**** BEGIN LOGGING AT Fri Nov 01 03:00:13 2019 Nov 01 03:21:58 Hi everyone, Nov 01 03:23:41 I would like to run the closest linux image of beaglebone on a Virtualbox, which one would it be? Nov 01 03:33:59 Rusty68: debian Nov 01 03:57:10 Is there a tool like visual studio I could use to create GUIs? Nov 01 04:00:12 What would be the best alternative to Visual Studio for Linux Debian? Nov 01 04:03:17 that's a really vague and generic question, and not something that's really beaglebone related. qt creator would be one option for example Nov 01 04:04:15 I'm pretty sure you can even run it on windows to create a gui application for the beaglebone Nov 01 04:04:20 Sorry for not being more descriptive, let me explain what I have in mind. Nov 01 04:06:01 I am building a portable pet patient monitor. I have a bunch of sensors that acquired, processed on a micro. Nov 01 04:06:50 The micro send serial data (UART) to the beaglebone which then renders it on a 800x600 display. Nov 01 04:08:05 In the meantime, the product sends packets over BTLE to a computer that also display the data. I used Visual Studio for the Windows application. I am hoping I could use something similar (high-level) to create a GUI on Debian. Nov 01 04:08:56 I have ordered the beaglebone but it will not show up for at least one week. In the meantime, I installed debian on VirtualBox with the intent to start creating the GUI. Nov 01 04:09:47 Now I am wondering what IDE I could use to create that GUI, it will have a few real-time charts, varying text boxes. Nov 01 04:09:48 I mean, that doesn't really change your question nor my answer Nov 01 04:10:26 qt creator is the first thing that comes to my mind, but I'm sure there are alternatives Nov 01 04:10:28 "I'm pretty sure you can even run it on windows to create a gui application for the beaglebone" what do you mean? Nov 01 04:10:40 what's unclear about that? Nov 01 04:10:56 you can run the IDE on windows to build windows applications but also to build embedded linux applications Nov 01 04:10:57 Well, AFAIK, I can not just push an executable on Linux. Nov 01 04:11:09 qt creator can cross-build applications Nov 01 04:11:37 I have not heard of qt creator. I will give a try. Nov 01 04:13:16 I don't really do any gui building myself, but a colleague of mine uses qt creator on Windows to build applications for the beaglebone... I don't think he ever got cross-building to work, but he just designs and tests the application on Windows and then copies the source code to the beaglebone and compiles it there Nov 01 04:14:07 Compiling on debian makes sense, thanks for the tip. Nov 01 04:15:51 I am an electrical engineer, I am thinking of doing my own pcb board. Nov 01 04:15:56 (we use distcc to speed up the build: distcc allows compilation to be distributed over the network from the beaglebone to a much beefier x86 debian machine where the gcc arm linux is installed) Nov 01 04:16:22 ehh I could have phrased that better Nov 01 04:16:46 My application will be fairly, I would not expect the compilation to take much time. Nov 01 04:16:53 'fairly simple' Nov 01 04:17:40 compiling stuff on the beaglebone is pretty slow though Nov 01 04:18:34 Ahah... Would that work if I compile on a debian virtual machine on my desktop? Nov 01 04:21:11 sure, why not. if you use the same debian release (e.g. stretch if you're running stretch on the beaglebone) and install the g++-arm-linux-gnueabihf package on it then you should have the same compiler available on beaglebone and in the debian vm hence you can use distcc Nov 01 04:21:39 I don't really have a guide for setting it up, just a short doc that shows my setup => https://pastebin.com/sYYjHghJ Nov 01 04:22:09 (I'm using debian buster