Implement deep sleep between readings

amotl commented 5 years ago

People have been asking [1,2] for Deep Sleep on FiPy. While @einsiedlerkrebs and others [3] tried already,

I have tried deep sleep, but without success. Tried 1s sleep but device didn't wake up.

I have to admit that I did not do any debugging on this yet. see: hiveeyes/hiveeyes-micropython-firmware@191a46b

it looks like it would be harder than expected. As usual, the devil might just be in the details.

We will try to dedicate some time to that exceptionally important detail.

[1] https://community.hiveeyes.org/t/bob-platine-im-lipo-betrieb-ohne-deep-sleep/2055 [2] https://github.com/Hiverize/FiPy/issues/2 [3] https://github.com/pycom/pycom-libraries/issues/26

amotl commented 5 years ago

Please recognize there's also the Deep Sleep Shield [1-4]. However, while Pycom and Espressif might have had their times [5-7], the devices might now be ready [8] to eventually feature the promised ultra low power operation (~1uA in deep sleep) [9], even when using only the RTC module for implementing Deep Sleep.

[1] https://docs.pycom.io/datasheets/boards/deepsleep/ [2] https://docs.pycom.io/.gitbook/assets/deepsleep-schematic.pdf [3] https://media.digikey.com/pdf/Data%20Sheets/Pycom%20PDFs/Deep_Sleep_Shield_Web.pdf [4] https://forum.pycom.io/topic/3977/power-supply-issue-by-vin-on-sipy-deep-sleep-shield

[5] https://forum.pycom.io/topic/543/lopy-deep-sleep-mode [6] https://forum.pycom.io/topic/1022/root-causes-of-high-deep-sleep-current-lopy1-wipy2-and-sipy-all-the-new-modules-do-not-have-deepsleep-issues [7] https://forum.pycom.io/topic/989/pycom-oem-modules-are-here-4-mbytes-of-ram-and-deep-sleep-test [8] https://pycom.io/l01-oem-module-deep-sleep-test/ [9] https://www.kickstarter.com/projects/pycom/fipy-the-worlds-first-5-network-iot-dev-board

amotl commented 5 years ago

Please recognize there's also the Deep Sleep Shield.

After some quick research, it appears to us that the Deep Sleep API documentation @ClemensGruber was referencing the other day is solely suitable for talking to the Deep Sleep Shield. On the other hand, another part of the documentation just called Sleep tells us that the respective routines for controlling sleep mode there would be embedded into the Pytrack and Pysense libraries.

So, as the native MicroPython's machine.deepsleep will probably not be available on the ESP32-based Pycom platform, we should nevertheless bring in the use of pycoproc.py in order to see what we can get from that running on a bare FiPy module without any of the Deep Sleep Shield or Pysense or Pytrack extension boards.

the devices might now be ready [8]

[8] https://pycom.io/l01-oem-module-deep-sleep-test/

Please also recognize Pycom is only talking about the WiPy and LoPy OEM modules in this video.

amotl commented 5 years ago

as the native MicroPython's machine.deepsleep will probably not be available on the ESP32-based Pycom platform

However, they actually are using machine.deepsleep() in this video.

So, we will turn away from our last argument about favoring pycoproc.py and will try the vanilla way first, still not being sure whether this will actually work with our current hardware configuration.

ClemensGruber commented 5 years ago

The PyCom deep sleep shield is only necessary for older WiPy / LoPy devices! It was necessary due to a hardware bug and was a workaround to get this older boards in a power saving operation mode!

It is NOT necessary for the current WiPy, LoPy or FiPy version!! See https://forum.pycom.io/topic/1022/root-causes-of-high-deep-sleep-current-lopy1-wipy2-and-sipy-all-the-new-modules-do-not-have-deepsleep-issues/88

ClemensGruber commented 5 years ago

Ok, the "DeepSleep library" is for the deep sleep shield only as @amotl mentioned! I found this remark for the new boards:

The WiPy3 has its own deep sleep functionality, which is part of the machine module https://docs.pycom.io/chapter/firmwareapi/pycom/machine/

Link is outdated, current is https://docs.pycom.io/firmwareapi/pycom/machine/ and here you can find

machine.deepsleep([time_ms])

Stops the CPU and all peripherals, including the networking interfaces (except for LTE). Execution is resumed from the main script, just as with a reset. If a value in milliseconds is given then the device will wake up after that period of time, otherwise it will remain in deep sleep until the reset button is pressed.

