#milkymist IRC log for Monday, 2011-07-18

kristianpaullekernel: mm1 bios is aware of endianess aligment right?00:53
awmorning ;-)01:02
kristianpaulhey there01:04
awrc3 keeps rendering until this morning 9:00, so totally is 8 hours passed. ;-) It's stop once my laptop's usb is ON. (this sounds a bug still there heard before though)01:04
awif I am wrong, correct me. ;-)01:05
kristianpaulah,you mean you let the jtag cable connected to your laptop, and once the laptop turned on, the  m1 it..01:06
kristianpaulbut stop mean no crash? just render stop?01:06
awkristianpaul, yes, just rendering stop. is it to be acted that?01:07
awif I remembered correctly, Jon had have this discovery before. Hope i am not the first one. ;-)01:09
kristianpauldunno, sorry had not test m1 in redering too much in my side, but my seems a easy bug to be solve later01:09
kristianpaulah,well. so you not alone, thats good01:09
awkristianpaul, no problem. I wrote firstly, later may get an answer. ;-)01:09
kristianpaulbut is good read from you the 8 hrs render went okay01:10
awyes, i am going to keep do test of boot-to-rendering for 70 times. Before this I need to discuss with factory. ;-)01:11
awxiangfu, morning, i have a question.01:50
xiangfuaw, morning01:50
awhave you known that rendering will be stop if my jtag cable connected to my laptop then once I turn on laptop?01:51
awis it a well-known bug already?01:51
xiangfuaw, hmm.. it is not recommand do that. you better connect hose table when m1 shutdown.01:54
kristianpaulis that a yes? :)01:54
xiangfuit's not design that connect while poweron. so the error is random01:54
xiangfuyes. :)01:54
awwhat's "hose table"?01:55
xiangfuI connect them while m1 poweron. make m1 freeze.01:55
xiangfuaw, this cable , I mean01:55
awxiangfu, hmm..so this is well-known for few people already?01:57
awi see, tks. so will this put into https://github.com/milkymist/bugs/issues ?01:58
awor no needed as u know? ;-)01:58
awor haven't discuss too much though. ;-)01:59
kristianpauli must said i was unknow of this, well, may be because rarelly i connect m1 to laptop02:01
xiangfuaw, we can write down this message to jtag board wiki page. no needs to create a issue I think. it's not a bug, it just not design to do that.02:02
awkristianpaul, yes, me either.02:02
xiangfuaw, write down to wiki page, with highlight.02:02
awxiangfu, i see.02:02
kristianpaulso the theory is that mm1 serial port if perturbed with ramdom data, bad thing happens? or may be jtag as well?.. well could make sense..02:11
xiangfukristianpaul, aw http://en.qi-hardware.com/mmlogs/milkymist_2011-06-21.log.html#t07:0802:16
kristianpaulwell, gpsd is not crap, but make sense now02:17
awhi NO, my status is not you met before. ;-) I didn't plug or replug my jtag board.02:22
xiangfuaw, you are doing that while m1 poweron right? that kind of same :)02:23
awmy condition was: 1. m1's jtag board connected to my  laptop and let m1 to keep rendering 2. then i turned off my laptop. 3, I turn on my laptop again then m1 stops rendering.02:24
awso you meant my case is the same behavior as you just posted that irc link?02:25
xiangfuI think it's kind of same. it only stop rendering not freeze?02:26
awyes, just stop rendering, but I forgot to check if freeze, then I power off m1. ;-)02:27
xiangfuI think we don't need look into detail. just do not connect  JTAG while m1 poweron. and do not reboot your pc while connect a poweron m1 :)02:29
kristianpaulbut power on/off laptop do same with jtag pod, and at the end generated magic garbage ;)02:29
awxiangfu, hi sorry , i correct my last sentences: screen freeze! not stop rendering.02:29
awhmm..got it. stop looking at this.02:30
xiangfuyes. just as kristianpaul said. "at the end generated magic garbage ;)"02:30
awokay..confused you..done. ;-)02:32
kristianpaulalso i think altought jtag pod is included with mm1, is not acesible easilly, so this is not going to happen at all02:32
xiangfukristianpaul, yes.02:34
awjust sent reworks info to factory about http://downloads.qi-hardware.com/people/werner/m1/rc3/fix2.pdf , and also talked with supervisor; they don't want to do adding a wire job and one new diode without a solder paste. :(02:42
awonly can accept a rework with easy adding parts, changing parts or taking apart in solder paste existed. thus R60(remove), R30(change), R157(change), C238(add)02:44
kristianpaulyeah, less troubles, imagine just trouhwing that wire from one side to the board to another ;)02:45
kristianpaulwell for they..02:45
kristianpaulbtw, you have pics from the rework in both rc2 and rc3 boards?02:46
awwill be noticed soon if factory wants to do easy jobs after evaluating current internal loads.02:46
kristianpaullarsc: new lm32 branch, nice !!02:46
awkristianpaul, second. I'll take pictures and give another fix2 with parts reference. before this , i need to finish some else and remaining 70 times tests.02:47
kristianpaulah, yes i forgot 70 times tests..02:48
larsckristianpaul: hm, that was not on prpose02:51
larscnow it's correct02:52
kristianpaullarsc: so, what is you plan?02:54
larscworld domination ;)03:00
larsckristianpaul: my plan regarding to what?03:00
kristianpaullarsc: he,to the branch you created on tmatsuya/linux repo03:02
kristianpaulahh, was deleted..03:03
kristianpaulokay nothing else to ask you then :)03:03
larsci wanted to push to master, just forgot to mention it on the cmdline03:03
awrc3 power up -> boot -> rendering 15 secs -> power off: 35 times03:05
kristianpaulbtw, now i can ask you quickly, there is a place in wich i can find a uclinux updated with all the patches you and mwalle were working on?03:05
larsckristianpaul: mwalle's patches are in the milkymist wiki03:07
kristianpaullarsc: but those patches are agains wich linux repo?03:09
kristianpaulalso do i need mwalle ulibc isnt?03:09
kristianpaulsorry, i never compiled or ran linux on mm1, just asking to be aware  of current status03:09
larscall the linux patches are in the repo at github03:11
larscif you want to build your own userland you'll need mwalles uclibc patches03:11
kristianpaulyou mean, are already aplied in repo at github?03:12
aw40 times03:12
kristianpaulah, okay just uclibc patches, not linux specific patches, i think i undertand that03:12
aw45 times03:18
wpwrak_(jtag while laptop power on) probably it froze. in my experience, the ft2232 likes to act up. not much you can do.03:19
wpwrak_(aw) the scope traces would be useful to have. we'll need them to confirm the value of C238.03:22
awwpwrak_, yes, i think so.03:23
aw50 times03:24
awwpwrak_, so program_b vs RP#, and program_b vs init_b, two scopes, else?03:24
wpwrak_aw: the test would be: connect scope (probes set ti 1:10) to TP36 (PROGRAM_B_2) and TP37 (FLASH_RESET_N). set to 100 ns/div, falling edge trigger on FLASH_RESET_N, trigger level ~1.65 V.03:25
awso my prober get touched again with program_b , not sure this influence to add more equivalent capacitors on C238.03:26
wpwrak_aw: then power up. wait until FLASH_RESET_N is at ~3.3 V. then touch FLASH_RESET_N with a resistor 47-100 Ohm to ground. that will pull FLASH_RESET_N down.03:27
wpwrak_aw: of course it will change the capacitance a little :)03:27
awalright, yesterday i tired to use 1:10 to get but not success, I'll try.03:27
wpwrak_aw: i just need to know how much contamination is there from FLASH_RESET_N to PROGRAM_B_2. i can then make an estimate of how much it would be without the scope.03:28
awum..yes, then you know the difference it is.03:30
awgot it.03:30
awwpwrak_, this means if C238 value changed, i should do 100 times again I think, no need?03:32
aw55 times03:32
wpwrak_i think we'd need to do more testing if we have to change C238, because it could cause other, yet unknown problems03:33
awso i think that maybe i just stop tests? and probe first to you?03:34
wpwrak_i.e., there's a lower and an upper bound for C238. so far, we're mainly looking for the lower bound.03:34
wpwrak_it may make sense to change C238 to, say, 1 nF or such at some point in time, to see if we're well clear of any upper bound. but let's first confirm the lower bound.03:35
wpwrak_(tests vs. scope) dunno. you're almost finished with the tests anyway :)03:35
awwell...so now i think i stop tests, seems this not make sense to keep tests.03:35
wpwrak_hehe. they're boring anyway ;-)03:36
awat least I have 55 times already as a base. i stop. the design for checking lower bound is much than collecting an unknown data.03:37
awwaht time it is your side?03:37
wpwrak_about 00:37 (am)03:37
awgive me 10 ~ 20 minutes.03:38
wpwrak_just had a nap after dinner :)03:38
awif u feel asleep, go ahead . I'll upload though. ;-)03:39
wpwrak_yeah, i'll be up for a little while. but not the whole night :)03:40
wpwrak_well, maybe, but not online. busy reading a book :)03:40
kristianpaulyou need buy ebooks more often :)03:41
wpwrak_kristianpaul: i actually have found an electronic copy :) but it's still much nicer to read on paper03:42
kristianpaulpaper is nice i must confess i still printing stuff to read later, also using it right now to take notes,may be epaper will change this, but not for  me soon03:44
awwpwrak_, now m1 is sitting in standby mode (thus the stage after reconfiguration) to ready capture falling edge.03:44
