Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Somone managed to run it on a M1 MBP with Big Sur? #52

Closed
Peebuddy43 opened this issue Nov 20, 2020 · 276 comments
Closed

Somone managed to run it on a M1 MBP with Big Sur? #52

Peebuddy43 opened this issue Nov 20, 2020 · 276 comments

Comments

@Peebuddy43
Copy link

Somone managed to run it on a M1 MBP with Big Sur?

@JiaaackY
Copy link

I tried on my m1 air. It doesn't work. I installed the helper tool and reboot my air but it still keeps charging.

@davidwernhart
Copy link
Collaborator

@JiaaackY thank you for this information! I already suspected that it would not work on M1 macs since they have no SMC like the intel ones have. Maybe I will be able to find a similar solution for M1 macs in the future, but for now I don't have a device to test. Best regards, David

@FreeWoRLD83
Copy link

@JiaaackY thank you for this information! I already suspected that it would not work on M1 macs since they have no SMC like the intel ones have. Maybe I will be able to find a similar solution for M1 macs in the future, but for now I don't have a device to test. Best regards, David

You can't imagine how many folks out there awaiting eagerly for the M1 version as this is the most amazing app you could ever install on your mac if you are careful about battery of your device. Love your work.

@asd123cqp
Copy link

I believe SMC should still exist on ARM Macs, it's just that the key BCLM has been renamed or removed.

I complied the smc tool from smcFanControl on an ARM Mac.

  • smc -l still shows a ton of stuff
  • smc -l | grep BCLM shows nothing (on an Intel mac it should give something like BCLM [ui8 ] 60 (bytes 3c))

@DevNulPavel
Copy link

DevNulPavel commented Nov 26, 2020

@asd123cqp
Can you show your output of smc -l command?

@asd123cqp
Copy link

Sure, here's the smc -l output on a M1 Macbook Air: smc-l-arm-mba.txt.

@davidwernhart
Copy link
Collaborator

Thank you for dumping the smc keys @asd123cqp ! That is very interesting to me, I suspect Apple simply renamed some keys for the new M1 macbooks. From looking at your dump, I would guess that the key "BCMV" (maybe "battery charge max value"?) could be the right one. If you are brave enough, you could maybe try to write to it and see what happens. Although I must warn you that you can also brick your MacBook by writing to the wrong keys and I do not take responsibility if that would be the case.
Anyways, thank you for sharing your findings!
Best regards, David

@DizzyFop
Copy link

Thank you for dumping the smc keys @asd123cqp ! That is very interesting to me, I suspect Apple simply renamed some keys for the new M1 macbooks. From looking at your dump, I would guess that the key "BCMV" (maybe "battery charge max value"?) could be the right one. If you are brave enough, you could maybe try to write to it and see what happens. Although I must warn you that you can also brick your MacBook by writing to the wrong keys and I do not take responsibility if that would be the case.
Anyways, thank you for sharing your findings!
Best regards, David

How could you brick it? Would resetting the SMC not fix an issue caused by this?

@davidwernhart
Copy link
Collaborator

@Superjack78 because we don't know what those keys do for now. Maybe BCMV actually stands for "battery charge max voltage" and modifying it actually messes with the charging circuit, having the potential to do some permanent damage to the battery even if you reset the SMC afterwards.
The point is that we just don't know yet and there is always some risk attached to messing with the SMC. I got lucky with finding the original BCLM key, but I cannot guarantee that this is the case for everyone.

@asd123cqp
Copy link

asd123cqp commented Nov 26, 2020

From looking at your dump, I would guess that the key "BCMV" (maybe "battery charge max value"?) could be the right one.

@davidwernhart I doubt BCMV is the renamed BCLM. Its current value is BCMV [ui16] 23312 (bytes 5b 10) and I am not sure how to make sense of this number. Maybe this is the new charge limit and it is now an absolute value instead of a percentage. Maybe it's totally unrelated. I don't know.

There are two values that stand out to me though. I searched the dump for keys that start with an "B" and have the value "100" and found these two: BMSC [ui16] 100 (bytes 64) and BTRS [ui8 ] 100 (bytes 64). Again, not sure what they stand for and might be totally unrelated. I don't know.


Update:

  • BCMV already exists on Intel Macs, so it's unlikely to be the new key for charge limit.
  • BMSC and BTRS are new. Could be the chosen one.

@smoothdvd
Copy link

BTRS is for Bluetooth?