on both beaglebone and the server where I'm running distcc) Nov 01 04:22:14 Thanks a lot for the link Nov 01 04:23:16 btw, why use a microcontroller? can't the beaglebone directly interface with the sensors? Nov 01 04:23:35 You really want to know the whole story eheh! Nov 01 04:23:54 now I'm scared :P Nov 01 04:24:04 One of my colleague accepted a job he cannot fulfill. Nov 01 04:24:46 He passed me the hot potato 3 weeks begging I work on it, while I was moving to a new place.... Nov 01 04:24:55 lol Nov 01 04:25:01 sounds like fun Nov 01 04:25:10 He decided to use an stm32 nucleo board if you know about it. Nov 01 04:25:41 I've worked with stm32 microcontrollers, I'm not sure if I've heard of the nucleo board before.. maybe Nov 01 04:25:47 It has a 2.4" LCd display. I was stuck with it because he promised the company it would work. Nov 01 04:25:56 I finally gave up and told him I would do it my way. Nov 01 04:26:45 The Nucleo is just a dev kit of a given stm32 micro. Nov 01 04:27:04 oh yeah I've seen an stm32 demo board with a GUI... it looked pretty nice, and of course no boot time whatsoever Nov 01 04:27:10 At the moment, I am using the stm32 as a sensor hub, nothing more. Nov 01 04:27:51 what sort of sensors? Nov 01 04:28:21 There is an auto GUI generator for the stm called TouchGFX but it is really difficult to integrate it with a project generated by the STM32CUBE. Nov 01 04:28:42 so the problem is basically he couldn't figure out how to build a gui on that thing, and now you're just going to use a BBB as the GUI for this thing Nov 01 04:28:47 yep Nov 01 04:29:08 So I got frustrated wasting tens of hours with the stm32, and told my colleague I would like to use a beaglebone and write up a GUI. Nov 01 04:29:24 Yeah, that is the story. Nov 01 04:29:41 not very elegant, you'd probably want to use *either* a beaglebone *or* the stm32, not both, but yeah that's just the legacy of the project so far ;) Nov 01 04:29:57 I am an electrical engineer. I am good at bare metal (fpga, assembly, c) and good at high-level (c#) but really bad at RTOS. Nov 01 04:30:36 heh, maybe you'll appreciate this little thing I once made: https://github.com/mvduin/bbb-asm-demo Nov 01 04:30:38 I have to do a demo in two weeks to the customer. My number priority is to have something. Nov 01 04:30:46 The sensor hub works, I just need the display to work. Nov 01 04:31:00 'number one' Nov 01 04:31:32 In the end, the stm32 will be taken off the equation, the beaglebone will do it all. Nov 01 04:31:35 unfortunately gui programming is not trivial on any platform, especially if you still have to learn a toolkit and IDE Nov 01 04:32:41 I already generated the GUI on Visual Studio for Windows, I was hoping I could use a tool such as Wine or Mono to run a windows executable ahah, is that possible? Nov 01 04:34:05 https://www.mono-project.com/docs/gui/winforms/ Nov 01 04:34:28 also yuck :P Nov 01 04:34:46 Ahah. Nov 01 04:35:14 What about python on Linux? Can I generate simple GUIs? Nov 01 04:35:42 dunno, have you google "python gui" or something like that? Nov 01 04:38:24 Google is good at providing neutral information. It is hard to rank things on google. Nov 01 04:38:40 I was hopping I could get some feedbacks from real people. Nov 01 04:40:35 I don't develop GUI applications at all, at the only GUI project on a beaglebone I've had any involvement in was a Qt5 application (as mentioned earlier developed by a colleague on Windows) running directly on the framebuffer (using eglfs), no X11 Nov 01 04:41:24 By frame buffer, you mean the graphic card memory buffer? Nov 01 04:41:40 This is wild! Nov 01 04:41:41 lol, "graphic card" Nov 01 04:41:51 I am an EE! Nov 01 04:42:03 I do my best at understanding. Nov 