P20151: Satellite Localization
/public/

Systems Design

Table of Contents

Team Vision for System-Level Design Phase

Our vision for the systems design review is to do a feasibility study for the TDoA and raster scan methods to understand the benefits and drawbacks of each. Once we decide on a method, we will edit our model, and begin a preliminary BOM to understand how much money we will need. We plan on reaching out to local companies for potential sponsorship

Plan:

Actual:

Functional Decomposition

 Functional Decomposition

Functional Decomposition

 Input Output Flow Diagram

Input Output Flow Diagram

Benchmarking

 Ground Station Benchmarking

Ground Station Benchmarking

 Antenna Benchmarking

Antenna Benchmarking

 SDR Benchmarking

SDR Benchmarking

 Processor Benchmarking

Processor Benchmarking

 Oscillator Benchmarking

Oscillator Benchmarking

Concept Development

Function: Locate Receivers

The best way for us to locate each of the receivers, is by using GPS. It’s the most precise method of any that we have brainstormed (landmarks, using a map, etc.). However, there are some drawbacks associated with this method. The biggest drawback is the cost. In order to be as precise as possible, we are going to need an expensive GPS. Our budget is our biggest risk at this point, so that is worrisome.

Function: Reconstruct Signal

In order to reconstruct a signal, the satellite signal must be sampled at 874 MHz due to the Nyquist rate. Samples can be collected using a Software Defined Radio (SDR) or manually sampling and converting it to IQ Data (magnitude and phase). Because using an SDR would decrease the complexity of the project, it will be used.

The signal must be sampled at a high enough rate to accurately detect peaks and valleys. To trigger each sample, SDRs may have an internal clock or required an input clock. If the SDR has an internal clock, it can operate without support.

If the SDR needs an input clock, two main attributes are important: frequency and frequency drift. If the clocks at all the stations are resynchronized often, frequency drift can be corrected. Otherwise, frequency drift will need to be very low. However, no clock has low enough frequency drift to make resynchronization unnecessary.

In case the selected SDR needs an input clock, three main options were considered: atomic clocks, GPS, and a crystal oscillator. Though atomic clocks are very precise and have low frequency drift (0.001ppb), they are out of budget and would still require resynchronization the clocks. The GPS is a low cost, simple solution, but not accurate enough since it can off by 40-100ns. Lastly, a crystal oscillator is relatively cheap and has comparably low drift (typically 0.5ppm). The crystal oscillator would still need to be resynchronized.

Frequency drift is often characterized in two ways.

  1. ppm/ppb. The clock drifts by some parts per million or parts per billion each second. For example, a clock with 2 ppm would drift

(2*10-6)*(60 seconds)*(60 minutes) = 7.2 ms/hr

  1. dB/Hz. This metric requires more research.

Function: Synchronize Clocks

We need to synchronize the clocks between the 3 ground stations so we can accurately determine the position of satellites. Exact relative time on any point in the satellite signal, accurate to < 3ns. This can be found by a) Getting the exact absolute time on all 3 stations. b) Calculating the time it takes to propagate the signal from station A to station B.

The main ways we plan on doing this are by using GPS, a wired connection, or Two-way time and frequency transfer (TWSTFT). Those are by far the most accurate solutions, and our system will be very sensitive to time. GPS clock synchronization is good because it is a simple, cheap solution. GPS would eliminate the need to manually set the clock. Accuracies typically range from 1 microsecond to a few milliseconds. We would need a really good GPS to eliminate error in our orbit determination. A wired connection would provide an accurate time, but restrict the stations from being placed far from each other. Stations that are closer together would result in smaller time differences and require higher precision data processing. TWSTFT is currently the most accurate way of synchronizing time, with results that are regularly below 1 ns. TWSTFT is out of our scope because this method requires parabolic dish antennas, capabilities to transmit and receive signals at each station, it is very expensive, and it is much more complex.

Solutions:

Option 1. Absolute Time with GPS

Getting time via a GPS receiver typically has an error of 40-100 ns... What if GPS error is << 40 ns since the path from a GPS satellite to the stations are very similar? Note that there is not a lot of research to backup this assumption, which is a strong disadvantage of this option.

