00:27 | sickness | if anyone wants to post here: http://www.solid-run.com/phpbb/viewtopic.php?f=6&t=1253 I would like say that this verbatim 8Gb stick also works fine for u-boot and install http://www.soundbasemegastore.com/27953-34209-home/verbatim-8gb-retractable-usb-flash-drive-blue.jpg |
00:27 | sickness | (I've tried to registered but still waiting for activation mail...) |
00:27 | sickness | tnx |
05:26 | MikeSeth | oy, a newsletter |
10:48 | rabeeh | sickness: activation of what? |
10:51 | _rmk_ | Shawn's taken the hummingboard patches this morning :) |
11:00 | rabeeh | :) |
11:00 | rabeeh | _rmk_: have url? |
11:05 | _rmk_ | not to hand, Shawn replied by email |
11:35 | sickness | rabeeh: forum subscription |
11:35 | rabeeh | sickness: which one? |
11:41 | sickness | phpbb on solid-run.com |
13:40 | skybach | hi |
13:40 | skybach | any of you received the cubox-i already? |
14:17 | mhoney_work | nope |
18:01 | crts1973 | Does anyone have suggestions for getting started with Yocto on the C1 board? Should I be looking at some other build environment? |
18:32 | skybach | hihi |
20:06 | rabeeh | crts1973: jnettlet already demonstrated it; dv_ also has it working |
20:06 | rabeeh | yocto on c1 should be already working (in my mind) |
20:43 | crts1973 | Okay, I am just trying to find the fastest way to get a development environment up and running for Linux. Yocto seemed promising. Any suggestions? |
20:48 | rabeeh | crts1973: there is ubuntu 13.04 floating around |
20:49 | crts1973 | that would be great. With the boundary devices board I just recompiled the kernel on the board. It is fast enough to not have to cross compile on my desktop. |
20:49 | rabeeh | it's working perfectly fine; but the main issue is that there is no a script on how to reproduce the exact binary. this worries me since this is a manually baked rootfs and can't be maintained |
20:50 | rabeeh | hste and jas-hacks_ are masters on ubuntu stuff; but as i mentioned we don't have a clear script that shows where the sources are dragged from and where they are installed |
20:51 | rabeeh | needless to say that it's debootstrapped; then probably apt-get installed some head virtual package; but the gpu stuff is still mysterious. |
20:52 | crts1973 | okay. well I will need the VPU for our application |
20:54 | crts1973 | Maybe I can try to dig more into what was done for the boundary devices builds |
20:55 | hste | there is a armel image witch work with gpu and vpu but thats based on the older 3.0.35_4.0.0 |
20:56 | crts1973 | I am fine if it is a bit outdated. |
20:56 | hste | yes but u need a kernel for it |
20:57 | hste | the newer kernel use aother vivante drivers |
21:01 | crts1973 | hmm okay. Also, Rabeeh is aware of this, but I am trying to test the CSI MIPI camera interface. I had this working on the boundary devices board running Ubuntu. |
21:02 | rabeeh | crts1972: to get the CSI working you don't really need the VPU |
21:02 | crts1973 | That is true |
21:02 | rabeeh | the vpu will be only useful if you want to encode |
21:02 | crts1973 | which I do want to |
21:03 | rabeeh | ok. |
21:03 | crts1973 | so that would be step 2 |
21:03 | rabeeh | ok. so my guess is that you can use the bondary device rootfs; although the gpu won't work for X but you can still encode with the vpu |
21:04 | crts1973 | okay, I will head in that direction. I have to admit that I am a bit of a n00b when it comes to this stuff, I have more of a low-level software/electrical background. |
21:06 | jas-hacks_ | crts1973: what do you plan to use it for? |
21:06 | crts1973 | HD video streaming over a network |
21:07 | crts1973 | I am going to put the SOM into the head of a robotic snake we are developing |
21:08 | crts1973 | (this is non-commercial, I work for a university) |
21:08 | jas-hacks_ | crts1973: ... sounds interesting, so you need rootfs with vpu encoding/decoding |
21:08 | crts1973 | yes, that would be very helpful |
21:10 | crts1973 | Also if you are curious about our project you can look at: http://biorobotics.ri.cmu.edu/index.html |
21:12 | jas-hacks_ | pretty cool, what os/kernel does the snake use? |
21:13 | crts1973 | Well we are building a new snake that has small arm processors running Chibios for the motion control. I am building the camera module for the end of the snake which will have one of the Solidrun SOMs |
21:14 | rabeeh | crts1973: please send the videos :) |
21:14 | rabeeh | a link to them |
21:16 | rabeeh | crts1973: i'm looking at Chibios; they don't seem to have i.MX6 support. |