The products with LTE connectivity (FiPy, GPy, G01), require the LTE radio to be disabled separately via the LTE class before entering deepsleep. This is required due to the LTE radio being powered independently and allowing use cases which require the system to be taken out from deepsleep by an event from the LTE network (data or SMS received for instance).

amotl commented 5 years ago

Introduction

With one of the recent commits 3e73a3fe, the Hiveeyes MicroPython Datalogger Firmware should support deep sleep.

Screenshot

It looks like this guy is sleeping well

...
27.5273 [terkin.datalogger        ] INFO   : Telemetry status: SUCCESS
27.7685 [terkin.datalogger        ] INFO   : Entering deep sleep for 5.0 seconds

and even wakes up again.

4.7042 [terkin.datalogger         ] INFO   : Reset cause and wakeup reason: {'reset_cause': {'code': 3, 'message': 'DEEPSLEEP'}, 'wakeup_reason': {'code': 2, 'message': 'RTC'}}

Please note

that

we upgraded to the most recent Pycom MicroPython 1.20.0.rc11 firmware with no effort [1] to stay up to date.
we switched from FatFS to LittleFS with no effort [2] to stop filesystem corruption in brownout conditions.

[1] https://github.com/hiveeyes/hiveeyes-micropython-firmware/blob/master/doc/pycom-firmware-upgrade.rst#install-and-acquire-prerequisites [2] https://github.com/hiveeyes/hiveeyes-micropython-firmware/blob/master/doc/getting-started.rst#switch-to-littlefs

Happy testing!

amotl commented 5 years ago

@ClemensGruber confirmed this would also work for him, thanks for testing! So we are closing this but feel free to reopen.

einsiedlerkrebs commented 5 years ago

Yip, looks good!

Am 7. Juni 2019 21:30:20 MESZ schrieb Andreas Motl notifications@github.com:

It looks like this guy is sleeping well and even wakes up again.
4.7042 [terkin.datalogger        ] INFO   : Reset cause and wakeup
reason: {'reset_cause': {'code': 3, 'message': 'DEEPSLEEP'},
'wakeup_reason': {'code': 2, 'message': 'RTC'}}
We will be happy to find any testers out there. Thanks already, @ClemensGruber! ;]

-- encryption? welcome!

gpg: 63e5d348

-----BEGIN PGP PUBLIC KEY BLOCK-----

mQINBFiDKSkBEACjAoZDjZ2GyPrW919ozieezixTX9+k9meyyDOxS+koA3YCdUMk L7+xwvPgCc/ZY6Ug3EF6+wJxVd+xZcAAGc4Lo8b9Fdrrg/YNbqjitIMA4gatBnEJ gDO4qrYmwGzA214yINLpSR+Pv6mMv4Y628FltkOnFrVDND3DnqOTLzq8tQHskaLQ KHgUYNUq93yVjSvquaJgpaJJQbGYyhSehvNLqPRdgC6rvJqAxh4InsclfaGWybxz cDPDLdUQEefg/kyeY0+IfYGQZgbr0jNRGZLbwFUSaCDpJlIAb3IABUUPj/HKwlP6 UDWyNTbXMdSedXaeRHpdUki33gbD8Q8CqVB5sDyR2nZiXGBuM9v6drQRCRrGalm+ eFTLWPajM53kCLnnNDYv4upiS+2jqKHtQ/txyVqIDkNme5zCzxs63O8OJA57Go53 r4ULfSU0pUDo8c8K58JBXLeovIsXUkdC+KlBXVMNEd/HxUhGIsTyZcs2a/j/2mrr gkzE2wAY2H5Q0rjMp0OCjomYUgG0/fNL5JZP11m3N84Au7H6qGwXoQbSe9YwK3S3 hZW64Grr6Twi5cB3+fe8QaWuwZt8lHbyySF9CSsTSI2wtAcmllNv4+BYu3foF9Z2 YlZJ8nPZkuajkhMnLIabaps3bsPhSlxD17/PizDOE0WM7GxazoJfNCUBIQARAQAB tCRteCBtc3RybW5uIDxyaWNoYXJkQGdpbm51bmdhZ2FwLm9yZz6JAk4EEwEIADgW IQRYiB2ee4h0BkApXNhM0u7DY+XTSAUCWIMpKQIbAwULCQgHAgYVCAkKCwIEFgID AQIeAQIXgAAKCRBM0u7DY+XTSMJeEACYk6slxE1ktg1fR2Nrv/NU6zvPWUS28BVX FplxYFTNEKZIS7I2ztL6BRnxtqoJoiMhTbzSINGW7lMk13r8D7r/VjtorcDVY6Fa Tmn2b1SemmQVnlAaW1rQr1Zf+kyuShROrUbQMMibhV/Im5f6fkciFIjlgX4Ejyo0 ItuvOQUGXBfjXk78Q0j856u888PMj0QGOaiR1lZGu3Vv/Z0+JmwlYDIsVo7H4hPO Nj+x0HWktjAMfoewMUbDi27k0UWU3pYDFcJ/xxgZQKVlk4iL7RpskSk9VvcW9Vns MngEHDeovUxzx4/hz4B+O54CUmgYOBVhPoI+lBzKOejbeyGbb6qxOSjPoffjdhzo hCYaxbIHSthzZ1o848mFYLaNnirYrbyWjGXNzBcuqdvXCic8nKAH9zNoImDEKsGF Nem8iaClcL9vKVELRJ8VEffnTuooa3UPYBZfx53IyDvyrlKDQRH8pVEbr8ce3InB XbNT19M8QiKggA+vI7h0YcBGMguhBQqSocGNzxE+EuX+xgsacR2GsrqwknhFLCHA zZZEzxa7QsJtb1Meyo6JnXm0RvzKXYueGn4CM1LqGyXB25imf4TuDu6quZ7sDWu1 qLJ/I2qzXAC+ayDmJoNi52T1Oarw+HxtIFbQC8SrA9ArUfsZOBho6JahvdXPbFel JZEDcWqyTrQvZWluc2llZGxlcmticyA8ZWluc2llZGxlcmtyZWJzQGdpbm51bmdh Z2FwLm9yZz6JAk4EEwEIADgWIQRYiB2ee4h0BkApXNhM0u7DY+XTSAUCWIMqgwIb AwULCQgHAgYVCAkKCwIEFgIDAQIeAQIXgAAKCRBM0u7DY+XTSDibEACJWr5TWA4u 6jlv6G1w4RZh7GR/rebd8D6/SpwAnyLRgYoicMRWMp/1QPSXMzTvLPk6SqmqwBSE avD7v6FmjAuUwhWCPVSPE38qU3BjxMsZaEMLhSPQRuL0HFGO9WUE2rGSlQyuXMPF inGAnaHgDoxtG0mtsVWvm3kgZ7bn1lQxwncfPu4fr8kXJSLXM3vhZrqwpze/F0Q5 q8YpVeLZPiK7LzxIDmiEt9FY7Z5ZIu3GCDu/uR8+GEDlnD2AoL7JnOK9ECxEXpUy 9cvr8ubOpocrshrqpIYWZLF1A+8KCUrMfTgPOCe5a8djrZGpelrlyVQkoW2W6L/3 D7Yj5j4H1c32J+OBc/VjASKHIwsIjaeyqttKtcv4TD3513YworWrSZskEQWzgtAK 6Rhr9FHMpw5ySxNgEMpIfwDyz8WrY5ymamDRVB1ZP898xJ2KlCd2HnN7NgKCylGO jQiUN3je8PiSk6qERCw62lS/hFyxXDypVNMzHYenWyJm9vdd1md0VZGxM12UONeQ DhkYF6sjOcN9D/2NOdQ2ySi664dBUlC+MfOOM5DTAxAYDyf6/eDMkZ470+ko5J2G iG7Oj5mcHaAJH9pz+/SDCDf2XZzIIKKASOolpjKtP5QbICkK6cvZxZ5fxqKBQPka neeol2r+5q7Nsn3Bsk0tq0tGzK7eDN0Qh7kCDQRYgykpARAA1hiPo3fYTLXMv8te UJkyDXQjr+czb59yvHXWeRMjcp3PkgLCOPzI1Hj4ESYT6uCrcaiAy3bw4TUpHHKE 05+s/a9T23/OR46+gzkDbC0zHPGqvSY2vMMQpbgo5MAwOQLBvn65q5gH+yI9M6ZN MHMHondUr8Xdsbv/CALMEDSNam3pgssLAL09Mo0b/vyC+uIDJxXNePOboYFGoZbH VUrL1cmaW0OLVQbiyTK/dBZFPmY0ertSue7UqvyrCSgYCVnCEiF6fYpUFS9jeKO2 bP7Q23Kjie+IfeDP4GWJI8AqaEwOL1B3n5OqUA4vQrH4pIPnJXPkDE2j3969xiq5 d5tbjy1XrgyXE9vICJ8OghtIZPEf1z6j+N/296jnjFt/xfUZr57Y5VX3mLERbINH pE25HF3RVY3wx8QfTjR4akHHK0JPiu/u8EiDcRL8HhC0ow5MMTQclG5I8OxrQi/R 7T06VeXonncYHFWqLClbasuL8cqGwaORVs2q+5wXIVQcEEcP5EkIJNfckVL9ihyy uQj/q8+t+saGUVO8xPfnM/Aqn+XSnZbwkNk9TozEgGHl0gSP/WjRaWxHxeBEfGJ3 PoQDpyUYFFt/Y3zCicO+Seo9OD1pYkB2fQzyDxbwvcZr39kXboXNpP68qrLgcarg aWMXcl0hpRUFzb9Ws1DnRCZmkUEAEQEAAYkCNgQYAQgAIBYhBFiIHZ57iHQGQClc 2EzS7sNj5dNIBQJYgykpAhsMAAoJEEzS7sNj5dNIGeMQAJOCa6hgeEUdZX+2eAYQ SE/RHnXESekjiZAizu83FZ+m1UdKzBXERnB/SC9pBMhmM3QnQg1Ni/tWLSVA83gN 6osw5/gmLciYzQawcC91iWyksoPbsUeecku4S5GfDsSpJexXbn0oabWGBf7L73yh shJ2DpbRohjZf/gTn1d7gyrB3h7WJIlCEbM/aHPe5a2hr8HDdm/YSxyyJI8ZN4mt JbQ7pFXF29N9BtrCj7pOcog2rC72x68v1YJw7EVBhq+CeLZ/suRnVKhX17hlKgA3 uQFxPGfQtUlVNdwmgkTWJ8FaZzbXzpZPxa+0TnOQJpCLqWgFFYaaGytrQGuAmhvY asSvunxkoq/nhu5rphSkJKuWYfMDE8Ux7rvpupkzSfGAraxNxBFVLQ8AYll81iD/ O7JNcHqupBFq7taPjLU6UwXYmQWptNx2pel6kokPhsV2TezM/oO5nMkgMot6hoaH gZ7mw+0tWuWXcQW+GtWzMX7y/5vELleUOrUYsV76KJ3yY+u7LRm5nw0gH3Jjc1e0 TDQNviaUI6bbWPA2+bQhQMva3r7SyQH2BNdoijAMC+pQdJJseEHkLIrVDpaxQ+mH cC2R3J27GAXJDagPilUARrOxD71gnFxJGiKAEvbtLORoxMsLl3uhosG3vyDL5nxy +cbOghoUEZPyzUyfOFxMG3hy =81oF -----END PGP PUBLIC KEY BLOCK-----

amotl commented 5 years ago

@ClemensGruber told us at [1] that

I believe deep sleep still does not work as intended. When configuring
'interval' 60.0,
'deepsleep': True,
it feels like the WiPy is only sleeping for 5 seconds.

[1] https://community.hiveeyes.org/t/sammelthread-kontinuierliches-bugfixing-der-hiveeyes-micropython-software/2115/3

Thanks for the report!

amotl commented 5 years ago

We still had a glitch in there but it should work now.

ClemensGruber commented 5 years ago

It's working now as a kind of delay after doing sensor reading and submitting. I had to do make install to get the fix running. make sketch-and-run is not sufficien only!

amotl commented 5 years ago

Dear @ClemensGruber,

this issue was really just about making machine.deepsleep() work at all ;]. Sorry that you are having problems updating your program. The documentation is gradually improving and we will be happy about appropriate contributions.

Also, we would like to tell you that we are tracking the appropriate computation of the delay as a real interval with #4.

With kind regards, Andreas.

hiveeyes / terkin-datalogger