Option 2. Radio Tower

Get an identifiable portion of a signal from a radio tower and reset the relative time on all stations at that point. Since the NIST radio towers are not accurate enough (1-150 ms range), a different radio tower would need to be identified and used. In addition, it would need some predictable, unique signal portion.

Option 3. Signal Propagation Time

This method can use a dedicated station as the emitter, or one of the three ground stations as an emitter. By determining the signal propagation time, the time on each station will be set.

 Signal Propagation

Signal Propagation

 Signal Propagation Steps

Signal Propagation Steps

Option 4. TWSTFT (Two Way Satellite Time and Frequency Transfer)

 Two-Way Satellite Time and Frequency Transfer

Two-Way Satellite Time and Frequency Transfer

This method uses a transmit and receive antenna locally at each ground station that needs to be synchronized. Each synchronization antenna is pointed at a geostationary satellite and does both transmit and receive operations with the satellite antenna simultaneously. This method is the most accurate way to compare the clocks at two distant points.

Option 5. Wired Connection

With a wired connection, the time delay of a signal can be accurately calculated. The master station sends all slave stations the time, the time sent. Each station will set its time to the time sent plus the calculated propagation time.

Function: Locate Satellite

The best ways for us to locate satellites are by using Time Difference of Arrival method, or the Raster scanning method. Of all the concepts, these two are the most feasible. The Raster method would only require one ground station. However, the antenna would have to be pointing at the satellite to receive signal. If the beam width is too small, it will be almost impossible to find the signal. If the beam width is too big, we won’t be able to pinpoint the exact location of the satellite. TDoA uses three ground stations to acquire signals, and triangulate the position of the satellite. This method is likely to be more expensive, and more complex due to the number of stations, and time synchronization.

Function: Stabilize Station

When mounting the antennas/stations to roofs, they need to be fixed in place. We need to stabilize the ground station so there isn’t any unwanted motion. Of all our ideas, there are three that are the most promising: having a very heavy base, drilling or bolting the ground station down, or clamping the station to some surface. Ideally, we’d like these ground stations to be permanent fixtures, so drilling or bolting them down to surfaces on rooftops would be best. However, we need permission from building staff to make that decision. As it stands now, some non permanent way of clamping our stations to the top of the building, keeping it stable, is better than having a heavy base.

Function: Protect Electronics

Generally, we have only one way of protecting the electronics from the weather, and that is by creating a “housing” structure for the electronics. The materials we use have to be waterproof. We have thought about using plexiglass, sheet metal, or plastic. Plexiglass and sheet metal are more durable than plastic, but more expensive. Plexiglass would be ideal, so we can show the insides during ImagineRIT. Sheet metal may be the easiest to work with however. We have not decided on a final material as of yet.

Function: Sealing the System

Along with the housing, sealing the housing is important for weather proofing our electronics. We need to have a good seal along any seams of the house to keep weather out of the system. Our concepts include caulking, using some kind of epoxy, neoprene padding, or some form of waterproof tape. That decision has not been made yet either, and will likely require some prototyping and testing before finalizing our solution.

Function: Connect to Internet

We currently have three concepts for internet connection: Ethernet, wifi, and cellular data. Ethernet is our best option since a wired connection is more reliable than a wireless method. Ethernet and wifi are viable options only if our ground stations are all RIT’s campus. If we decide to move the base stations further apart and off of RIT’s campus, we would have to use cellular data for out internet connection.

Function: Transmit to Master Station

We need to transmit data back from the slave stations, to the master station. We can do this through internet connection, RF repeaters, or by gathering it manually. Manually is really not an option. This would require the user to actually go to each slave station, gather the downloaded data, and bring it to the master station. This is a major inconvenience to any operator that uses LASSO. RF repeaters require the station to have both a receiver and a transmitter. This is not in our scope, as we are not planning to transmit any signal

Function: Power Electronics

There are a few different concepts available for powering our electronics, but they could change depending on how far apart our ground stations are. If the stations are on buildings at RIT, we could have access to ethernet cables, and run the power through the ethernet cable to make our system a little simpler. If we decide to have a large distance between our stations, then power through the ethernet will not work because we will be using cellular data. If there are any outlets, we could simple plug in to the out at whatever location we’re at. However, it’s not very common to have outlets on the roofs of buildings. The final option could be to have a battery at each station powering the electronics. This way we can have power no matter the location.

