During fermentation of juice to wine, the conversion of sugar to alcohol produces carbon dioxide (CO2). For a single liter of juice that is fermented, approximately 64 liters of pure CO2 is produced. The OSHA permissible exposure limit of CO2 is 5000 ppm for 8 hours of exposure, approximately 10 times the CO2 content of the atmosphere. It would take 200 liters of fresh air to dilute a single liter of pure CO2 to permissible level. Considering a typical fermentation tank can hold approximately 5000 liters of juice and a winery will ferment in many tanks at the same time, the levels of CO2 produced poses a significant safety hazard and must be controlled. Since CO2 is denser than air, a vertical separation will exist, in addition to the inherent variability in CO2 emission and distribution across a floor plan. Existing CO2 monitoring devices for wineries typically consist of a single sensor mounted on the wall, providing only a point measurement, however, the spatial dependency of CO2 lends itself to a multi-sensor system.


For each actor in the mesh network (coordinator, router and end node) a printed circuit board (PCB) was created with the optimal components and form factor to minimize price. For the end node two PCBs were created: one for measuring temperature, humidity and CO2 and one for measuring temperature and humidity. All nodes utilize a microcontroller with an integrated multi-band sub-1 GHz and 2.4 GHz radio frequency (RF) transceiver (CC1352r; Texas Instruments, Dallas, TX). The end nodes implement a low-power architecture with block diagram shown below.

The low-power architecture is based on a Texas Instruments reference design, and completely disconnects the microcontroller and sensors from the battery between measurements by using a nano-power system timer (TPL5111; Texas Instruments, Dallas, TX) and a low-leakage load switch (TPS22860; Texas Instruments, Dallas, TX). A DC-DC boost converter (TPS61291; Texas Instruments, Dallas, TX) was used to boost the battery voltage of 3 V to an operating voltage of 3.3 V. On the end nodes, the microcontroller uses UART to communicate with a non-dispersive infrared CO2 sensor (COZIR-AH-1; Gas Sensing Solutions, Glasgow United Kingdom) and I2C to communicate with a combined temperature and humidity sensor (HDC1080; Texas Instruments, Dallas, TX). 

The current drawn from an end node during operation was measured to confirm expected power consumption and battery life performance. A portable power supply, voltage and current measurement device (Otii Arc; Qoitech, Lunden, Sweden) was used for current measurements. The measured current on an end node device through two measurement cycles is shown below.


For the end nodes, an inverted-F (IFA) PCB trace antenna was implemented to lower bill of material (BOM) cost, as the antenna is built directly in the PCB. Additionally the IFA design is compact with a footprint of 26 mm x 8 mm, allowing the overall size of the PCB to be reduced. The antenna was simulated in HFSS (high frequency structure simulator), a three-dimensional electromagnetic simulation software for designing and simulating high-frequency electronic products, to confirm the peak operating point of the IFA design was 2.4 GHz. The return loss at 2.4 GHz was simulated to be -38 dB. Simulations of this antenna show good power delivery and low directivity, resulting in flexible positioning of the end device while maintaining antenna efficiency. The simulated characteristics confirmed the IFA design to be an attractive choice for low cost and low power applications.

After the boards were manufactured, return loss measurements were taken on the antenna in the enclosure with a vector network analyzer (MS2027C: Anritsu, Atsugi, Japan). The measured versus simulated return loss is shown below. The measurement shows performance shifting up approximately 100 MHz from the simulation, with the return loss at 2.4 GHz increasing from...

Read more »