| vagrantc | josch: i thought i would end up reassigning them ... but now i am not so sure\ | 00:00 |
|---|---|---|
| vkoskiv | The max brightness duty cycle is only ~2.5%, which doesn't leave much wiggleroom | 00:00 |
| - Gooberpatrol66 (QUIT: Ping timeout: 260 seconds) (~Gooberpat@user/gooberpatrol66) | 00:28 | |
| + Gooberpatrol66 (~Gooberpat@user/gooberpatrol66) | 00:40 | |
| - vagrantc (QUIT: Quit: leaving) (~vagrant@2600:3c01:e000:21:7:77:0:20) | 00:55 | |
| - mjw (QUIT: Ping timeout: 252 seconds) (~mjw@gnu.wildebeest.org) | 01:52 | |
| - jn (QUIT: Ping timeout: 272 seconds) (~quassel@user/jn/x-3390946) | 02:24 | |
| + jn (~quassel@2a0a-a549-8313-0-20d-b9ff-fe49-15fc.ipv6dyn.netcologne.de) | 02:24 | |
| - jn (QUIT: Changing host) (~quassel@2a0a-a549-8313-0-20d-b9ff-fe49-15fc.ipv6dyn.netcologne.de) | 02:24 | |
| + jn (~quassel@user/jn/x-3390946) | 02:24 | |
| + ndufresne (~ndufresne@apple.collaboradmins.com) | 02:39 | |
| - NanoCodeBug (QUIT: Remote host closed the connection) (~NanoCodeB@c-67-170-15-47.hsd1.wa.comcast.net) | 02:56 | |
| - paperManu (QUIT: Ping timeout: 265 seconds) (~paperManu@107.159.243.8) | 03:02 | |
| - aloo_shu (QUIT: Ping timeout: 248 seconds) (~aloo_shu@90.166.193.48) | 03:38 | |
| + aloo_shu (~aloo_shu@90.166.193.48) | 03:48 | |
| - colinsane (QUIT: Quit: bye) (~colinunin@97-113-93-25.tukw.qwest.net) | 04:14 | |
| + colinsane (~colinunin@97-113-93-25.tukw.qwest.net) | 04:17 | |
| - Bertl (QUIT: Ping timeout: 252 seconds) (herbert@IRC.13thfloor.at) | 07:07 | |
| + Bertl (herbert@IRC.13thfloor.at) | 07:07 | |
| + networkstring (~ac130@185.104.123.26) | 09:55 | |
| - hairu (QUIT: Remote host closed the connection) (m-uotkmd@user/hairu) | 10:20 | |
| + hairu (m-uotkmd@user/hairu) | 10:21 | |
| + mjw (~mjw@gnu.wildebeest.org) | 10:30 | |
| minute | reform spotted at electronica https://hostux.social/@tanty/113480399764820740 | 10:36 |
| josch | minute: would you like a new reform-tools release with the new rk3588 u-boot? anything else? | 10:37 |
| minute | josch: yes, that would be cool | 10:37 |
| josch | minute: can you tag the latest commit? | 10:38 |
| josch | oh you did | 10:38 |
| josch | sorry :) | 10:38 |
| josch | 2024-11-13b it is | 10:39 |
| josch | nice, the u-boot build is even bit-by-bit reproducible thanks to SOURCE_DATE_EPOCH \o/ | 10:50 |
| hramrach | the other u-boots as well? | 10:58 |
| hramrach | I think not tagging the latest change is a problem with all of them | 10:58 |
| josch | they should be tagged once they are tested and ready -- are they? | 10:59 |
| josch | hramrach: i didn't press the merge button yet, which other u-boot would you like to see included? | 10:59 |
| josch | imx8mplus latest tag is from May this year... | 11:00 |
| josch | right, a new imx8mplus u-boot would bring the fdtfile and fbcon fixes | 11:01 |
| - GNUmoon (QUIT: Remote host closed the connection) (~GNUmoon@gateway/tor-sasl/gnumoon) | 11:04 | |
| josch | hramrach: did you flash some of the latest imx8mplus u-boot to your pocket? | 11:06 |
| + GNUmoon (~GNUmoon@gateway/tor-sasl/gnumoon) | 11:11 | |
| hramrach | no, I did not flash an u-boot ever | 11:39 |
| hramrach | And if the source is pulled fresh each build as is don on rockchip each build can produce vastly different results, including way more changes than what has changed in the repository | 11:41 |
| josch | hramrach: the source is pulled fresh but a pinned git commit is checked out | 11:42 |