2018-08-07 09:16:29.919311+0100 0x7b Default 0x0 0 0 kernel: (IOBluetoothFamily) **** [IOBluetoothACPIMethods][SetBTRS] -- mACPIDevice->evaluateObject ('BTRS') successful -- took 23211 microseconds ****

https://github.com/khronokernel/DarwinDumped/blob/b6d91cf4a5bdf1d4860add87cf6464839b92d5bb/MacBook/MacBook9%2C1/BootLog_Kernel/Kernel_Messsages_BootLog.txt

@thloh85-intel
Copy link

thloh85-intel commented Nov 30, 2020

@asd123cqp Wonder if you have managed to try out writing to BMSC?
Note: I can't test it yet as the M1 Macbook isn't available in my home country yet. Just checking if I'm still able to limit the charge when I buy the new M1 MacBook since my Macbook sits on my charger all day

@timension
Copy link

I find it very strange that so many people (including me) want this feature (which is very sound, considering the sustainability etc.) but Apple cannot provide it natively. :-/

@lasharor
Copy link

lasharor commented Dec 1, 2020

I find it very strange that so many people (including me) want this feature (which is very sound, considering the sustainability etc.) but Apple cannot provide it natively. :-/

I believe it to be even more relevant now as with the increased battery life it becomes somewhat difficult to get this laptop to "healthy" battery levels in regular office use.

@tairasu
Copy link

tairasu commented Dec 1, 2020

I find it very strange that so many people (including me) want this feature (which is very sound, considering the sustainability etc.) but Apple cannot provide it natively. :-/

I believe apple doesn't want to spread the fact that keeping your battery at 60% is keeping it healthy. if too many people know it, the "user experience" might suffer, because then people are afraid of charging it up to 100%. they rather hide it as a "battery optimization" feature, that is out of control of the user

@thloh85
Copy link

thloh85 commented Dec 4, 2020

Did anyone tried out changing BMSC? I want to buy the new M1 MacBook, been using AlDente for a while, and wish to know if I can still use it for M1 MacBook.

@stefandesu
Copy link

I'm slightly worried that this feature was removed from the M1 MacBooks. The support article "About battery health management in Mac notebooks" has two versions:

The latter doesn't talk about the "battery health management feature" at all. 🤔

@lasharor
Copy link

lasharor commented Dec 8, 2020

I'm slightly worried that this feature was removed from the M1 MacBooks. The support article "About battery health management in Mac notebooks" has two versions:

The latter doesn't talk about the "battery health management feature" at all. 🤔

Screenshot 2020-12-08 at 21 48 09

M1 Macbook Air

@stefandesu
Copy link

Screenshot 2020-12-08 at 21 48 09

M1 Macbook Air

What I meant is the option "Manage battery longevity" under Battery Health (see screenshot in the Intel article). That feature is missing on M1 Macs (at least on my MBA).

@thloh85-intel
Copy link

thloh85-intel commented Dec 9, 2020

Well as long as the hook to enable "Optimised battery charging" is there, there should be a way to limit the charge.
However, if the hook is now placed in the firmware (ie. ARM's SMC, secure monitor call, not system management controller, callback handles Y/N to optmize battery charging instead of the percentage), then we're out of luck.
If the Secure Monitor Call still accepts battery percentage value instead of a typical Y/N, then we're still able to do this, with proper privilege.
I hope they didn't kill this.

@stefandesu Are you planning to try out changing BMSC's value?

@stefandesu
Copy link

@stefandesu Are you planning to try out changing BMSC's value?

I'm not willing to take the risk, sorry. 🙈

@mattiasottosson
Copy link

I don't really get how the optimised battery charging works. Had that checked since day one on both my iPhone and M1 mac, but haven't seen it stay at 80% once.

I use my macbook at work plugged in to a screen, so it stays at a 100% at all times now unfortunately.

Maybe start a BMSC insurance campaign at gofundme? 🤓

@celsoazevedo
Copy link

celsoazevedo commented Dec 10, 2020

I tried on my m1 air. It doesn't work. I installed the helper tool and reboot my air but it still keeps charging.

With Big Sur, I had to set the % max and reboot my Intel MBP 16" twice for the limit to be applied.

Not saying that AlDente works on the M1, but I would try set the limit and reboot more than one time. Also, if we set it to stop at 80%, sometimes it goes over that (usually 1 or 2% above the value we set).

@Evertt
Copy link

Evertt commented Dec 10, 2020 via email

@thloh85-intel
Copy link

No it just doesn't work on M1 MacBooks.

It won't work without proper changes definitely. As per @asd123cqp's check, the BMCV SMC value is no longer there, thus we need to look for the proper value to change (if it exists)

@Evertt
Copy link

Evertt commented Dec 11, 2020 via email

@davidwernhart
Copy link
Collaborator

I don't really get how the optimised battery charging works. Had that checked since day one on both my iPhone and M1 mac, but haven't seen it stay at 80% once.

I use my macbook at work plugged in to a screen, so it stays at a 100% at all times now unfortunately.

Maybe start a BMSC insurance campaign at gofundme? 🤓

This sounds like a good idea, but I honestly doubt to get enough backers to afford a ~1100€ M1 MacBook Air to experiment with. I guess it's much more likely that someone who already owns one finds the right key in the meantime.

