*4.1. Parameters of Data Rate*

The crystal oscillator used in this paper had a crystal oscillator error of 10 ppm. Texas Instruments provided the SmartRF Studio tool, which can set and select the appropriate frequency deviation and the receiving channel bandwidth at a specific data rate. The node update period of 5 s was used in this paper, the design goal was 255 nodes. Therefore, the slowest data rate could only be set to 26 kbps. In view of the above factors, the data rate of this paper was divided into nine segments, from fastest (250 kHz) to slowest (50 kHz), to obtain the parameters in Table 1. The paper adopted this method to set the required data rate and the correlation coefficient.

**Figure 1.** CC430F6137 development board.

**Figure 2.** Star network architecture.

**Figure 3.** (**a**) The placement of the bridge; (**b**) the placement of node 2 and node 7.

#### *4.2. Receiver Sensitivity and Transmission Time in Di*ff*erent Data Rate*

A sensitivity experiment was proposed in this paper, and the detail is shown in Figure 5. Two CC430 RF devices were used in this experiment: The transmitter and the receiver. On the transmitter, adjusting the transmission power is used to change the RSSI, 1000 packets are sent in a fixed data rate, and then average RSSI and PER are calculated on the receiver. In this paper, the corresponding RSSI when the PER was 1% was called sensitivity, and each data rate had a sensitivity. The relationship between data rate, RSSI, and PER is shown in Figure 6. The RSSI closest to one percent PER at each data rate was taken as the receiver sensitivity of the data rate, as shown in Table 2. The above Figure 6 relationship and Table 2 sensitivity table were used to analyze the wireless performance of different data rates.

**Figure 4.** Diagram of TDMA architecture.


**Table 1.** Parameters of different data rates.

#### *4.3. Current Consumption in Di*ff*erent Transmission Power*

The transmission powers of 121 segments are provided by the CC430F6137, Sub-1GHz wireless communication chip, while only 41 segments were selected in this paper. Due to the transmission power table provided by the original manufacturer, the actual value did not correspond to the 920 MHz band used in this paper. Therefore, the 121-segment transmission power values were measured at 920 MHz by the Rohde & Schwarz RTO2044 digital oscilloscope with a bandwidth of 4 GHz and a sampling rate of 20 GSa/s, and the more suitable 41 segment settings were chosen in this paper.

**Figure 5.** Experimental architecture of receiver sensitivity.

**Figure 6.** The relation between RSSI, data rate, and PER.


**Table 2.** RSSI when the PER was 1% at different data rates.

Figure 7 is a chart comparing the measurement results of the current consumption corresponding to each transmission power with the original manufacturer. When the transmission power was larger, the difference between the values of the datasheet and the measurements was larger. In the interval where the transmit power was −9 dBm to 6 dBm, regardless of the values of datasheet or the measurements, the consumption variation was suddenly increased. Therefore, the appropriate 41 segment settings from the set value of 121 were selected, as shown in Table 3.

**Figure 7.** Comparison of the datasheet and measured values in the 121 segment transmission power.


**Table 3.** Current consumptions in 41-segment transmission power.

*4.4. Total Power Consumption of Data Rate and Transmission Power*

In Equation (1), I(TxPower) is the current consumption of the transmit power, and the relationship between the transmit power, data rate, and power consumption is shown in Figure 8 on the condition that the packet length is 33 Bytes. Figure 8 results were used for the method of energy efficiency comparisons in Section 5.

$$\text{Power Consumption} = I(\text{TxPower}) \times Bit \,\text{number} \times \frac{1}{\text{Data Rate}} \tag{1}$$

**Figure 8.** Relation of data rate, transmission power, and power consumption.

#### **5. Control Algorithm**

#### *Algorithm of Transmission Power and Data Rate Hybrid Control*

The transmission power and data rate hybrid control algorithm was proposed in this paper, and this algorithm was used to balance wireless quality and low power consumption to achieve a PER less than 1% and more power-saving. The control architecture between the bridge and sensor nodes is shown in Figure 9, and the hybrid control algorithm was run in Bridge. In Figure 9, the bridge received the RSSI feedback from sensor nodes' transmission signals and calculated the PER using the packet error interval algorithm. After the hybrid control algorithm was complete, the new transmission power and data rate control command that generated by adaptive algorithm was sent to the sensor nodes from the bridge. The detailed flow of the hybrid control algorithm is shown in Figure 10 and its pseudocode is shown in Figure 11. In this system, input includes the real-time RSSI feedback, PER record, sensitivity table of different data rates, and power consumption table, and output includes the data rate control and transmission power control.

The communication quality target of this paper was set at a PER below 1%. In the algorithm of the packet error interval, the threshold of data rate and transmission power is adjusted by 128 packet durations. If there are no errors in the continuous 128 packets, it means that the PER is less than 1%. The data rate will be increased, or the transmission power will be reduced. However, if there is only one incorrect packet in a 128-packet period, it means that the PER is 1%, and the data rate and transmission power will not be changed.

Then, if two errors have occurred before 128 packets have been completed, the data rate will be reduced or the transmission power will be increased. In this algorithm, when the error interval is short, the larger amplitude of the transmission power is set, because it is necessary to react immediately when an error occurs. Otherwise, when the error interval is long, the lower amplitude of the transmission power is adjusted, because the environment may be stabilized and the error does not occur easily.

Finally, by the error interval method, we determined how much to set the N-grade of data rate and transmission power for the next transmission. After the above N-order adjustment, based on the database of Figures 6 and 8, the lowest power consumption combination of data rate and transmission power was selected as the result of the final adjustment. The method flow is shown in Figure 12.

**Figure 10.** Flowchart of error interval algorithm controlling data rate and transmission power.


**Figure 11.** The pseudocode of control algorithm.

**Figure 12.** Flowchart for selecting the best energy efficient combination of data rate and transmit power.

### **6. Result**

#### *6.1. Experimental Method*

The experimental location of this paper is the Sixth Hall of Engineering, National Yunlin University of Science and Technology, Taiwan. A total of one bridge and ten nodes were set up in the experiment, as Figure 13 shows.

**Figure 13.** Experimental bridge and nodes placement map.

Every two nodes were placed in the same position. One node had the ADR and TPC control algorithms, and the other node fixed the data rate to the lowest (50 kbps) and the transmission power to the maximum. Through long-term testing, the power consumption and PER of nodes placed at the same position were compared to verify the effect of the algorithm. Nodes 1–5 were the nodes that had the algorithms, and nodes 6–10 comprised the experimental control group. A total of ten nodes were located in five different locations. Note that all nodes ran in the TDMA mode to save a lot of power for the sensing node. However, the power consumption of TDMA is not discussed later.
