How's everyone's [55Wh] battery health and wear looking?

Adrian_Joachim · April 9, 2023, 4:21pm

Yeah that probably explains the variance, you using windows or something else?

I’d bet it doesn’t discharge all the way down but just a little further each time which would explain your initial downward trend.

Framework really should implement the online re-calibrate routine in the ec to eliminate that (hell it’s open source I might take a crack at that if they don’t).

That’ll just give you even messier data, the dc-dc efficiency changes a lot more than the charging efficiency does. You could use a slow charger to reduce the charging efficiency differences even more if you want to do that.

Good enough I suppose since that is how it has been done for ages now but it is certainly not perfect not that perfect really is required.

Since the wear curves are usually not linear idk if you can extrapolate like that, they also afaik didn’t define what a cycle means and your battery may actually have more than 55Wh (even factory new cells have varying capacities) from factory which would hide the actual wear rate for a bit.

With the 18650s the roule of thumb was about double the cycles for every 0.05v less max voltage (of course you loose capacity initially but you’ll keep it longer), so if a 4.2v max cell is rated for 80% capacity after 300 4.2->2.5v cycles it would get 80% capacity after 600 4.15-2.5 cycles and 1200 4.1-2.5 cycles and so on. Given that there usually wasn’t all that much capacity at the higher voltages this tends to be a good tradeoff.

I have not seen too much data on the spicy 4.40 or 4.45 chemistry the framework battery uses but I’d bet it has similar wear characteristics so your only charging to 4v per cell should increase your longevity dramatically, assuming you aren’t grilling the battery cause temperature causes wear too independently.

Anyway, point is you should be careful drawing conclusions from data that is more noise than signal, humans are made to see patterns so we often tend to see them where they aren’t or misinterpret stuff.

Would be neat to get a datasheet for the cells in the framework batteries but I doubt that’s going to happen.

amoun · April 9, 2023, 4:39pm

The exercise is more of providing some background of what and how the readings are representative of charge limits and charging, as the topic is a question.

That the results may not be truly representative of actual battery wear they can be representative of use.

If the reading of coulomb count is fairly accurate then I can compare what the capacity is relative to when it was received, that gives me an idea of wear.

UPDATE: May 1st: I have another way of showing the wear by comparing the recorded cycles and the values wear.

I hadn’t managed to get a record of the cycles until a specific BIOS performed that, I think it was 3.09, so it is from then data was available

The wear was 3.4% and dropped to 2.9 however it wasn’t for a few months until I took daily reading when one day in May I saw the ‘wear’ as nearly 7% and the cycles around 50

So I have compared the cycles to the wear based upon Frameworks claim that the battery will loose 20% capacity in 1000 cycles, or 1% every 50 cycles on average.

So the [Blue] line on the graph is the cycles divided by 50 hence 1 and up, given I started at 50 cycles.

The [Red] line is the wear minus 3.2 (as a figure between the 3.4 and 2.9 that were shown via powercfg in the early days)

Ideally these lines would coincide or at least run parallel. It is also worth noting that wear or capacity reduces quite quickly in the first 100 cycles although this is greatly influences by temperature and time. Ref to be included, or see a link in one of my previous posts

Adrian_Joachim · April 9, 2023, 5:38pm

Just don’t read too much into a couple percent swings and you should be good.

amoun · May 11, 2023, 7:00am

Just an update on how fickle the readings are.

I usually use plugged in and have the limit set to 66% but inadvertently the mains wasn’t secure and it dropped to 60%, so I engaged the plug and then thought I 'd check the ‘wear’ again, having only done it 40 minutes before.

It could have gone either way but it dropped from 4.1 to 2.8
UPDATE: 16th May The wear is recorded at 2.4 ???

Adrian_Joachim · May 11, 2023, 7:35am

So much for not reading too much into a couple percent, especially without a full discharge and charge cycle.

