**** BEGIN LOGGING AT Wed Apr 20 02:59:58 2016 Apr 20 06:40:28 zmatt: thanks, I'll take a look at it Apr 20 06:43:30 zmatt: for example this https://github.com/beagleboard/linux/tree/4.4-ti-rt-r6 Apr 20 10:10:05 Hello, How can I store the data of sensor in file in the beaglebone black? Apr 20 10:11:57 Hello, How can I store the data of sensor in file in the beaglebone black? Apr 20 10:12:11 1. how you can get the data? Apr 20 10:12:16 or how are you getting it Apr 20 10:12:58 I am using LM35 to sense the temperature and I want to store it in the file Apr 20 10:14:51 how it's connected and how are you accessing it? Apr 20 11:01:40 Hi guys and gals, I'm having trouble working out what the sharealike license means for using beaglebone black design. Does the statement: "If you remix, transform, or build upon the material, you must distribute your contributions under the same license as the original." mean a derivative design has to be shared or simply if it is shared it has to be under the same license? Apr 20 11:09:40 Guest31459: it means if you base some design on it, then you MUST share it under same terms Apr 20 11:10:15 for more information reread the applicable license or consult your IPR lawyer Apr 20 11:13:56 Thanks. Apr 20 11:37:50 Hi, I try to exe a program on ssh but there is an error x11 proxy ( Apr 20 11:37:57 PuTTY X11 proxy: Unsupported authorisation protocol widtest: cannot connect to X server localhost:10.0 Apr 20 11:38:20 I enabled X11 in setting Apr 20 11:39:44 and I still have this problem... Apr 20 11:40:03 you're running X11 on windows? o.O Apr 20 11:41:12 or are you trying to run a program on the X11 display on the BBB itself? Apr 20 11:41:38 zmatt: sorry I'm running putty on windows Apr 20 11:41:51 yes that I understand, but that wasn't my question Apr 20 11:43:05 zmatt: running x11 on win Apr 20 11:44:02 zmatt: I supposed ;) I installed xming don't know if it helps Apr 20 11:44:08 xming, ok Apr 20 11:45:01 no experience with it, but clearly something wonky is going on with authentication. a quick workaround might be to just disable authentication altogether (I'm assuming you're not sharing your LAN with malicious individuals) Apr 20 11:45:13 in a local xterm type: xhost + Apr 20 11:45:29 (local = in xming, not on the BBB) Apr 20 11:45:54 alternatively, if you're running xming anyway, might as well use a proper ssh client instead of putty Apr 20 11:46:58 the problem might be that X11 proxy over ssh requires that you run ssh from inside your X11 session (i.e. DISPLAY must be set correctly) Apr 20 11:47:40 if putty is started separately it might not know hot to access Xming correctly Apr 20 11:48:52 what you're doing is quite uncommon, hence some fiddling may be required :) Apr 20 11:55:09 also, in my experience X11 over ssh often has quite crappy performance even on a fast LAN... running Xvnc on the BBB and connecting with a vnc client might actually be faster (although I'm not 100% sure since it'll put a heavier load on the BBB, which isn't particularly fast itself) Apr 20 11:56:09 zmatt: ok ok, thanks for your help Apr 20 11:56:12 hmm, there are actually some notes on Xming and PuTTY here... http://www.straightrunning.com/xmingnotes/ Apr 20 11:56:38 apparently "PuTTY is Project Xming's preferred and integrated X terminal emulator for Microsoft Windows" ... so I revoke my earlier comment on ssh client choice then I guess Apr 20 11:56:51 I honestly know nothing about Xming :D Apr 20 12:04:20 gonna give it a try when I'm at the office though, this could be potentially useful Apr 20 12:05:47 I use cygwin for X. Xming is simpler to set up but is more laggy in my xp. Apr 20 12:06:02 