awwpwrak_, i remembered those two test points are normal high03:45
wpwrak_aw: they should be, yes03:45
wpwrak_after you pull FLASH_RESET_N towards ground, the M1 will be in an unknown state. maybe it'll crash/freeze. i don03:49
wpwrak_'t know it well enough to predict.03:49
awwpwrak_, the falling edge is captured @2.5 ms/div not 100 ns/div, be noticed03:55
awis this that your supposedly thoughts? hmm..let me confirmed: two probers set X1, oscilliscope also set X103:58
wpwrak_that seems too slow. the contamination should recover within a few us03:58
wpwrak_probes set to X1003:59
awif I set my probers at X10, I can't trigger and capture.03:59
awset again.03:59
wpwrak_eh ? why can't you trigger with X10 ? is the scope broken ? scope should be set to X10 too, of course04:00
wpwrak_i think you also need to set the trigger after changing X1 -> X10. it may become 16.5 V or such after the change, not 1.65 V04:02
kristianpaulyes, well, with the scope i have here is that way wpwrak_ points04:03
kristianpaulearly morning roh04:03
rohwtf? your scopes dont calculate the displayed trigger themselves?04:03
kristianpaulyes, but that s tricky04:03
kristianpaulfor this situation, at leat04:04
rohno. the scope either knows (by contacts on the probe plug) which mode you set (x1/x10) or probe you use. so it can just display it correctly.04:04
awhey..sorry that I got it. ;-)04:04
rohif it doesnt know it, you can set it in the input section of the scope usually04:05
awconfused you again. ;-)04:05
wpwrak_kristianpaul: (scope) you mean that the trigger changes when you change X1 <-> X10 ?04:05
kristianpaulah, well, one i have dont, i have to set mode manually04:05
kristianpaulno i dont wpwrak_04:05
kristianpauli remenber04:05
awwpwrak_, but 10 ms/div got it04:05
wpwrak_roh: they do. but they may not adjust it when you change the probe. so your previous x V may become 10x V, or 0.1x V. but when you set to 1.65 V again, it'll be 1.65 V04:06
wpwrak_aw: 10 ms/div is meaningless ;-)04:06
rohwpwrak_: thats why you need to tell your scope which probe you use (if it cant detect it)04:06
wpwrak_aw: maybe you have too much filtering on the trigger04:06
wpwrak_roh: correct :)04:06
rohmaybe there is some 'denoise' or 'low pass' enabled on the input04:07
wpwrak_roh: adam has his probes set to X1. i have no idea why ;-) i'm telling him that this measurement is better done with X10. that's all04:07
awwpwrak_, no, I scales both channel again, also my two probers are set @ X10 now. ;-)04:07
rohyeah. else they will influence the measured device much more04:07
wpwrak_roh: (mess up the measurement) yup04:08
wpwrak_aw: your scope shouldn't have any problem at all capturing this at 100 ns/div. i have a TDS1002. it has even less trigger problems than the rigol i normally use ;-)04:09
kristianpaulaw: are you sampling at full sampling rate?04:10
kristianpauli had problems with triggering because of that sometimes..04:11
awmoments. let me upload picture firstly04:11
wpwrak_kristianpaul: i'm not sure if the trigger should even depend on the sampling rate ...04:12
kristianpaulyeah, neither me, but is what i tell from my short scope oepration experience04:12
kristianpaulat least if you want a correct visualization i think04:13
kristianpaulbut i may wrong04:13
wpwrak_kristianpaul: is your acquisition set to "normal" or "peak" ?04:13
kristianpaulusually normal, but i dint noticed any different with peak to be honest04:14
kristianpaulnot yet, i havent catched pesly signals a Adam now :)04:15
wpwrak_that doesn't make sense at all ;-)04:16
awwpwrak_, i set 'Normal', the duration of falling is longer that you expected. ;-)04:16
awkeep calm a bit...the sharp curve is done by reset iC04:17
wpwrak_the FLASH_RESET_N you touch with 47-100 Ohm should drop sharply04:17
awlet's check its reset IC's behaviour, then you won't be surprised at all.04:17
wpwrak_ah, wait ... how did you trigger all this ?04:18
wpwrak_you have to have the M1 powered up, then force FLASH_RESET_N down04:18
wpwrak_since i don't know how to do this via software and lekernel isn't around, i suggest to use a 47-100 Ohm resistor connected to ground04:18
awhmm...you wanted me to manually pull low with touching with 47 - 100 Ohm???04:19
wpwrak_if you don't have one, a direct connection to ground would do as well. but via a resistor is a little safer04:19
wpwrak_yup :)04:19
awhmm..i see now.04:19
awbut can you realize what i said that FLASH_RESET_N normal behavior though. ;-)04:20
wpwrak_you can set the trigger to "SINGLE". the M1 may reset or do other weird things afterwards, so it may trigger again. but that would all be meaningless.04:21
awwell...i need to find a 100 O ohm now. :(04:21
awyes, i used "SINGLE".04:22
wpwrak_(normal behaviour) hmm. not quite sure what this actually shows ;-)04:22
wpwrak_i see that the two are about 200 mV apart. so the diode's Vf isn't too bad. good to known.04:22
awyes, the diode's low is good though04:23
wpwrak_yup. diode works as a diode should :)04:24
kristianpauldetectflash 0 in urjtag may be a software alternative, but i dont think it will pull down that flash reset signal for ever04:24
wpwrak_you have a little glitch on FLASH_RESET_N. cute :) doesn't matter here, though.04:24
wpwrak_if you set acquisition to "Peak Detect" (instead of "Sample"), you'd see those glitches more clearly04:25
wpwrak_the TDS1xxx series has very good peak detect. it's worth using it. on my rigol, it's much less useful. but then, it has about 400-800 times the memory :)04:26
awpage 10 of reset ic shows when our 3V3 reachs "Release Voltage" so that we can realize this normal falling edge it is. ;-)04:26
wpwrak_aw: i mean the glitch around -8.0 ms, on CH204:27
awand the slow slope may somewhere discharges.04:27
kristianpaulok, so whare are you going to measure init_b ?? or is not needed anymore?04:28
wpwrak_the glitch may actually exist on both channels. we just don't see it on CH1 due to the slow setting and not using peak detect04:28
wpwrak_kristianpaul: we're after FLASH_RESET_N -> PROGRAM_B_2 contamination04:28
wpwrak_kristianpaul: INIT_B is still used, but not nearly as critical as PROGRAM_B_204:29
kristianpaulah, okay, i tought PROGRAM_B_2 wasnt aware of contamination, yes thats important04:29
wpwrak_kristianpaul: anything that would be even a small problem on INIT_B would be apocalyptic on PROGRAM_B_2 ;-)04:29
awour reset ic's Release Voltage is rough at 2.63V which we can see that waveform acts alike.04:29
wpwrak_aw: (reset ic) thinking of it, it seems a little strange that it doesn't pull more. but then, i don't quite know what your setting was.04:31
awi set ACQUIRE at "Sample" not "Peak Detect", you wanted me set to "Peak Detect" again?04:33
wpwrak_aw: ah yes, you may want to set the scope to "peak detect" in general. it helps to find things :) and you don't seem to have much trouble with interference, so it won't add much noise04:33
awcapture again now.wait...04:34
awto see if difference there. ;-)04:34
wpwrak_aw: "peak detect" should work very well on your scope. and, given that you have almost no memory, it helps a lot with finding glitches and such04:35
wpwrak_kristianpaul: i think yours has a bit more memory, but also not a lot (maybe 10x). so you may want to experiment with "peak detect", too04:35
wpwrak_aw: the difference should be quite significant ;-)04:36
awno, seems it's not. wait..uploading04:37
kristianpaulis same?04:37
wpwrak_kristianpaul: can't be :) unless he's not doing what i have in mind04:38
kristianpaulaw: your scope have a single freq button or mode ?04:38
awkristianpaul, single mode04:38
kristianpaulha werner !04:39
awPeak Detect this time. ;-) without manually touching 100 ohm.04:39
wpwrak_aw: ;-) peak detect is nice. now we know that the glitch doesn't go deep04:40
kristianpauli had similar experience when testing that peak mode on the scope i have..04:40
wpwrak_aw: BUT .. what the heck is this showing ? ;-)04:40
wpwrak_aw: what you should do is this:04:40
awwpwrak_, go on: listening...04:40
wpwrak_1) power the M1 up. let it boot, etc. until it's stable04:40
wpwrak_2) then arm your scope04:40
awhmm..i didn't let it boot04:41
wpwrak_3) then pull FLASH_RESET_N low04:41
wpwrak_i think you have to let it boot or it will pull FLASH_RESET_N something like 0.5 s after powering up. you could also measure this, but it may be difficult to catch04:42
awwpwrak_, i quite no sure if I manually pull FLASH_RESET_N low, what will be happened? since I only have rc3 one now...my last rc2 is temporarily sleeps now. ;-)04:42
wpwrak_i don't know what will happen if you manually pull FLASH_RESET_N. the m1 may simply crash/freeze. or maybe it doesn't even notice. doesn't matter. all we need to see is how PROGRAM_B_2 reacts04:43
awyes,...should be rough 0.2s~0.5s I think thus is reset IC's delay time + else after powering up.04:44
wpwrak_the transition on PROGRAM_B_2 will be short. a few microseconds. and the part that's actually interesting is < 1 us.04:44
awbad...so now i still can't answer to let you know a lower bound evaluate. :(04:45
wpwrak_(the rest of the PROGRAM_B_2 transition will be C238 charging, which should take about 1-2 us. not very interesting to watch.)04:45
wpwrak_aw: naw, you really need to force FLASH_RESET_N down :)04:46
