tmlind | freemangordon: maybe use regmap_read_bulk() to read values to buffer, then process the buffer? | 06:25 |
---|---|---|
tmlind | freemangordon: also, maybe you also should call triggered_buffer_cleanup()? | 06:26 |
freemangordon | tmlind: I use p, li { white-space: pre-wrap; } devm_iio_triggered_buffer_setup | 09:36 |
freemangordon | so triggered_buffer_cleanup is not needed, iiuc | 09:36 |
freemangordon | I cannot find regmap_read_bulk in 6.x kernels | 09:39 |
freemangordon | oh, it is regmap_bulk_read() :) | 09:40 |
tmlind | oh ok :) | 10:09 |
freemangordon | ok, calculating Ri by using values from iio is way more stable | 10:20 |
tmlind | nice, so what's the call from the charger driver to get the data? | 10:25 |
tmlind | i guess you're still playing with your ocv user space app | 10:26 |
freemangordon | yes | 10:26 |
freemangordon | but now it looks way better | 10:26 |
tmlind | still no idea though how long sampling the 2x4 values takes? | 10:27 |
freemangordon | unfortunately battery is not calibrated so I cannot compare with coloumb counter, but ... | 10:27 |
freemangordon | I guess 1 second | 10:27 |
tmlind | oh so 4 * 250 ms? | 10:27 |
freemangordon | IIUC sampling period is 250 ms | 10:27 |
freemangordon | yes | 10:27 |
freemangordon | that's what my gut feeling tells me :) | 10:28 |
tmlind | ok makes sense, might be possible to complete that with the adc interrupt | 10:28 |
freemangordon | I can measure how long it takes till interrupt arrives | 10:29 |
tmlind | yeah might be worth checking | 10:29 |
freemangordon | https://pastebin.com/S2wBgJKU | 10:29 |
freemangordon | looks like real values | 10:29 |
freemangordon | "filtered 90" is SoC | 10:30 |
tmlind | ok | 10:32 |
tmlind | do the values look sane if you query after a longer period, like every 20 seconds? | 10:32 |
freemangordon | I do 16 measurements every 30 seconds | 10:33 |
freemangordon | I guess 30 should become 60 | 10:33 |
tmlind | ok, presumably longer delays don't produce weird values either | 10:33 |
freemangordon | and 16 should become 4 | 10:33 |
freemangordon | but this is the first try after changing U/I source | 10:34 |
tmlind | yeah great | 10:34 |
freemangordon | yeah, Ri ~= (0.100 .. 0.120) Ohm sounds sane | 10:35 |
tmlind | ok | 10:36 |
freemangordon | uvos said he measured about .15, but not sure @ what SoC | 10:36 |
tmlind | maybe the coulomb counter has longer term drifting because it does not account for the battery internal resistance? so.. the difference between the adc and coulomb counter for a period should be usable for calculating the ocv? | 10:39 |
freemangordon | after first cycle it should be accureat, my idea is to compare my algo with coloumb, to see how accureate it is | 10:39 |
freemangordon | *accurate | 10:40 |
tmlind | yeah the coulomb counter measurements should be accurate, but the driver may not account for external consumption so i suspect it may drift over longer periods of time | 10:40 |
freemangordon | right, that's why I want to use voltage-based estimation | 10:41 |
tmlind | and there could be still some bugs lurking there in the coulomb counter code, and the hardware may have bugs too since it's not used for later pmics | 10:41 |
freemangordon | mhm | 10:45 |
freemangordon | I plan to write user-mode helper that will push battery SoC table to kernel every few minutes. run-time calculated based on the adc U/I readings | 10:46 |
freemangordon | that should make everybody happy, assuming estimation algo is correct | 10:46 |
freemangordon | ok, seems 2 4x2 measurements are enough | 10:53 |