01 04:43:15 the AM335x has a very simplistic display controller (lcdc) which basically just reads a framebuffer in memory and shoves the data out a parallel lcd interface (with configured timing) Nov 01 04:44:24 I see. I have written a few simple app using direct draw when I was a teenager. Nov 01 04:44:41 I should be able to create my own graphics for charts, text boxes, etc. Nov 01 04:45:03 (it also has a GPU but that's entirely unrelated: the GPU renders OpenGL ES/ES2 graphics into a framebuffer in memory, the display controller sends a framebuffer in memory to the LCD panel or HDMI framer, the kernel and userspace are responsible for gluing these together. the GPU is not supported when using X11 Nov 01 04:47:43 yeah a framebuffer is just a 2D array of pixels you can map into the address space of your program (if you have suitable privileges) and directly write to. the legacy fb API uses a single fixed framebuffer, the newer drm API lets you allocate and map any number of framebuffers and submit one to be displayed on screen (which is how you achieve tear-free video/animation) Nov 01 04:48:24 That sounds like double buffering. Nov 01 04:48:34 yep, double, triple, whatever you want Nov 01 04:49:06 the kernel driver will basically just want until vsync and then quickly update the framebuffer pointer in the display controller Nov 01 04:49:12 Like I said earlier, I like low-level and high-level programming, but nothing in between. I would be ok with working with the display memory buffer. Nov 01 04:49:47 Plus, this is the most efficient way, no cpu waste.... Nov 01 04:50:00 That brings me to my second question. Nov 01 04:50:29 yeah if you want a single-window-fullscreen application anyway (which is typical) then using a windowing system like X11 is kinda silly Nov 01 04:51:11 I will build my own pcb and I noticed that Texas Insutrments has hundreds of Sitara processors from a couple of dollars to almost $100. Nov 01 04:52:16 making a nice example of how to use the modern parts of the drm api (ignoring the hideous legacy stuff) to configure the display and display graphics is still on my to-do list... I have small tests, but nothing I'd want to show Nov 01 04:52:30 uhuh, yes Nov 01 04:53:27 My application will be fairly simple and I am hoping I could optimize debian so that I can pick a lower cost processor and less ddr memory. Nov 01 04:53:59 even lower cost than the am335x ? Nov 01 04:54:22 One of the AM335x runs at 400MHz and only costs $5, I suppose that could run debian, right? Nov 01 04:55:05 https://www.digikey.com/product-detail/en/texas-instruments/AM3351BZCE30R/296-47425-2-ND/7707709 Nov 01 04:55:10 sure, just make sure it has the feature you need. and you can probably force a beaglebone to 400 MHz to simulate the performance Nov 01 04:55:17 Oops 300MHz, Nov 01 04:55:27 I was about to say, 400 MHz sounds very unfamiliar Nov 01 04:55:51 'and you can probably force a beaglebone to 400 MHz to simulate the performance' Great I was hoping I could emulate. Nov 01 04:55:59 anyway, you don't have a working project yet, isn't it a bit early to optimize for high-volume production :) Nov 01 04:56:40 Like I said earlier, my colleague promised them a Bom cost which I cannot meet with the processor on the beaglebone. Nov 01 04:57:25 I need to show them I have a path to meet their requirements. Nov 01 04:57:48 I mean there's a trade-off between component cost and development time. stm32 is cheaper, but harder to develop a gui for Nov 01 04:58:01 lower processor might mean more time needs to be spent on optimization Nov 01 04:58:02 etc Nov 01 04:58:04 *slower Nov 01 04:58:56 I understand. Nov 01 04:59:25 They want a low recurring cost, even it means high