zmatt: I appreciate thanks Apr 20 12:06:09 Ragnorok: that's interesting, you'd expect that Xming ought to be faster Apr 20 12:09:17 I'm particularly interested in its support for AIGLX: "Xming can exploit the capabilities of an OpenGL enabled device, on a Windows machine, by use of its -wgl option{ Apr 20 12:09:17 Why? Apr 20 12:09:20 { Apr 20 12:09:42 Ragnorok: because you cut an API emulation layer out of the middle Apr 20 12:10:43 Ah. Dunno. I just know Cygwin "feels" snappier than Xming. It also renders cleaner. I don't recall the details, but the truth is they are roughly equivalent. It's not like Cygwin was some obvious winner. Apr 20 12:11:24 I actually use msys2 on windows machines, although I don't remember what convinced me to use it instead of cygwin Apr 20 12:11:27 I primarily run CCS 6 & dolphin with it, so that's not a big sample. Apr 20 12:11:41 ehh Apr 20 12:12:08 you run CCS in cygwin? what? CCS is available natively for windows Apr 20 12:12:21 (and generally works better on windows than on linux in my experience) Apr 20 12:12:50 I decided doing cross compile from linux was better for BBB development. (shrug) It works fine as frog hair. Apr 20 12:13:43 heh Apr 20 12:14:31 must be tons of alternative IDEs available for doing that though? I would personally not voluntarily use CCS unless I really need to do JTAG debugging of a TI target Apr 20 12:15:20 At the time I thought JTAG may be necessary, but that's because I was a rank n00b at BBB dev. (shrug) Now it's set up that way I see little impetus to change it. Apr 20 12:15:55 (if I just need to upload, poke around in a few registers, or do memory dumps, I typically use CCS debugserver scripting from the commandline rather than launching the whole IDE) Apr 20 12:17:14 I've been using Eclipse for years and years. I see nothing wrong with CCS. Apr 20 12:17:34 JTAG could still be useful to have watchpoints and be able to step into syscalls, but to avoid halting irqs and other processes it would require kernel support that afaik ain't there Apr 20 12:17:40 ok, I guess it's what you're used to Apr 20 12:18:35 Back In The Day it was one of the few cross-platform IDEs, so I started using it. (shrug) It more that a I'm used to it at this point. Apr 20 12:19:56 yeah I should perhaps still browse around for an IDE I can tolerate, but usually I'm okay with vim and makefiles (even though make sucks, at least I know most of its quirks) Apr 20 12:20:26 something like the inability to easily customize syntax highlighting would already be a deal-breaker for me Apr 20 12:21:46 I really like vi, but I like tabs better, which is what drew me to IDEs. I detest make like no other, which was another thing that drew me to IDEs. I'm generally turning off "helper" crap in any environment; minimal syntax coloring and I'm good to go. Apr 20 12:22:27 I dislike the opacity of the build process in many IDEs Apr 20 12:22:45 Eclipse is very, very flexible but I've never tried to figure out how to change it's highlighting other than to disable parts. Apr 20 12:23:03 i.e. I know what I want to accomplish, but have no idea where the fuck I'd need to configure it Apr 20 12:23:09 Eclipse shows what it's doing in a command line window. It even uses make under the hood, and I can edit that file. Apr 20 12:23:11 mazes of settings Apr 20 12:23:28 plus CCS is really bloated imho Apr 20 12:23:41 i've actually had a case where it get axed by the OOM killer Apr 20 12:23:49 I don't like the gobs of dialogs in IDEs, to be sure. Eclipse at least has a search function in settings. Apr 20 12:24:38 I've never seen that happen. Apr 20 12:25:13 Well with Eclipse. I only started using CCS