tmlind | ok | 11:58 |
freemangordon | on non-calibrated d4, my applet reports SoC 36%, battery applet: 62% :) | 13:59 |
Wizzup | freemangordon: I assume 36% is more accurate? :P | 14:05 |
freemangordon | mhm | 14:06 |
freemangordon | waiting for the battery to become flat to be able to compare with coloumb | 14:06 |
Wizzup | I think the BL-5J batteries I bought are quite crappy (they were not polarcell) | 14:08 |
Wizzup | I'm getting the polarcell ones, but they wouldn't ship to me, so I have to pick them up in a week | 14:08 |
freemangordon | I wonder what would be the results on an old battery | 14:10 |
Wizzup | which battery are you testing with | 14:10 |
freemangordon | e960 | 14:10 |
Wizzup | I see | 14:10 |
Wizzup | I think there might be some eb41 in initlab, if you don't have any | 14:11 |
freemangordon | I was thinking about getting one from there | 14:11 |
freemangordon | btw, if you have some spare time, could you enable sysfs triggers in omap kernel and rebuild? | 14:11 |
Wizzup | I think this would be for uvos, he typically does the kernel builds | 14:13 |
Wizzup | but if you know what the options are, I can try | 14:13 |
freemangordon | I guess CONFIG_IIO_SYSFS_TRIGGER=m | 14:14 |
Wizzup | if you need it now/soon I can do it, otherwise I'd prefer to wait for uvos | 14:15 |
freemangordon | no hurry | 14:15 |
freemangordon | Wizzup: PM seems to become so good I cannot deplete the battery :) | 14:46 |
Wizzup | freemangordon: :) | 14:48 |
Wizzup | that will change even more with OFF mode | 14:49 |
freemangordon | yeah | 14:49 |
freemangordon | if we have it some day | 14:49 |
Wizzup | :p | 14:49 |
freemangordon | stupid chromium seems to pause video playback when minimized | 14:59 |
freemangordon | what's wrong with those guys?!? | 14:59 |
freemangordon | they want to turn desktop to android or what? | 14:59 |
sixwheeledbeast | could it not be the website itself | 15:43 |
freemangordon | how would YT know that I am bringing the browser out of focus? | 15:47 |
freemangordon | in the tasknav that is | 15:47 |
sixwheeledbeast | javascript | 15:47 |
sixwheeledbeast | ? | 15:47 |
freemangordon | well, javascript is a scriping language, no? :) | 15:48 |
freemangordon | but what do I know | 15:48 |
sixwheeledbeast | i have had a few videos embeded in content do this annoying behaviour | 15:49 |
freemangordon | which browser? | 15:49 |
sixwheeledbeast | any | 15:49 |
freemangordon | hmm | 15:49 |
sixwheeledbeast | firefox chromium | 15:49 |
sixwheeledbeast | firefox you can use the pop out playback window to make it continue | 15:49 |
* freemangordon tries with wtwebbrowser | 15:50 | |
freemangordon | ok, the battery went flat, finally :) | 15:52 |
sixwheeledbeast | tbf i have never had it with youtube, but other video content within articles on other sites. | 15:55 |
sixwheeledbeast | it is something people detect to autoplay and then autopause if not in view. it's suppose to be helpful but it's annoying. I want to read the article while listening to the video .... | 15:56 |
hiddener | does intel atom have any support in mainline? have a zenfone, i'd play with it | 15:57 |
freemangordon | hmm, does not happe in the VM | 15:59 |
freemangordon | yes, it happens | 16:03 |
freemangordon | when I open a new tab | 16:03 |
freemangordon | but, only on m.youtube | 16:03 |
sixwheeledbeast | Is that user agent detected? | 16:06 |
sixwheeledbeast | Desktop can't seem to get to m.youtube | 16:06 |
freemangordon | yep, I am using iphone6 UA | 16:07 |
freemangordon | sixwheeledbeast: it turned out I had some old extension (AutoplayStopper) that was doing this | 16:16 |