Best Regards,
David

@thloh85
Copy link

thloh85 commented Dec 12, 2020

Honestly, I suspect they moved it to firmware level callback (ie. the ARM SMC callback) instead of the SMC value.
I've been working on ARM bootloader and are familiar with SMC callback's functionality, and this is one of the implementation that I personally would put in as an SMC call. As for how it is called, what value, it really is up to the kernel level driver. In Linux, it is open source, thus is pretty open and accesible how one can set it, but in MacOS, not so much. However, Darwin kernel is open source, so, in that department, we might find something :)

@Ryanfsdf
Copy link

I've figured out some more clues in case anyone is willing to risk their their laptop.

When I first got my M1 MBA, there were 4 values in the SMC which had a value of 100. These were BMSC, BNSC, BTRS, and BUIC.

Upon checking now, I see that BMSC and BNSC are both at 99 instead of 100. They seem to be related in some way, and the fact that they dropped to 99 seems to suggest that they are percentages.

However, even though both of these values are 99, my laptop still charges all the way up to 100%. This may not mean much though, because I've noticed on all my MacBooks that even if my laptop battery is ~97% full based on actual current battery capacity divided by the maximum capacity, the laptop doesn't charge beyond that and the battery percentage reported is still 100%.

@thloh85
Copy link

thloh85 commented Dec 13, 2020 via email

@joelucid
Copy link

joelucid commented Feb 23, 2021

@joelucid Going to try make some more sense of this myself now.
According to BetterBattery2, date of manufacture was Oct 12, 2020 (4.4 months old).
Started using this MBP M1 on Dec 5, 2020 and currently on 15 charge cycles. Used to charge between 50/60 - 80% before Eternal Power and AlDente were released one week ago.

Yeah it's weird - I've trying to make sense of what the gas gauge does in MacBooks. Especially when it updates qmax (the no load capacity of a cell). It doesn't look like it behaves like it should according to the technical manual of the chip.

So far it seems that qmax is only updated after the first discharge after a full charge. This would be bad since it requires periodic full charging to keep track of battery health. And resistance updates get disabled after some time until full charge as well. Given apples "optimized" battery charging it's possible that a 80% charge counts as full charge for this. Currently testing if that's true.

I think getting to the bottom of how to get a realistic battery health measurement is essential to make use of a tool like AlDente or EternalPower intelligently. We don't want to optimize the coconut bat full charge capacity if our regimen prevents it from updating to reality :).

@tomschlenkhoff
Copy link
Contributor

@tomschlenkhoff:

00 7e 00 6c 00 78 00 84 00 96 00 6c 00 6b 00 72 00 77 00 7d 00 77 00 8a 00 ca 01 58 02 10 
00 72 00 64 00 6f 00 7c 00 8a 00 63 00 61 00 66 00 6c 00 77 00 73 00 8d 00 e1 01 ad 02 8e 
00 77 00 66 00 74 00 7e 00 91 00 6a 00 6b 00 6e 00 73 00 7a 00 78 00 8d 00 cd 01 67 02 2a 

This one seems different - less of a difference in cell 3. Is this also an air or a pro 13?

This is an Mac Book Pro 13.

@tomschlenkhoff
Copy link
Contributor

This is the same MBP again, now after spending 12h being charged to 100% to check if this changes anything:
{"Ra03"=0,"Ra10"=0,"CellWom"=(0,0),"RaTableRaw"=(<0000007e006c007800840096006c006b00720077007d0077008a00ca01580210>,<000000720064006f007c008a006300610066006c00770073008d00e101ad028e>,<0055007600650073007d00900069006a006d0073007a0078008d00cd0166022a>),"Qstart"=0,"AdapterPower"=963952690,"DailyMinSoc"=86,"Ra04"=0,"Ra11"=0,"CellVoltage"=(4085,4088,4084),"PackCurrentAccumulator"=18446744073707152949,"PassedCharge"=669,"Flags"=16777217,"PresentDOD"=(15,15,15),"Ra05"=0,"Ra12"=0,"FccComp1"=5176,"ChemID"=9216,"iMaxAndSocSmoothTable"=<0000000000000000000000000000000000000000000000000000000000000000>,"FccComp2"=5176,"PackCurrentAccumulatorCount"=10591,"DOD0"=(528,512,544),"Ra06"=0,"ResScale"=0,"Ra13"=0,"WeightedRa"=0,"FilteredCurrent"=0,"RSS"=0,"CellCurrentAccumulatorCount"=0,"Serial"="D860505A9FPL3L2ED","DailyMaxSoc"=99,"Ra07"=0,"Ra14"=0,"MaxCapacity"=100,"ChemicalWeightedRa"=0,"Ra00"=0,"BatteryHealthMetric"=0,"DesignCapacity"=5103,"Ra08"=0,"BatteryState"=<000000000000000043e4000205000002>,"CellCurrentAccumulator"=(0,0),"AlgoChemID"=9216,"MfgData"=<00000000100200012400000003313533033030330341544c013f000000000000>,"ManufactureDate"=52987853617202,"ISS"=18446744073709551046,"Ra01"=0,"Soc1Voltage"=0,"QmaxDisqualificationReason"=0,"ChargeAccum"=0,"SimRate"=0,"Qmax"=(5509,5531,5507),"PMUConfigured"=0,"ITMiscStatus"=0,"StateOfCharge"=86,"Ra09"=0,"GaugeFlagRaw"=192,"CycleCount"=20,"Voltage"=12243,"SystemPower"=1089035415,"LifetimeData"={"Raw"=<00002b7a0007ff5200000000000000000060d61f0080e467b240000000000000016a004511050adf3302243210d8f2b71491f0bbfc06fb6700cf00006e2e000b>,"UpdateTime"=1614073115,"ResistanceUpdatedDisabledCount"=0,"TimeAtHighSoc"=<03000000910000000000000000000000>,"TemperatureSamples"=28206,"TotalOperatingTime"=1762,"MaximumDischargeCurrent"=18446744073709548215,"MinimumPackVoltage"=9266,"MaximumPackVoltage"=13058,"MaximumChargeCurrent"=4312,"AverageTemperature"=18446744073709551567,"MinimumTemperature"=69,"RDISCnt"=0,"MaximumTemperature"=362},"Ra02"=0}