wpwrak_if you want to, you can also try to find the point where the firmware forces FLASH_RESET_N down. but that's more work.04:46
wpwrak_you would do this with a pulse trigger, maybe < 1 ms. but i don't know the correct value. lekernel would :)04:48
awgood reminder04:48
awseems gui have an icon can let it reboot which may let FLASH_RESET_N to low to activate.04:48
awlet me try now..;-)04:48
awgood . when i press middle btn, i see low pulse on RESET_N pin04:50
wpwrak_now let's see that falling edge :)04:51
awjust very tiny duration... ;-)04:51
wpwrak_at 100 ns/div :)04:51
wpwrak_(tiny duration) very good :)04:51
awyes, the "reboot" also make that pin have a low duration04:52
wpwrak_great ! :) let's see what happened :)04:52
awhere you are. ;-)04:57
awi should ahve marked as 100pF for C23804:57
awyeah...C238 has a charging curve.04:58
wpwrak_okay. we're in the "yellow" range. the drop is down to 1.9 V. without the scope, it would be a little lower.04:59
wpwrak_can you set the scope to a faster setting, maybe 100 or even 50 ns/div ?05:00
awyes, also up to 4.5V roughly05:01
wpwrak_4.5 V ? oh. that's weird.05:01
awso you only want to see the program_b's falling edge with a super tiny setting?05:01
wpwrak_yup :)05:02
wpwrak_4.5 V is very very strange.05:02
awwpwrak_, it can be happened, we added C238 and no protection on both polarity clamping.05:02
wpwrak_naw, the little overshoot at the end is okay. what's weird is that you seem to drop to -1.0 V when CH2 is being reset05:04
wpwrak_that doesn't make sense at all. lekernel will have fun with it ;-)05:04
awthen we need to see if flash spec can face it as a full low though even having a negative pulse. just to see its maxmum value. ;-)05:05
wpwrak_your channels are DC coupled, correct ?05:05
awsure sure05:05
awyes, DC coupled.05:06
awmoments. i go to capture a new one.05:06
wpwrak_-1.0 V is quite unexplainable to me. note that it looks quite stable over 1.5 us. it's not just overshoot. as i said, lekernel will have fun with this :)05:09
wpwrak_very nice !05:11
awit's quite late you there. ;-)05:12
wpwrak_okay, we have 1.9 V. that's almost good. and much better than my simulation predicted, which is also good.05:12
awI already talked to factory about this C238 @ 100pF, but haven't got their reply if accept to rework. ;-)05:13
wpwrak_to err on the safe side, it may be better to use something like 120 pF. that should add roughly 150 mV.05:14
wpwrak_right now, you have about 250 mV before the forbidden zone. without the scope, this may shrink to 100-200 mV. i don't know what component tolerances will do.05:16
wpwrak_(or temperature variations, etc.)05:16
wpwrak_do you have air conditioning ?05:16
awyes..go on05:17
wpwrak_okay, so your M1 has about 25 C then ?05:17
rohwpwrak_: there will be ambient testing on the ccc-camp ;)05:18
rohwe can put it in a dataloo (up to 60deg C in there)05:18
awthen captured again..so you can compare curve with datasheet and calculate @ 25 degree, right?05:18
wpwrak_roh: *grin*05:18
kristianpaul60, wow05:18
wpwrak_aw: the data sheet doesn't tell us anything about the effect of temperature on the capacitance :)05:19
rohkristianpaul: 8 years ago we learnt that cisco has only monitoring up to 96deg C via snmp/whatever. from there on the value stays stuck05:19
awwpwrak_, must be somewhere we can know. ;-)05:19
wpwrak_aw: but i think qucs can. lemme see :) my guess would be that lower temp = larger capacitance05:19
rohstill switches packets then.05:19
awwpwrak_, let's focus a bit, you are trying to know a lower bound..05:20
rohs/cisco/some cisco devices/g (also may have changed in the meantime)05:20
wpwrak_ah no. higher temp increases the effect.05:21
awthen if using 25 degree by air conditioning..then I can use heat air and measure to 60 degree to capture scopes then you can know the suitable capacitor we should be have.05:21
kristianpaulroh: haha05:21
wpwrak_yes, a hot test would be good. also in general, to see if there's anything else the M1 doesn't like.05:22
rohaw: sure.. if you have a heatgun which you can set so low, try it out. 60 dec C should to no harm to any components05:22
rohirgh. should NOT do any harm05:22
awseems that I have to make some decision to make this C238 value done before factory tell me they can do.05:22
awroh, yeah..sure05:22
wpwrak_aw: you could replace it with a 1 nF or add a 1 nF in parallel, whatever is easier, to see if we're close to the upper bound05:23
awbut like I said wpwrak_ need to let me know what exactly step i have to provide him..then we can a better value of C238 we need. ;-)05:23
wpwrak_aw: if the M1 boots and works normally with 1 nF or 1.1 nF, then we're safe with 100 pF, 120 pF, etc.05:23
rohhm. 7:23 and already worked 2 hours. this will be a bad monday.. shower now.05:24
wpwrak_roh: if you have time for a shower already at 7:23 am, you're having a relaxed day ;-)05:24
awwpwrak_, oaky...now i go to add 1nF on 0.1nF to see if works well. ;-)05:25
rohwhen i am up at 5am on a monday it means something is wrong. usually i stay in bed on mondays till 150005:25
wpwrak_aw: temperature should only have a relatively small effect on C238. i simulated at 125 C and the difference if maybe 10-20 pF.05:25
rohor make that 5pm if it was a good sunday05:26
Action: kristianpaul will take shower in 4 hrs05:27
awwpwrak_, but I remembered that i used fix2 + rc2 then got a C238 1nF then rendering a couple seconds then screen OFF yesterday.05:27
awbefore at that time, my rc2 with C238 (47nF) and worked(rendering) well.05:27
awbtw, now soldering.05:28
wpwrak_aw: 47 nF or 47 pF ?05:28
aw47pF // 1M ohm ;-) I tried to simulate prober then. ;-)05:30
wpwrak_so the 47 pF result only means that the real lower bound may be even a bit lower than 100 pF. this is to be expected. the 1.65 V limit is according to the data sheet, the point at which the chip is still guaranteed to work properly. the real limit will be lower, because they want a safety margin too. maybe 1.5 V or such.05:30
wpwrak_yeah :)05:30
awfrom xilinx spec?05:31
wpwrak_hmm. checking a different data sheet (also from xilinx), i see less friendly values.05:34
wpwrak_the one i picked said "50%" (= 1.65 V). this one says 2.0 V.05:34
wpwrak_this would put us already deep into the forbidden zone05:34
awnow..C238 1.1nF..time to capture @ 10ns/div again05:36
wpwrak_so going from 100 pF to at least 150 pF may be a good idea. but let's see how things go with 1(.1) nF05:36
wpwrak_simulation predicts about 3.1-3.2 V05:37
wpwrak_with a veeeery slow recovery ;-) (not that we'd actually care)05:38
wpwrak_aw: does the M1 boot okay with +1nF ?05:38
awbooting up... ;-)05:39
awbut with my probers..wait..discconnect it to check again05:39
wpwrak_ah, leave them there05:39
wpwrak_they should only make things "worse"05:39
awdisconnected...now still can booting up...05:40
awwait...to see rendering...05:40
awalright...still good...05:41
awnow to get your bound. ;-)05:41
wpwrak_(boots) excellent !05:41
awgreat captured , much better!05:43
wpwrak_as it should :)05:44
awbut the CH2 is have -1V. ;-)05:44
wpwrak_sebastien will have kittens ;-)05:45
wpwrak_afaik, the FPGA doesn't have any negative voltage. not sure if it can generate any inside. so something looks extremely wrong about those -1.1 V05:46
awactually is -1.5V05:47
wpwrak_nice :) the drop is now down to only 3.1 V, like the simulation predicted05:47
awnot sure this is came from other circuits area and pfga can still work normally05:48
awso you had have simulation tool?05:48
wpwrak_(-1.5 V) okay, with overshoot. some of this also comes from the scope. it's not the overshoot i'd be worried about :)05:48
awokay...so we should use a 150pF~120pF?05:48
awyeah...probably some came from my scopes.05:49
wpwrak_aw: i think 150 pF would be better than 100 pF, yes. with 1.9 V, we're probably already in the "forbidden zone"05:49
wpwrak_aw: and the scope adds a few pF, so the unobserved M1 will be worse05:50
awso now I need to find a maybe 470pF (easiler) to test 100 times again.05:50
wpwrak_i would wait with those 100 tests until we have an explanation for those -1.1 V05:51
awyeah...btw, very much thanks about today's lessons. ;-)05:51
wpwrak_i think C238 is solved now and using 150 pF instead of 100 pF won't cause new problems05:51
awthe remaining is my jobs now. ;-)05:51
awyes, but a 470pF is easier to source. ;-)05:52
wpwrak_thanks for doing all the measurements ! :)05:52
awno problems I ought to.05:52
wpwrak_(470 pF) are you thinking of putting a 470 pF into all the M1 rc3 ? or just for the test ?05:53
awwait...so you said yo uwould wait a 100 times until?05:53
awi prefer to use few capacitor value catogories. ;-)05:53
awhmm...wait..checking  bom now05:54
awrc3 bom has 120pF and 220pF already.05:55
awwell...this is important function05:55
awlet's stick on 470pF though05:55
wpwrak_(100 tests) i think you don't need extra testing for going from 100 pF to 150 pF. the test with 1 nF already proves that you can survive with 10x the capacitance.05:56
wpwrak_hmm. i think 470 pF is very large.05:56
wpwrak_lekernel had concerns about rise/fall times, and a large C238 makes them worse.05:56
awyes, theory it is. but you said a 150 pF...05:57
awhmm..okay..I need to source a 150pF immediately. ;-) done?05:57