sixwheeledbeast | :nod: | 16:16 |
tmlind | freemangordon: any info so far on the battery charge state difference between the coulomb counter and the 2x4? | 16:16 |
Wizzup | freemangordon: btw, when I click reply all on Joel's mail from the mailing list, modest uses 100% cpu and goes up in ram a lot | 16:17 |
Wizzup | not sure what is up there | 16:18 |
freemangordon | Wizzup: strace/gdb? | 16:18 |
Wizzup | it doesn't crash | 16:18 |
freemangordon | tmlind: no, since 10 minutes ago I was trying to deplete the battery | 16:18 |
freemangordon | now charging | 16:18 |
tmlind | ah calibration data missing | 16:18 |
freemangordon | Wizzup: ok, but maybe it does infinite loop | 16:18 |
freemangordon | sixwheeledbeast: no, it was not the extension | 16:19 |
Wizzup | freemangordon: right | 16:19 |
freemangordon | so, if chromium is in fullscreen, it will pause if I go in tasknav | 16:19 |
freemangordon | tmlind: yeah | 16:19 |
tmlind | freemangordon: while charging maybe log your 2x4 data to check how much power gets charged for let's say 10% state increase :) | 16:20 |
sixwheeledbeast | I can't access m.youtube to even reproduce. My user agent tool can't seem to fool youtube in to serving mobile to me. | 16:20 |
freemangordon | sixwheeledbeast: it is called "User-Agent Switcher for chrome" | 16:22 |
freemangordon | this https://chrome.google.com/webstore/detail/user-agent-switcher-for-c/djflhoibgkdhkhhcedjiklpkjnoahfmg | 16:23 |
sixwheeledbeast | I am using the same for ff | 16:23 |
freemangordon | choose iphone6 | 16:23 |
freemangordon | and then m.youtube.com | 16:23 |
freemangordon | tmlind: yeah, I will play a bit once I have calibrated battery | 16:24 |
freemangordon | I am not even sure my current algo is the best | 16:24 |
tmlind | freemangordon: if you can log uAh charged based on your alog you can compare it directly to the uAh value difference read from /sys/class/power_supply/battery/charge_counter.. | 16:25 |
tmlind | and the missing part is hopefully the battery internal resistance :) | 16:26 |
freemangordon | no, the only thing I can get is SoC in % | 16:26 |
freemangordon | well, I am calculating 'the missing part' | 16:26 |
freemangordon | so it is not really missng | 16:26 |
tmlind | battery % change translates direcly to uAh if you know the capacity of the battery, might be best to check the capactiy with some external reliable charger though | 16:27 |
freemangordon | ok | 16:27 |
freemangordon | but that would mean a calibrated battery :) | 16:27 |
tmlind | right :) | 16:27 |
freemangordon | and I am aiming *uncalibrated* batteries | 16:28 |
tmlind | i think uvos has some fancy battery charger that could verify the capacity between between certain voltages | 16:28 |
freemangordon | because mAh is not better that % charge | 16:28 |
freemangordon | I think Wizzup has too | 16:28 |
tmlind | ok | 16:28 |
freemangordon | but I am not trying to come up with some 0 error algo | 16:29 |
freemangordon | and error up to 2-3 or even 5 % should be acceptable | 16:29 |
tmlind | yeah sure, i'm just wondering how much the charge_counter is off | 16:29 |
freemangordon | once I have verified I have stable algo, I will post it for anyone to play with it | 16:30 |
tmlind | i guess the external charger vs charge_counter capacity test can be done anytime with no changes | 16:30 |
freemangordon | mhm | 16:30 |
freemangordon | and my algo is not needed | 16:30 |
tmlind | right | 16:30 |
tmlind | then when both the charge_counter and 2x4 algo are accurate, we'll be happy | 16:31 |