Function: Removing Snow

RIT gets a lot of snow during the winter time. We need some way of removing the snow from our antennas, and our main concepts are with heat and vibrations. Heat would be the easier method, but it would require more power. This could be accomplished by having a few stationary space heaters directed towards the antennas, or by incorporating some wired connection throughout the antenna to heat them from the inside out. Our other option is to somehow incorporate an oscillatory motion in the antenna. However, to get the snow off, the frequency of vibration would have to be pretty high, and this risks damage to our equipment and could introduce a new form of errror in our Orbit determination. The easiest method appears to be the space heaters.

Function: Process Data in Real Time

The concepts for processing data in real time include having a full processing system at each ground station (no slaves) and using a central processing station that receives data from each station (all stations are slaves). Having a full processing system at each base station is not ideal because that would increase the cost of the project. The best concept for our project scope is having a microcontroller at each ground station to pre-process the data it collects by adding a timestamp to IQ data. This would then be sent on to a central location with a master microcontroller where the remaining processing will occur.

Feasibility: Prototyping, Analysis, Simulation

What time differences can we expect for TDoA in our application?

We estimate the expected time differences using a 2D model of the world and satellite.

Assumptions:
  1. The atmosphere has no effect on the satellite.
  2. The height above the earth (elevation) is constant between the ground stations.
  3. 2D model; the satellite goes directly overhead everytime.
  4. The Earth is a perfect circle...or sphere.
Model Inputs:
  1. Height of base station 1 and 2.
  2. The distance between stations
  3. The radius of the Earth
  4. The altitude of the satellite
  5. The elevation of the satellite with respecct to elevation.
Model Outputs:
  1. Distance from satellite to stations 1 and 2.
  2. Estimated Time Difference.
  3. Estimated coverage based on a time accuracy.
Model Summary:
We determine the distance from the satellite to each ground station using 3 triangles. The difference in these distances divided by the speed of light is the estimated time difference. This results in the "Valley Plot”. By choosing a time accuracy and assuming we cannot measure anything below that threshold, we can estimate the total coverage of the sky. This is displayed in the “coverage plot”
Model Implications:
The model shows both best and worse case for time difference. When the satellite is just on the horizon of station 2 results in the maximum time difference. When the satellite is directly in between two stations, there is 0 time difference. Expanding this to 3D will result in time differences between these two extremes.
Results:
We can expect maximum time differences of 3 microseconds at RIT and 100 microseconds at 30km. We can expect 80% coverage at an accuracy of 100ns for RIT and 99% at 30km.
 2D Schematic for Estimating time difference.

2D Schematic for Estimating time difference.

 Math Behind Estimating the Time Difference of Arrival. Essentially, we solve for 3 triangles.

Math Behind Estimating the Time Difference of Arrival. Essentially, we solve for 3 triangles.

 Each line represents a distance between ground stations. Highlighted are a selected elevation for the RIT option (~1 km) and the Far option (~30 km). The max time difference occurs near the horizon and decreases towards 0.

Each line represents a distance between ground stations. Highlighted are a selected elevation for the RIT option (~1 km) and the Far option (~30 km). The max time difference occurs near the horizon and decreases towards 0.

 Given some time accuracy, the coverage is equal to the number of time differences 10 times larger than that time difference over the total number of calculated time differences. This is done for several base station distances between 1 and 100km.

Given some time accuracy, the coverage is equal to the number of time differences 10 times larger than that time difference over the total number of calculated time differences. This is done for several base station distances between 1 and 100km.

Benchmarking Sheet(Excel)

What will be the projected uncertainties in TDOA on final output (time, azimuth, elevation)?

We are still working on answering this question.