non-recurring costs. And this is good for me! Nov 01 05:00:08 I wish I could share a picture to show you how simple my gui is. Nov 01 05:00:08 yeah then I guess you'll just be using the BBB to prototype for a custom board based on the cheapest am335x variant that suffices for your needs Nov 01 05:00:28 hope you have experience with pcb layout for ddr3 memory ;) Nov 01 05:01:07 Yeah, I work on high-speed board all the time. Nov 01 05:01:20 PCIe, MIPI, DDR4. Nov 01 05:01:31 ok yeah that makes a custom design much more attractive Nov 01 05:02:04 maybe with a bit of luck you don't need to cram so much in such a small space as the BBB does Nov 01 05:02:06 The electrical side is the fun part. Nov 01 05:02:37 I will have the processor, a ddr memory, a flash and a few bucks. Nov 01 05:02:52 and you'll have an opportunity to fix known issues, like the ethernet phy not working on a small percentage of power-ups ;) Nov 01 05:03:08 Ahah, I don't need ethernet! Nov 01 05:03:14 problem solved! Nov 01 05:03:18 flash? eMMC? Nov 01 05:03:34 All I need is 3x uart, ddr, and an LCD. Nov 01 05:03:40 and a flash. Nov 01 05:03:59 Maybe an ADC over SPI. Nov 01 05:04:09 don't forget the jtag port Nov 01 05:04:43 Good point/ Nov 01 05:05:32 also you may want to consider the identification eeprom... it's not strictly required, but it can simplify things by allowing u-boot to detect the board (it's why the same image works on so many beaglebone variants) Nov 01 05:07:03 Yeah, I've seen the EEPROM on the schematic. Nov 01 05:07:43 how will the device be (re)programmed ? you'll either need a way to boot from something external (e.g. SD or network) or a way to directly access the eMMC (I think using raw nand flash is pretty rare nowadays since it's just obnoxious to use) Nov 01 05:07:53 See link to a picture of the GUI Nov 01 05:07:54 https://ibb.co/sb1Q59r Nov 01 05:08:10 that seems super simple Nov 01 05:08:28 this doesn't look like I'd use a GUI toolkit for it at all Nov 01 05:08:40 "how will the device be (re)programmed " Should I use the ethernet for that? Nov 01 05:09:11 "that seems super simple" There will be a menu and a few more things (fitler, analysis, etc.), but nothing crazy. Nov 01 05:09:50 and that's up to you.. using ethernet means getting netbooting working :) Nov 01 05:10:04 using an SD card is simple but awkward for large volume Nov 01 05:10:14 What is the easiest way to program the flash then? Nov 01 05:10:37 for mass production you can pre-program eMMCs before soldering, but that doesn't solve how you program your prorotypes or how you reprogram a device when needed Nov 01 05:11:29 I don't have to think about that for now. Nov 01 05:11:30 I think in principle you could even use jtag or uart or that's going to take a while ;) Nov 01 05:11:52 "this doesn't look like I'd use a GUI toolkit for it at all" => Good to hear that, what would you use instead? Nov 01 05:13:37 What about using USB2.0 to program the flash? Nov 01 05:14:12 dunno, maybe something like SDL? or clutter? or just fully custom using some library to draw text? (I'm pretty confident I can draw lines without a library :P ) Nov 01 05:14:29 like I said, I don't really do gui programming Nov 01 05:14:47 if I did, I'd probably just use whatever I happen to be comfortable with Nov 01 05:16:06 anyway, yeah don't forget jtag and a debug serial port, otherwise you're going to have a problem when u-boot or the kernel crashes and you have no idea why ;) Nov 01 05:16:48 Ahah, I will keep in mind. I will user the bb schematic as a reference and won't remove things that are free! Nov 01 05:17:13 Can the USB2.0 be used to program the flash? Nov 01 05:18:21 it can boot from the usb device port by presenting itself as an RNDIS (usb