@joelucid
Copy link

joelucid commented Feb 25, 2021

Kind of shocking how relevant a good cycling routine is to battery health measurement. My 2018 MacBook pro 15" used to flagged as needing battery service with ~77% of battery health. A couple of cycles between 75% and 10% with about 2 hours wait time at each level and I'm back in good standing with battery health of 82%, sometimes up to 85%. Qmax is still increasing each cycle:

maxQ0: 6965
maxQ0: 7107
maxQ0: 7297
maxQ0: 7368
maxQ0: 7443
maxQ0: 7523

I think the core problem which caused the bad battery warning was cycling the battery with external 4k display connected. The discrete graphics card just draws too much to work well with a battery. These new cycles were done without external display at a power draw of about 10W.

The laptop can draw 56W if its busy doing graphics. I think that pushes the battery internal resistance into the non linear range and messes up the calibration. And of course sitting at 40 C and 100% SOC for two years can't have helped.

BTW there's also some evidence that capacity lost by sitting or cycling at very high SOC can be recovered by resting or even cycling at lower state of charge. Not sure if that was a factor in the rejuvenation of this laptop. If it was then running a couple of calibration cycles might have unmasked the improvement.

image

image

@joelucid
Copy link

joelucid commented Feb 25, 2021

Much worse news for the MacBook Air M1. It's battery sucks. It's only been at 100% battery health since I never cycled it in the region where the underlying calibration data is updated. After a couple of cycles like above it's now at 96% battery health - after 12 cycles total. Looks like I got a rather bad battery. I'll keep cycling it until the calibration steadies and then I'll try to get a replacement.

@joelucid
Copy link

Long term I think one calibration cycle per month is absolutely necessary to keep track of its battery health. Good that 75% max charge is enough and you don't have to torture the battery at 100% for it.

@bigyanphobia
Copy link

Long term I think one calibration cycle per month is absolutely necessary to keep track of its battery health. Good that 75% max charge is enough and you don't have to torture the battery at 100% for it.

Sorry for the stupid question but I'm wondering to what percentage must we discharge for calibration?

@joelucid
Copy link

I've had success getting qmax updated when charging to 100% and then discharging to 0% or 55%. And 75% to 10% also works. Important is to let the battery sit both charged and discharged for a long time - like two hours. Once the DOD0 updates you can move on. Important is you need to charge to >= 70% and then discharge to >=~50% or <= ~10%. And the difference between both levels needs to be >40%.

If your battery has been at a constant charge level for years you'll have to do this a couple of times. qmax needs to be updated and the RaTable as well. And both interact: for example when qmax is increased the gauge also decreases the values in the RaTable. So it takes a couple cycles to converge.

Of course for true knowledge of capacity you need to do 0% - 100%. Did this with my iPhone and that really is still at 105%. For my laptop I'll probably just monitor how 75% to 10% changes with time. Should give a good indication of aging.

@stanleydesu
Copy link

Turned off charge limiter overnight to see if MacOS's BHM would work. Have been on AC at least 95% of the time in the past 2 weeks (device is ~3 weeks old).
Woke up to 100%, seems MacOS still hasn't learned my charging patterns.
Screen Shot 2021-02-28 at 8 34 21 am

@joelucid
Copy link

You might need to have the laptop sit at 100% for it to kick in. At first I thought the “machine learning” they use in it was just marketing but the PowerUI agent really does run neural networks to somehow decide what to do with your battery. It’s baffling.