wpwrak_with 470 pF, the drop would be to ~2.7 V. with 220 pF, about 2.45 V. so if we can't have 150 pF, we may consider 220 pF05:58
awyo just done with your simulated tool again?05:59
wpwrak_120 pF would also be an improvement over 100 pF. but it would leave no safety margin05:59
wpwrak_of course :)05:59
wpwrak_edit value, click on "Simulate", 2 s later, i have the new values :)06:00
awso seems that now I go back to factory to bring few 120pF then try m1 again.06:00
awalways smart you are AUTO man.06:00
wpwrak_if you don't know qucs yet, you should have a look at it. it's very friendly to use.06:00
wpwrak_you mean to bring 150 pF ? :)06:01
wpwrak_i thought they already have 120 pF ?06:01
awi don't have 150pF now.06:01
wpwrak_"<aw> rc3 bom has 120pF and 220pF already."06:02
awrc3 bom have 120pF parts but they are now at factory side.06:02
awnot on my hand here. :(06:02
wpwrak_aah, you want to go to the factory to take some 120 pF from them !06:02
wpwrak_well, we can already predict how 120 pF will perform :)06:02
awand also may see how current 89pcs going on...06:03
wpwrak_you've tested with 100 pF, so we know that 100 pF "works". we also know, from the measurements, that the signal still violates xilinx's specification with 100 pF06:03
awif that's said that a way..we think 150pF is better then I directly order on Digkey right way.06:04
wpwrak_i'm confident that my simulation correctly predicts that the signal will not violate the specification with 150 pF (and we'd have a little safety margin as well. so you could heat the board to 70 C and you'd still not violate the spec)06:04
awalright...from few scopes we had have and simulations, I think I need to order a 150pF right now...this is some sort of delay a bit.06:06
awbut it seems much worthy though definitely.06:06
wpwrak_i would recommend 150 pF. i would consider 120 pF a little risky, but probably okay. 220 pF would probably be okay. it would definitely produce a very good signal as far as the drop is concerned, but we may then have to worry about rise/fall time. lekernel would know more about that issue.06:07
awwolfspraul, hi good afternoon06:08
wpwrak_aw: (order 150 pF now) i don't think you need to rush this. it would be too late for the fab anyway, no ? so just tell them to leave C238 unpopulated06:08
awwpwrak_, maybe we wait a bit until how lekernel think this rise/fall time06:08
wpwrak_and i want to see how he likes those -1.1 V. too bad we don't have videochat ;-)06:09
awwpwrak_, yeah..06:09
wpwrak_let's hope it's something normal and benign. i would suspect your scope setup, but all the other levels look okay, so i can't think of a scenario where your scope would cause this06:11
awi can pretty sure my scope on  settings...but this shouldn't have caused that negative level.06:12
wpwrak_well, bad ground or such. but again, that would also upset other results06:13
awyeah..could be06:13
wpwrak_so i don't know what we're seeing there. if the FPGA can output negative voltages, maybe it's just a misconfigured I/O buffer.06:14
wolfspraulyes I read the backlog, looks like a lot of work06:16
wolfspraulI was a bit worried about the 1nF test because I thought that's where the problems started in rc2 yesterday, but now it seems it can boot even with that. alright then.06:16
wolfspraulyes, don't rush buying 150pF until the negative voltage is looked at06:17
wpwrak_wolfspraul: the good news is that 100 pF is almost right. a little more and the signal will be good. not perfect but good. plus, we already know that we have plenty of headroom.06:17
wolfsprauldo we know the upper boundary?06:17
wpwrak_wolfspraul: (1 nF) i think that was also around the time the M1rc2 developed its fatal flaw. anything after than was meaningless.06:18
wolfspraulthat's why I'm saying - hopefully not caused by the 1nF :-)06:18
wpwrak_(upper boundary) no. we know it's > 1 nF. which is plenty :)06:18
wpwrak_ah ;-)06:18
wpwrak_naw, i don't think 1 nF can do this06:18
wpwrak_too much rework can, though06:19
wolfspraulhow big of a problem is the negative voltage?06:19
awsorry my laptop crashed. ;-)06:19
wpwrak_wolfspraul: how much of a problem is a sparrow doing deep see diving ? :)06:20
wpwrak_wolfspraul: it's something absurd06:21
wolfspraulif we are so little worried about upper boundary, why not just take the 200pF instead of 150pF then?06:21
wpwrak_(220 pF) lekernel seems to worry about rise/fall time. making C238 big would increase these times. i don't know what the problem patterns are with PROGRAM_B_2 would be. could be that slow rise/fall just increases the risk of sporadic failure. or that some chips have a problem and others not. etc.06:23
wpwrak_(220 pF) so it's up to lekernel to decide what value he'd still consider to be safe06:23
wpwrak_aw: ah, in your rc3, the pull-ups on PROGRAM_B_2 and INIT_B have already been changed from 4.7 k to 10 k, correct ?06:24
awnot yet.06:25
wpwrak_wolfspraul: (220 pF) if he thinks 220 pF are still okay for the rise/fall time, then i'd agree to go ahead with it06:25
awbut I talked to factory this morning already.06:25
wpwrak_aw: i mean your rc3 board, the one we're testing now06:26
awsure sure my rc3 are 10K. ;-)06:26
wpwrak_excellent :) have you also already removed the pull-up on FLASH_RESET_N (R60)06:27
awi may order/prepare 150pF in advance not just wait. ;-)06:27
awsure R60(removed), also talked to factory.06:27
wpwrak_aw: naw, let's not waste wolfgang's money :)06:27
wolfspraulwpwrak_: it was lerkernel who immediately suggested 1nF instead of 47pF. When the test result came in from Adam ('poweroff after short rendering'), that's when he started to talk about the rise/fall time.06:27
awwpwrak_, ;-)06:27
wolfspraulbut if he now finds out that 1nF works, who knows06:28
wpwrak_aw: if you need some other component because of the -1.1 V problem, you'll be happy to be able to put that into the same order06:28
wolfspraulif it would have worked yesterday, he would have never mentioned the rise/fall time, I'm sure06:28
wolfspraulso it may have been an attempt at explaining the 1nF failure, which we now thing wasn't even caused by 1nF (possibly)06:28
awwolfspraul, maybe just my rc2 board was not pretty good at well-condition yesterday.06:28
wpwrak_wolfspraul: if you want to put something large (470 pF or more), adam should test with >> 1 nF, to make sure we still have enough headroom. maybe 4.7 nF or 10 nF06:29
wolfspraulbetter not06:30
wpwrak_why not ? afraid of too much rework ? :)06:33
wolfspraulno because I have no reason to ask for that, and I don't want to complicate the issue06:34
wpwrak_ah :) well, if you were to use a large C238, then the 1 nF test adam just did wouldn't be enough. but as long as we stay around <= 220 pF, it's plenty06:35
awhmm...i don't have 150pF, even 220pF on hand now. although i can parallel two 100pF but not good.06:39
awchecking if have 10nF though.06:41
wpwrak_aw: you'll need the caps only for rework. i don't think we need any additional testing with values in the 120-220 pF range. i had a lot of faith in my simulation results :)06:41
wpwrak_aw: (10 nF) yeah, that could be useful to see how much headroom we really have. not really urgent, but not a waste of time either.06:41
awwpwrak_, i have 10nF 0603...let's see if m1 still can boot well..and capture!06:44
wpwrak_but i'm very curious about those -1.1 V :) they're a totally unexpected finding06:45
aw10nF...booting... ;-)06:53
awtime to capture.06:53
wpwrak_nice :)06:55
awnow good we have more hard data. ;-)06:58
wpwrak_a flat line, as expected :)06:58
awso this proves that my rc2 + fix2 was somewhere is wrong already to cause not rendering...let's forget about yesterday. ;-)06:59
wolfspraulyou are brave06:59
awall related to rc3. ;-)06:59
awmaybe just cuased by my soldering. :-(07:00
awyeah..this one is better than 1.1nF07:01
awso i guess wpwrak_ now you are checking your simulation tool again. ;-)07:01
wpwrak_wolfspraul: (brave) naw, C238 can't do much harm. even if you shorted it and drive FLASH_RESET_N or INIT_B permanently high (i think both are open-drain), it's extremely unlikely to cause damage07:02
awbut it's really late you there. :)07:02
wpwrak_aw: (simulation) naw, i don't need simulation to know that 10 nF eliminates the drop ;-)07:03
wpwrak_aw: (late) oh, i'm not really synchronized to the sun07:03
awwpwrak_, ;-)07:04
wpwrak_wolfspraul: (brave) most chips don't mind if you short their outputs to ground and let them burn 100 mW or whatever on that pin. not that anyone would recommend doing that, but it's usually quite safe. e.g., consider the lovely LED transistor without base resistor we had in GTA02. that was as good as a straight short to ground. and the GPIO survived this quite well.07:06
wpwrak_(there's we burnt something like 20-50 mW, don't remember the exact current. in any case, a lot. we could just have driven the LED directly, without the transistor. it would have been easier on the pin ;-)07:08
awwpwrak_, will be greatful if you can plot capacitor value vs voltage drop from your tool then set two boundaries as well?07:09
awif it's hard, forget it. ;-)07:09
wpwrak_hmm, a parametric plot .. it can do that. there's probably a min(x) function somewhere, too07:10
wpwrak_the simulation can't really give you an upper bound, though07:11
