3V battery for door sensor node.
-
@skywatch said in 3V battery for door sensor node.:
CR2032
if you power everything directly from 3V battery and run arduino at 1MHz, CR2032 battery should last more than a year (or even 10 years).
-
and set bodlevel to 1.8V
-
Thank you for the fast reply! :)
I have already set fuses for 1.8V brownout but read somewhere that 8MHz actually uses less power than 1MHz as the 'time alive' sending data is 8 times longer at 1MHz and therefore uses more battery (the clock has no bearing during sleep it seems?)....
Do you (or anyone) have any thoughts on that?
Thanks
-
Thank you for the fast reply! :)
I have already set fuses for 1.8V brownout but read somewhere that 8MHz actually uses less power than 1MHz as the 'time alive' sending data is 8 times longer at 1MHz and therefore uses more battery (the clock has no bearing during sleep it seems?)....
Do you (or anyone) have any thoughts on that?
Thanks
-
Good call! - Yes I had forgotton that (to much festive spirit)! - Thanks for the pointer.
So the next question then is would it be better to use 1MHz and connect direct to the batteries or would it be better to run at 8MHz and use a boost converter to 3.3V? By 'better' I mean longer battery life ;)
Thanks
-
In your scenario the most power hungry module is the nrf24 module when it sends data. I don't know how mysensors handles this but I believe that the time that nrf module is in the send mode does not depend on the arduino clock because nrf module has its own clock source. Someone correct me if I'm wrong.
-
In your scenario the most power hungry module is the nrf24 module when it sends data. I don't know how mysensors handles this but I believe that the time that nrf module is in the send mode does not depend on the arduino clock because nrf module has its own clock source. Someone correct me if I'm wrong.
@rozpruwacz yes that sounds reasonable.
@skywatch a really good booster should work, but most cheap ones draw way too much power when there is no load, which will drain battery while sleeping. Check the datasheet for "quiescent current".
As an example, ADP1610 uses 1-2mA with no load:exclamation:
-
Thanks to you both! :)
I was thinking of using these....
I have these already for other things, but I will look into measuring the quiescent current (my old fluke meter won't go down to the levels of nA).....Maybe I will get find a histab resistor and put it in series with the supply positive and see what voltage I get across it on the meter or scope. A little Ohms law and I should be done!
It's taken a long time - but I am getting there now! :)
-
Thank you for the fast reply! :)
I have already set fuses for 1.8V brownout but read somewhere that 8MHz actually uses less power than 1MHz as the 'time alive' sending data is 8 times longer at 1MHz and therefore uses more battery (the clock has no bearing during sleep it seems?)....
Do you (or anyone) have any thoughts on that?
Thanks
@skywatch said in 3V battery for door sensor node.:
I have already set fuses for 1.8V brownout but read somewhere that 8MHz actually uses less power than 1MHz as the 'time alive' sending data is 8 times longer at 1MHz and therefore uses more battery (the clock has no bearing during sleep it seems?)....
Do you (or anyone) have any thoughts on that?
My experience is it's not true when using NRF24, especially with "switch" sensors taking no time for measurement and needing no complex operation/calculations. Most of the time active is when sending data through the radio, and frequency of atmega doesn't change sending speed of the NRF24. And if you have radio link problems and your sensors has to resend message a few times, then it's better to save a few mA by running at 1mA all that extra time.
It seems it's a different story with RFM radios though, that's why other users using those radios here or on other forums seem to favour the 8MHz option.@skywatch said in 3V battery for door sensor node.:
So the next question then is would it be better to use 1MHz and connect direct to the batteries or would it be better to run at 8MHz and use a boost converter to 3.3V? By 'better' I mean longer battery life ;)
Do not use a boost converter, even the very good and expensive ones consumes a few uA, and if you optimise your sensor that's more than what it will consume in sleep. In addition they create noise and decrease the radio performance and stability of the node. Just use one CR2032 it lasts a very long time with basic sensors, my temp/hum sensors with si7021 are on the way for something like 2 years of battery life, and my door sensor with NO/NC reed switch will probably last 3 years.
Using 2 CR2032 in parallel is a bad idea, it seems the one with highest voltage will try to "charge" the other. Then its voltage will lower below voltage of the one it's trying to charge and the will switch the "charge"/discharge positions until they are completely depleted.Just use one CR2032 and apply the following things to make it last :
- put a capacitor in parallel, 100-200uF ceramic (real capacity nearly cut in half due to voltage bias) or 100uF electrolytic (but it has more leakage)
- add short sleeps between radio TX/RX sequences, so your battery can "rest" and recharge the capacitor. You can look at what I did in the scripts for the NModule TH/Door sensor board on openhardware.io, it might not be perfect but it works for my sensors.
This is necessary because lithium button cells have a high internal resistance and their voltage drops quickly when you apply a high load. CR2032 can only sustain 2mA so the capacitor is needed to help it power the radio (15mA). Without capacitor a lot of energy will be wasted as heat and you will only be able to use a fraction of the real capacity of your battery.
