & ack_o;' for combinatorial actors23:15
larsc_ this could either mean my model is still to generic, it is wrong, or there are some special rules on how to handle self contradicting logic23:25
larsc_and if you build the truth tables it show that for the cases where the logic is coherent it is equivalent to ack_o = ack_i;stb_o = stb_i23:29
wpwrakhmm :)23:33
wpwraki'm actually a bit surprised that you can get away with assign a = f(b); assign b = f(a); i.e., a cyclic dependency23:34
larsc_i'd expect it to settle to a fixed point. if it has one23:36
lekernelyeah, combinatorial loops can be messy23:43
lekernelyou can actually build sequential systems with them :)23:43
wpwrakdid you check the logs for suspicious warnings ? like "confusing logic set to 0" or such ? i noticed that xst rather likes to do such things23:43
lekernelthere's even a generic method to build any (asynchronous) FSMs using combinatorial logic only23:44
lekernel(which I forgot, because it's of little use in FPGAs)23:45
lekernellarsc_: I think you can resolve the problems without combinatorial loops... afaik all you need is a way to track partial acks when you have several outputs23:47
lekernelwhich you can do with normal registers23:47
lekernelFallenou: X's may or may not be normal23:52
--- Wed Feb 15 201200:00

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