| josch | i reached out on the forums with a plea for help: https://community.mnt.re/t/how-to-flash-latest-u-boot-on-pocket-reform/2737 | 11:55 |
| - mjw (QUIT: Ping timeout: 252 seconds) (~mjw@gnu.wildebeest.org) | 12:10 | |
| * Guest3791 -> mjw | 12:20 | |
| + paperManu (~paperManu@107.159.243.8) | 12:35 | |
| + gustav28 (~gustav@c-78-82-55-219.bbcust.telenor.se) | 13:02 | |
| - sevarat (QUIT: Quit: sevarat) (~sevarat@2601:207:600:d90::d573) | 13:55 | |
| - kensanata (QUIT: Quit: Ping timeout (120 seconds)) (~alex@user/kensanata) | 13:56 | |
| + kensanata (~alex@user/kensanata) | 14:02 | |
| + chomwitt (~chomwitt@2a02:587:7a1a:7a00:1ac0:4dff:fedb:a3f1) | 14:31 | |
| - AbortRetryFail (QUIT: Quit: housekeeping! housekeeping!) (~arf@146.ip-149-56-132.net) | 14:43 | |
| minute | josch: the rk3588 uboot build is so far only validated on microsd card on pocket reform. | 14:46 |
| josch | minute: i thought rk3588 didn't read u-boot from sd-card? we had the topic here a few days ago | 14:57 |
| josch | we should have a table detailing which som prefers what in what order... | 14:58 |
| - amospalla (QUIT: Ping timeout: 245 seconds) (~jordi@user/amospalla) | 15:14 | |
| + amospalla (~jordi@user/amospalla) | 15:15 | |
| josch | ooooh my local copy of the irc logs was bodged!! | 15:15 |
| josch | it was this one: https://mntre.com/reform-irc-logs/2024-11-09.log.html | 15:15 |
| josch | minute: so we had this confusion 3 days ago with NanoCodeBug and from past irc logs i read that rk3588 will prefer u-boot on emmc | 15:16 |
| ryukazou | If I want to check uboot Linux commandline, which file should I check? | 15:22 |
| minute | josch: it's a bit more complicated than that: 1. it will look on emmc first. 2. if that fails, it will try SD. 3. SPL will be loaded from emmc if it's there, but that SPL (at least the old 2017 version, well we don't ship that) loads the rest of uboot from SD | 15:34 |
| minute | 4. emmc uboot will try to boot (ext)linux/boot.scr from sd first | 15:35 |
| ryukazou | I cannot find fbcon.rotate in boot.src, where is this setting placed? | 15:36 |
| hramrach | ryukazou: /etc/default/flash-kernel | 15:39 |
| hramrach | I think the SPL loading u-boot from different device is needlessly confusing, and should be avoided. Both because the SPL and u-boot can be out-of-sync and boot may fail, and because it gives more different ways the system can boot which then need to be considered. | 15:42 |
| minute | screws now available: | 15:47 |
| minute | https://shop.mntre.com/products/m2-x-5-mm-phillips-countersunk-screws | 15:47 |
| minute | https://shop.mntre.com/products/m2-x-4-mm-phillips-cheese-head-screws | 15:47 |
| - chomwitt (QUIT: Ping timeout: 264 seconds) (~chomwitt@2a02:587:7a1a:7a00:1ac0:4dff:fedb:a3f1) | 15:51 | |
| ryukazou | It seem like my /etc/default/flash-kernel has been altered, fbcon.rotate=3 is gone but my display still behaves the same, is this normal? | 15:53 |
| + mark_ (~mjw@gnu.wildebeest.org) | 16:00 | |
| - glu (QUIT: Ping timeout: 276 seconds) (~glu@178.115.78.40.wireless.dyn.drei.com) | 16:02 | |
| + glu (~glu@178.115.78.40.wireless.dyn.drei.com) | 16:03 | |
| hramrach | ryukazou: you need to run flash-kernel after changing the file. You can inspect the result in boot.scr The format of the options can be seen in the u-boot sources, eg https://source.mnt.re/reform/reform-rk3588-uboot/-/blob/main/mnt-pocket-reform-rk3588_defconfig?ref_type=heads#L70 which shows fbcon=rotate:3 | 16:06 |