awyes, that's what i meant a parametric plot. so we and others can easily know capacitor value behavior, thanks. ;-)07:11
awno problem about upper bound. ;-)07:11
awmeanwhile, although i don't have suitable capacitors on hand, but from scope results, the m1 rc3, the range of choosing C238 seems that can be 150pF ~ 470pF for safe?07:14
awwpwrak_, if i was wrong, pls correct me. since the 10nF is to know a enough headroom.07:16
wpwrak_yes, 150 pF to 470 pF seem reasonably safe, provided that lekernel doesn't have any horror scenario for slow rise/fall times07:18
awmeanwhile i am testing rendering at least 10 times @ 10nF07:20
awsince this is really no need though. becuase the curve is much better. ;-)07:20
wolfspraullet's not become obsessed about this value07:21
wolfspraulwe have enough data and to me it sounds like we can use the same 220 pF we have already in the bom07:21
wpwrak_here's a preview of C238 variation: http://downloads.qi-hardware.com/people/werner/m1/rc3/fix2-c238-variation-preview.ps07:21
wolfspraulthe only thing left open to look at is the negative voltage07:22
wpwrak_still need to find a way to plot it in a nicer way, though07:22
awsure, but just confirm a full test on rendering07:22
awwpwrak_, great.07:23
wpwrak_aw: you should take the time to learn to use qucs. it's very easy. the only difficult part is that sometimes the simulation hits a problem and loops forever. then you need to tweak some parameters or add parasitic resistances or such.07:24
wolfspraulfully supported by me, but maybe after the rc3 run :-)07:25
wpwrak_wolfspraul: he can do this while waiting for lekernel to wake up :)07:25
awwolfspraul, ha...i was though you won't let me spend too much time on this. ;-)07:26
wpwrak_aw: qucs will serve you with a lot of other problems :)07:26
wolfsprauloh, I totally support this. first it's a free tool, and then of course simulation can help us be faster and more effective, once the tool is well understood.07:27
wpwrak_aw: and this is an easy problem to simulate, so it's a nice things to get started with07:27
awwpwrak_, alright..so just google "qucs"?07:27
wolfspraulin theory we can simulate the entire board and all chips, right? then no more small runs, no more production uncertainty - wonderful!07:28
awi save related link first. ;-)07:28
wolfspraulaw: can you upload pictures of the reworked rc3?07:29
lekernelaw, when the M1 stops at JTAG USB connection, it's usually because your distribution runs some daemon (e.g. gpsd) on the serial port that pushes some crap data which makes the M1 enter the debugger07:30
lekernelnothing to worry about07:31
awlekernel, ;-) alright. good to know for me.07:31
lekernelwhere are those large under/overshoots measured?07:31
awrc3 board07:32
wpwrak_lekernel: here's the worst one: http://downloads.qi-hardware.com/people/adam/m1/pic/rc3_2f_CH1-TP36_CH2-TP37_falling_PeakDetect_tinyDuration_C238-100pF.JPG07:32
awch1 is program_b, ch2 is RP#07:32
lekernelflash_reset is actively driven by the fpga after configuration, don't short circuit it07:32
lekernelyou can make it pulse low by doing a software reset07:33
wpwrak_aw: (qucs) something like  aptitude install qucs   then just play :)07:33
awlekernel, i was afraid of it, so i didn't do. ;-)07:33
awthen used "reboot" btn in gui to capture this tiny pulse.07:34
lekerneler, program_b should not drop07:36
lekernelwhen did you measure that?07:36
wpwrak_lekernel: PROGRAM_B_2 drops because of the diode's capacitance07:36
awwpwrak_,  this? http://qucs.sourceforge.net/07:36
wpwrak_aw: yup07:37
lekernelbut it drops from 4.1V to 0?07:37
wpwrak_lekernel: no. CH2 is on top07:37
wpwrak_lekernel: CH2 is FLASH_RESET_N. CH1 is PROGRAM_B_2. what you see is the reset pulse on FLASH_RESET_N and the contamination it causes on PROGRAM_B_207:38
lekernelok, so I understand the PROGRAM_B waveform and it doesn't look too bad07:38
awall those measured today.07:38
wpwrak_lekernel: yes, and we made it even better :) that's as good as solved07:38
lekernelbut why the heck does the FLASH_RESET drop have a 4.1V amplitude?07:39
wpwrak_lekernel: what's weird is that FLASH_RESET_N drops to something like -1.1 V07:39
wpwrak_exactly :)07:39
wpwrak_i'm glad you're as puzzled about this as i am :)07:39
lekernelaw, does the power supply on your board work?07:39
lekernelis the 3.3V really 3.3V, or is it more like 4.1V?07:40
awlekernel, no i used DC adapter07:40
lekerneli'm talking about the power regulators on the pcb07:40
wpwrak_lekernel: can the FPGA output negative voltages ? (e.g., if the I/O driver is configured for something funny, differential or such)07:40
wpwrak_lekernel: note that the high voltage is a nice 3.3 V. the drop is from +3.3 V wrt ground to -1.1 V wrt ground. and it's fairly stable - no significant change over 1.5 us07:41
lekernelI think this is a measurement/ground problem07:41
awlekernel, power regulators?07:41
lekernelaw, what is the "3.3V" voltage on your board?07:41
awmy common ground pin is TP34.07:42
wpwrak_lekernel: (measurement/ground problem) let's hope. otherwise, it's something very very interesting :)07:42
lekernelaw, and what happens when you measure the flash power supply pin with the ground on TP34?07:42
wpwrak_lekernel: is there a way to drive FLASH_RESET_N low for more than 1.5 us ? e.g., 1 ms or just permanently ? (doens't matter if the M1 crashes/hangs)07:43
lekernelyes, keep the reset buttons pressed07:43
lekernelall three pushbuttons07:43
wpwrak_aw: this would be a way to keep FLASH_RESET_N low for, say, a few ms. then we can see if it recovers (goes back to 0 V) or if it constantly stays at -1.1 V07:47
wpwrak_aw: you may want to connect the other channel to some reference voltage. maybe the 3.3 V rail.07:48
wpwrak_naw, that doesn't show anything ;-))07:50
awch2 is TP1(3V3), before I do all previous scopes, I coupling GROUND at central axis as well.07:50
wpwrak_aw: what is CH1 ?07:51
awwpwrak_, i know. just wanted to say my zero axis I checked before i probed. ;-)07:51
awch1 is nothing07:51
wpwrak_alright. now, connect CH1 to TP1 and CH2 to FLASH_RESET_N. scope to falling edge of CH2, 100 us/div07:53
wpwrak_then push all three buttons (lekernel, correct ?)07:53
wpwrak_hmm. could be worse.08:02
wpwrak_but that's the only good thing i have to say about it :)08:03
lekernel1V overshoot for 100us??08:03
wpwrak_(about the signal, not the measurement. the measurement is perfect)08:03
wpwrak_quite a lot08:03
lekernelhow is that possible?08:03
lekernelabsolute maximum rating for the flash is -2V, so this is still acceptable, even if weird & a lot08:05
wpwrak_now, where is that inductor :)08:05
wpwrak_or maybe a wayward cap ?08:06
lekernelI wonder what the other flash signals look like08:06
lekernelaw, can you try with R60 = 10K?08:06
lekernelsame measurement08:06
wpwrak_lekernel: maybe you can check on one of your boards ? or has FLASH_RESET_N changed a lot ? (besides the diode)08:07
lekernelI don't have a scope with memory08:07
wpwrak_oh dear08:08
wpwrak_well, even an analog scope should do08:08
wolfspraulI will review leaflet and brochure today, and send off for printing08:09
wolfspraullekernel: how many brochures should we print?08:09
lekernelnot too many i'd say, < 10008:09
wolfspraulthat few?08:09
wolfspraulI'm in China here they will laugh at me.08:10
wolfspraulthey will think I forgot the 'K'08:10
lekernelI don't know, what do you want to do with them?08:10
wpwrak_wolfspraul: maybe ask them about their price structure. find the "knee" :)08:10
wolfspraulsimilar to the notebook stickers, hand out to people who go to events, conferences, etc.08:10
wolfspraulthe question is how long they will remain unchanged08:10
rohwolfspraul: send me one together with a empty box (for my 'collection') ;)08:10
wolfspraulso I need to review the text and see how quickly it will age :-)08:11
wolfspraulmaybe I print 500 or 1000 or so08:11
wolfspraulI also plan to send them to would-be shops and distributors, so they can put some in their store and see whether any customer reacts08:11
lekernelok, then print more08:11
wolfspraula lot of shops would do that for us, but it's a bit of work for us to mail it all out...08:11
lekernelhow much are they anyway?08:11
wolfspraulno idea08:12
wolfspraulvery little08:12
lekernelmaybe 5k if we distribute them around like crazy08:12
wolfspraulwell :-)08:12
wolfspraulyou are a bit extreme08:12
wolfspraulfirst you say 100, now 500008:12
wolfspraulthe problem is:08:12
wolfspraula) how quickly does the text age08:12
wolfspraulb) how much can we realistically distribute without recipients discarding them or treating them badly08:12
wolfspraulI'll start with 500 or 1000 or so08:12
wolfspraulwhen I mail them to shops, I guess if I mail 20 or so each time that's enough08:13
wolfspraulalright, I got my answer - thanks :-)08:13
wolfspraul100 leaflets, 500-1000 brochures08:14
wolfspraulI'm also thinking about making new notebook stickers with the new design from Christopher Adams, but I still have a lot of the old ones08:18
wolfspraulwell, we get it all distributed...08:18
lekernelaw, also try to measure flash A1 (pin 28) and see if you get large overshoot too when this pin toggles (e.g. when you apply power, during FPGA configuration)08:19
lekerneland during boot08:19
lekernelaw, so:08:19
lekernel1) overshoot on RESET with 10K pull up present (R60)08:19