networking) device to the host and then performing netbooting over that (equivalent to netbooting via ethernet) Nov 01 05:21:21 Thanks for all the information, this is really appreciated. Nov 01 05:21:33 Is there a way to save that conversation in a nice format? Nov 01 05:22:10 uhh that depends on your irc client I guess Nov 01 05:23:26 I see. Nov 01 05:23:31 mine logs to plain text, and there are also online logs in plaintext: http://logs.nslu2-linux.org/livelogs/beagle/beagle.20191101.txt (our conversation should be part of the 20191102 log once it's up) Nov 01 05:23:47 making it "nice" is left as an exercise for the reader ;) Nov 01 05:25:08 Ahah, that is good enough. The information is not lost. Nov 01 06:39:07 nucleo is their next gen of discovery Nov 01 11:27:24 I'm tring to set up a BBB to do PRU stuff with a handmade cape. The dts file is at https://pastebin.com/7Azc8ZTu - can anyone see what's wrong? The uEnv.txt file is at https://pastebin.com/g55HeEbW//pastebin.com/g55HeEbWttps://pastebin.com/g55HeEbWhttps://pastebin.com/g55HeEbW Nov 01 11:55:55 hi Nov 01 14:19:01 I'm tring to set up BBB with the new uEnv.txt but can't get the PRU to work. Nov 01 14:23:44 The .dts file is at https://pastebin.com/7Azc8ZTu, the uEnv.txt at https://pastebin.com/g55HeEb https://pastebin.com/g55HeEbW Nov 01 15:19:58 Hi there.. is beaglebone AI at any store at Europe now? preferable Sweden, France, or Germany? Nov 01 15:52:37 WhileTrue: store? As you go there to pick it up? Nov 01 16:26:40 I'm hoping for some help with recovering my BBB. Board tries to boot the EMMC flasher, but will not successfully copy an image over. Board initiates that without button press. I'd like to be able to completely clear the EMMC. I am connected to the BBB via Serial. Nov 01 17:16:41 tony_: well erasing the eMMC is the first step performed by the flasher Nov 01 17:17:14 tony_: can you share the error it's producing on the serial console? (copy-paste it to a paste service like pastebin.com) Nov 01 17:39:15 zmatt: Here is the debug serial output. This one I pressed the flash button and have a 2Gb image on the sdcard with Debian 9.11 Nov 01 17:39:28 zmatt: https://pastebin.com/yr8YrAcc Nov 01 17:40:50 zmatt: Just to note that whether or not I press the flash button, it actually runs through this. Nov 01 17:42:07 there is no "flash button" Nov 01 17:42:21 if you boot a flasher card, it will flash eMMC Nov 01 17:43:12 the SD boot button is only needed in rare cases to force bootrom to load the bootloader (u-boot) from SD card instead of from eMMC, e.g. if the one on eMMC is corrupted or incompatible with the system on SD card Nov 01 17:43:35 and it looks like the image you're trying to flash onto eMMC doesn't fit Nov 01 17:44:18 zmatt: Yes. it is only a 2Gb Debian image. I also tried an older small Angstrom image. Same issue. Nov 01 17:45:12 zmatt: is there a really tiny image out there I can try to flash on it? Nov 01 17:45:25 zmatt: something less than 1Gb? Nov 01 17:45:32 tony_: rsync: write failed on "/tmp/rootfs/usr/share/vim/vim80/syntax/csdl.vim": No space left on device (28) Nov 01 17:45:33 sure, a console image Nov 01 17:46:02 tony_: looks like we blew thru the 2GB size again.. just use a microsd ;) Nov 01 17:46:06 rcn-ee[m]: recently someone also ran into out-of-space while trying to flash a 4GB lxqt image onto a 4GB eMMC beaglebone Nov 01 17:46:24 or, well, it didn't run out of space while flashing Nov 01 17:46:28 zmatt: yeap that's the BeagleBoard.org default image.. Nov 01 17:46:31 but it left so little free space that X11 failed to boot Nov 01 17:46:38 it was fixed a on Sep 25th.. :( Nov 01 17:46:41 ok Nov 01 17:47:16 ti-opencl/ti-tidl bloated the image, that's why it has it's own