recently and that's only being used for BBB dev, so it's not starved in any way. Apr 20 12:25:38 I agree CCS is bloated though. Normal Eclipse is much cleaner. But there it is. Apr 20 12:25:42 I never used it for anything more complicated that what fits in internal SRAM :P Apr 20 12:25:51 (on the target) Apr 20 12:33:19 wow, I learn something new all the time... you can actually have the X font server running on a separate machine to share your repository of fonts across multiple machines? Apr 20 12:33:34 zmatt: Yes Apr 20 12:33:35 Sure. Apr 20 12:33:39 zmatt: That's the original design of XFS Apr 20 12:33:47 Fonts were large, and had per-install licensing Apr 20 12:34:00 yes, makes perfect sense in 20/20 hindsight Apr 20 12:34:10 it just never occurred to me Apr 20 12:34:20 I remember having to split out the xfs server to it's own hardware cause it was using a fair bit of resources when ran off the other server. Apr 20 12:34:26 Meh Apr 20 12:34:30 Suddenly I feel old ;P Apr 20 12:34:48 that's another thing about X11.. reading its documentation tends to feel like I'm doing archeology :P Apr 20 12:34:57 You are, in a sense Apr 20 12:35:06 "twm or other popular window managers" ... ... Apr 20 12:35:15 xhost the ICE authentication cookies Apr 20 12:35:26 Many designs that _assume_ that you are nfs sharing /home Apr 20 12:35:48 I already had some run-ins with xauth Apr 20 12:35:48 Between the machine your display is on, and the machine you run your X application on. Apr 20 12:36:05 Otherwise you have to manually (or scriptedly) copy cookies over Apr 20 12:36:40 Knowing all this, about how X is "display agnostic" "network transparent", and listening to the whine about Wayland.. I just want to go over and strangle the neckbeards with their own beard. Apr 20 12:37:01 ( This coming from someone who's worked with it, and who's got the grey unixbeard ;) Apr 20 12:37:16 Well, still not mostly gray, but point stands. Apr 20 12:38:00 Now, I'll go join the Wayland future and play with desktop containerized applications Apr 20 12:38:01 the deepest I've dug into X11 was when trying to make a single X11 application run as startup service without display manager or window manager Apr 20 12:38:12 https://github.com/alexlarsson/spotify-app <-- fancy shiny stuff! Apr 20 12:38:20 Hahaha Apr 20 12:38:31 You always had one of that "xinit" ;) Apr 20 12:38:32 My initial research into Wayland wasn't positive, but perhaps I missed something. Apr 20 12:38:35 X Apr 20 12:38:41 Spidler: doesn't work Apr 20 12:38:56 you see, a service isn't a (PAM) session Apr 20 12:39:07 and doesn't have access to the VT Apr 20 12:39:17 Haha Apr 20 12:39:19 That's modern crap Apr 20 12:39:23 normally the display manager sets up the session Apr 20 12:39:26 Yea Apr 20 12:39:35 You'll have to spawn a VT'd session from something Apr 20 12:39:49 I've got an ansible recipie here that uses openbox session Apr 20 12:39:50 but I only found one "nodm" and it was unmaintained and didn't work Apr 20 12:39:55 With a single fullscreen app Apr 20 12:40:04 another thing that bothered me, I wanted to get rid of VTs Apr 20 12:40:06 I can share that if you want. It's not _what_ you want, but it works :) Apr 20 12:40:16 X11 doesn't actually need it, it uses drm for the actual graphics Apr 20 12:40:21 Mmmm Apr 20 12:40:27 libinput uses it though Apr 20 12:40:42 and I think modern xinput uses that Apr 20 12:40:46 Could be wrong Apr 20 12:40:58 doubtful, it opens the /dev/input/event* device Apr 20 12:41:17 xinput directly, or does it fork out to libinput to do that? Apr 20 12:41:23 Well, fork. "delegate" Apr 20 12:41:30 no idea Apr 20 12:41:42 either way, the only