lekernel2) overshoot on A1 (pin 28) during FPGA configuration08:20
lekernel3) overshoot on A1 (pin 28) during software boot (immediately after BIOS starts, as soon as you get something on the serial console)08:20
lekernelmaybe it's the inductance of the long wire that causes the overshoot on reset?08:22
aw1) use rising when power up?08:22
lekernelno, same measurement as before, when the 3 pushbuttons are pressed after the board booted08:24
awto eliminate consideration of inductance of long wire, i can use sharp prober again. then we know. ;-)08:24
lekernelit's also weird that there is this large overshoot but no ringing at all08:24
lekernelit looks like a RC circuit08:24
lekernelaw, leave the long wire in place, just add R6008:25
awlet me directly touched TP37, then we know if a long wire caused that. yes i added 10K and leave long wire.08:25
awbad, no three hands..thinking.. ;-)08:27
lekernelaw, I just want a trace of FLASH_RESET_N, when the three buttons are pressed, and with R60 present08:30
lekernelone channel is enough08:30
awno problems, captured.08:31
wpwrak_lekernel: i get something like this in simulation if add a 100 pF in series between my "FPGA" and FLASH_RESET_N08:35
lekernelok, still lots of overshoot...08:35
awalmost the same previous one.08:35
lekernelwpwrak_, sure, but this capacitor isn't there08:35
wpwrak_lekernel: yes, but it gives us an idea of the magnitude08:35
wpwrak_my recovery is MUCH faster, though08:36
wpwrak_hmm, can't find a useful equivalent circuit for this mess :-(08:49
lekernelaw, do you have the same problem on the A1 (28) pin?08:56
lekernelor is it only the reset pin?08:56
lekernel(see items 2/3 above)08:57
awnot yet...soldered. ;-)08:58
awcleaned to be ready..wait ;-)08:59
lekernelI can change I/O properties of the FPGA pad after configuration09:00
lekernelatm it's09:00
lekernelNET "flash_rst_n" LOC = P22 | IOSTANDARD = LVCMOS33 | SLEW = FAST | DRIVE = 8;09:00
lekernelI think it could definitely use a slower slew rate and less drive strength09:00
wpwrak_aw: btw, on which side of the long cable to INIT_B is the diode ?09:01
lekernel(unit for DRIVE is mA)09:01
wpwrak_aw: i.e., near R157 or near the reset IC ?09:01
awwpwrak_, both sides have the same length09:02
wpwrak_aw: so it's reset IC -> cable -> diode -> more cable -> R157 ?09:02
wpwrak_interesting approach :) how long are those cables ?09:03
awabout 10cm.09:04
awwire wrapping wire AWG2409:04
awitems 2 captured09:07
wpwrak_hmm, simulation isn't too happy if i insert a transmission line09:07
wpwrak_so the same misery09:23
wpwrak_and INIT_B is off the hook09:23
awrising edge during configuration09:23
lekernelaw, what's that?09:24
awitem 209:24
lekernelcan you zoom on the pulses?09:24
lekernelbut it seems they go rather low already ...09:24
awyeah...so don't need to zoom in though09:25
lekernelthe bottom curve is the 3.3V supply right?09:25
awNO, ch2 is TP3709:25
awcorrect sorry09:25
lekernelcould you still zoom the pulses?09:26
awtyped too quick. bottom is CH1 (TP37)09:26
wpwrak_good :)09:26
awzoom in A1, right?09:26
wpwrak_lekernel: lots of heavy overshoots also after the first one.09:26
lekernelyeah, but no ringing, why?09:27
wpwrak_i don't claim i understand what's happening there :)09:27
lekernelusually, heavy overshoot is accompanied by at least some ringing09:27
wpwrak_aw: oh, maybe do a probe compensation, just in case09:27
lekernelah, yeah, maybe all this happens inside the probe09:28
lekernelit definitely looks more than a RC discharge than ringing or other transmission line effects09:28
wpwrak_you'd need a rather long transmission line for this ... no ringing within ~500 us ...09:29
awoah..yes...maybe compensation on prober itself...man!09:30
wpwrak_well, about 50 km minimum. not too bad :)09:30
lekernelyeah... I did some measurements on the SDRAM signals (which are the most critical) with a high end 6GHz scope and there was negligible overshoot. so it's weird that it would happen that much on the flash.09:31
wpwrak_seems too extreme for being just a probe compensation problem, though09:31
wpwrak_but we'll see ...09:31
lekernelhttp://www.milkymist.org/wiki/index.php?title=RC1_signal_integrity_measurements#DDR_clock and below09:33
lekernelit'd be weird that the flash has 1V overshoot, from the same FPGA and similar PCB traces09:34
wpwrak_yeah. the whole thing screams "weird"09:35
wpwrak_nice scope by the way. i want one of these too ;-)09:36
awyes, one of my prober doesn't make a good compensation. :-( I fucked out half day though. and confused you!09:37
awincluding A1, forget it now.09:37
awnow I am going to prober ch1-program_b & TP37 again. :-(09:38
wpwrak_ah, if it's just one probe that's off, then this would also explain why the other signal looked clean.09:38
awsorry that I now a little sad myself though to say out bad words. not to you. You made me really helpful today. Really sorry.09:48
awI apology here.09:50
wpwrak_hmm ? where are the bad words ? :)09:50
wolfspraulaw: relax, all fine. it seems we are close to being in full control again.09:52
wolfspraulso it seems that was the last little bit of uncertainty? and now it's clear?09:53
wpwrak_we still need lekernel's verdict on the size of C23809:54
wolfsprauloh, I see09:54
wpwrak_aw: does FLASH_RESET_N look okay now ?09:54
wolfspraulmine is 220 pF, just to throw one number on the table :-)09:54
wolfspraulthe least qualified person makes the first bet :-)09:54
awwpwrak_, hi re-capture again..09:55
wpwrak_mine is 150 pF. lowest "safe" value :) but i could live very comfortably with 220 pF09:55
wolfspraullekernel: 150 or 220? we have 220 already in the bom, 150 we need to order09:56
lekerneltest 220, if it doesn't cause reboots or other problems, use that09:57
wolfspraulI'm sure it won't cause reboots, today we tested 1nF and no reboots either09:57
lekernelit's still quite a mystery for me that 1nF caused reboot09:57
wpwrak_we even tested 10 nF :)09:57
wolfspraulthe results yesterday with 1nF were flawed because at some point a bigger problem on rc2 emerged, which masked the other test results09:57
lekernelor maybe the test was not carried out correctly?09:57
wolfspraulyes correct, it may not have been the 1nF perse09:58
wpwrak_lekernel: i think the board was already dead or dying due to excessive rework at that time09:58
lekernelok, 220pF then09:58
wolfspraulwe tested 1 nf and 10 nF today and no reboots (on rc3)09:58
wolfspraulok, 22009:58
awno more -1V overshoot caused by didn't compensation.10:01
awso previous on program_b just confirmed surely no doubt. :)10:02
wpwrak_hmm. there are still a few rather interesting excursions10:03
wpwrak_i wonder what they are ...10:03
lekernelaw, zoom on the pulses10:04
lekernelmaybe just a scope problem - from my experience, digital scopes tend to do all kind of weird things when the sample rate is not high enough10:04
wpwrak_aw: you should be able to see them as follows: get rid of CH1. set CH2 trigger level to -0.5 V. Auto/Normal trigger. then zoom in.10:04
lekernelgood old analog ones are better in this respect10:05
wpwrak_lekernel: also note that he's in peak detect mode now. in general, this shouldn't produce artefacts. but it may exaggerate some problems.10:05
wpwrak_lekernel: in general, peak detect works very well with that kind of scope (i have almost the same). and you need it to see things because it has very little memory.10:06
wpwrak_aw: after zooming in, lower the trigger until it stops triggering. then we'll have the worst glitch/whatever10:07
lekernelthere are 2.5Gsps ADC chips under 100E (in single qty) now10:07
lekernelwe should hook one of those to a FPGA, and add 1GB of DDR3 SDRAM and a USB connection10:08
wpwrak_lekernel: WANT ! :) how many bits ?10:08
lekernel14 iirc10:08
wpwrak_lekernel: (hook up to FPGA) YES :)10:08
wpwrak_lekernel: my dream is a scope with, say, 8-16 channels. doesn't have to have a lot of bits, 6 would be enough.10:09
lekernelthis thing hooked up to a 77GHz RF frontend from a cruise control radar could do nice things too10:10
wpwrak_lekernel: tons of memory, of course. and a display with touch screen, for easy setup. going via PC is messy.10:10
scrts2how much such ADC costs?10:10
scrts2I am thinking of software defined radio, but my idea was 4ADCs of 250MSPS10:10
scrts2each sampled same clock but +90deg10:10
lekernelmom... I just received some ADI spam about that10:10
lekernellet me find it again10:11
scrts2250MSPS ADC costs ~60-100USD @ digikey, but I found the same for 10USD10:11
awwpwrak_, okay10:11
lekernelargh, no, it's a DAC10:12
lekernelstill a nice beast though10:12
scrts2ah.. DAC :)10:12
wpwrak_hmm. the transients may actually be positive ones, on CH1. not negative from CH2.10:13
wpwrak_aw: maybe first get rid of CH1, then repeat. that'll make it clearer10:13
scrts2http://www.analog.com/en/analog-to-digital-converters/ad-converters/ad9286/products/product.html the fastest You can get from AD10:14
wpwrak_not too expensive10:14
lekernelscrts2, I guess you need additional and non-trivial sample and hold circuitry for the 4x ADC trick to work, no?10:15
scrts2that was brainstorm idea, I've not checked how it should work, but basically I belive not10:15
scrts2because the input clock is the sample clock at the same time10:16
scrts2each ADC has its own sample&hold10:16
