[Nut-upsuser] Weekend Puzzle: computer posing as an UPS

[Tim Dawson]

That, or use a low current source such as the fan header to control a 5v switch feeding the pi from a different source. Same net result, and no excess loads on your mobo. (Might be able to work this into the case using a spare power supply connector to feed it . . .)

Still have not figured out why though, but that's not relevant to what you are trying to accomplish.

On September 28, 2024 5:57:41 AM EDT, Jim Klimov via Nut-upsuser <nut-upsuser at alioth-lists.debian.net> wrote:
>I guess I should scratch the idea about fan sockets as the power source:
>they are rated typically at 0.2A each, so fan headers should be expected 1A
>max (3A in some vendors/models), well under the 5A that the RPi5 wants.
>
>Jim
>
>
>On Sat, Sep 28, 2024 at 11:52 AM Jim Klimov <jimklimov+nut at gmail.com> wrote:
>
>> Follow-up:
>>
>> * Powering the Raspberry Pi5 from an USB-C port wired on the motherboard
>> was much more promising, it survived over 8 hours building NUT in a loop
>> (in a tmpfs). And in the morning I found it turned off (red light on the
>> Pi).
>>
>> * Per
>> https://stackoverflow.com/questions/4702216/controlling-a-usb-power-supply-on-off-with-linux
>> it seems not possible to programmatically truly power-cycle USB port/hub,
>> as 5V pins are "always on"; and per
>> https://stackoverflow.com/a/16316401/4715872 - at least not on MoBo ports
>> (some hubs may have invested into being fully up to spec and control power
>> fully). The referenced uhubctl <https://github.com/mvp/uhubctl> on my PC
>> said "No compatible devices detected!" so I guess that's it.
>>
>> * Whatever I could find with (suggestions welcome) `grep -r . `find
>>  /sys/devices/pci0000\:00 -name '*usb*port*'` /sys/bus/usb/devices/` did
>> not expose any differences beside timestamps with the RPi5 plugged and off,
>> unplugged, and plugged back in again (with auto-boot). I hoped for some
>> power draw statistics to at least learn which port it lives on, to try
>> managing that somehow.
>>
>> * Overall, housing the Pi inside a cooled and somewhat dust-protected PC
>> case seems a neat idea, but for powering the Pi, it seems I would need to
>> use a real wall-power adapter (as noted many times on the net, ideally
>> Raspberry's own one as others tend to vary in actual voltage provided under
>> load).
>>
>> * But before that, I'm thinking if I could tap into the ATX power supply
>> though (using "HDD" or "FDD" plugs) or fan sockets (could be individually
>> manageable? gotta exempt one from OS/HW temperature-based mgmt then).
>>
>> Did anyone trod these side paths yet, any learnings? :D
>>
>> Jim
>>
>>
>>
>>
>>
>>
>>
>> On Fri, Sep 27, 2024 at 4:43 PM Jim Klimov <jimklimov+nut at gmail.com>
>> wrote:
>>
>>> FWIW, a few lessons learned:
>>>
>>> * Different USB-A sized ports (even if marked USB-3.2) did not prove a
>>> stable source, with Pi5 occasionally turning off or rebooting. Sort of
>>> behaved well for days, but as soon as I added load like package installs or
>>> NUT builds, it did not survive 5 minutes...
>>>
>>> * Might be the MoBo turning off or cycling the port due to "overload"?..
>>>
>>> * Tried the `usb_resetter` script (referenced in NUT contribs) and host
>>> `dmesg` did show re-detection of keyboard etc., but a turned-off Pi did not
>>> boot up. Did not check much further, but did harbor hopes that a funny NUT
>>> driver could "shutdown/reboot" the USB port acting as an UPS for Pi...
>>>
>>> * The Pi power socket is dumb-USB so the host does not "see" any IDs
>>> about it.
>>>
>>> * Currently trying with a USB-C port on the MoBo, and a USB-C to USB-C
>>> cable rated for 60W -- and so far it survived a few loops of NUT fightwarn
>>> builds. Complains of undervoltage fairly frequently (every 30-60s), but now
>>> did not crash yet.
>>>
>>> Jim
>>>
>>>
>>> On Mon, Sep 16, 2024, 13:57 Greg Troxel via Nut-upsuser <
>>> nut-upsuser at alioth-lists.debian.net> wrote:
>>>
>>>> Kelly Byrd <kbyrd at memcpy.com> writes:
>>>>
>>>> > With USB-C ports and cables, there are a ton of profiles, I don't know
>>>> what
>>>> > the new Pi's support, but likely something like 3A @ 5V, 9V, or 12V
>>>> over
>>>> > USB-C
>>>>
>>>> Up to the RPI4, I was pretty sure there wasn't PD, just 5V and it drew
>>>> what it drew, and you hoped that the supply was big enough.
>>>>
>>>> It seems the RPI5 will use PD if given a capable supply.  Looks like 5V
>>>> 5A, and it won't negotitate higher voltages.   There's an official
>>>> supply that does PD
>>>>
>>>>
>>>> https://www.newark.com/raspberry-pi/sc1153/power-supply-usb-c-5-1v-5a-white/dp/82AK3955
>>>>
>>>> and the output spec is
>>>>
>>>>   5A at 5.1V, 3A at 9V, 2.25A at 12V, 1.8A at 15V
>>>>
>>>>   looks like a TUV seal
>>>>
>>>>
>>>>
>>>> I find Jim's way of using this interesting, but my approach is totally
>>>> different.  First, when I'm using a Pi, it's because I want a low power
>>>> computer that I can leave on all the time, or can place in a different
>>>> physical location I don't particularly want to do things on a Pi instead
>>>> of a desktop.  And then I want it to be reliable.
>>>>
>>>> That leads me to plug a power supply into a UPS, or to use POE (from a
>>>> POE switch which is plugged into a UPS).  For RPI3, I found a POE
>>>> ejector that splits the POE ethernet cable into ethernet only and a
>>>> micro USB.
>>>>
>>>>
>>>> _______________________________________________
>>>> Nut-upsuser mailing list
>>>> Nut-upsuser at alioth-lists.debian.net
>>>> https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/nut-upsuser
>>>>
>>>

-- 
Sent from my Android device with K-9 Mail. Please excuse my brevity.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://alioth-lists.debian.net/pipermail/nut-upsuser/attachments/20240928/79e3b235/attachment-0002.htm>