thing they might need from the VT is disable it Apr 20 12:41:48 Mmm Apr 20 12:41:52 and handle notifications about VT switching Apr 20 12:41:58 Anyhow, I'm off to deal with some web related JS crap. Apr 20 12:42:01 none of which would be necessary if VTs are disabled Apr 20 12:42:06 but anyhow Apr 20 12:42:13 QPA=linuxfb worked like a charm Apr 20 12:42:16 bye bye X11 Apr 20 12:42:17 If anyone here run some NFS server, I've got stuff I'd love to see you try Apr 20 12:42:19 CONFIG_VT=n Apr 20 12:42:21 haha Apr 20 12:43:26 I am still considering setting up NFS4... I already once got kerberos briefly up but then other stuff got prioritized Apr 20 12:43:57 I found problems with current kernels default rsize/wsize setting on both nfs 3 and 4, tcp and udp. Apr 20 12:44:20 Basically, a random 4k write series will cause high IO wait but almost no actual IO, either network or disk Apr 20 12:44:36 To the point where the systems, both client and server, become _really_ unhappy with life. Apr 20 12:44:40 $ fio --name=randwrite --ioengine=libaio --iodepth=1 --rw=randwrite --bs=4k --direct=0 --size=256M --numjobs=8 --runtime=60 --group_reporting Apr 20 12:44:53 fun :/ Apr 20 12:45:09 It goes away if you manually force down rsize/bsize on either server or clients. Apr 20 12:45:13 which reminds me, I need to check how to limit the amount of writes being buffered to eMMC on the BBB Apr 20 12:45:18 But with the default settings, that will just _kill_ everything. Apr 20 12:45:42 I've had a case where the kernel memory filled up to the point some PF_ATOMIC allocation failed in the block layer, leading to a very very hosed filesystem Apr 20 12:45:44 We're talking about a few k's of traffic going, loads in the 30+ range, and servers + clients grinding to a halt. Apr 20 12:45:54 Ohhh. fancy Apr 20 12:46:17 I don't know what idiot thought returning ENOMEM is ever acceptable in the block layer Apr 20 12:46:24 defer it, do it more slowly, just DO IT Apr 20 12:46:25 Fancy Apr 20 12:46:35 Well Apr 20 12:46:37 error = data corruption, pretty much guaranteed Apr 20 12:46:37 lol Apr 20 12:46:38 overcommit Apr 20 12:46:58 I generally dislike overcommit, but so much real code sucks so badly, I completely understand it's existance Apr 20 12:46:59 no the problem was allocating a (probably large) DMA descriptor, for some reason with PF_ATOMIC Apr 20 12:47:06 Ohh Apr 20 12:47:22 1. I don't see why that would need to be PF_ATOMIC. just fucking block Apr 20 12:47:37 Fuck yeah. I've got a containerzied Spotify running on my desktop! Apr 20 12:47:39 2. if dma allocation fails, do it oldfashioned if you need to Apr 20 12:48:12 but an error coming from the block layer is the end, period Apr 20 12:48:22 there's nothing the fs can do Apr 20 12:51:10 http://pastebin.com/nTunqn7u <-- snippets from kernel log Apr 20 12:51:28 (trouble starts at line 233) Apr 20 12:54:08 I managed to recover the system (via USB mass storage mode) but it wasn't easy... especially since I initially couldn't enter the system with nspawn since libc.so contained garbage (this incident happened right in the middle of a system upgrade including rather essential stuff, such as libc) Apr 20 12:59:26 zmatt: ouch. that looks so bad. Apr 20 12:59:36 zmatt: how do you use nspawn on the devices? Apr 20 13:02:30 make sure qemu-user-static is installed, attach bbb via usb ( http://pastebin.com/x5QzB18E ), systemd-nspawn --bind-ro=/usr/bin/qemu-arm-static Apr 20 13:02:54 binfmt_misc takes care of the rest Apr 20 13:03:43 with a bit more creativity I can even "boot" the system, although this requires binding some more stuff from the host system which isn't supported by qemu-user, such