| - glu (QUIT: Ping timeout: 252 seconds) (~glu@178.115.78.40.wireless.dyn.drei.com) | 16:08 | |
| ryukazou | Yeah saw it, it appear in `setenv bootargs` | 16:17 |
| ryukazou | Now I know how to fix sometime console font become so small | 16:18 |
| ryukazou | I got this symptom after update to kernel 6.11.4-mnt-reform-amd64 | 16:19 |
| ryukazou | I guess add fbcon=font:TER16x32 to Linux commandline should fixed the problem. | 16:22 |
| - mark_ (QUIT: Ping timeout: 252 seconds) (~mjw@gnu.wildebeest.org) | 16:22 | |
| + glu (~glu@178.115.78.40.wireless.dyn.drei.com) | 16:24 | |
| ryukazou | It worked | 16:49 |
| + chomwitt (~chomwitt@2a02:587:7a1a:7a00:1ac0:4dff:fedb:a3f1) | 17:04 | |
| josch | minute: uff that indeed adds quite a bit of complexity to the whole process... And in your setup you don't have u-boot on emmc and that's how you tested rk3588 u-boot from sd-card? | 17:23 |
| - glu (QUIT: Ping timeout: 272 seconds) (~glu@178.115.78.40.wireless.dyn.drei.com) | 17:25 | |
| minute | josch: correct. later i can flash it to the emmc | 17:30 |
| + glu (~glu@77.119.199.138.wireless.dyn.drei.com) | 17:39 | |
| + mark_ (~mjw@gnu.wildebeest.org) | 18:44 | |
| - chomwitt (QUIT: Ping timeout: 248 seconds) (~chomwitt@2a02:587:7a1a:7a00:1ac0:4dff:fedb:a3f1) | 18:48 | |
| josch | minute: i think it would be interesting having instructions of how to "delete" u-boot from emmc. Would one just flash enough zeroes to the correct location? | 18:49 |
| josch | minute: would you like me to release reform-tools with new u-boot for rk3588 now or would you like me to wait for the new tag for imx8mplus u-boot to land first? | 18:51 |
| minute | josch: yes, for example in uboot i used the mmc erase command to erase the first 5242880 blocks (10 MB) | 19:05 |
| josch | minute: i'll probably come back to you once i have my rk3588 here :) | 19:07 |
| - mjw (QUIT: Killed (zinc.libera.chat (Nickname regained by services))) (~mjw@2001:1c06:2488:1400:4fd:39a7:74ac:7bae) | 19:10 | |
| * mark_ -> mjw | 19:10 | |
| + Guest9036 (~mjw@2001:1c06:2488:1400:4fd:39a7:74ac:7bae) | 19:10 | |
| - networkstring (QUIT: Ping timeout: 252 seconds) (~ac130@185.104.123.26) | 19:11 | |
| + maybeqbit (~maybeqbit@2601:283:c400:36c0:2920:f804:3e3e:1b7d) | 19:25 | |
| - maybeqbit (QUIT: Remote host closed the connection) (~maybeqbit@2601:283:c400:36c0:2920:f804:3e3e:1b7d) | 19:25 | |
| + maybeqbit (~maybeqbit@2601:283:c400:36c0:2920:f804:3e3e:1b7d) | 19:26 | |
| vkoskiv | I still don't understand why the backlight brightness goes up so rapidly at a really low duty cycle | 20:58 |
| vkoskiv | I'm now running it in fast PWM mode at ~7.8kHz, so now it's (more) silent, but the +width of the individual pwm pulses is still far below the minimum spec of 5us, they now range from 500ns to ~3.5us | 20:59 |
| vkoskiv | I'd like to have the lowest brightness be at least 5us, but that would already be above our max brightness :( | 20:59 |
| + vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:20) | 21:02 | |
| vkoskiv | If I choose a /64 divider instead of /8 to get 976Hz PWM, the signal is more in spec, but I get blinky lights even at the lowest brightness. | 21:07 |
| vkoskiv | I really just don't understand why this chip is so inconsistent. I'm sending a valid PWM, as far as I can tell. | 21:07 |