Running plugged in (especially with a charge limit) leads to drift in the coulomb counter over time, that plus maybe a bit of battery self discharge would lead the battery to estimate it’s capacity a bit higher on a partial charge. That wear level is an educated guess of the battery and with it never reaching one of the extremes it tends to drift more and more from educated towards guess over time.

Look at it this way, it’s like you are measuring the distance between work and home by counting steps, that can be quite accurate as long as you reach home or work from time to time and start counting from there. If you just move a couple steps forwards and backwards over and over again, it turns less accurate.

Alex_K1 · May 25, 2023, 6:01pm

Hi everyone!

Hoping to get some help here.

Running an 1135G7 that I got in March 2022.
Using HWinfo64, my battery shows 10.2% wear after only 138 cycles.

I’m not really happy about this to be honest, as the claims by Framework that the battery is designed to retain 80% capacity after 1000 cycles seems to be a bit misleading. I’m not anywhere close to halfway there!

I’m not really using this laptop for heavy tasks. Light gaming while pugged in, but on battery it’s for document writing, web browsing, and watching videos. I don’t really use it off power that much.

So yeah, hoping to get some input. I’m not ready to be thinking about upgrading the battery, it’s still practically brand new!

amoun · May 25, 2023, 8:49pm

Hi Unless you have a reading from the beginning it’s hard to say how much of the wear is down to initial capacity. The battery may say 55W nominal which is used as a base line but could have been only 52.5 and appeared as nearly 5% from the start.

Mine was apparently 2.9% and then rose to 7.4% in maybe 30 cycles. (29.6% in 120 cycles ) So I thought I’d see if it was a calibration issue. I will never really know for a few years

Then I tried calibrating by discharging and a full recharge. Every time I get a high reading I recalibrate and over a year I have only 64 cycles and an average of 4.5%
or maybe 9% for 128 cycles. In that sense what you have is comparable.

So I have a graph that compares recorded wear ( I check a few times most days and take the lowest) and compare that to the wear that would/should have been seen by cycles

On the graph I’ve divided the cycles by 50 to get them on the same level.

The red is the expected wear given the number of cycles ( 1000 for 20% [50:1] ) the other the wear as recorded.

As you can see the ratio is getting better. Data line was aligned to the cycles[red] that would be in tune with expectations. You can see how it is getting better and below the red means it is better than expected.

But note this data is from checking every day, so if you don’t do that the odd recorded measurement can be way of by a multiple at least 4 in my case 7.4% at 30 cycles.

All my data is at
http://217.155.51.23/newdevices/framework/battery/may23.php
graphs at
http://217.155.51.23/newdevices/framework/battery/graph.php

Adrian_Joachim · May 25, 2023, 9:25pm

Have you done a full 0 to 100% charge cycle lately?

amoun · May 25, 2023, 10:09pm

Just to note a full discharge isn’t really possible as there are safety issues, the OS/BMS won’t allow such.

Set battery charge limit to 100%
Run down till the laptop switches off
Restart an hour later and maybe use for another 30 min until auto shutdown
Restart again after 1 hour and maybe another 10 min
Further restarts won’t work etc.

Recharge to 100% check wear
Keep at 100% for the rest of the day check wear occasionally.
Next morning set Battery charge limit to personal pref, mine is 66%
Check wear for a few days.

Adrian_Joachim · May 26, 2023, 5:54am

The point where the bms taps out is the defined 0% point, the os on the other hand is another issue.

Framework really should implement the lenovo style online recalibration mode, would make this a lot less confusing.

Alex_K1 · May 26, 2023, 11:50pm

Appreciate the help!

I just did a 100% to 0% and charged back to 100%.
I’m showing 10.4% wear after 140 cycles now. Seems like it’s pretty consistent here.

On the FW marketplace, the 55Wh battery has a “typical capacity” of 56.6 Wh, and a Rated Capacity of 55Wh. So that would make sense that the battery comes with wear from the factory, but this is sort of a manufacturer guarantee that it at least should contain 55Wh. If a new battery has less out of the gate, then the manufacturing standard is not upheld (but feel free to correct me ).

