Mysensors RFM69W multisensor node(CR2032)

Description:

This is a Mysensors node using RFM69W radio module, Si7021 temperature/humidity sensor and MAX44009 ambient light sensor.

Key Features:

• Si7021 temperature/humidity sensor
• MAX44009 ambient light sensor
• A3212 micropower hall sensor(this is not a low voltage sensor as it requires 2.5V min to work - there's the A1171 variant but it's a DFN package)
• ATSHA204A for signing
• AT25DF512C-SSHN-B SPI flash module for OTA
• MySensors ready(Atmega328p + RFM69W)
• DC-DC boost converter(NCP1402 based) - can be bypassed using the on board jumpers
• Small board(36mm diameter) using a CR2032 coin cell

Work progress:

1. Test/verify the pcb board correctness(from both electrical and mechanical point of views) - TO DO
2. Produce the initial PCB's using some external PCB manufacturer
3. Components assembly(soldering)
4. Upload Mysensors temperature/humidity node sketch code
5. Verify the real board with running code
6. Code optimizations for low power operation
7. Create project enclosure using a 3D printer

** Tested and it works! **

HW Revisions:

• 0.1 - initial design
• 0.2 - added option to bypass step-up converter using jumpers
• 0.3 - added MAX44009 ambient light sensor
• 0.4 - added A3212 hall sensor

Credits:

All credits go to MySensors team for the great work and to the opensource community

• 

  canossa commented 6 years ago
  

  Hello.
  Any news from this project?
  Do you made the changes for the timer TLP?

• 

  scalz commented 8 years ago
  

  @DavidZH
  that's exactly what i said depends on wakeup freq needed. So i agree of course :).
  I use some coincell with my 328p multisensors, pir onboard so i need the wdt for some tasks.
  and for new coincell board i've in progress, or low power wdt.. i'm not using anymore 328p, but arm mcus

• 

  DavidZH commented 8 years ago
  

  @scalz
  You also use coincells for timed nodes? As the watchdog timer uses 4,5uA that's quite a lot over 500 or 600 nA with only interrupts waking the MCU. I have some door sensors which I would like to have send a heartbeat every 12 hours or so, because these doors sometimes will not be opened for a few days. That TI chip would be nice to use as a heartbeat timer on an interrupt on only 55nA.
  But maybe I'm nitpicking now...

• 

  mtiutiu commented 8 years ago
  

  As I have the step-up converter already in there I was thinking what if I use it to power only the sensors and act as a switch too? The rest of the components(radio + MCU) will be powered from the battery.

  So my idea is to use that dc-dc converter to power the sensors when the MCU is awake in order to read data and shut it off in sleep mode. This way I can overcome the situation when I have sensors that require min 2.5V to work AND more important I can cut power from most of the parts from this board when needed using the converter as a power switch. I recently added the A3212 hall sensor which requires min 2.5V to work(because it's the cheapest one that I found and in a user friendly package too - and it's micropower also).

  One possible issue that I see here is about the logic levels as when the battery voltage drops the MCU will work at a lower voltage(so the logic levels will drop too) and the sensors will work from the same 3.3V provided by the converter(and logic levels).

  The NCP1402 device will accept 0.9V minumum on the CE pin as stated in the datasheet so I see no problem here. The off state current is 0.6uA so it's pretty low.

  I need to change the schematic a little bit and re-route the board to accomplish this but it's not a big issue at all.

• 

  scalz commented 8 years ago
  

  I have played too with TI's nano ultra low iq step up and these timers, (i like their design&app notes). but finally thought it was overkill for coincells. i also made my own ulpnode for experiments too (sequence with stepup+voltage supervisor, microchip app note).
  Finally i choosed the simple no reg way for those batteries. plus it saves the need of filtering the switching reg for radio (which is good for better snr).

  I thought the same for the reset cycle, not great, regarding startup time+internal res of coincell .

  This ic is still a nice ultra low power wdt but i thought i could use rfm69 listenmode and save money&footprint because between 1ua or 2uA no big diff, depending of the wakeup frequency..
  Always compromise everywhere

• 

  mtiutiu commented 8 years ago
  

  @DavidZH

  You're right about the finding parent + presentation part...forgot about that. Anyways it's still an interesting approach in order to reduce power consumption and maybe some other combination can be thought in order to profit from this.

• 

  DavidZH commented 8 years ago
  

  I too came across this last week. It's really neat!
  Another drawback I see is that the node resets every time when wired in this way. That means the whole finding-parent and presentation section is run every time. That eats a lot of power as the radio is powered for more than a second, especially when the node runs at 1MHz.
  I think that in a MyS application this IC should be used to trigger an interrupt to send the data to the controller and, in case of hungry sensor types, switch the power to the sensor.
  That does mean you will be running at about 600nA, but that is still 1/10th of what a watchdog-timed sensor will use.

• 

  mtiutiu commented 8 years ago
  

  Hi,

  I just found an interesting approach and idea to use for CR2032 battery powered sensors in order to reduce power usage. Well..it's not my idea but I borrowed it from an application example provided by Texas Instruments(all credits go to them). So the thing is that in the end all battery powered sensors only report data at some regular interval right? So what about using an extremely low power timer like this one TPL5111 (it takes only 35nA to work !!!) and gate the power for the rest of the circuit(MCU + radio) at a regular interval which can be set using a external resistor?

  Instead of using the internal watchdog from the MCU to do the sleep we're using an external timer to do just that. But in this case we're cutting the whole power from the board keeping only the TPL5111 timer active. This combined with a low quiescent 3.3V boost converter like TPS61291 which will be put into shutdown mode when not needed by the external timer should give a "sleep mode" of only 50nA !!! (the shutdown current of the converter is around 15nA).

  TPL5111 is built in a such a way that after MCU finishes it's work it can reset it using an external pin in order to put the system in shutdown mode again.

  All of this sounds great and the advantages that I see here are:

  1. don't need to sleep the MCU + radio anymore(no code optimizations are needed for low power operation as long as the ON time is pretty low - and it should be as we only do some readings and send them quickly)
  2. the standby current is extremely low(around 50nA in theory)
  3. long life for the cr2032 battery
  4. no need to think about code being stucked as the external timer will restart it each time(so no watchdog needed as we already have one)

  Disadvantages:

  1. The sleep interval is fixed - this shouldn't be such a big issue as in general it is a FIXED value
  2. OTA and other mysensors features - I don't know how it will be influenced by this approach
  3. any other???...

  Reference schematic for usage(it uses TI components for mcu+radio but the idea is the same)
  

  This can be improved further by using an analog switch also:
  

  Reference documentation:
  TPL5111 datasheet
  TPS61291 datasheet

• 

  jeti commented 8 years ago
  

  @mtiutiu
  got that too

• 

  mtiutiu commented 8 years ago
  

  @jeti

  Great. I forgot to mention about the rfm69w module which can be found on AliExpress too. I use this adaptor for rfm69w testing: https://oshpark.com/shared_projects/TKNcHTRl and it works great.