lekernelyes, but what is the width of the sample window?10:16
scrts2the delay between sample take and data output is fixed (e.g. 18 cycles)10:16
lekernelthis has nothing to do with the width of the input sample window10:16
scrts2I am not sure of the sample window10:17
scrts2it depends on ADC I suppose10:17
lekernelthat's the one parameter to check when using multiple ADCs with phased out clocks10:17
scrts2anyway, 4ADCs is not my idea, I've seen such hardware in cheap oscilloscope10:18
awyes, after this, i tried to trigger at -0.5V, then I can't get anything.10:21
wpwrak_uuh. ugly things.10:21
lekernelaw, could you reduce the time/div setting?10:22
lekernelnah I don't believe this10:22
wpwrak_aw: what's the trigger coupling ?10:22
awyes, i did..but once i reduce one segment time/div, then I can't get anything. seems that my scope memory not enough.10:23
awDC coupling10:23
awuse level?10:23
lekernelwhat do you mean "can't get anything"? scope won't trigger?10:23
wpwrak_aw: you should let the board just run. no need to reset. these glitches should happen all the time.10:23
lekernelno, the flash is not accessed after boot10:24
awsince this A1 have a lot of info in front of current one I showed you.10:24
lekerneleverything gets in the sdram10:24
wpwrak_lekernel: oh, damn. there goes my idea for "catching" the gliches :-(10:24
lekernelI don't believe there are such glitches10:25
awi can slow down to 250us /div only then if 100us/div then screen shows nothing10:25
lekernelthey are over the power supply voltages, and the only thing that could cause such voltages are transmission line effects. but transmission line effects do not cause sporadic glitches.10:25
wpwrak_i'd also be surprised if they're real. but the scope must have picked up *something* ...10:26
wpwrak_aw: yeah, your scope has very shallow memory. something like 1 kSa. barely enough for two zoom steps.10:27
awi think so10:27
awthat 500us/div i can zoom on 250us/div to see, surely the waveform is strange. but if I need to trigger A1 @ 250us/div then no pulse i can get. :(10:29
awbtw, i need to call factory about C238. it's bad that this morning I told a 100pF firstly.10:30
wpwrak_(silly me. scope shows the trigger setup. could have seen myself hat the coupling is DC ;)10:30
wolfspraulaw: we have settled on 220 pF now10:31
errordeveloperhello there ..10:31
awif they do fast, then..bad10:31
lekernelhi errordeveloper10:31
errordeveloperactually I got a question down to the point -10:32
wpwrak_lekernel: shall we file the "glitches" under "UFO sightings - no action required" ?10:33
errordeveloperwhy lm32 and not openrisc or .. hm (i won't metnion the opensparc just yet)10:33
lekernelerrordeveloper, this is explained in the documentation online10:33
errordeveloperoh .. sorry10:34
lekernelhttp://milkymist.org/thesis/thesis.pdf should answer most questions about the soc architecture, including this one10:34
wpwrak_lekernel: how about printing this FAQ and hanging it across, say, the brandenburger tor ? ;-)10:34
errordeveloperthanks for the link  :)10:34
errordeveloperlekernel: llhdl is your project, right ?10:35
errordevelopervery interesting btw10:35
errordeveloperlekernel: do you have a roadmap somewhere on-line ?10:37
wolfspraulerrordeveloper: the roadmap will be largely defined by your own contribution :-)10:37
wolfspraulwant to help?10:37
lekernela roadmap for what? llhdl? flickernoise? mm soc?10:37
wolfspraulthe milkymist project is a fantastic project with huge potential - but - spread scarily thin across its wide ambitions10:37
errordeveloperyep, see what I can do10:38
wolfspraulso we are seriously welcoming any new contributor10:38
errordeveloperI have to admit, I'm fairly new to the HDL world10:38
errordeveloperbut I got a few projects that I'd like to implement10:38
errordeveloperI'm mostly into audio DSP stuff10:39
awjust called and confirmed that: they are doing fast more than i thought10:39
wolfspraulcan they change to 220 pF?10:39
lekernelatm i'm working on implementing http://graphics.stanford.edu/papers/texture_prefetch/ into the milkymist tmu10:39
lekernelthis should make it go 3 to 4 times faster10:40
awR60(removed), R30/R157(10K), C238(100pF) , now have 45 pcs reworked done, then I just told to remove that 100pF10:41
lekernelor you can just solder another 100pF on top of the existing one during rework ...10:41
awsince I'll do a soldering for wire and one diode for sure, no need to tell them rework on 220pF now.10:42
awI don't want them to dislike us and back and forth to get pain for them.10:43
wolfspraulso you do the 220pf10:43
wolfspraulwe are doing design work in parallel with the run :-)10:43
awthey helped half of reworks, then half of it  belongs to me. ;-)10:43
wolfspraulaw: yes, all good10:44
awwpwrak_, lekernel do i need to still re-capture considerable signal concerned?10:45
lekernelwhat signal?10:45
lekernelno, you shouldn't recapture anything imo10:45
wpwrak_"UFO sighting" then10:46
errordeveloperwell, I'm at work and here I have an apple machine, so ..hm10:46
errordeveloperlet's see if I'll get llhdl to compile10:46
Action: wpwrak_ goes back to his book. and soon a little nap :)10:48
lekernelerrordeveloper, are you a C programmer?10:48
errordeveloperlekernel: to be honest still learning10:52
errordeveloperwell at the moment I know C better then Verilog10:52
errordeveloperbeen working on a few smallish projects10:52
errordeveloperin terms of llhdl, i'd be up for doing some testing to start with ...10:54
lekernelit's way too early to test anything. in fact, most things do not work and I know it.10:54
errordeveloperdamn, I still didn't get any fpga hw10:54
lekernelplease no not report llhdl bugs or missing functionality unless they're accompanied with a patch. there are too many.10:55
errordeveloperwell .. I'll start by trying to compile llhdl on darwin :()10:55
wolfspraulerrordeveloper: you can buy the Milkymist One fpga hardware in about 2 weeks or so, 499 USD + shipping. it's not the cheapest dev board you can find but it will give you a great starting point and something that actually works and is fun to use.11:00
wolfspraulif it's too expensive, there are dev boards of all kinds, going down to 50 USD or so11:01
errordeveloperoh, well .. I would, thought it's not the kind of money I'm able to spare right now :(11:01
errordeveloperyeah, I've seen some very cheap cpld boards from poland11:02
wolfspraulthere may be cheaper dev boards that will still allow you to contribute to Milkymist effectively11:02
wolfspraulwell you just have to set realistic goals first - focus. you can learn for 100 years...11:02
errordeveloperhowever those where really just break-out boards11:02
errordevelopersure thing11:02
wolfspraulobviously from a Milkymist perspective it would be great if you could contribute something to the bigger unified goals of Milkymist. there is a lot to choose from :-)11:03
wolfspraulestablishing focus for yourself should probably be your #1 priority now :-)11:03
wolfspraulmilkymist one is a product, we spend a lot of time and effort to bring it up to 'product' level, in hardware quality, manufacturability, software features, accessories, etc.11:04
errordeveloperwell, I don't have a problem with that ... despite a couple of very disturbing things which happend very recently11:04
wolfspraulthe cheapest fpga dev boards are essential marketing tools for the fpga makers. nothing wrong with that, like I said first you need to know what you actually want and what your focus is.11:05
wolfspraulwhat disturbing thing happened?11:05
errordevelopera) I kind of really need to change the job and  & b) my uni is giving me a bit more trouble11:05
errordeveloperI got my project done quite well, though happend to fail a math module (which I chose myself)11:06
errordevelopervery very silly11:06
errordeveloperhow could I know that error correcting codes is totaly not my thing !?!11:06
errordeveloperwell ..11:07
errordeveloperI mean, I totally understood how they work11:07
errordeveloperbut just not able to do all the proofs etc11:08
errordevelopersorry don't want to complain11:09
wolfspraul:-) we hear you11:10
wolfspraulMilkymist for all its greatness cannot help you with that :-)11:10
errordeveloperlol :)11:10
errordeveloperlekernel: milkymist-ruby, is it just a patched MRI  to run on lm32 ?11:13
errordeveloperMatz's Implementation of Ruby11:14
lekernelit's the small ruby interpreter, with very minor modifications to run on rtems and lm32 yes11:14
lekernelif you're planning to run that on your FPGA board, keep in mind this needs SDRAM, and all non-professional people I've heard of have failed to get that to work when trying to port Milkymist (or other softcores) to their FPGA boards11:15
lekernelif you buy a M1, you get usable SDRAM out of the box11:16
errordeveloperwell, I'm just generally interested how far Ruby will get in the embedded world11:17
errordeveloperI quite like Ruby as a language11:17
errordeveloperhm .. is the ac97 core capable of 96kHz sampling rate ?11:19
wolfspraulerrordeveloper: if you have friends who are also developers, tell them about Milkymist :-)11:19
wolfspraulthat'd be a really cool thing you could do to help the project11:19
errordevelopersure, I mean .. at the moment I really don't know any hw people11:19