-
Thank you for taking the time to post that and help me out - All posts here should be a reference for the current 'best practice' for battery powered nodes!
I was just looking at the chip on the boards I posted (ADP2018A) - and whilst advertised as 'low quiescent current' devices, I guess that 'low' is a subjective term!
All you said makes sense, so that is the route I will go. Now I have to find the battery holders......they're here somewhere! ;)
CHeers and Happy New Year. :)
-
Hi,
of course I agree with all comments above. imho 6months, even a year is too short for maintenance..
Regarding frequency, in general, the radio modules run their own crystal so they can't be very impacted on their side by changing 1/8mhz (or maybe spi comm, but that's not the case here, this is usually working).
But MySensors lib "could be". I'm not saying it will!
As changing freq can also change some timeout, delays etc in the lib. So far there is no time slot syncing mechanism, which wouldn't help at all here..
So that may depends on features needed (just thinking to delays in radio drivers for collisions, signing, heartbeat etc when it's about a main door node), and may or not add some nacks, then more transmit, more power etc.. -
Hi,
of course I agree with all comments above. imho 6months, even a year is too short for maintenance..
Regarding frequency, in general, the radio modules run their own crystal so they can't be very impacted on their side by changing 1/8mhz (or maybe spi comm, but that's not the case here, this is usually working).
But MySensors lib "could be". I'm not saying it will!
As changing freq can also change some timeout, delays etc in the lib. So far there is no time slot syncing mechanism, which wouldn't help at all here..
So that may depends on features needed (just thinking to delays in radio drivers for collisions, signing, heartbeat etc when it's about a main door node), and may or not add some nacks, then more transmit, more power etc.. -
@mfalkvidd as long as there is no precise timing involved in lib, this should be ok. who can do more can do less. well this is always good to remember 1mhz mode is not as precise vs 8/16Mhz or faster espressif.
There is another point I'm thinking, when it's about security nodes, lowpower (like a main door node for example). When using 328p deep sleep without time, you lose the watchdog feature. I think, in general, for a reliable security node, wdt may be a good idea to prevent any lockup (especially when using cheap, outofspecs ?? 328p clone??).
Sure it's important to optimize runtime power consumption, but in case of a node which doesn't wake up often, the most important point to improve for lifetime is the sleep power consumption of the device.
I agree with gohan for lifepo4, too bad there are not lot of variants yet.
I think there are no definitive answer for lowpower node and batt lifetime. It depends on the device requirements, always compromise. As a primary lithium AA/AAA+booster will always last a lot longer than a coincell. But that needs nano quiescent current with high efficiency during lightload boosters that you won't find already assembled at aliexpress, +filtering (depends on the part but doesn't hurt)
AA/AAA compromise is a bigger profile than coincell, or need more parts so more expensive for example. Here I guess, a lower profile is needed for a door node. -
@mfalkvidd as long as there is no precise timing involved in lib, this should be ok. who can do more can do less. well this is always good to remember 1mhz mode is not as precise vs 8/16Mhz or faster espressif.
There is another point I'm thinking, when it's about security nodes, lowpower (like a main door node for example). When using 328p deep sleep without time, you lose the watchdog feature. I think, in general, for a reliable security node, wdt may be a good idea to prevent any lockup (especially when using cheap, outofspecs ?? 328p clone??).
Sure it's important to optimize runtime power consumption, but in case of a node which doesn't wake up often, the most important point to improve for lifetime is the sleep power consumption of the device.
I agree with gohan for lifepo4, too bad there are not lot of variants yet.
I think there are no definitive answer for lowpower node and batt lifetime. It depends on the device requirements, always compromise. As a primary lithium AA/AAA+booster will always last a lot longer than a coincell. But that needs nano quiescent current with high efficiency during lightload boosters that you won't find already assembled at aliexpress, +filtering (depends on the part but doesn't hurt)
AA/AAA compromise is a bigger profile than coincell, or need more parts so more expensive for example. Here I guess, a lower profile is needed for a door node.@scalz said in 3V battery for door sensor node.:
But that needs nano quiescent current with high efficiency during lightload boosters that you won't find already assembled at aliexpress, +filtering (depends on the part but doesn't hurt)
Any good suggestion for that ? I'm making a breakout for TPS61098 to test, but if you have other suggestions of similar chips I'm interested :)
-
@gohan said in 3V battery for door sensor node.:
I still believe a 2€ LiFePO4 battery is the cheapest solution :D
Where do you get it at this price ?
-
@gohan said in 3V battery for door sensor node.:
look for soshine AA batteries on aliexpress, I paid 8€ for 4 AA batteries plus 2 connectors
Yes I have those, paid more as shipping was expensive. But they are too big :(
-
@gohan said in 3V battery for door sensor node.:
look for soshine AA batteries on aliexpress, I paid 8€ for 4 AA batteries plus 2 connectors
Yes I have those, paid more as shipping was expensive. But they are too big :(