@manuelmenzella
Copy link

manuelmenzella commented Mar 20, 2021

These are fantastic tools, thank you!

@joelucid, I was wondering -- is there any downside to repeatedly writing to the CH0B key? Is this actually writing to some kind of NVRAM that may have limited number of writes? Can this degrade the non-volatile memory?

@cjmaria
Copy link

cjmaria commented Mar 20, 2021

@manuelmenzella My guess would be that the "NVRAM" cells in the M1 chip are still capacitor-based memory that does not wear out like NAND flash does. At least on the M1 mac mini, there is still a NVRAM battery that could support this functionality. On the M1 MBP, perhaps they just use the device's main battery to do this.

@inprealpha
Copy link

@manuelmenzella Would you say that the risk is substantial? Doesn't the system itself write far more often to the NVRAM when charging and doing other stuff?

@manuelmenzella
Copy link

@inprealpha I think you're probably right; I don't really have a good intuition for how often writes happen, though.

@cjmaria good to know, thank you. It would be great to have some confirmation but I suspect this is safe.

@MatthiasKerbl
Copy link
Collaborator

Hi Guys,

AlDente Pro is finally out. You can check it out on our new website: https://apphousekitchen.com/
As promised, everyone who has helped reverse-engineering the new keys will get a 50% discount on AlDente Pro.
Just send us an email to support@apphousekitchen.com with your request for the discount.

Best regards,

David & Matthias

@krackers
Copy link

krackers commented Jan 5, 2022

@joelucid Excellent insights, and your conclusion that SOC of ~30% is good for storage matches with my research as well. See [1], [2], [3]. I think you already covered this with the 20% lower bound, but since the battery is used for burst current draw even when plugged in, another factor in the mental model should be the battery resistance at varying SOC.

Luckily since as you mentioned newer laptops expose this information in the SMC we can query that directly. I'm not sure precisely how the intervals are broken up, but assuming each interval is the same there seem to be 16 intervals so I'll assume that each region is about 7% SOC. On my laptop the resistance was roughly the same (0x33, not sure what units though) from 40-20, rose after 20 (and especially sharply after 10), was the lowest (~0x27) around 40-80, and rose after 80+.

So anywhere from 40-20 seems to be a fine region for regular use (low calendar aging, relatively low resistance during active use, and low resistance after aging as seen in one of your charts). If you are always plugged in and never use battery, then 30% seems like a decent value to use?

I also agree that people probably cite ~40-50% storage SOC to account for discharge buffer, but for a battery that's plugged in this is irrelevant. Another factor might be what they're considering 0% SOC though? The first paper you linked stated that they considered 20% SOC as 3.6V I believe. You can also query this information from SMC to try to match it up (I'm getting 3.7-ish V at 30%).

One question: What are your thoughts on maintaining a target charge level (i.e. just setting BCLM directly and having the laptop trickle-charge as needed to maintain that level) vs. setting a lower and upper bound (what AlDente calls "sailing mode")? The former has lower DoD, but most charge limiters from Manufacturers (e.g. Lenovo) implement the latter approach, so I wonder if it has any advantages?

[1] https://old.reddit.com/r/batteries/comments/qu60z2/halfcharged_vs_halfdischarged_lithium_ion_battery/hkq6gy6/?context=999

[2] https://old.reddit.com/r/batteries/comments/pyn6fr/tip_how_to_increase_battery_life_and_battery/hf6psuu/

[3] https://old.reddit.com/r/batteries/comments/owjt2j/the_4080_rule_myth_or_truth/h7gmhlu/?context=999

@joelucid
Copy link

joelucid commented Jan 5, 2022

@krackers: the internal resistance table isn't scaled linearly since more resolution is needed at the lower end where internal resistance changes more. Here's how to interpret the table (from https://www.ti.com/lit/an/slua511b/slua511b.pdf?ts=1641374769571&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FBQ20Z45-R1):

For CellN R_aM where:

  1. if 0≤M≤8: SOC=100%–(M×10%)
  2. if 9≤ M≤14: SOC = 100% – [80% + (M – 8) × 3.3%]

I log all this data in my current version of EternalPower.

I don't have an opinion on sailing vs trickle charge - don't remember seeing anything on that. My guess would be that trickle charge is better unless you're at a very high SOC (say 100%) and sailing lowers the average SOC.