freemangordon | if 2x4 is accurate we won;t need charge_counter, no? | 16:32 |
freemangordon | well, unless we want to show mAh | 16:32 |
tmlind | right, no need for calibration | 16:32 |
tmlind | algo can just populate charge_now | 16:32 |
tmlind | hmm capacity will still be missing? | 16:33 |
freemangordon | mhm | 16:33 |
freemangordon | so it can only come=up with % | 16:33 |
tmlind | i'm getting about two days on the 960 battery fyi | 16:33 |
freemangordon | on leste? | 16:33 |
tmlind | no.. this is still my alpine with shells scripts.. | 16:34 |
freemangordon | leste/chimaera is more than a day, for sure | 16:34 |
tmlind | about 75 - 85 mW idle consumption sampled once a minute, 3g data connected, wlan off | 16:34 |
freemangordon | nice | 16:35 |
tmlind | power consumption goes bad to around 400 mW always when 3g data connection is lost, then i have to ifdown wwan1; ifup wwan1 | 16:35 |
freemangordon | I think on leste it is a bit higher | 16:35 |
tmlind | yeah it's the timers | 16:35 |
freemangordon | but didn;t verify after we moved to chimaera | 16:36 |
freemangordon | where uptime definitely increased | 16:36 |
tmlind | ok, planning on setting up my m-l d4 with 960 battery too | 16:36 |
tmlind | very bad battery on that one right now | 16:36 |
freemangordon | maybe take pictures of the process of converting e960 to eb41 | 16:40 |
tmlind | heh | 16:40 |
tmlind | i'll try | 16:40 |
freemangordon | Wizzup: same to you ^^^ | 16:41 |
tmlind | oh, and what i'm still seeing is lcd not coming back on after about two weeks of uptime and i need to reboot | 16:41 |
Wizzup | yeah, I will make photos if I manage | 16:41 |
freemangordon | tmlind: on leste? | 16:41 |
tmlind | freemangordon: no, this with sway, seems to be some dss bug though | 16:42 |
freemangordon | ah | 16:42 |
Wizzup | this might be the drm bug, but we'd have to verify | 16:42 |
Wizzup | probably the same we see in charge mode I guess | 16:42 |
tmlind | ok | 16:42 |
freemangordon | mhm | 16:42 |
freemangordon | hard if not impossible to repro though | 16:42 |
Wizzup | I think it happens often in charge mode, no? | 16:43 |
tmlind | yeah, i wonder if it can be reproduced with doing blank/unblank in a loop though | 16:43 |
freemangordon | I never hold my device in charge mode | 16:43 |
Wizzup | tmlind: possibly | 16:44 |
tmlind | hmm only around 150 blank/unblank cycles if i usually blank/unblank 10 times a day | 16:45 |
* tmlind starts a test | 16:45 | |
freemangordon | :) | 16:45 |
* freemangordon starts a rest | 16:46 | |
tmlind | enjoy & ttyl | 16:48 |
freemangordon | seems charging is close as well https://pastebin.com/04xbPHRn | 18:15 |
freemangordon | also, it seems charging does not account for Ri, and terminates a bit early | 18:17 |
freemangordon | OCV is ~4160 mV when charging gets terminated | 18:17 |
sicelo | https://mastodon.social/@LanceWex/110084766693974205 | 20:15 |
buZz | wonder if that article writer was aware that N900 wasnt the first 'linux in pocket' device :) | 20:16 |
sicelo | where does the article say it was the first? | 20:17 |
buZz | they also seem to paint the picture that Leste is just for n900s :) | 20:19 |
Wizzup | *shrug* | 20:20 |
Wizzup | still cool! | 20:20 |
buZz | yes! | 20:23 |
buZz | did they end up installing any os on a phone in the article? seems they just did some VM on a desktop? | 20:24 |
sicelo | maybe you're overthinking it all :-) | 20:25 |
tmlind | freemangordon: there's also a bit of a voltage drop across the shunt resistor if you're trying to add up things | 20:33 |