now.. the base lxqt doesn't have ti-opencl anymore.. Nov 01 17:47:20 None of my images have X11. I just need basic cli and ssh. Nov 01 17:47:39 tony_: then use a console image Nov 01 17:48:09 tony_: use this one: https://rcn-ee.net/rootfs/bb.org/testing/2019-10-01/stretch-console/bone-eMMC-flasher-debian-9.11-console-armhf-2019-10-01-1gb.img.xz Nov 01 17:48:13 https://elinux.org/Beagleboard:BeagleBoneBlack_Debian has a debian stretch console image and a debian buster console image Nov 01 17:48:43 . Nov 01 17:49:50 zmatt: what is the diff between the IoT image and the console image? (the first one I tried was the 9.5 IoT image Nov 01 17:49:59 rcn-ee[m]: of course it doesn't help that "2GB" is only 1.79G and "4GB" was originally 3.66G, then 3.60G, then 3.56G Nov 01 17:50:48 tony_: iot image lacks X11 but otherwise has everything and the kitchen sink Nov 01 17:51:03 webserver with cloud9 web-based ide, node-red, etc Nov 01 17:51:29 zmatt: got it. This is just going to be a remotely deployed honeypot, I don't need most things... Nov 01 17:51:39 then start with the console image Nov 01 17:51:59 zmatt: thanks, will give it a shot. Nov 01 17:52:58 Humpelstilzchen: sure.. do you know any store? Nov 01 18:02:35 Has anyone successfully installed machinekit on to beaglebone ai ? Nov 01 18:06:05 anoop: well the machinekit image's are "am3" only.. Nov 01 18:08:53 that's an unfortunate abbreviation for the am335x considering the am35xx also exists Nov 01 18:10:05 Well that's covered: "(AM335x based BeagleBone's and PocketBeagle)" : https://elinux.org/Beagleboard:BeagleBoneBlack_Debian#BBW.2FBBB_.28All_Revs.29_Machinekit Nov 01 18:10:30 zmatt: The console image worked. Thanks! Nov 01 18:10:56 on the machinekit website it does appear machinekit can run on basically any debian system? Nov 01 18:11:38 and our machinekit image we generate, has no maintainer anymore.. Nov 01 18:46:30 argh, using the blkio cgroup to account for which services are responsible for writing how much data to eMMC turns out to be basically useless... unless you use O_DIRECT nearly all writes are accounted to the kernel itself Nov 01 18:54:01 trying to figure out how another eMMC managed to wear out even though there shouldn't be any significant amount of eMMC writes and the eMMC is even configured in SLC mode :( Nov 01 19:06:29 zmatt I presume you consulted the data sheet first? what FS are you using on it? Nov 01 19:14:51 ext4, what data sheet? Nov 01 19:17:19 you mean the eMMC? it doesn't really have one, just a one-page flyer that I'm not going to call a datasheet Nov 01 20:07:17 rah: hmm, so with just a reset extender I got 9/344 fails, with reset extender and 1K pull-up I got 6/689 fails... so that might imply that the slow rise time (caused by the excessive capacitance on the reset line) is part of the problem (which was already a suspect since someone reported success with removing the caps), although I'm not sure if the difference is statistically significant, I'd need to ... Nov 01 20:07:23 ...look up the math for that Nov 01 20:08:36 I'm now testing a bit more without pull-up (the testing is automatic but not very fast, about 10s per attempt) Nov 01 20:11:21 if the slow rise time is indeed the issue, I also wonder if maybe the AM335x is released from reset much earlier than the phy, although I don't think that should cause any problems even if it is the case Nov 01 20:26:42 Is it possible to reliably use rs485 with BBB? Nov 01 20:26:53 uhh, sure? Nov 01 20:26:57 what do you mean "reliably" ? Nov 01 20:28:22 My (insufficient) understanding that rs485 is half-duplex and requires some reliable real-time magic to turn the connection around. Nov 01 20:28:47 So it is more than converting signals to