as systemd Apr 20 13:03:53 (qemu-user doesn't support netlink sockets) Apr 20 13:04:07 gives a really weird-looking uname -a Apr 20 13:04:23 Oh. fancy. Apr 20 13:04:29 Thanks! Apr 20 13:04:38 I still have on my to-do list to augment this with a cross-compiler that pretends to be the native compiler Apr 20 13:05:13 then you get most of the performance of cross-compilation, but in an environment where e.g. configure scripts will work correctly Apr 20 13:16:26 afk Apr 20 13:57:27 zmatt: could you point me to any *recent* xenomai solution for BBB? Apr 20 14:55:26 I'm trying to get usb networking working with an arm board and a usb tethered phone. I'm building the kernel with the usb gadget drivers and usb network peripherals. However, when I do lsub -t my device associated with each port shows driver=. ie. blank. When I modprobe g_ether I get a dmesg about unknown UDC and on the console could not insert 'g_ether" no such device. Any help is much appreciated. Apr 20 15:15:33 davis: wait what, you have a phone connected? so what is that setup actually Apr 20 15:27:57 thank god, it is rndis and activesync Apr 20 17:17:45 I have a quick question. Is there any reasons why the BeagleBone Black RevC is out of stock almost everywhere? Apr 20 17:20:17 Popularity? Apr 20 17:22:41 Guest47449: look for the Element14 version. that should have decent supply Apr 20 17:22:52 also the beaglebone green is an alternative Apr 20 17:25:52 Yes, I found that Element14 has BBB in stock. So, the BBB will not be produced anymore? The replacement product is the BB Green? Apr 20 17:30:00 wrong Apr 20 17:30:41 there are at least 2 companies producing the BBB. one of them is E14. the other is circuitco. Apr 20 17:30:59 if then it might be, again, that circuitco has supply problems Apr 20 17:31:10 E14 supply seems to have always been quite stable Apr 20 17:31:33 Perfect, thank you very much Apr 20 17:44:31 Cco can't handle high volume. . where ele14 can Apr 20 17:45:09 it would be interesting to know what numbers ele14 have produced actuall Apr 20 17:47:26 As long as it is still in production, it is perfect with me. I just do not want to start a project with a discontinued product. Apr 20 17:48:06 I don't think there's any plans at all to EOL it Apr 20 17:48:34 for as long as the am335x is in production Apr 20 17:48:55 Thanks Apr 20 17:48:57 it is trivial enough to run your own build Apr 20 17:49:16 For large values of "trivial". Apr 20 17:52:25 it is not that hard Apr 20 18:01:40 I wouldn't hot-air rework a bga .. I just wouldn't Apr 20 18:02:20 its not cost-effective for me to organise a board run of less than ~1000 boards .. that's not 'small' and what would I do with the other 995 Apr 20 18:07:29 why not hot air rework a BGA? Apr 20 18:07:50 < 10 boards seems viable Apr 20 18:07:55 because, No. Apr 20 18:08:12 that's what the CMs use Apr 20 18:08:17 tqfp is bad enough .. and undoing a bridge .. ick. Apr 20 18:08:41 if the board is layed out, it should just work Apr 20 18:08:43 I've never successfullly unbridged a bga yet :) Apr 20 18:08:59 but then I don't have a solder-paste printer Apr 20 18:09:02 bridging isn't that big of an issue; again - layout Apr 20 18:09:14 stencils are cheap (<$10) Apr 20 18:09:25 not here Apr 20 18:09:32 do you want a customized board in quantities of < 100? Apr 20 18:09:37 nope :) Apr 20 18:09:49 veremit: worse case - vinyl cutter on mylar Apr 20 18:09:57 interesting Apr 20 18:10:23 if all goes well, I should have a process for low quantity builds Apr 20 18:10:42 it just isn't that hard Apr 20 18:11:13 now a BBC/BBX is a totally different story Apr 20 18:11:26 I think if you get the right kit .. its