deprecated | # 301 | 2016-12-02, 14:42 | Report

@mautz, I'm ordering a new mid frame housing for my N5 in red, as well as a couple of LG G2 batteries to do this same hardware mod. I wanted a backup mid frame in case I screw something up.

May I ask what voltages, clock speeds, gpu settings, governor and hotplug driver combination you're using to get ~2 days of usage?

Thanks!

BluesLee | # 302 | 2016-12-02, 17:00 | Report

Can you both share your battery mods once finished? Thanks.

Also interested in two days of battery life, not more than one day here. I think battery consumption was better with older sfos versions, can't proof it.

minimec | # 303 | 2016-12-02, 17:25 | Report

I absolutely agree with you!!! For me the change happened here...


... and I also complained about higher power consumption, but I seemed to be the only one.



@mautz: Would it be possible to make a kernel without that 300Hz timer rate? Is that even used in your kernels? A quick search showed me, that it should be a kernel setting.


UPDATE:

I digged a little further. Looks that on older Android versions the 'Kernel Tick Clock' was set to 100Hz. I would really love to test a kernel with these 'old' settings.

There is a long discussion here about that change. I would like to quote some posts of the discussion...

2nd UPDATE

We are definitely running with 300Hz with the mautz kernels! You can check that in the kernel config file in /proc/config.gz. Search for 'CONFIG_HZ'!

@mautz: So I guess it's worth a try to switch back to 100Hz.

---

Last edited by minimec; 2016-12-02 at 21:39.

deprecated | # 304 | 2016-12-02, 18:43 | Report

This is absolutely true for me as well. In 2.0.1.11 it was fairly good on those earlier CM builds. Immediately upon going to the later 12.1 builds and on to 2.0.2.x/2.0.4.x, the drain increase is dramatic. Sometimes as high as 1% per minute with the screen on. I commented on it in this thread and others, but it seemed to be a known issue that most were OK with as we couldn't really pin it down. I'm almost certain @minimec is on the right track with the 300Hz timer rate.

I'd like to see if we can solve it with that change!

deprecated | # 305 | 2016-12-02, 18:45 | Report

Absolutely, will do. There's an image guide of the mod found here:

https://www.reddit.com/r/Nexus5/comm..._battery_from/

The picture guide at the top of that post is most helpful, and that's the route I'm planning to take once all the parts come in.

Edit: I spent less than $8USD on the genuine LG G2 battery, and about $30 for the red front housing/midframe INCLUDING a new LCD, all in one piece. Both of these purchases were from eBay, domestic sellers. I did this because my existing LCD has some bad light bleed and I figured I'd just take care of everything all at once. I'll take photos of my process and put a guide together as I go, just for the sake of clarity.

---

Last edited by deprecated; 2016-12-02 at 18:52.

mautz | # 306 | 2016-12-02, 20:04 | Report

At the moment i'm using zzmoove governor, but still testing others. But 2 days sounds much, but it's still bad. I barely use the phone during the day, with a stock installation and default kernel i get 1 1/2 days. So basically every hotplugging governor that puts the unused cores to sleep when the phone is suspended, will do the trick for me. My old Jolla 1 lasted 5 days with one charge...

I 'm trying to build a new kernel with some backported features from newer CM kernels, but at the moment the Nexus reboots directly when put to sleep or during boot or in the first few seconds after start...hard to debug

deprecated | # 307 | 2016-12-02, 21:29 | Report

Yeah, I get 5 days on the Nexus 5 with copperheadOS, a stripped down security-centric Android revision without Google Play and the like. At this moment I'd be thrilled with over one day of battery without the 1% drain per minute with the screen on (web browsing, etc.) on Sailfish. I LOVE the OS, and the battery was pretty good in the older runs of 2.0.1.11, what changed is what's ruining the battery life I think.

We'll see what happens, but I'd be interested to see how removing the 300Hz timer thing would impact battery life. Also interested to see what you come up with regarding the backported features!

Edit: I just thought about it. The Jolla 1 used a 2100mAh Li-ion cell, where the Nexus 5 uses a 2300mAh Li-Po cell. I doubt the battery chemistry could be causing this issue, and the capacity is a little more... Oh well, we'll figure it out eventually!

mautz | # 308 | 2016-12-03, 16:45 | Report

I got some good and some bad news...

I've got SFOS booting with cm14.1 kernel, but it's not fully working. Lipstick is crashing on start, reboots and everything is fine. Wifi is not working, Telephon app does not start...maybe this problems could be solved...i'll first try a kernel based on CM13...

deprecated | # 309 | 2016-12-04, 01:08 | Report

I'm in agreement here. We'll see what mautz has in store for us coming up. Maybe we can build a test kernel with the 100Hz timer instead, see if that makes a difference - which I'm almost positive it will.

Excellent find, @minimec! And great work @mautz! I'm starting to get excited.

minimec | # 310 | 2016-12-04, 12:45 | Report

So I set up that cross compile build environment and wanted to build a 100Hz kernel.

I tried with the 'gcc-arm-linux-gnueabi' toolchain from the ubuntu repos, but also with 'android-ndk-r13b' and 'arm-eabi-4.8'. I followed different build instructions and also tried the 'make_defconfig.sh' script, found in the 'mautz' kernel sources...

After the most painful and frustrating two hours in my whole Linux career, I had to take a break, otherwise ... https://www.youtube.com/watch?v=JtEkUmYecnk&feature=youtu.be&t=589

I can make the defconfig and run 'make menuconfig', but the building process would always fail with different error messages, depending the toolchain I use...

@mautz: Believe me... I tried... Maybe you can explain me (us) your 'kernel compile work-flow' and kernel build environment. It would help me rebuild my broken ego...