Firmware 4.33.19608

[rtiangha]

Originally Posted by rashkae

If the speculation is correct that there is some instability related to processor weirdness when clock speed is too low, there can easily be a great variety of individually affected units and trigger conditions. Someone will probably have to create some easy to apply patches for people to test that theory.

If that's the case, I wonder if changing the CPU governor from Interactive to something different (ex. Ondemand) would make a difference? No idea what the effect on battery life would be, though.

echo ondemand > /sys/devices/system/cpu/cpuX/cpufreq/scaling_governor

where X is the number of your processor core, starting at 0 (are the newer devices multi-core? If so, run the command once for each core; if not, just run it once using 0). Of course, you can substitute whatever supported CPU governor your device supports instead of ondemand (like maybe conservative).

You can input the above command in a terminal via telnet/ssh or turn it into a nickelmenu command with cmd_spawn. Note that it will reset back to Interactive whenever you reboot or switch to a different application like koreader.

[NiLuJe]

performance would be the first choice to validate the theory *at all*, battery life be damned (hell, it might still be more efficient than a stock Sage anyway ;D) .

Nobody's answered me on that front in the related KOReader GH issue, FWIW.

As for the sterile back and forth before: It's definitely NanoClock, trust me.

[rtiangha]

Originally Posted by NiLuJe

Nobody's answered me on that front in the related KOReader GH issue, FWIW.

I just read this comment and that's the EXACT same behavior I get on my Glo HD sometimes, both in nickel and in koreader. Funny thing is that I've had this device since 2016 but only really noticed this happening more frequently during the pandemic after a firmware upgrade (can't remember which one; I think it was in the 4.20s). And it looks like the Glo HD kernel was updated in that timeframe, so maybe there's a connection? In the meantime, to work around it, I disabled dithering in the developer settings which helps a bit, but not all the time.

Anyway, I recently imported a couple of books whose covers or images manage to consistently power off my device. Inspired by this thread, I tried changing the governor to ondemand and I was able to get past the cover on one of those books. With the other one, I couldn't get past things unless I also disabled dithering. If I had more samples, I'd see if performance would work consistently with dithering on, but I don't.

No idea what that all implies though.

[rtiangha]

Well, for those who just want to copy/paste, here's a simple nickelmenu entry to change the governor on one CPU core:

menu_item :main  :CPU Governor :cmd_spawn :quiet : echo governor > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor

where governor can be one of:

interactive
ondemand
conservative
powersave
performance

interactive is default. Here's what they do.

However, it looks like Elipsa and Sage are quad core devices, so that may not work because you should apply the same governor to all cores. You may be better off creating a script (I'll call it governor.sh):

#!/bin/sh

echo governor > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
echo governor > /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
echo governor > /sys/devices/system/cpu/cpu2/cpufreq/scaling_governor
echo governor > /sys/devices/system/cpu/cpu3/cpufreq/scaling_governor

And use nickelmenu to call the script instead:

menu_item :main  :CPU Governor :cmd_spawn :quiet : /path/to/governor.sh

[frostschutz]

if you prefer a one-liner you can also use tee

echo ${governor} | tee /sys/devices/system/cpu/*/cpufreq/scaling_governor

I haven't had time to experiment with governors yet, so I don't know if it would help...

[rtiangha]

OK, looks like I was wrong about changing cpu governors solving my problem. It was more having the device plugged into power that made things work, regardless of governor. It's quite reproducible and I have no idea what's going on.

If anyone wants to take a look at the kepub, let me know (no judging on my choice of reading material!). It forces my Glo HD to turn off as soon as you open the cover, so testing things is quick. Open to any other ideas too.

[JSWolf]

Originally Posted by rtiangha

OK, looks like I was wrong about changing cpu governors solving my problem. It was more having the device plugged into power that made things work, regardless of governor. It's quite reproducible and I have no idea what's going on.

If anyone wants to take a look at the kepub, let me know (no judging on my choice of reading material!). It forces my Glo HD to turn off as soon as you open the cover, so testing things is quick. Open to any other ideas too.

Using Calibre, try copying the cover to a different eBook that you know works. Then try that and see if you get the same problem. If you do, it's the cover.

[NiLuJe]

@rtiangha: The GloHD quirks on specific patterns is a separate issue that's known by Kobo (IIRC, there's even a reply by George about it in this very forum. It might even be referenced in one of the issues in the message you quoted).

(It's tied to specific patterns which require the eInk tech to draw more power to switch. Said behavior depends to some extent on the tech's generation, but it's usually checkered patterns (i.e., specific contrasts over a specific amount of consecutive pixels)).

Dithering patterns are usually full of those, hence the link with dithering .

[rtiangha]

Originally Posted by NiLuJe

(It's tied to specific patterns which require the eInk tech to draw more power to switch. Said behavior depends to some extent on the tech's generation, but it's usually checkered patterns (i.e., specific contrasts over a specific amount of consecutive pixels)).

Dithering patterns are usually full of those, hence the link with dithering .

...So the solution is to figure out how to over-volt this thing?

I didn't know it was a known issue. Is there anything else that can be done to mitigate it? Setting aside in-book images, would converting all covers to black-and-white help reduce that power draw? It's a really annoying problem but as long as I can open the book, I can use the slider to skip over problem images inside.

[JSWolf]

Originally Posted by rtiangha

...So the solution is to figure out how to over-volt this thing?

I didn't know it was a known issue. Is there anything else that can be done to mitigate it? Setting aside in-book images, would converting all covers to black-and-white help reduce that power draw? It's a really annoying problem but as long as I can open the book, I can use the slider to skip over problem images inside.

I wonder. Does ePub have the same problem as KePub with these images?

[NiLuJe]

It's a hardware issue. Displaying an offending pattern with no software running at all would trigger the issue.

[rashkae]

Originally Posted by NiLuJe

It's a hardware issue. Displaying an offending pattern with no software running at all would trigger the issue.

Considering how inferior the dithering on Epub is compared to Kepub, CBZ, I think it's still a valid question

[JSWolf]

Originally Posted by rashkae

Considering how inferior the dithering on Epub is compared to Kepub, CBZ, I think it's still a valid question

Can you show us some photos showing the difference in dithering for ePub and KePub? I've never found a problem with the ePub dithering. in the 4.x firmware versions.

[rtiangha]

Originally Posted by xyclonei

I combed over the settings page but could not locate that option. Any idea where it might be?

For those that were curious, I figured out how to enable Dyslexic Mode and posted instructions in the 4.33.19611 thread.

[Quoth]

Two Sage models now rolled back to 4.32.x as either version of 4.33.x makes the Advanced Notebooks intermittent for writing or export.