errordeveloperthough .. hm, I know some people in VJ circles11:20
wolfspraulanything 'below' kernel is still out of sight for many software folks, whether that's dsp, fpga/ic design, data converters, synthesis, and so on11:20
wolfsprauljust tell your friends about the project, that'd be an awesome help11:20
errordeveloperyeah, some only care about the webapps11:21
errordeveloperand may be arduinos :)~11:21
wolfspraulnot bad, that's a start11:21
Action: lekernel is reading a hardware hacker journal atm with _thermodynamics_ in it. but that was 1993 ...11:21
wolfspraulsome may only care about fishing, or basketball, or ... that'd be a bigger gap :-)11:22
lekernelthose days, this has turned into http://hackaday.com/2010/02/21/steorn-orbo-motor-replica/. sigh.11:23
errordeveloperlekernel: as far as I can see you most concentrated on video part and the only component in the audio path is the ac97, right?12:41
lekernelyes ....12:41
errordeveloperI can imagine that in the current implementation the FPGA resources are alsmost fully utilised  ...12:44
errordeveloperI am asking cause I'd be interested to implement an audio mixer and filters ..well13:01
errordeveloperthere only two ballance audio channls on the milkymist, so not sure about what sort of mixer it could be13:02
errordevelopererr .. my version of Xcode doesn't have clang++13:13
errordeveloperonly clang without '++'13:13
errordeveloperlooks like clang++ is a symlink ..let's see13:32
awjust throw some rc3 hardware patches pictures first: http://en.qi-hardware.com/wiki/Milkymist_One_run_3_schedule#Hardware_Patches_of_Boot_up13:46
awwill keep giving each picture some descriptions13:46
awhaven't applied glue, so need to add that picture.13:48
rohaw: we need to run the cable between the usb sockets or so14:02
rohthere is no space between the pcb and the acryllic on the dmx socket side14:02
awroh, goes through two usb sockets? will this more not influence to acryllic?14:04
wolfspraulnice pictures14:05
awactually I've not decided to go through what area.14:05
awwe can think more carefully now.14:05
wolfspraulaw: do you have a complete assembled case?14:06
wolfspraulif not, please assemble a case, you have enough of the old purple color cases14:06
rohaw: there is 1mm space between the pcb and the acryllic on the side with the buttons14:07
awor do you think that somewhere the acryllic inside case you can scrape endophragm of case?14:08
wolfspraulaw: please assemble a full case, you have to learn the assembly anyway. then you can also see what space we have.14:08
wolfspraulaw: do you have a full case assembled?14:08
wolfspraulyou have to answer that question first :-)14:08
awsure I had have assembled once14:08
wolfsprauldo you have a case there now?14:09
awbut not enough14:09
wolfspraulthen you can put the board inside and see yourself14:09
wolfspraultake one of the purple cases, and assemble it for yourself, make it your own case14:09
wolfspraulyou can use that case for experiments14:09
awso have all 1mm offset to acryllic?14:09
wolfspraulroh is saying that the wire must run on the _FRONT_ side of the pcb, where the buttons are14:09
wolfspraulonly where the buttons are14:09
wolfspraulplus - assemble a case and try yourself. but for now we say: cable must be on the side where the buttons are14:10
awcable "MUST" be on the side where the buttons are?14:10
wolfspraulyes :-)14:11
wolfspraulthat's what everybody is saying14:11
awso this means that I need to solder a longer wire then to test boot again. ;-)14:11
awokay. ;-)14:11
wolfspraulyes, longer14:11
awso ha moments...soldering again... ;-)14:12
errordeveloperlekernel: so you are using glpcver as I can see .. how do you find it ?14:12
errordeveloperlekernel: I have used icarus and wanted to try veripool,14:13
awi can imagine I need to play "soldering" these two weeks. ;-O14:16
wolfspraulyes not so good. to rework the 89 boards will probably take several full days.14:17
wolfspraulbut it's ok14:18
errordeveloperlekernel: I'm actually reading your thesis, wonder if you have a version with hyperref enabled, it would be much easier to navigate!!14:18
wolfspraulyou will slowly accelerate :-) the first 10 will be slow, the next 10 a little faster, and so on :-)14:18
awyeah..it's okay.14:21
awthe new longer cable I cut is 15 cm. seems the best route is to go between SW1 and SW2 then surrounding dram chips then go to D16.14:22
awneed upload pictures again. ;-)14:23
wolfspraulaw: I think between the buttons is not so good14:25
wolfspraulmaybe if you are really in the middle between SW1 and SW2 is OK14:26
awbut roh said it MUST?14:26
wolfspraulno, you need to read more carefully14:26
wolfspraulfirst we say it has to be on that _side_14:26
wolfspraulthe front side14:26
wolfspraulbecause on that side, we have a gap between pcb and acrylic14:26
awcause i can see yes a little like 1mm from push btn's metal and push pod.14:26
wolfspraulon the entire side14:26
wolfspraulyes correct, if you are exactly in the middle between SW1 and SW2, maybe it's good14:27
wolfspraulbut if you are close to sw1 or close to sw2, you may block the push-down action14:27
wolfsprauldo you have an assembled case there?14:28
wolfspraulthen you can see it right away14:28
wolfspraulbetween the USB would probably be safest, but also longer?14:28
wolfspraulor maybe between SW1 and USB?14:30
wolfspraulclose to the USB connector14:30
wolfspraulhow about that? where C177 is...14:30
wolfspraulaw: how about go down where C177 is? or go down between the two USB? or go down where U8 is?14:35
rohhm. that stuff is nasty15:04
Action: roh just tried gluing with that evil special stuff... and it just vapourizes in no time15:05
rohmeans i have one try per piece and i have not time to move it afzer contact.. irgh15:07
rohand one piece at a time.. not a rig with a lot of them. *sigh*15:08
wolfspraulbut does it function as a glue? how does the button look like? (in terms of cleanliness)15:08
awi am back ;-)15:09
wolfspraulhi :-)15:09
rohit looks nice if its done right (completely clear)15:09
rohbut it makes the surface 'misty' if spilled where it shouldnt15:09
rohthats what makes it so difficult to handle.. timing (few seconds, <5) and precise contact and short pressure (3-5sec)15:10
rohseems solid after just some moments. (havent tried timing it yet)15:10
awwolfspraul, hmm...let's see what i have now...need to assemble mine here firstly ;-) and take pictures.. :)15:10
wolfspraulsure, take your time15:10
wolfspraulalso route the cable so it's less likely to cause electrical trouble15:10
rohi could pull both parts from the tape by pulling on the upper one after <60 sec15:10
wolfspraulbut mechanically, maybe close to USB is better than between the buttons...15:11
wolfspraulbecause the wire down must not block the button push action15:11
rohthey should have written on the label that not only inhalation but also the mad timing constraints of that stuff make headaches ;)15:11
wolfspraulaw: and as you can see, roh is working on gluing the buttons15:11
kristianpaulroh: is silicone not proper forbutton glugin?15:13
kristianpaulwell, evil special stuff, still sound interesting :)15:14
awha, great . alright..firstly i supposedly to be familiar case assembly methods again. ;-)15:14
wolfspraulyes, please. you need to know the case well, play with it a little...15:15
wolfspraulhow many cases do you still have? you should have at least 5 or so15:15
wolfspraulyou can take some out for your own use / lab use. so don't worry about scratches or so, we will never sell those15:15
kristianpaulwolfspraul: rc3 ships with assembled cased?15:16
wolfspraulyes of course, full product15:16
wolfspraulassembled case, accessories, box, everything15:16
kristianpaulnice !15:16
rohkristianpaul: glue breaks. the stuff i use is making the surfaces of the acrylic parts bond together15:17
kristianpauli see15:17
wolfspraulroh: what is the chemical?15:22
kristianpaulcalling fedex? :)15:30
wpwrak_wolfspraul: (first 10 slow,next 10 faster) and then boredom sets in and with it come mistakes :)15:46
GitHub112[extras-m1] yizhangsh pushed 1 new commit to master: http://bit.ly/mVa4D015:47
GitHub112[extras-m1/master] modified EVA design - Yi Zhang15:47
awwpwrak_, boredom sets failures too, i agreed though. so I need to take this care. ;-)15:48
wpwrak_(cable routing) my vote would be for running it between the two USB receptacles, if possible. no moving parts in the area.15:49
Fallenouwin 2015:49
wolfspraulyes I agree, between USB looks great15:50
awwpwrak_, (cable routing) I would try tomorrow. ;-)15:50
wpwrak_roh: is it the stuff with the good cyanide fumes ? if yes, you should try to heat it a little. i've heard the effect is quite impressive ;-)15:53
wpwrak_roh: ah, DCM is a different beast. pity.15:56
rohwpwrak_: cyanide? i dont hope so.15:58
awjust set to keep rc3 rendering through night... ;-) cu16:15
GitHub53[milkymist] sbourdeauducq pushed 1 new commit to master: http://bit.ly/oQk9Fb20:52
GitHub53[milkymist/master] TMU prefetch (WIP): tag memory (untested) - Sebastien Bourdeauducq20:52
Vaati_oh awesome, milkymist source is in verilog22:00
Vaati_anyone tried loading it on altera devices?22:00
--- Tue Jul 19 201100:00

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