Introduction

The first tranche was all about doing the hardware design and firmware. The focus was on doing a design that can operate on batteries and provide local networking capabilities. The goal we set ourselves was to deploy a bunch of these devices and collect data from a serial interface. The data collection was local only and KPI was to get to a design that can operate on batteries for a long time. We also wanted to package the device neatly so it can be handed to others for real life deployments.

Goals

1. Development of Low Power sensor device that can operate on mobile size battery for 6 months.
2. Demonstration of Low power sensor device with temperature and Humidity sensor. Data will be collected and shown on a laptop.
3. Packing of the sensor device in a professional package

Development

The first problem was to design for low power. We selected the 8-bit ATMEGA 1284P as the micro controller at the center of our design. The MCU and other circuits will be powered from a 2600 mAH 3.7V Li-Ion battery. During radio transmissions the board will take approx. 50mA current. considering a 2 second transmission window every 10 mins, we have to budget approx. 144 x 50x (2/3600) mAH, or 4mAH/ day for radio transmissions. So we need to budget about 1460 mAh for radio transmissions for a year (365 days). At 80% battery utilization, this leaves about [(2600 x 0.8) - 1460] = 620 mAH for other activities.

so the circuit should not use more than 70uA of current when not doing radio transmissions. 70uA x 24 Hour/day x 365 days = 613 mAH. The challenge then is to design the PCB with sensor interfaces and voltage translations etc. that do not consume more than 70uA for normal operations.

we considered AVR and ARM cortex M0 processors for design. The choie of an 8-bit MCU can be controversial in todays world. However in the final analysis, we felt that we have to really get down and dirty with processor details to do the kind of power optimizations we want and our use case was not speed oriented. we were already familiar with AVR chip data sheets having used them before. We do not have experience with ARM processors to make any expert comments vis-a-vis our choice.

The way to succeed is to double your failure rate.

The processor takes about 5uA in the deep sleep mode. We managed to get the boards to 40uA average idle current consumption. Most of the information is in the data sheet and AVR libc pages. The importrant thing is to get the details right and that is where you may end up doing lot of bare metal debugging. The low power recipe is to shut down all peripherals and only run an internal watchdog timer to reduce power consumption. The watchdog timer interrupt will overflow every second and then we either need to do sampling and transmission or go back to sleep.

The issues one should be aware of are,

• watchdog needs to be disabled on a reboot otherwise you will end up in an infinite reboot cycle because there is no code to kick the watchdog.
• Fuse settings like BOD and clock speed matter. It is better to run off an external oscillator and turn off BOD to save power.
• keep GPIO pins in known state instead of floating and reduce leak currents.
• choice of components matter. our LDO uses about 30uA in leak current of the 40uA power budget in shut down mode. A normal LDO would have caused mA of current.
• when you are in shut down mode, nothing runs, not even TIMER0. All functions like millis() that depend on TIMER overflow can no longer be trusted with time arithmetic.

The summary is that there is no single put this to low power mode function. if you are chasing micro amperes and a general purpose PCB then even the leak currents would matter.

Deployment

With the board in hand and confidence that it can last upto a year on 2600 mAH single Li-ion cell, we started looking for suitable deployment opportunities to test the nodes in real world conditions. After our talks, we decided to deploy the nodes in two warehouses, one in Huballi (run by KAPPEC, Karnataka State Agriculture produce processing and export corporation limited) and another in Guntur. The features of Yuktix wireless sensor nodes are,

1. Battery powered nodes that can operate for more than 6 months without charging.
2. Operating on Sub GHz low noise spectrum. No licensing required.
3. Capable of creating local wireless network.
4. On-board Temperature and Humidity sensor

The company in Guntur was storing commodities in warehouses that needed to be monitored for quality. The current setup was to do 2 manual readings every day at the ground floor. The cost of wiring in such places can be prohibitive because to do right sensing, we have to run wires to each floor. Yuktix provided a wireless sensing solution that is long range battery powered radio devices not requiring wires. Multiple wireless sensor nodes can be installed on different floors and a local network is created to gather data.

KAPPEC – Karnataka state Agriculture produce, processing and export corporation limited graciously agreed to try the Yuktix sensor nodes for a limited trial at one of their Hubbali warehouses. The company was storing fruits. The challenge was to do continuous temperature and humidity monitoring under very high humidity environment.

Enclosure

A 3D printed enclosure was created to house the sensors and PCB