I think where this is coming from (for me as a consumer), if I read “designed to have 80% capacity after 1000 cycles”, then the message or meaning that I get is “This battery degrades slowly” or “this battery has a long lifespan”. And when I see that I have 89.6% capacity after 140 cycles…that leaves me scratching my head as to whether there is an issue.

amoun · May 27, 2023, 8:52am

You will probably have to check a few more times, not reset and try reducing the max BCL to 80% for example.

I’ve just had mine go up a percent, so I reset the BCL to 78% from 66% and the wear went down 1%.

The readings are fickle, but I have an average since Feb 2022 of around 4.5% after 64 cycles. If that is representative then 9% after 128 cycles which is equates to about 10%

However what I also have is a graph showing the percentage difference between a presumed wear by cycles ( at 20% for 1000) and the recorded wear. Both sets of data are adjusted to align on the graph.

From over 7% at 50 I am now getting 4.5 @ 64

On the graph an expected wear would follow the red line a figure off that shows a 100% per point error so the worst was 350% above the best 100% below.

But again I must emphasise I check every day and adjust the Battery Charge Limit, or a full reset at least every month.

So I have a wear has been nearly four times what I expected and the odd occasion where it is half what was expected.

I can only suggest you try changing the battery charge limit a few time in the immediate future an check the wear even twice a day to see how it may change.

All the best.

Updated 24th June 2023

Design wear is the ration of 20% wear in 1000 cycles as specs i.e. 1 per 50 cycles

Other graphs and data at
http://217.155.51.23/newdevices/framework/battery/graph.php

Wolf_Trabeitt · June 12, 2023, 8:03pm

Great to see that you are still continuing this Amoun.

As for me, the use case of my Framework has greatly changed from a work machine used on battery 5-7 hours a day to being plugged in and used nearly 24/7, aside from during my drive to and from work. Battery wear doesn’t worry me nearly as much as it did when I got it in Sept '21 and used it in a classroom without an outlet near the board. Now I have a new campus that has outlets everywhere. 1.75 years later, I’m at around 13% wear, which is the same as my last posting here in Nov '22, but never use it off the charger more than a few hours at a time since then, and haven’t seen it under 50% charge in … well, recent history.

I’d love to add to the data, but I don’t have the foresight or consistency to check that you do. I also feel like anything short of 15% wear is getting overly pedantic for most people’s use case, anyway. When the battery has more than 20% wear, I’ll get worried. It looks like we’re getting more than a couple years use out of the lithium batteries, and that batteries of a higher capacity are now in the pipeline. I think I’m very happy with how things have turned out overall.

Watching your exercise in battery monitoring has been great over the last year. Just thought I’d throw a couple more cents at this.

Wolf

anarcat · July 21, 2023, 4:46am

I must admit I haven’t read the whole thread here, but I was about to create a topic on this and figured it was better to join the flock than shout on my own soapbox…

Anyway. I found my fork of the battery-stats program to which I contributed a script that will generate a nice graph of the life of your battery, provided that you already have the collector writing to the CSV file once in a while.

Here’s my rendering of the battery right now, after slightly less than a year of use:

The red line is the evolution of the battery health. According to the current levels, it’s at 3286000 out of 3572000 design capacity, which is 91.99%. You can clearly see “drops” in that line where the battery was somewhat “damaged” in one shot, which is, I think, the most interesting part of that graph.

I wish I could remember what happened mid-april or mid-may there! Those are two 5% drops that are basically the reason why the battery is ~10% used. I actually keep a life journal here (on paper, believe it or not), and didn’t not anything particular on those days, for what that’s worth. There’s a somewhat deep discharge before the may event (<10%) but nothing unusual, there’s plenty of complete discharge cycles out there that didn’t affect the battery that much.