tmlind | hmm my test system rebooted after blank/unblank loop with 1 second sleep in between | 20:41 |
Wizzup | after how long? | 20:42 |
tmlind | did not print out date.. noticed after a few hours, nothing in the logs based on dmesg -w over ssh, don't have pstore right now | 20:43 |
Wizzup | was this the leste one, or? | 20:46 |
tmlind | no, just plain v6.3-rc3 kernel running a minimal install of alpine, no pvr patches | 20:47 |
Wizzup | ok | 20:48 |
Wizzup | makes sense :) | 20:49 |
tmlind | i'll try to reproduce and debug when i can, ttyl | 20:54 |
Wizzup | ttyl | 20:58 |
arno11 | Wizzup: i found that some modules are missing from boot to get nokia modem working normally | 21:06 |
arno11 | cmt_speech and omap_ssi and probably omap_ssi_port if it has not been merged with omap_ssi | 21:07 |
arno11 | it doesn't solve the issue with pm=1 but enabling new modules creates new interactions according to lsmod | 21:09 |
arno11 | and power cons decrease a bit on idle | 21:10 |
arno11 | maybe it's just a bias for idle but new interactions are real | 21:11 |
arno11 | interact with it i think pulseaudio is the root cause of many issues | 21:15 |
arno11 | lol something is missing at the beginning | 21:16 |
arno11 | of the sentence | 21:16 |
arno11 | Wizzup: freemangordon: sicelo: is pulseaudio-module-cmtspeech usable ? | 21:20 |
buZz | sicelo: natural trait of being me ;) | 21:26 |
Wizzup | arno11: no, pulseaudio-module-cmtspeech was compile tested on beowulf, but I have to first build it all for chimaera, and when I tested it on beowulf it used a lot of cpu but id nothing | 21:41 |
Wizzup | we need to look at the code and figure out how it works | 21:41 |
Wizzup | what interactions are you talking abut? | 21:41 |
arno11 | interactions between cmt_speech omap_ssi nokia-modem hsi ssi_protocol hsi_cart | 21:43 |
Wizzup | do you see oopses in dmesg? or? | 21:46 |
arno11 | nothing interesting | 21:46 |
arno11 | in dmesg syslog ofonod.log | 21:47 |
Wizzup | so what do you observe? | 21:47 |
arno11 | a bit long to explain with n900 keyboard lol | 21:49 |
Wizzup | aight | 21:49 |
arno11 | anyway i tried many things following what has been prevously done with nokia modem and this module seems impossible to load on boot with no troubles elsewhere | 22:03 |
arno11 | otherwise it works great if it is loaded from desktop | 22:04 |
Wizzup | ok, well, if you're at a non-n900 keyboard let me know what problems you see | 22:07 |
arno11 | ok no probs | 22:08 |
Wizzup | arno11: as in, I get modem on startup, I killed sphone just to make sure, but and I can receive sms when sphone is running | 22:10 |
Wizzup | without touching ofono | 22:10 |
arno11 | oh ok i'm doing differently | 22:12 |
arno11 | after booting i run options nokia-modem pm=1 then | 22:13 |
arno11 | killing sphone and restart it | 22:13 |
arno11 | and close it | 22:13 |
arno11 | and sending receiving calls and sms works | 22:14 |
arno11 | like this | 22:14 |
arno11 | if i add modem in etc/modulesblabla i can't get online | 22:17 |
arno11 | otherwise even if sphone is closed sms and calls work | 22:18 |
Wizzup | hm, maybe it's racy, I have it on modules-load.d and it works ok for me, with pm=1 | 22:18 |
Wizzup | of course there were some ofono problems I guess, but sms works | 22:18 |
arno11 | ok | 22:19 |
arno11 | all of this is a bit tricky but it works great finally | 22:22 |
arno11 | just difficult to get OOTB | 22:22 |
buZz | pavel's batman is kinda cool :) from the ML | 22:22 |
Wizzup | arno11: yes, we need to get it to work ok, what I also noticed is that restarting ofono makes it not see the modem, and restarting again makes it re-appear | 22:23 |