differential signls Nov 01 20:29:06 you'll need an rs485 transceiver of course Nov 01 20:29:53 whether kernel-managed driver-enable (via the rts pin) is adequate depends on the turn-around time I guess (i.e. how soon after transmission you need to disable the driver and start listening) Nov 01 20:30:48 zmat, what would be an example of the tranceiver I can buy? Nov 01 20:30:58 using an RT kernel with properly configured priorities would also help to reduce the risk of the kernel not reacting to end of transmission fast enough (e.g. due to another interrupt handler) Nov 01 20:31:22 zmatt: I'm going to use a software workaround of rebooting the machine until the Ethernet works :-) Nov 01 20:31:29 if you need really tight and perfectly guaranteed turnaround timing, use PRU Nov 01 20:31:35 rah: reboot does nothing Nov 01 20:31:59 rah: the reset line needs to be pulled low for >= 25 ms Nov 01 20:32:43 no sorry, for >= 100 μs but after the supplies have been stable (> 80% of nominal) for at least 25 ms before reset deassertion Nov 01 20:32:49 rah: rma it and swap in a replacement.. it doesn't affect 100% of boards.. Nov 01 20:32:52 zmatt: https://github.com/bigjosh/bbbphyfix Nov 01 20:33:10 rah: that "fix" is a recipe for destroying your hardware Nov 01 20:33:13 rcn-ee[m]: I bought it nearly four years ago, it's out of warranty Nov 01 20:33:20 (although it'll depend on external connections) Nov 01 20:33:35 zmatt: how is that a recipe for desroying my hardware Nov 01 20:33:35 ? Nov 01 20:34:31 what harm would it do? Nov 01 20:34:33 zmat, I am interfacing with a 3-rd party device (small pump). I don't think that they ever published this turn-around time. Getting to PRU would be above my head, especially that I even do not know this turnaround time. Nov 01 20:34:58 rah: the SYS_5V and 3V3B supplies (which are what's available on the P9 header for external electronics, and the 3V3B is also used for various on-board components) will remain active while nearly all the supplies to the AM335x, including the 3.3V I/O supplies, are cut for a second Nov 01 20:35:24 rah: this create a situation that's at high risk of driving 3.3V signals into the unpowered I/O cells of the AM335x Nov 01 20:36:10 zmatt: you're saying rebooting the machine risks killing it? Nov 01 20:36:11 e.g. if you'd connect a serial console cable you'll end up powering the VDD_3V3B net via the protection diode of the UART0_RXD pin of the processor :P Nov 01 20:36:18 rah: he's not "rebooting" the machine Nov 01 20:36:38 rah: he's doing a really obscure trick to sort of power-cycle the system using test-mode registers of the PMIC Nov 01 20:36:53 O_o Nov 01 20:36:54 is he? Nov 01 20:37:13 (it's not really a power-cycle, it temporarily enters "RTC only" mode, which has been defeatured by TI because of unspecified problems) Nov 01 20:37:25 zmatt: it runs a program, /usr/sbin/bbb-long-reset Nov 01 20:37:53 you're saying this program isn't safe? Nov 01 20:39:06 that's exactly what I'm saying and I also explained why :P Nov 01 20:39:43 well, it could just as easily run "reboot" Nov 01 20:39:46 note that it's a shell script that pokes a PMIC register using i2cset and then uses /dev/mem to poke RTC registers Nov 01 20:39:49 no it couldn't Nov 01 20:39:53 because reboot has zero effect on the phy Nov 01 20:40:09 :-/ Nov 01 20:40:48 the AM335x does try to drive reset low (briefly) during a warm reset (i.e. reboot), however the capacitance on the reset line is so big that it doesn't make any sort of dent in it Nov 01 20:41:32 (which also means that driver basically experiences a brief short-circuit every time you reboot, yay :P ) Nov 01 20:41:36 this is no good :-( Nov 01 20:41:57 Zmat, so is there a transceiver something