not too bad .. but capital cost usually prohibitive for small volume .. unless you're lucky with second-hand Apr 20 18:11:48 or you've got plenty of time spare :D Apr 20 18:11:49 what capital costs are you thinking of? Apr 20 18:12:06 machines Apr 20 18:12:10 reflow ovens Apr 20 18:12:17 wave soldering Apr 20 18:12:20 reflow ovens aren't that expensive Apr 20 18:12:25 pick-n-place Apr 20 18:12:39 can't wave a BGA; and PnP is overkill for small volumes Apr 20 18:12:58 basic ones .. depends what you consider 'expensive' and whether you've got a surplus of ex-pcb house equipment in garage sales :D Apr 20 18:13:08 no, you reflow bga's -sigh- Apr 20 18:13:27 pnp is much easier than hand-place for anything >10 boards. Apr 20 18:13:43 if you get the PCB right, hot can reliably do BGAs Apr 20 18:13:56 I have hot air'ed 0.4mm QFNs w/o issue Apr 20 18:14:06 how many pins? Apr 20 18:14:26 if you got solder mask .. probably ok .. no good on a home-brew pcb Apr 20 18:14:29 veremit: how many are you talking about? I was looking at the < 10 case... with > 10, I'd bring in a CM Apr 20 18:14:37 48 IIRC Apr 20 18:14:45 ah not many Apr 20 18:15:01 blah...that's what I mean by properly designed PCB Apr 20 18:15:02 cm? Apr 20 18:15:08 contract manufacture Apr 20 18:15:36 lots of designs need tqfp with 64+ pins .. sometimes you just can't get anything smaller .. or don't want it Apr 20 18:16:01 tqfp is harder to deal with in a lot of ways Apr 20 18:17:33 it is a pain to make sure the pesky pins on a tqfp stay straight and flat Apr 20 18:18:00 in contrast, the QFNs just get sucked in unless you screwed up the foot print Apr 20 18:18:06 if they come in a tray and nobody's butchered them .. lol Apr 20 18:18:26 they're all good if you have the right paste and temp profile Apr 20 18:19:21 for me, the TQFPs tend to jump to be one off and don't selfcenter as nicely as QFNs *shrug* Apr 20 18:19:52 w/QFNs, even wire solder works (tin, add a ton of flux, and reflow) Apr 20 18:20:11 but this is getting OT. Apr 20 18:22:45 but should know more of the viability of low volume custom boards in a month or two Apr 20 18:24:28 good luck :) Apr 20 18:27:56 veremit: I'd expect the am335x to be around for a while... industrial users tend to frown on "end of life" Apr 20 18:28:16 zmatt: indeed Apr 20 18:28:30 unless they get a year/two's notice .. then they can 'design' out Apr 20 18:28:59 we still have boards in production that were designed in 2002 Apr 20 18:31:36 with a whole megabyte of external SRAM ! Apr 20 18:31:40 XD Apr 20 18:32:28 zmatt: a data point for you - I have heard some comments suggesting the power sequencing issue with the battery + regulator in shutdown mode is a non issue Apr 20 18:33:57 ds2: I have extensively tested and seen with my own eyes that it very much is an issue Apr 20 18:34:08 got a metric fuckton of scope pics to go along with it Apr 20 18:34:30 zmatt: no disagreeing with you; just a data point. I am designing with the assumption you are right Apr 20 18:34:55 now if you have a ton of dead boards to go with it ;) Apr 20 18:35:14 one of them showing something disturbing that may have been a latch-up (fortunately limited by the fact that it occurred on the I/O that was only its only source of supply power) Apr 20 18:36:36 ds2: it's quite possible that if not too much current is flowing that, at least on short term, nothing bad will happen apart from draining your battery Apr 20 18:38:28 zmatt: I wonder if there is some unpublish spec Apr 20 18:39:24 but even with no external connections, the various pull-up resistors on the BBB itself inject sufficient current to cause the 3V3A to converge to about 1.3-1.4V, despite