This is the current tlp-stat output:

--- TLP 1.5.0 --------------------------------------------

+++ Battery Care
Plugin: generic
Supported features: none available

+++ Battery Status: BAT1
/sys/class/power_supply/BAT1/manufacturer                   = NVT
/sys/class/power_supply/BAT1/model_name                     = Framewo
/sys/class/power_supply/BAT1/cycle_count                    =    151
/sys/class/power_supply/BAT1/charge_full_design             =   3572 [mAh]
/sys/class/power_supply/BAT1/charge_full                    =   3286 [mAh]
/sys/class/power_supply/BAT1/charge_now                     =    648 [mAh]
/sys/class/power_supply/BAT1/current_now                    =   2600 [mA]
/sys/class/power_supply/BAT1/status                         = Charging

/sys/class/power_supply/BAT1/charge_control_start_threshold = (not available) 
/sys/class/power_supply/BAT1/charge_control_end_threshold   = (not available) 

Charge                                                      =   19.7 [%]
Capacity                                                    =   92.0 [%]

Interestingly, I haven’t used the laptop as my daily driver until about March 2023. Before that, I used it to do extensive tests so it spent almost half of its life being the subject to my torture, but none of that time actually caused problems to the battery. If anything, my normal daily use is more of a problem.

If I would venture a guess, my bad habit is to take the laptop from work, slap the lid down (so it suspends, presumably), shove it in a bag and forget about it. It happened a few times that I opened the lid after a few days and realized I forgot to plug it in, and the battery drained out on suspend.

I really wish Linux had some built-in (hello systemd?) way of waking up from suspend to do an emergency shutdown when the power is low. Surely there’s something that can be triggered somewhere to enable this instead of catastrophically failing the way it is now…

Anyway, that’s my data! happy to share the actual CSV file with Framework staff if that’s useful. Data is logged every minute or so.

Adrian_Joachim · July 21, 2023, 7:11am

Those are likely not when the damage happened but when the damage is noticed, which do seem to coincide with some of the closer to full discharges.

They way the battery measures its capacity is imperfect and will drift a little over time (imagine walking 10 steps forwards then 10 steps back with your eyes closed over and over, you’d probably drift a bit too XD), you might want to give it a full discharge (like as full as you can, basically until the battery taps out not the os) followed by a full charge to see how it actually looks.

amoun · July 21, 2023, 9:51am

Yes similar thoughts: and like wise an increase after 6 months ??

Thanks for providing your data: Can you show the cycle count too ?

All the best

anarcat · July 23, 2023, 1:47am

That could be when I started keeping the laptop mostly plugged in at work.

This?

I read one hundred and fifty one (151) cycles… is that good or bad?

anarcat · July 23, 2023, 1:49am

Interesting. I have done basically this recently, and now the full size went down:

/sys/class/power_supply/BAT1/charge_full_design             =   3572 [mAh]
/sys/class/power_supply/BAT1/charge_full                    =   3279 [mAh]

amoun · July 23, 2023, 7:02am

Yes it’s a bit ‘random’.

From the 151

According to specs 100 should equate to a 2% drop (variables) so 3% in your cycle case. (20% in 1000 cycles)

My cycles are only 71 so would expect a wear of 0.71% but what I have, this week is an average of 4.7%, but I take into account that it was around 3% down when I started.

There is also general degradation. So maybe 1000 when all done quickly, how do they do these tests.

My Laptop is now 17 months old.

Here’s from Feb 2022 to 23rd July 2023

Here’s the latest (ist Aug 2023) of wear compared to expected as a par on cycles, currently nearly 300% over what is spec.

Adrian_Joachim · July 23, 2023, 6:07pm

Full discharge then charge is the only way the battery can find out it’s actual capacity so it may naively look like full discharges directly damage the battery when it is actually only when it notices the damage.

Did they claim a linear decline in capacity? Battery wear tends to not be linear.