This might be a good opportunity to share some other things I've learned about the M1 and its battery management:

  • full charge capacity estimate depends on maxQ and internal resistance. maxQ is only updated under certain conditions and one of them its that after a charge SOC > ~70%. So if you always keep the machine below 50% you never update maxQ and full charge capacity does not degrade in the eyes of the Mac. Calibration cycles are necessary to get a true reading.
  • Apple seems to have used a poorly matching OCV to SOC curve for the chemistry. This results in different maxQ results depending on the depth of the calibration cycle. 80% to 5% gives much lower maxQ than 100% to 5% for example.
  • After some time Apple limits the max charge SOC - not sure based on what data. Mine no longer charges beyond 96% (based on DOD readings). When this happens the reported max capacity is also reduced. Mine now reports fccComp: 3828, 3901 - macOS uses the lower one while the higher one is the real max charge capacity based on maxQ and internal resistance.
  • My machine is down to 87% (based on Coconut) or 94% (battery management pref pane). This is after 247 cycles most of which were automated cycle from 20%-30% and 10 months. I think the battery came around 5% below design capacity when I received the laptop. And then some 2% is due to the lowered reported full charge capacity. So I might realistically be down to around 94% now - similar to what macos estimates.

Other data:

  • A battery which has been at 100% for a long time always plugged in will recover a significant amount of capacity when cycled below 50%.
  • Stored an iPhone 8+ with 90% battery health for a year cycling between 19 and 21% (around one such tiny cycle per day). After calibrating battery health had not degraded at all over the year.

@krackers
Copy link

krackers commented Jan 6, 2022

@joelucid Thank you for the link to the manual, it was an interesting read and it means that the 20-30% soc region has even lower resistance than I originally thought, and the bulk of the resistance increase happens from 15% downwards. That would also explains why dropping to 20% SoC is not at all sufficient for calibration since half of the table is dedicated to SoCs < 20%.

As for

After some time Apple limits the max charge SOC - not sure based on what data.

This might possibly be related to the "Manage battery longevity" feature, which can be optionally disabled on Intel machines but seems to be always on in M1. This feature is orthogonal to the 80% charge limit ("optimized charging" feature), and as you noted it effectively reduces the max capacity the system sees which I guess would have a similar effect of preventing charge to 100% of true SoC.


I also have to correct myself when I mentioned "trickle charge" since it appears that term as defined in the literature applies only to non-lithium ion batteries, and with LiIon batteries the behavior approaching 100% is to reduce current, then once 100% is reached the BMS will cut charging until it falls below a certain threshold voltage, at which point a top-up charge is applied.

This in principle seems just like sailing mode, just with a delta soc threshold on the order of few percentage and reduced current. In the TI doc that you linked, I see there is a section for "Termination Config" which specifies the voltage at which to reduce current as approach 100%. I'm also not certain, but since I didn't see anything in there about specifying an RSOC at which to terminate charging I assume the BCLM SMC key does not actually set anything on the BMS and instead involves only the SMC which means we wouldn't get any sort of tapering behavior when setting charge limit via BCLM.

I'm not sure how the SMC actually behaves when BCLM is set – does it attempt to maintain the charge at the fixed level exactly (i.e. effectively a 1% SOC window), which could theoretically lead to rapid switching between charging/discharging under heavy draw (or a really shot battery)? Or is it smart enough to apply some hysteresis?

I also recall the behavior (at least in past versions of osx) at 100% SOC was to wait until 95% discharge before beginning charging; I do not know if this is implemented at the SMC or BMS level though. This makes we wonder if there is an SMC key that controls this behavior [1], which could be used to natively implement "sailing mode", or if the tapering threshold of the BMS is exposed.


Another point of minor note from the TI doc is that the displayed battery percentage is RSOC instead of ASOC ( / fcc) instead of (/ design capacity), and there is indeed a reserve even at 0% SOC which I think corresponds to PackReserve as seen in ioreg (~200, or 3% of total). I assume in the literature when they mean SOC they are referring to RSOC (since otherwise charging to 100% ASOC would be terrible).


One more question, @joelucid – I noticed a lot of sources suggest storing at 3.6-3.7V, which I presume to mean 3.6V when not under load. Correlating SOC with the CellVoltage key though, 3.7V would be under 20% soc (on my machine at least) when not under any load (i.e. when connected to the mains). So then that leads to a natural question: should optimal storage condition be based on SOC or cell voltage? On one hand voltage curves will differ between compositions and additives and so may not be generally applicable, but on the other voltage seems more indicative of the underlying chemistry compared to SoC where manufacturers may set their own cutoffs.

For instance, the study you linked which discussed resistance after aging at different storage SoC used a 3.0V cutoff as 0% SOC I believe. I have not actually checked on my machine, but I'd suspect on the macbook 0% might be 3.3 or slightly higher. The CellVoltage on my mac is ~3.8V @ 30% SOC and ~3.7V @ 20% SOC (both plugged in, not charging). If we wanted to use voltage instead of SoC as the guideline then the optimal would seem to be around 10%-20% (lower bound needs to be checked carefully since |dV/dSoC| rises sharply at those lower levels).


