My testing on 5.6-rc4 with mainline ATF seems to suggest that BM818 gets powered off in suspend
It sounds like the Linux drivers for the BLM818 cellular modem don’t work correctly if the modem shuts off when it should go into suspend mode or the BLM818 has some serious problem in the hardware. This should be tested with the software stack that BroadMobi provides, to determine if it is a software issue with the free/open driver used by the Librem 5.
Thanks for this update! HW related relatively simple things are described within the NXP document AN12147: “In cases where the thermal properties cannot be altered, a lower operating voltage can be used to increase the lifetime of the device.” As well, on page 54 from this Embedded Artists document is recognized: “A lower die temperature will result in lower power consumption due to lower leakage current.” And in addition (hardware optimizations): “The i.MX 8M SoC and PMIC together implement DVFS (Dynamic Voltage and Frequency Scaling) and Thermal Throttling. This enables the system to continuously adjust operating frequency and voltage in response to changes in workload and temperature. In general this result in higher performance at lower average power consumption.” Sorry, if bothering again, just couldn’t resist. Is there something wrong with using NXP PF4210 PMIC? Is it L5 Dogwood still using ROHM BD71837MWV or perhaps new BD71837AMWV PMIC?
From what I read in the technical description of the Gemalto modem, it can run entirely on its own and wake up a computer either via usb or via a dedicated connection.
Returning from deep sleep on a Pinephone seems to need less then 2 seconds. I’d suspect that that is good enough to still catch a call (if the wake-up by the modem works).
P.D. wake-up from ‘systemctl suspend’ on pinephone works in PureOS. Compared to an active android phone the ring came about 3-4 seconds later. Fine for me if it would give me that much more standby time.
BTW, did anybody find any information about the power consumption of the ram in deep sleep?
Is there any metrics on which components draws the most power?
I would be interested in underclocking and lowering the voltage of the CPU for more battery life if that made a significant difference. Then maybe clocking it back up to stock speeds when docked once that works would be cool.
Thanks, I’ve just switched to PureOS and you’re having right systemctl suspend works with EG25-G out-of-the-box, as you described. I’ve took and installed 1.2-Community-Edition PureOS image. Direct to eMMC over Jumpdrive microSD, it made me happy, great tool.
I thought a bit more about my use of a smartphone.
I do not need any time critical services like important messages from social networks. I like to have my phone available, be reachable via SMS and calls and have the option to look at stuff (like social networks, forums, whatever) when I decide to take the time.
This leads for me to the question whether the optimization we’re talking about is what I really need and want. I guess I’d be fine if the computer part of the phone would be switched off completely (suspend to disk or even shutdown) and would be woken up the moment I get a call or SMS.
If I’d miss the call due to the wake-up time of the computer part and would find the number of the caller afterwards in my history of calls I’d be fine.
These thoughts rise two questions for me:
Is it possible to wake the imx8-part of the phone by the modem from suspend to disk or even from a shutdown (like wake-on-lan)?
For my use case: Wouldn’t it be a valid design for a phone to have two parts: One dump phone only part that can be used as a modem by a computer and that works completely on its own and another computer part. Both in the same housing.
That’d lead to the situation I have now: A notebook that uses another device to connect to networks, e.g. a phone which can be used as a modem. And a phone which is dump and usable as a modem.
Such a design compared to the librem5 would move some function from the imx8 part to the modem part (calls, sms, driving a (second) display).
The advantage is obvious: I can use the phone without the computer being switched on. The battery runtime of such a phone would in phone only mode would probably be some weeks.
If I switched on the computer part I’d be able to use all functions of the Librem5 (if I don’t miss any possible problem).
If I’d cut off the phone part from the computer I’d have an airgapped computer (at least as long as the connection is cut of between the two parts).
If I’d switch off the phone part completely I could still use the computer part without being logged onto a mobile network at all.
Does anybody know such a project? Wouldn’t it be a possible improvement on the Librem5 general design? Dreaming of a future version with an additional e-ink display for the dump phone (formerly modem) part .