similar to CAN https://www.amazon.com/SN65HVD230-CAN-Board-Communication-Development/dp/B00KM6XMXO/ref=sr_1_1?keywords=SN65HVD230+CAN&qid=1572640824&sr=8-1 Nov 01 20:42:06 dreamhiker-2: CAN is quite different Nov 01 20:42:12 I know Nov 01 20:42:53 I meant a similar conceptually device that I plug to uart pins (instead of can) and get RS485 Nov 01 20:43:00 but there are lots of RS485 transceivers, https://www.analog.com/en/parametricsearch/11052#/p4583=|RS-485 Nov 01 20:43:10 yes Nov 01 20:43:42 3 signals: txd, rxd, and rts (used for driver-enable) Nov 01 20:45:26 Those seems as chips. Is there a small board to it similar to https://www.amazon.com/SN65HVD230-CAN-Board-Communication-Development/dp/B00KM6XMXO/ref=sr_1_1?keywords=SN65HVD230+CAN&qid=1572640824&sr=8-1 Nov 01 20:45:47 rah: I guess one solution might be an external open-drain pulse generator connected to the reset line (capable of discharging the 2.2 uF cap) triggered by a gpio being driven high when the phy problem is detected Nov 01 20:47:29 or maybe a 100 ohm pull-up resistor works, but then there's the potential issue that the reset pulse may be too short (iirc 20ms instead of 25ms) Nov 01 20:48:02 or you could remove the 2.2 uF cap, but that doesn't sound like fun in production Nov 01 20:48:51 brb Nov 01 21:04:42 dreamhiker-2: is google down? Nov 01 21:08:40 dreamhiker-2: https://lmgtfy.com/?s=g&q=rs485+breakout+board Nov 01 21:09:09 (be sure to pick one that uses 3.3V I/O) Nov 01 21:10:17 (or preferably simply powered from 3.3V) Nov 01 21:14:20 Thanks zmatt! Nov 02 00:21:20 rah: with reset extender and no pull-up: 34/1431 failures (2.4%), with reset extender and 1K pull-up: 12/1189 failures (1.0%) Nov 02 00:22:33 maybe I'll try even stronger pull-up and just risk violating the abs max ratings of the reset extender for testing Nov 02 01:02:06 oof, now the reset-extender only manages to drive it down to 0.96 V, while the phy specified a rising-edge threshold of min 0.81 typ 1.65 max 1.90 V ... hopefully it's closer to typ than to min :P Nov 02 01:07:31 oh but I'm actually not even violating its abs max ratings, hurray (4 mA vs abs max 5 mA) Nov 02 01:17:02 (scope pics in case anyone is curious: https://photos.app.goo.gl/jSMLyep3Hohvn3Ro7 ) Nov 02 01:17:11 wonder how more work is it to cut the trace and reroute it to a GPIO Nov 02 01:19:41 looking at the density of traces near the phy that idea sounds distinctly unfun Nov 02 01:34:49 this isn't a BGA so there is always, remove, kaptop tape and tack wire Nov 02 02:13:14 I looked and that meant nothing to me. Nov 02 02:13:38 I am dealing w/ the Replicape people b/c I am switching controlers right now to the Replicape. Yea boy! Nov 02 02:14:28 Most of them are friendly but not heavy on chatting like I thought people would be in the field. Well, it is Fri. Fun Day! Nov 02 02:19:40 Is anyone supporting the replicape here? Nov 02 02:21:27 https://photos.app.goo.gl/1eRbSBBC988TfAxK9 is my issue now. Should I clip the barrel jack and use the power and GND wires w/ the same input from my power supply? Nov 02 02:21:46 Let me draw a schematic. Nov 02 02:35:43 This will take a while. The brain is slow right now. Nov 02 02:41:20 Okay. I think I got it. Please hold. Nov 02 02:41:27 Messy but nice! Nov 02 02:42:16 If this dang phone would send the photo. Nov 02 02:44:42 https://photos.app.goo.gl/FDfqBGf6zHtkBoM38 if that makes sense. R is PWR and B is GND. Nov 02 02:47:35 The right side of the schematic has three or four lines. Those are to signify that that line has been marked out. Nov 02 02:47:48 I did not mean to draw that line. Nov 02 02:58:03 Forget it. I need to find the schematic of the particular older board. **** ENDING LOGGING AT Sat Nov 02 02:59:57 2019