By the way, out of curiosity was the first version of EternalPower (the one that didn't use the SMC key and required SIP disable) based off the PowerUI private framework? I see some functions for stopCharging and startCharging in there (assuming they're the same as in ios since I have not dumped the dyld shared cache on mac). If so, shouldn't you be able to link against and invoke it without needing to disable SIP?

[1] If you haven't already seen it, there was a talk by Ionescu about dumping SMC firmware, and others in the hackintosh community have made a turnkey solution to flash an smc with auth bypassed from which you can dump a memory image and analyze in Ghidra. Would be nice to reverse engineer this and attempt to work out what other useful SMC values exist, or if we can talk to the BMS and get it to do anything useful (e.g. lower charging current).

@krackers
Copy link

krackers commented Jan 6, 2022

Actually there's definitely a difference between setting BCLM and letting it stay at that max charge limit versus lowering BCLM once the target is reached (i.e. implementing sailing mode). In the former I'm getting IOPS callbacks once a minute, but in the latter I'm not getting any callbacks. So in the former there is some battery state change happening once a minute. I need to investigate this further to see why this callback is triggered.

Edit; I just tried again and I cannot replicate. Hm very strange. It's possible it had something to do with time remaining computations causing the trigger.

@krackers
Copy link

After looking at the literature a bit more i'm fairly convinced that the SOC should be taken in relation to the lower and upper voltage limits. See this paper which albeit not on the same chemistry has the quote

An important factor that needs to be pointed out is the definition of the SOC window. Schmalstieg et al. [3] defined the SOC window from 2.5–4.2 V, compared to the 2.8–4.15 V used for the here studied cells. If we consider the corresponding voltage level for the SOC intervals instead, the results produced by Schmalstieg [3] then prove to be similar to the ones presented in this study. From this, we can conclude that how we define the SOC window is important since the ageing is strongly linked to the SOC level, i.e. the voltage level of the cell.

which explicitly states that ultimately the voltage level is what is crucial, and the SOC is merely a user-facing presentation of that. I've seen a trend that in the literature most setups use a lower bounds of ~3V @ 0% SOC, but in most consumer electronics this seems to be set a more conservative 3.3V @ 0% SOC.

With this in mind, we can see that the recommendations of storage at 3.6 - 3.8V storage would correspond to anywhere between 10 - 40% SOC depending on how the lower bound is defined, which I think also plays a role in the variation between what people say online.

For the specific case of the macbook pro, I've found that for my non M1 machine the voltage vs. SOC very roughly corresponds with one shown here (3.3V lower bound), but take note that any point measurements will vary over time due to different loads, temperature, etc.

We also have another datapoint on which to base our analysis for the MBP: the resistance table mentioned in previous discussions which stays pretty flat around ~0x30 from 40% to 15% SOC and only starts rapidly increasing below 10% SOC.

Based on this, I think it's reasonable to conclude that on battery use when optimizing for longevity a hard-upper bound of 40%, suggested-upper bound of 30% suggested-lower bound of around 15%, and hard lower bound of around 10% would be fine. Note that the lower bound is lower than the threshold cited by joelucid but from what I saw the study showing increased resistance below 20% SOC was based on a 3.0V cutoff; a 10% lower bound also being "safe" also accords with the resistance not really changing much up to that point. Take note though since both voltage and resistance start to rise sharply after 10% the soft-lower bound is probably better to use in practice. And of course the upper bound can be reduced for better longevity at the expense of reduced runtime, although the studies seem to indicate that the marginal benefit from doing so is much less than the benefit from going from 80->60->40->30.

For cases where you will be almost always plugged in, seems like a target SOC somewhere between 20% - 15% seems like a good choice.

@Robiemaan
Copy link

Robiemaan commented Jan 10, 2022 via email

@actuallymentor
Copy link

The values are: CH0C and CH0B (C - H - zero - B).

@joelucid I see you referencing CH0C but see @davidwernhart's AlDente code using CH0B, was there a consensus on the difference between these keys?

@joelucid
Copy link

@actuallymentor yeah I'm using CH0C. I have no idea what the difference between the two is. Long time since I looked at it. CH0C controls whether the laptop charges the battery from the power adapter. If it's 0 then only the operating power will get taken from the charger.

@joelucid
Copy link

@krackers

I'm also not certain, but since I didn't see anything in there about specifying an RSOC at which to terminate charging I assume the BCLM SMC key does not actually set anything on the BMS and instead involves only the SMC which means we wouldn't get any sort of tapering behavior when setting charge limit via BCLM.

It actually does taper current as it reaches target charge.

I'm not sure how the SMC actually behaves when BCLM is set – does it attempt to maintain the charge at the fixed level exactly (i.e. effectively a 1% SOC window), which could theoretically lead to rapid switching between charging/discharging under heavy draw (or a really shot battery)? Or is it smart enough to apply some hysteresis?