active discharging of that rail by the PMIC Apr 20 18:39:53 unpublished spec of... ? Apr 20 18:39:58 the am335x Apr 20 18:40:09 about what? Apr 20 18:40:28 how much current can be driven into the am335x w/o powering it Apr 20 18:42:05 45 mA or the amount of current that causes the voltage to exceed abs max limits, whichever is less Apr 20 18:42:45 and no 3.3V supply may ever exceed any 1.8V supply by more than 2 V Apr 20 18:42:46 that abs number may not be firm or something along those lines Apr 20 18:43:46 they are the specs... of course you can often push an individual chip beyond those boundaries; but you lose any guarantees Apr 20 18:45:02 it makes a difference whether you want to push one chip to its limits, or want to produce 10000 boards and not see some significant percentage RMA'd Apr 20 18:46:09 especially since TI states an estimated 100000 operating hours (= a little over 10 years of 24/7 operation) Apr 20 18:46:18 iirc Apr 20 18:46:38 *shrug* Apr 20 18:46:55 overvoltage is cumulative Apr 20 18:47:00 so it may seem to be okay at first Apr 20 18:48:58 note also that without 3.3V supplies available, the AM335x I/Os are _not_ 3.3V-compatible Apr 20 18:49:32 think that is assumed Apr 20 18:50:07 the process used tolerates abs max 2V across a transistor... to make 3.3V I/O they have to do creative stuff with cascaded transistors; that mechanism needs a bias voltage ("PBIAS") to operate correctly Apr 20 18:50:41 i.e. without supply that mechanism won't be functional Apr 20 18:51:06 I wonder if all these problems would go away with a 1.8V IO setup Apr 20 18:51:11 most of them yes Apr 20 18:51:24 the am335x really prefers 1.8V I/O Apr 20 18:51:58 that's also evident from the electical specs... e.g. the specification for the internal pull-up current is hilarious for 3.3V I/Os Apr 20 18:51:59 know if you lose USB w/a 1.8V IO? Apr 20 18:52:07 USB has a separate supply Apr 20 18:52:24 (two actually, one at 1.8V for the internal logic and 3.3V for the I/O) Apr 20 18:53:42 yikes more voltages :( Apr 20 18:53:44 the general-purpose IO is also partitioned into 6 voltage domains; you can choose the voltage (1.8V or 3.3V) independently for each of the six domains Apr 20 18:54:07 e.g. to make 3.3V for the SD card Apr 20 18:54:41 if you look at the first tab of my spreadsheet the voltage domain is indicated for each pin Apr 20 18:55:22 and lol, have you never seen how many different supplies a SoC like the am335x has? Apr 20 18:55:36 on a tangential topic... know of any am335x pmic-LESS designs? Apr 20 18:56:04 you need a pmic of some sort, although there are alternatives to the tps65217 Apr 20 18:56:21 TI currently lists three PMICs compatiblew with the am335x Apr 20 18:56:50 but you need to make a lot of different supplies, and sequence them in the right order Apr 20 18:56:53 PMICs bring sequencing and multi voltages to the table Apr 20 18:57:11 exactly Apr 20 18:57:13 sequencing can be handled in other ways (say a MSP430) Apr 20 18:57:21 and voltage adjustment Apr 20 18:57:36 that's only if you want the SR stuff Apr 20 18:57:41 no Apr 20 18:58:00 what must have voltage adjustment besides SR? Apr 20 18:58:04 it's needed for cpu frequency scaling Apr 20 18:58:42 there is a subset that works fine in a fixed frequency Apr 20 18:59:03 well, not strictly needed... I guess you could just fix vdd_mpu at 1.35 V and ignore the wasted power Apr 20 18:59:17 or fix at 1.1 V and limit yourself to 500 MHz Apr 20 18:59:29 the smaller package forces some of the voltages to be fixed Apr 20 18:59:39 yes, it ties core to mpu Apr 20 18:59:45 which limits it to 1.1V Apr 20 18:59:48 (hence 500 MHz) Apr 20 18:59:53 *nod* Apr 