Qball's Weblog

NRF24 10th month update.

Tags atmega328pb  nrf24  Sensors  Low Power 

Battery life time

This one is very hard to answer, the batteries Voltage is very temperature sensitive, so I have no clue if they are empty or not. This is clear from the voltage graph below. The red one is placed outside, the other is in different places that is somewhat more stable.

Voltage sensors

For one sensor, the one in the fridge, the battery ran out. This is not unexpected given that its in a shielded metal box and at low temperatures. Even this battery would still measure 3.8V if not under load. I had to attach it to my dummy load and measure it under 20mA to get a clear reading. On the logged voltage I got a weird pattern, but an almost empty battery should be detectable.

Empty Battery

I am curious to see when the first (non-fridge) sensors will go out because of empty battery.

Sensor PCB

As I mentioned in a previous post, I wanted to switch to the SMD version of the NRF24 unit. I redesigned the PCB so you could solder one on the back. This worked well and it turned out the SMD modules have a lot less issues with voltage dips, and I did not need to add extra capacitors on the power rail.


Image front PCB front


Image back PCB back

Stressing the network

As a test I also added three current sensors that report constantly via the NRF24 network. This, without any active arbitration, seems to work well. Maybe if needed in future, after adding more sensors, I will put these on the separate channel. So far this is not needed.

NRF current

This plots 3 current sensors I build to track power from a small single panel solar setup. This is a small PCB I build that uses a hal current sensor and a 19bit DAC to implement a +-30 Amp sensor (+-50Amp max).

current pcb

The sensor and DAC are pretty good, but have an offset. This is a sweep from 0 to 10 Amp plotting the offset between this sensor and a 6.5 Digit multimeter.

Cur plot

Even while not being ideal, I can keep better track of the battery state using this compared to previous solutions.


I’ve been looking at the ESP32-C6 (or H2 if I can get one), to see if I can add sensors to my zigbee network. I think this in the end might be the nicest solution. So far I can add switches, lights and sensors to the network, but the whole thing is very unstable. After a while, I get reports from the esp that the calibration data checksum fails, it then also stops connecting to the network and finally it never manages to transmit new data points. I also notice the devkit and a seeed studio module behave differently, both esp32-c6. I hope this will stabilize in the future.

Sodium-Ion batteries

I’ve been playing with sodium-ion batteries to see how they work and how to use them in the future. I was a bit early to this, because after I did some initial tests several electronic related youtube channels reviewed the same cells. Making my experiment superfluous. I got some mixed feelings about them. The first batch of 16 batteries I received had very unevenly matched cells (1500mAh spec, but some did not get above ~1300mAh). I recently got some bigger (10Ah) cells to play with and these seem better. This is the discharge curve of one of the batteries:


This gives some doubts on how to start using them for sensors, the biggest problem is the very large voltage range. Full is almost 4V, they are empty at 2V. Compare this to LifePO4 that sits around 3.2 V for most of their charge level. The large range it travels during discharge is ideal for estimating the remaining capacity, but hard to run a microcontroller on. 4V is to high and 2V is to low for most sensors/microcontroller, so we need to add a voltage regulation that can both step down and step up the voltage or use 2 batteries with a step down. To add a step down/up regulator cost extra components and used to be expensive, both in extra energy consumption as cost. I need to look what is currently available on the market. From what I found from a quick search, a buck-boost chip with low quiescent current doubles the BOM cost, not a viable solution imho. The TI tps62743 buck converter I found when looking is interesting as it works from 2V-5.5V, can provide enough current and has a low quiescent current, still leaves the 3 down to 2V unusable, this is where around 1/2 the capacity lies, but if we can find hardware that can work down to 1.8V? In theory I can run the atmega328pb down to 1.8V @ 4MHz, but that leaves the radio and the sensor. The NRF24L01 should be able to work down to 1.9V, a sht30 sensor should work down to 2.15V. Maybe it is .

tps62743 Graph from the TI TPS62743 Datasheet.

comments powered by Disqus