Team Vision for System-Level Design PhaseOur team planned to use this phase to develop an overview of the project as a combination of the various subsystems necessary to obtain the desired result. We also planned to outline the technology and engineering requirements for each of these subsystems as well as to begin designing them.
During this phase we were able to mostly solidify our desired technology and engineering requirements pending additional testing based on our feasibility analysis. We are already underway with testing of our ultrasonic localization concept as well as some conceptual design for the RF system.
Functional DecompositionA functional decomposition was created by determining the broad groups of features that our project will support.
The drawback of this system is that it relies on anchor nodes which are different from the standard nodes. Additionally, the positioning was not accurate to under a meter, with one reference giving a value of around 1.6 meters on average. Otherwise this is a very cost effective system and the technologies applied should not be ignored.
Pozyx https://www.pozyx.io/The Pozyx system is a primarily enterprise positioning system for large scale use (hundreds of anchors, thousands of nodes). The benefit is that we are also attempting to implement this technology for potentially large scale use with over 100 individual nodes.The Pozyx system is also accurate to 10-30 cm which is our ideal RF accuracy, though we expect the ultrasonic accuracy within its range to be much better.
An important note mentioned on the Pozyx website is that complexity is "kept at the gateway and anchor level... [keeping] the total cost of ownership within bounds." Because we are employing relative positioning without anchor nodes, we will rely on each node to have a high level of computing capability or the ability to perform computations in a shared manner to keep costs down.
RF Phase TriangulationThis concept makes use of the phase difference at three points in an equilateral triangle to determine the angle of arrival (AoA) of the signal. The difference in phase between each of the antennas is found using a phase detection system consisting of a phase-locked loop, frequency to voltage converter, and a lookup table. The basic schematic for this can be seen below.
RF RSSIIn addition to the AoA, a measurement of distance is required to determine the position of the node. RSSI (Received Signal Strength Indexing) makes makes use of attenuation during propagation to determine the distance traveled. This can provide an accurate distance measurement in free space. However, obstacles which attenuate or reflect RF signals such as metallic objects, people, bodies of water, and thick walls will introduce error in the measurement without an existing knowledge of the obstacles.
Ultrasonic TriangulationTriangulation for ultrasonic works similarly to triangulation for RF. The difference is that light propagates at 3x10^8 m/s while sound propagates at 343 m/s. Because of this, we are able to use a time difference rather than a phase difference, simplifying the process. The resulting time difference between multiple microphones gives us an accurate estimate of the angle of the received signal.
Ultrasonic TDOATo determine the distance of the transmitting node using ultrasound, we use a combination of RF and ultrasonic signals. The RF signal is treated as an instantaneous trigger due to the speed it will travel in short distances. The arrival of the ultrasonic signal then stops this timer giving us an accurate measure if distance.
Feasibility: Prototyping, Analysis, Simulation
A feasibility analysis was performed to determine the estimated accuracy of the Ultrasound triangulation relative angle calculation.
A bit-accurate Simulink model was built to test the PDM microphone capability and performance. The model was implemented using a Sigma-Delta ADC to emulate the PDM microphone output. A 2d spacial model was then determined to accurately calculate the transport delay seen by sound across an air medium. 314m/s was chosen for the speed of sound through air.
The Simulink system is shown below.
The PDM output was reconstructed using FIR lowpass filters. Noise was generated and added to the system for more accurate simulation modeling. The TDOA signal was then generated and the filtered reconstructed signal was captured, shown below.
The system was architected using 71 coefficient FIR low-pass filter. By applying this filter through a delay line, a trivial FPGA architecture can be used, and is well within the resource limitation of even many small FPGA devices. While high-frequency noise is limited, the system is designed to use a constant 4.8MHz clock for the entire processing subsystem to have the highest possible temporal resolution.
Numerical analysis was done on the 2d time-delay system, and it was determined that a temporal resolution of less than 4.83 microseconds was needed for 1 degree relative angle determination. The theoretical resolution of this system is 208.3 nanoseconds.
A physical device was designed in addition to the simulations. The schematic of the device is shown below. A PCB allows for greater precision in placing the microphones in the correct triangle pattern, and allows the team to see how the devices might behave on the final node design.
A rendering of the PCB is shown below.
Ideally all three microphones would be placed the exact distance from each other, but small rounding and calculation errors resulted in less than a 10th of a millimeter error, which is acceptable for initial testing a 6.5cm distance.
Microcontrollers were evaluated based on cost, size, power consumption, and more. Power consumption, IO pins, and programming environment were deemed to be the most important features. Based on the analysis, the Beaglebone Black and Giant Gecko are the best choices. Notably, the Beaglebone is a system-on-a-chip (SOC) meaning that it runs a full Linux operating system. Whether a SOC or microcontroller is used will be determined through further analysis and benchmarking
Batteries were evaluated against standard metrics. Each node needs to be small, easy to debug, and safe to use. The size of nodes will be dominated by the battery, so an energy dense chemistry will determine maximize time-per-charge. Additionally, a chemistry that is easy to regulate will keep the supporting BOM smaller. Batteries that cannot hold a charge for a long time will make developing on the board challenging and the demo frustrating. Additionally, batteries are expensive. Rechargeable batteries are more expensive to purchase but they will not need to be replaced for the lifetime of the project. Carbon-Zinc batteries are inexpensive but will need to be purchased frequently.
Lithium Polymer and Lithium Ion technologies meet requirements, but Lithium Ion are less volatile when charging and therefore safer.
Designs and Flowcharts
Plans for next phaseFor the next phase we plan to have fully solidified concepts for ultrasonic and RF localization schemes in addition to preliminary detailed designs or highly detailed system designs when possible. This process will require discussion with professionals and researchers including RIT faculty and connections outside of the university. Specific attention will be paid to the RF system which is subject to greater potential for error and failure. This is in part due to its complexity but also the cost of highly accurate components which will need to be placed in every node.
All Embedded Documents: Systems Level Design Documents