21:16 | crts1973 | We don't have any good videos of the new snake, however it will have the same capabilities of the old snake. When we do release videos we stick them on our youtube channel: http://www.youtube.com/user/CMUBiorobotics |
21:16 | crts1973 | Oh sorry, only one module on the snake has an imx6 |
21:16 | crts1973 | the other 16 use an STM32F4 |
21:16 | rabeeh | so motion sensing MCU is on every part of the snake |
21:17 | jas-hacks_ | crts1973: planning on using opencv? |
21:17 | rabeeh | and i assume that the micro som would be on the head with a camera |
21:17 | crts1973 | Yes! |
21:17 | rabeeh | oh. nice |
21:17 | rabeeh | did u try opencv with gpgpu? |
21:18 | crts1973 | no. So the point where we are at right now is that our old snake just had analog video going back to a network video encoder |
21:18 | crts1973 | when then would just do image processing on a laptop |
21:19 | crts1973 | It would be great to be able to do vision processing right on the head of the snake (this is one reason for sticking the SOM in) |
21:19 | rabeeh | the different versions of the SOM are all footprint compatible |
21:20 | rabeeh | the solo and dual lite processors have dual CSI lane; where the dual and quad have quad CSI lanes |
21:20 | crts1973 | That is also great. I ordered the quad version |
21:20 | rabeeh | in my mind if you design with dual lane then you would be able to scale from the lowest end to the highest without issues |
21:21 | rabeeh | then you might start with a quad; and when there is more performance optimizations you must stick with lower end and thus burn less power too |
21:21 | rabeeh | is it going to be battery powered? |
21:22 | crts1973 | It could be in the future, it is tethered at the moment |
21:43 | crts1973 | Okay so this has been helpful. jas-hacks_ where would be a good place to start? Should I peruse your blog? |
21:52 | jas-hacks_ | crts1973: What's your first goal? |
21:53 | jas-hacks_ | To get the camera working on linux? |
21:53 | crts1973 | Well I would like to verify that the camera connection works, for this I don't need the VPU |
21:53 | hste | crts1973: I think yocto is a good way to start |
21:54 | jas-hacks_ | yocto is good as long as your patient for the build image to be produced |
21:54 | crts1973 | I am patient as long as I have some good instructions |
21:55 | jas-hacks_ | ubuntu may be easier if you want to quickly test the camera is working providing the packages are available |
21:57 | crts1973 | I will need to recompile the kernel to add in some lines to the board file (board-mx6q_microsom.c or board-mx6q_c1.c) |
21:58 | crts1973 | The camera driver already seems to be in the kernel (OV5640) |
22:00 | crts1973 | So maybe short term, I should use an Ubuntu image and long term use Yocto for development |
22:00 | jas-hacks_ | that may be easier |
22:01 | crts1973 | okay. do you recommend using your root-fs, xubuntu 13.04? |
22:06 | jas-hacks_ | you can try that, it works on the C1 |
22:07 | crts1973 | okay, thanks I will let you know how it goes |
22:13 | hste | crts1973: I have a script for downloading and putting jas-hacks_ rootfs on a sdcard http://stende.no-ip.info/c1/mkrootfsc1.sh |
22:14 | crts1973 | awesome! |
22:15 | hste | you have to change the kernel with one with the needed modules etc |
22:15 | crts1973 | good place to start though, thank you |
22:26 | crts1973 | hste: was the kernel that is downloaded using your script created using the patches that are linked on the wiki? |
22:27 | hste | yes |
22:30 | rabeeh | crts1973: how are you going to hook the camera? |
22:32 | crts1973 | I am using the boundary devices camera, I was going to make up a quick PCB adapter but I didn't want to wait |
22:33 | crts1973 | I just used 30AWG wire and jumpered their adapter board to the C1 |
22:33 | crts1973 | using very short, matched lengths, twisted pairs |
22:33 | crts1973 | I am not sure if it will work given the high speed of the MIPI interface. If it doesn't, I will make up a PCB |
22:33 | hste | crts1973: the kernel in the script has ov5640_camear.ko |
22:34 | crts1973 | yeah! I just need to enable a PWM output for the camera clock (it doesn't have its own oscillator) |
22:34 | crts1973 | also a reset line, I think the I2C is already enabled |
22:35 | rabeeh | crts1973: there is PWM or CCM_CLKO2 |
22:35 | rabeeh | you can also use a pi camera |
22:35 | rabeeh | reset? |
22:36 | crts1973 | The camera has a reset line |
22:36 | rabeeh | you can hook the reset from the 8 pin header |
22:36 | rabeeh | look at the wiki - |
22:36 | rabeeh | http://imx.solid-run.com/wiki/index.php?title=Carrier-One_Hardware |