Absolute Time Uncertainty:
The effect on time will be solely based on the accuracy of the clock. Being off by 100ns for an object moving at 8 km/s will translate to an error of 0.8mm, which is more than acceptable. Absolute time accuracy is not essential.
Azimuth and Elevation Uncertainty:
One degree of uncertainty could translate to up to 9km cross-track for a 500km altitude satellite.

We will relate the estimated uncertainties in TDoA localization to input errors in time sync, location, and cross-correlation error in the next phase.

What sampling rate is required?

For direct sampling, Sampling Frequency >= 2 * Signal Frequency. For an SDR, using frequency signal tuning will correctly demodulate the signal within the range of the SDR.

Can we reconstruct signals and overlay them?

Signal demodulation will yield the information of the signal. Looking at this information in the time domain will yield a time difference.

What roofs do we have access to?

We have begun correspondence with the facilities managers at the RIT Inn and the Riverwood Tech Campus. We've explained our project goals and are awaiting further response from them. From our systems design review, Dr. Barbosu and Amber Dubill told us about potential roof locations at University of Buffalo and at Brockport University. We will pursue these options further in the next phase.

Is it possible to localize the signal better than the antenna FOV?

Can we find the entrance and exit on the cone of the FOV:
The entrance and exit of the cone will be narrowed down to the edges of the cone. For a beam width of 10 degrees, the cone has a radius of 65km at 750km of elevation. Parabolic antennas are used to minimize beamwidth and the most common application of parabolic antennas are fixed position for pointing at a geostationary satellite.
Can we use the gain of the signals to better localize it:
SDRs will amplify and filter signals within the beam of the antenna. The signals will be demodulated and sampled into digital data. The 8-14 bit resolution of the ADC will be an upper limit for sensing changes in signal strength within the beam. Looking at small changes in signal strength within the beam of the antenna will be difficult because of noise and sensitivity.
What about two pointing antennas together?
Two antennas that are 50 km apart would have approximately 4 degrees of separation from the pov of a satellite at 750 km away. If the signal is within the beam of both antennas, the radius of the cone is smaller by 200m if both antennas are pointed straight up. For best case, the satellite would be on the outer edge of both beams when they are not overlapping until the elevation of the satellite. In reality, the beams are lobes with a half power beam width without clear cut boundaries. Ultimately, increasing the number of pointing satellites can reduce error from a singular system if they are working together properly.

How often can a concentric raster scan detect an unknown satellite doing a random search?

Model Assumptions:
  1. Raster Scan follows a concentric circular pattern.
  2. Changing between radii is much faster than circling and is neglected.
  3. Drawing any concentric circle has the same period.
  4. Satellite can enter and exit any point of the horizon and moves between 0.5 - 2 deg/s.
  5. The antenna Field of View makes a perfect cone with some degree width.
  6. All we have to do is SEE the satellite once. We are ignoring the fact that now we have to guess it’s direction and try to follow it.
Model Inputs:
  1. Radius of the horizon
  2. Satellite Radius around horizon
  3. Initial Angle of Satellite
  4. Final Angle of Satellite
  5. Velocity of the Satellite
  6. Starting position of the Scanner
  7. Angular speed of Scanner
  8. Field of View of Scanner
  9. Number of iterations in simulation
Model Outputs:
  1. Estimated Coverage of Scanner
Model Summary:
When the simulation begins, the scanner begins at a random point in its search pattern. The satellite begins at a random location on the edge of the horizon. Both the satellite and scanner begin moving. The simulation ends when the satellite leaves the horizon or when it enters the field of view of the scanner.
Model Implications:
This model assumes a success if the satellite enters the field of view of the scanner. We ignore the fact that the scanner next has to guess the direction of the satellite to track it.

The rules of this game are no prior knowledge. This is true for satellite hunting. This is not true if we know the satellite we want to track. From this perspective, the model is a worse case.

Results:
STAR could detect satellites approximately 50% of the time with a beamwidth 53 degrees. A beamwidth of 2 degrees at 10% coverage.
 Schematic of concentric circle simulation start.

Schematic of concentric circle simulation start.

 Top down view of the horizon with input variables.

Top down view of the horizon with input variables.