| vkoskiv | Lemme try /256, though I vaguely remember I tried all kinds of permutations, and none were good. | 21:08 |
| vkoskiv | *but* on the bright side, at ~1kHz, the lowest brightness actually turns the LEDs fully off! So worst case, I can just hack in logic to tweak the PWM timing when at 0 brightness. | 21:08 |
| vagrantc | hrm. mnt/reform keyboard sliding all over the place ... need to set it on something slightly grippy\ | 21:11 |
| vagrantc | getting pretty well adapted already, though | 21:11 |
| vkoskiv | At /256 it's really inconsistent. Toggling between levels 0 and 1 repeatedly, level 1 is randomly either off, dim, bright, or something in between | 21:15 |
| vkoskiv | I can see on my scope that the pwm just toggles between the two duty cycles I'd expect to see, but the chip is just behaving seemingly randomly. | 21:16 |
| vkoskiv | 244Hz is withing spec, and the +width of these pwm pulses are now well within spec. | 21:16 |
| vkoskiv | Either my backlight chip is broken, or there is something fundamental here that I'm just not aware of, and can't find in the TI sheet | 21:16 |
| vkoskiv | I would expect 50% duty cycle -> 50% keyboard brightness. But as is, 2.5% duty cycle is max brightness, and the hardware behaves inconsistently in many cases, due to an unknown reason | 21:18 |
| - maybeqbit (QUIT: Remote host closed the connection) (~maybeqbit@2601:283:c400:36c0:2920:f804:3e3e:1b7d) | 21:39 | |
| vkoskiv | Hmm, the 5V rail dips quite low with the backlight brightness up, at the PWM frequency. And the effect is more pronounced when it's in that blinking fault state, so maybe that causes this? | 21:55 |
| vkoskiv | I'll try the hacky timing workaround, to see if that would work consistently, at least on my unit | 21:55 |
| jfred | vagrantc: I think mine came with adhesive rubber feet which helps a lot - if yours didn't come with some then they're cheap online, would recommend | 21:57 |
| jfred | I added some to my Pocket Reform for the same reason | 21:58 |
| - gustav28 (QUIT: Quit: Quit) (~gustav@c-78-82-55-219.bbcust.telenor.se) | 22:15 | |
| vagrantc | jfred: hah. i did not look in the box very hard ... just apparently grabbed the keyboard and starting using it :) | 22:16 |
| vkoskiv | workaround won't work, just verified and that too just sometimes works. Not really sure how to fix this, then. | 22:17 |
| vkoskiv | And yeah, turning PWM off doesn't actually turn off the output, except sometimes it randomly does :D | 22:18 |
| vkoskiv | Maybe a larger cap across the 5V rail would patch up those drops, hmm... | 22:19 |
| vagrantc | hrm. no rubber feet after all ... | 22:35 |
| + chomwitt (~chomwitt@2a02:587:7a1a:7a00:1ac0:4dff:fedb:a3f1) | 22:48 | |
| vkoskiv | I'm measuring peaks down to ~3.3V at C7 when the backlight does the flash thing, so I suspect that may be the root cause of that blink? | 22:58 |
| vkoskiv | And the drops in voltage get bigger and bigger at each PWM pulse, and then the backlight cuts out (and the observable flashing happens) | 23:08 |
| vkoskiv | No idea if this behaviour is normal/expected, this is where my knowledge ends and guessing begins :P | 23:09 |
| - aloo_shu (QUIT: Ping timeout: 252 seconds) (~aloo_shu@90.166.193.48) | 23:17 | |
| + aloo_shu (~aloo_shu@90.166.193.48) | 23:28 | |
| - mjw (QUIT: Ping timeout: 264 seconds) (~mjw@gnu.wildebeest.org) | 23:52 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!