20 19:00:06 it can still be scaled down to 0.95 V e.g. in suspend Apr 20 19:00:13 (and is) Apr 20 19:00:15 toying with the idea of slaving the AM335x to a much lower power MCU Apr 20 19:00:34 you'd need a lot more components to replace the pmic though, why would you want to? Apr 20 19:00:45 flexibility Apr 20 19:01:37 you need to make 1.1V (core/mpu), 1.35V (ram), 1.8V (a lot of stuff), 3.3V (usb and 3.3V I/Os) Apr 20 19:01:48 note btw that my pins spreadsheet actually has a Supply tab with that info Apr 20 19:02:18 thought there is a way to run either DDR3 or DDR3L at 1.8V? Apr 20 19:02:20 you could do without the 1.35V if you use mDDR at 1.8V Apr 20 19:02:58 DDR2 is also 1.8V Apr 20 19:03:09 so that boils down to about 3 voltages or less Apr 20 19:03:35 yeah the am335x is relatively straightforward Apr 20 19:03:44 note that the 1.1V can draw a rather lot of power Apr 20 19:03:56 just an amp or two :D Apr 20 19:04:21 1A actually at 1.1V Apr 20 19:04:55 (again see spreadsheet) Apr 20 19:06:11 it is really too bad that they need to all that "security" crap; would be so nice if that M3 on there was truely and completely uncommitted Apr 20 19:06:22 the M3 is freely programmable Apr 20 19:06:26 its firmware is in a public git repo Apr 20 19:06:56 yes but AFAICT, I cannot have that thing run by itself before the A8, etc Apr 20 19:07:31 that's mostly because it has no boot ROM Apr 20 19:07:40 you can actually make it run first... via JTAG Apr 20 19:07:45 something with the power and flexibility of an external M3 sequencing power Apr 20 19:08:10 might as well have an external M3 Apr 20 19:08:27 it's also not hard to have the A8 just load a tiny program that does nothing but load M3 firmware and transfer control to it Apr 20 19:09:09 a bit of a problem however is that the M3 has no access to the L3 interconnect, only to the L4WKUP Apr 20 19:09:16 that too Apr 20 19:10:53 but to get back to your earlier question... why not have the M3 control the PMIC ? then you already have the switching power supplies needed, correct sequencing, support by existing bootloaders and drivers, and still retain full control Apr 20 19:11:07 especially one of the other two pmics seemed quite suitable for that Apr 20 19:11:20 (the tps65217 kinda sucks) Apr 20 19:11:39 the PMIC combos don't have everything I want Apr 20 19:12:23 also not the TPS65910A31A1 ? Apr 20 19:14:17 I mean, I'm sure you can do without one Apr 20 19:14:37 there is also the problem being able to buy the PMIC Apr 20 19:15:14 https://store.ti.com/AddToCart_TI.aspx?p=TPS65910A31A1RSLT Apr 20 19:15:16 ;) Apr 20 19:15:23 :P Apr 20 19:16:05 but if you want to do it yourself, I say go for it... the sequencing requirements of the am335x are actually quite modest Apr 20 19:16:54 if the atmels can do it... Apr 20 19:17:03 1. don't allow any 3.3V supply to exceed any 1.8V supply by 2V or more at any time during power-up/down Apr 20 19:17:19 2. don't ramp stuff up too fast Apr 20 19:17:42 3. don't allow any 3.3V supply to exceed any 1.8V supply by 2V or more at any time during power-up/down Apr 20 19:17:52 1. - Slave 3.3V enable to 1.8V; 2. Careful choices of caps Apr 20 19:17:57 :) Apr 20 19:18:34 generally the order is: 1.8V up, 3.3V up, core up, wait a bit, release POR Apr 20 19:19:26 bbl lunch Apr 20 19:19:45 actually that's about it Apr 20 19:20:03 1.8 and 3.3 can go at the same time as long as you observe the 2V distance rule Apr 20 19:20:39 ram can also go at the same time as 1.8 Apr 20 19:22:03 max ramp is 10 us/V Apr 20 19:22:32 eh, 0.1 V/us Apr 20 19:23:02 the am335x actually seems to have pretty relaxed rules **** ENDING LOGGING AT Thu Apr 21 02:59:58 2016