Nah - it does connect the charger to the laptop for operation. So it would only start recharging once the battery voltage drops either if the charger couldn't deliver enough or by self discharge. Not sure how much hysteresis it uses.

I also recall the behavior (at least in past versions of osx) at 100% SOC was to wait until 95% discharge before beginning charging; I do not know if this is implemented at the SMC or BMS level though. This makes we wonder if there is an SMC key that controls this behavior [1], which could be used to natively implement "sailing mode", or if the tapering threshold of the BMS is exposed.

There is actually quite a bit of buffer between 100% and 99% in the old macs (IIRC 99% is more like 95%) - so that might be what leads to the hysteresis. But I'm not sure. I don't really think sailing mode is very useful unless you want to stress the battery. I have a cycling functionality in my app for that. It'll either wait for the BMS to record the OCV before switching for calibration - or just switch directly when reaching either limit (for stress testing).

Another point of minor note from the TI doc is that the displayed battery percentage is RSOC instead of ASOC ( / fcc) instead of (/ design capacity), and there is indeed a reserve even at 0% SOC which I think corresponds to PackReserve as seen in ioreg (~200, or 3% of total). I assume in the literature when they mean SOC they are referring to RSOC (since otherwise charging to 100% ASOC would be terrible).

Yeah they charge to a voltage obviously. BTW this is a bit of an issue generally: RSOC can vary by applied load and internal resistance. So a percentage may or may not be close to the voltage which makes sense from a longevity perspective. I had thought about allowing specifying the charge levels in voltage terms but never got to that. It's important for calibration: an old battery needs to go down much further than a new one to reach the OCV which enables calibration. Since the internal resistance disables the lower SOC range from use.

By the way, out of curiosity was the first version of EternalPower (the one that didn't use the SMC key and required SIP disable) based off the PowerUI private framework? I see some functions for stopCharging and startCharging in there (assuming they're the same as in ios since I have not dumped the dyld shared cache on mac). If so, shouldn't you be able to link against and invoke it without needing to disable SIP?

I had experimented with various ways to access the functionality. I think other than discharging with connected charger I had everything working without SIP disable in the end. But since then everything worked with SMC that just seemed like it was the easier path.

With this in mind, we can see that the recommendations of storage at 3.6 - 3.8V storage would correspond to anywhere between 10 - 40% SOC depending on how the lower bound is defined, which I think also plays a role in the variation between what people say online.

Yeah that's all true. Of course we don't really know what the best voltage to target is though. Clearly the chemistry has been tuned to allow higher voltages (I think iPhones now go up to 4.45 IIRC) likely using additives - not clear what the effects are overall. I'd go for maybe around 3.6V for storage.

We also have another datapoint on which to base our analysis for the MBP: the resistance table mentioned in previous discussions which stays pretty flat around ~0x30 from 40% to 15% SOC and only starts rapidly increasing below 10% SOC.

Right - also note that there is something like a local internal resistance maximum around 60% charge. From what I've read this is due to a phase transition in an electrode - I believe the li ions accumulate in layers and a new layer is opened then which leads to extra stress due to expansion. So I'm trying to avoid passing 60% generally even if I need more capacity.

@joelucid
Copy link

joelucid commented Jan 24, 2022

Right - also note that there is something like a local internal resistance maximum around 60% charge.

For illustration here is my 2018 MacBook pro. Discard first two data values. Then it's 100%, 90%, ... - notice the local maximum at 60%.

RaDataCell0: 00 55 00 56 00 4E 00 5A 00 5F 00 6E 00 63 00 62 00 74 00 85 00 9A 00 B5 00 F8 01 7B 02 77 08 31
RaDataCell1: 00 00 00 58 00 54 00 5B 00 5C 00 68 00 61 00 5D 00 6C 00 78 00 8F 00 AD 01 04 01 C2 03 07 0E 0F
RaDataCell2: 00 55 00 63 00 60 00 65 00 69 00 72 00 68 00 69 00 83 00 99 00 A3 00 B3 00 E4 01 63 02 43 08 07

You'll often see this effect in the literature even if its not discussed. That discharge cycles with a constant delta SOC around different average SOCs improve the lower you go in average SOC. But if 60% is included its bad. E.g. 50% - 70% is likely worse than 70% - 90%.

@actuallymentor
Copy link

actuallymentor commented Jan 26, 2022

For the CLI aficionados in here, I made a little open source tool in the same vein @joelucid's GUI that adds the following CLI commands:

battery charging on: sets CH0B to 00 (allow charging)
battery charging off: sets CH0B to 02 (disallow charging)

Open source and without warranty.

@double-thinker
Copy link

double-thinker commented Aug 14, 2023

In case this could be useful new firmwares has a new flag (CHWA) to set a 80% limit managed natively. This limit works while the mac is sleeping. I modified @actuallymentor's tool (PR) to implement this based on Asahi's reverse engineering efforts (kudos to Asahi team for this)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests