Saving Power: Approaches, results and questions

[psychogenic]

Hi,
I'm building two projects around the GPy (using LTE Cat M1--no NB-IoT in these parts yet) and have been experimenting with ways to keep power usage low.

I've summarized the tests and modeling in this video, with a bit of extra info on this page.

The short version is that I'm comparing sleeping while completely detached, attached but disconnected and connected. As expected, sleep power needs increase with increasing level of connectivity but my main questions are about the attach/detach and connect/disconnect process.

I've found them to be relatively slow--attaching usually taking around 6 seconds but sometimes going up to almost 30, and connecting being closer to 1.5s but getting up to 6 or more at times.

This overhead makes quick "hit & run" type wake-transmit cycles difficult and I'm wondering if they are normal... what is your experience in this matter?

Also, the sleep current requirements when I'm connected are somewhat higher than I expected as well. Are the readings I've collected (~25mA with spikes into the 100+ mA range) standard?

Finally, do you have any advice that might help extend battery power further than the approaches I mention in the resources above?

Thanks!

[psychogenic]

To anyone using the Bell Canada LTE Cat M1 network: just got word from them and it seems that at this point there is PSM support but: "eDRX feature is under consideration for future availability but there is no announced plan yet".

I'll be playing with the +CPSMS AT command to see what's possible.

[psychogenic]

@jasonriedy Yes, when I'm fully detached and sleeping with everything off, the entire GPy + board consumes around 0.5mA.

I don't know if its possible to go lower, but that level is fine for my application. The main issue there is that re-attaching and connecting can take a good while.

I did some further testing and it's not so bad if I'm outside: on average it takes a little more than 4 seconds to get back up. But sometimes that goes way up to over 12s and it also implies detaching before sleep, adding an extra 4-6 seconds.

So, deep sleep is fine if I'm going to do it for a while and don't care about SMS/other network messages, but if I want to wake every 30 seconds, say, then those extra 10 seconds of network setup/teardown really add up.

[psychogenic]

@danielm I didn't know about the eDRX and it does seem, from the linked article, that it can work with Cat M1. Looks like it could really help a lot.

I got the impression that the provider needs to implement it on their side, too, so you can tell the network "I'm going to sleep now, please get back to me in n hyperframes"--which would imply we need support on both the GPy and the network side of things.
I'll touch base with Bell (the network I've been testing with) to see what their plans are on this front, if any.

[danielm]

@psychogenic
Great article and video :)

I have an information that new modem firmware release for NB-IoT with eDRX and PSM features is currently being tested. However I am not sure if there is similar firmware release for LTE Cat M1.

Anyway I believe these features should allow to design devices with really long battery life:
https://www.link-labs.com/blog/lte-e-drx-psm-explained-for-lte-m1

[jasonriedy]

Did you include deep sleep in your detached measurements?