22:37 | crts1973 | That's true |
22:37 | hste | crts1973: it has a module ov5640_camera_mipi.ko also |
22:37 | rabeeh | crts1973: where do you do proto pcbs? |
22:38 | crts1973 | Well it depends. For cheap stuff where we have time, oshpark. Sometimes Advanced Circuits or Sunstone if we are in a hurry |
22:38 | crts1973 | for 6+ layers we are using quickturnpcb.co.kr |
22:39 | crts1973 | they can ship in a day or so from Korea to the states |
22:39 | rabeeh | which tools do you use? |
22:39 | crts1973 | I just switched over to Altium this last year, we used to use Eagle |
22:39 | rabeeh | Ado they have a free version now? |
22:40 | crts1973 | Altium has good educational pricing, but no free version |
22:40 | rabeeh | i recall they were beaten on eevb on not having a free version |
22:40 | crts1973 | kinda sucks, yeah I wish they did |
22:40 | rabeeh | btw - for length matching don't worry about it |
22:40 | rabeeh | it should work |
22:40 | rabeeh | i have on my desk single ended HDMI connection and it works |
22:41 | crts1973 | Ha! |
22:41 | rabeeh | well.. sometimes the green color makes some issues but i need a really long cable |
22:41 | rabeeh | :) |
22:42 | rabeeh | differential signals are made for EMI !!! |
22:42 | crts1973 | Okay yeah, I am not expecting it to be problematic |
22:42 | rabeeh | well.. not only for EMI; but 1Gbps traces and few inch wires will be only EMI |
22:43 | crts1973 | Do you guys use Altium? |
22:43 | rabeeh | nop; but I want to evaluate it |
22:43 | rabeeh | i think all PCB schematics and layout tools are designed by hardware engineers :) |
22:44 | crts1973 | Yeah seriously, I agree. I have to say that Altium works well once you get used to it. It is way more powerful than Eagle |
22:45 | rabeeh | i haven't tried it; but i think all of the tools are designed by hardware engineers that learned some C or C++ and started coding |
22:45 | rabeeh | i mean for instance; when you want to include a new chip in your design; you need to have it from scratch in your design |
22:46 | rabeeh | no one really thought about a universal language that you #include and you are done? |
22:46 | crts1973 | Yeah. Seriously I spend a lot of time making library files. It is so tedious |
22:46 | hste | rabeeh: why does the i1 only have 10/100Mbps? I thought i.mx6 solo also had 1Gbps and the c1 has 1Gbps |
22:47 | rabeeh | hste: a bonus :) |
22:47 | rabeeh | even CuBox-i1 has 1Gbps |
22:48 | _rmk_ | rabeeh: you're using the ar8035 on everything now? |
22:48 | hste | ok on the wiki it says only 10/100 |
22:48 | rabeeh | _rmk_: for now |
22:48 | rabeeh | hste: i know.. for now due to logistics, and compensating about the pre-order we have upgraded everyone |
22:48 | rabeeh | still we need to announce it.... but this is the bottom line |
22:48 | hste | :) |
22:48 | _rmk_ | when you switch over to the 8030, someone's going to have to work out the DT stuff for it... |
22:48 | rabeeh | we might change that in the near future.. |
22:49 | rabeeh | _rmk_: i know... a big headache |
22:49 | _rmk_ | which'll then be interesting to work out which DT file for which box |
22:49 | rabeeh | i don't know if auto-detect will even work... and i don't want to burn more fuses |
22:49 | _rmk_ | autodetect won't work with DT stuff |
22:49 | rabeeh | _rmk_: i know; for now i prefer not to think about it |
22:49 | _rmk_ | unless its done in the boot loader |
22:50 | rabeeh | _rmk_: fuses will be the same story too; proprietry only for SolidRun and non standard |
22:50 | _rmk_ | one of the issues of switching over to wonderful DT is that this kind of thing becomes loads more difficult... before we could just write code in a platform file to work it out and do the right thing |
22:50 | rabeeh | _rmk_: it has it's good and bad sides |
22:51 | rabeeh | i think DT was too extreme since it completely removes flexibility that every embedded software engineer is used to |
22:51 | jas-hacks_ | _rmk_: where should that type of code sit for DT? |
22:52 | _rmk_ | jas-hacks: it involves modifying the DT according to what's fitted on the board, and we don't have the kernel do any modifications to the DT blob |
22:54 | jas-hacks_ | _rmk_: I'm have similar issues as I'm starting to look at the DT for the Utilite |
22:54 | rabeeh | _rmk_: assuming detecting (or autodetecting) fast ethernet vs gig ethernet is done in u-boot; what is the best way to reflect that in dt? |
22:54 | rabeeh | completely load a new one? or there is something smarter? |
23:05 | _rmk_ | I don't know whether uboot can graft two fragments together |