arno11 | weird | 22:25 |
arno11 | or not weird: the common point of all of these modules is pulseaudio and i'm fed up to fight with him for 15 years... | 22:33 |
bencoh | tbh I have a few issues with pulseaudio as well, the most painful one is audio crackling and high cpu use when using twinkle with it ... I really wonder how hard it would be to try pipewire | 22:34 |
bencoh | would our custom pulse modules work against pipewire? | 22:34 |
arno11 | same idea | 22:34 |
arno11 | i've another leste install and i think i'll try switching to pipewire | 22:35 |
Wizzup | if it features full pa compat it might just work | 22:35 |
Wizzup | you could hopefully at least prevent some of the resampling | 22:35 |
arno11 | indeed | 22:35 |
bencoh | the question is whether they kept the module API or just the client API | 22:35 |
Wizzup | arno11: if I were you I'd be happy if we can get call audio to work at all, with or without pulse ;) | 22:36 |
Wizzup | bencoh: both I think | 22:36 |
Wizzup | bencoh: I use pipewire and I use pulseaudio bluez modules iirc | 22:36 |
bencoh | oh? | 22:36 |
Wizzup | (i think, not sure...) | 22:36 |
arno11 | Wizzup: yep :) | 22:37 |
bencoh | https://gitlab.freedesktop.org/pipewire/pipewire/-/wikis/Migrate-PulseAudio#modules-1 | 22:38 |
bencoh | so apparently it works with some modules only (dunno if that's up-to-date) | 22:39 |
Wizzup | I checked my module list and I think I have only libpipewire modules in pa module list | 22:39 |
bencoh | hmm? | 22:39 |
Wizzup | so yeah, I'd really recommend getting audio calls working with pa before trying pipewire | 22:39 |
arno11 | without pulseaudio-module-cmtspeech ? | 22:40 |
sicelo | you need *-cmtspeech | 22:41 |
arno11 | hi sicelo. please explain | 22:44 |
Wizzup | arno11: with the module | 22:45 |
Wizzup | or with some other path (if not PA) | 22:45 |
arno11 | ah ok | 22:46 |
Wizzup | uvos: thanks for merging the mce pr | 23:31 |
Wizzup | I will once again work on the tp sphone module and see if I can get it over the finish line this time | 23:31 |
Wizzup | and then look at conversations | 23:31 |
uvos | Wizzup: gotcha | 23:32 |
Wizzup | I might ignore the part where the calls are anonymous for the moment and at least get the rest set up, then poke the sfos guys again | 23:34 |
Wizzup | it always surprises me how much smoother the bionic is than the droid 4 | 23:36 |
Wizzup | I mean, I know why, but still, it surprises me | 23:36 |
uvos | yeah | 23:36 |
uvos | tmlind mentioned that omapdss has a register that can be used that jitters the pixel clock that might help with the undelivered frames | 23:36 |
uvos | tmlind every try that one? | 23:36 |
uvos | *ever | 23:37 |
uvos | i also wish we could have the bionics lcd firmware on d4 | 23:37 |
uvos | for the less aggresvie rgbw color conversion | 23:38 |
uvos | Wizzup: where shal i relase mce too | 23:39 |
uvos | on chimaera | 23:39 |
Wizzup | I think chimaera or chimaeara-devel, either is fine | 23:41 |
uvos | ok | 23:43 |
uvos | beowulf is officaly discontinued right? | 23:46 |
Wizzup | I mean | 23:50 |
Wizzup | we should probably announce chimaera before we do that | 23:50 |
Wizzup | but yeah, I don't push my changes to beowulf anymore | 23:50 |
Wizzup | for fear of breaking things that I don't test | 23:51 |
uvos | Wizzup: ok | 23:51 |
uvos | could you at least merge beowulf-devel and stable one last time then? | 23:51 |
Wizzup | yeah I suppose, not right now though | 23:53 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!