SDR and Signal Processing

3 different sources were used to collect data. An RTL-SDR was used to record an FM signal at 100 MHz. The SDR modulated the signal into audio data and baseband data. Both of these signals were processed through a low pass filter in Matlab to a lower frequency. This test was carried out as a proof of concept for SDR data being sinusoidal, manipulatable in Matlab and a controlled file size.

 Raw Baseband Signal at 100 MHz.

Raw Baseband Signal at 100 MHz.

 Filtered Baseband Signal from 100 MHz.

Filtered Baseband Signal from 100 MHz.

 Raw Audio Signal at 100 MHz.

Raw Audio Signal at 100 MHz.

 Filtered Audio Signal from 100 MHz.

Filtered Audio Signal from 100 MHz.

The third source was data taken off of SatNogs. An audio data file was uploaded by a ground station. This data had a reported carrier frequency of 435 MHz, so this is currently as close as we could get to actual data. This data was also processed in Matlab for proof of concept.

 Raw SatNogs Audio at 435 MHz.

Raw SatNogs Audio at 435 MHz.

 Filtered SatNogs Audio from 435 MHz.

Filtered SatNogs Audio from 435 MHz.

SDR and Signal Processing Next Steps:

-Prototype an antenna and SDR setup

-Familiarize with SDR capable software

-Further research limitations

-Acquire relevant data at our target frequency

-Analyze for noise and information

-Choose SDR to use for project

-Refine benchmarking and commit to a mode

What does 3D TDoA geometry look like?

A time difference between two base stations correlates to a hyperbola in 2D and a hyperboloid in 3D. Our next steps will be to produce a realistic set of time differences and create intersections between 3 hyperboloids to see our potential solutions.

For more information, see: Knott, Michael and Edge, Alasdair. “Passive Geolocation with 3D TDoA.” CFRS Inc. 2019.

 3 Hyperboloids from 3 time differences.

3 Hyperboloids from 3 time differences.

 Matlab computes the intersection (highlighted) in black between 2 hyperboloids.

Matlab computes the intersection (highlighted) in black between 2 hyperboloids.

MATLAB Simulation Programs

These are the Matlab scripts used to generate the TDoA plots for the feasibility study:

Equilateral Triangle Hyper New

Intersection.m

Right Triangle Hyper

Create Surface

Test Surfaces

Satnogs Audio files

Morphological Chart and Concept Selection

 Morphological Chart

Morphological Chart

Morphological Chart(Excel)

Concept Selection

Criteria Description Related Engineering Requirements
Team Experience The LASSO team should have enough experience in the fields required for the design concept to produce a viable product. Stakeholders
Team Interest The design concept's challenges should fulfill the team's interests and preferred engineering disciplines. Stakeholders
Overall Simplicity Programs for TDoA, orbit determination, signal processing, should be as simple as possible. User interface should be easy to work with. Mechanical and electrical build should not be overcomplicated. 23ER
Weather Resistance The design be functional in extreme hot and cold temperatures and withstand the rain and snow. 1ER, 2ER, 12ER, 13ER
Cost Cost of components should be as low as possible, given our uncertain sponsorship status. 20ER
Noise Noise should be limited to reduce uncertainties in OD and downlink data reception. 15ER, 16ER, 9ER
Field of View Field of View should be as large as possible. 17ER, 18ER
Modular Design should be able to add new functions (for future teams' project implementation) and scalable (to add new ground stations). 23ER
Processing Time Processing time should be as low as possible. 11ER
Maintenance The design should be easy to fix and easy to calibrate. 21ER, 22ER
Size It is better for the ground station's size to be smaller so we don't take too much rooftop space or require an oversized transport vehicle. 19ER
Accuracy The system should accurately determine a satellite's orbit. 3ER, 4ER, 5ER, 5ER, 14ER

System Level Pugh Charts (Excel)

Synchronize Clocks Pugh Charts (Excel)

This consists of two elements:

The real value in this step of the process is not the comparison matrix you generate to compare your concepts, but the analysis and discussion you do to support your evaluation.

Systems Architecture

 High Level Systems Architecture Flow Chart

High Level Systems Architecture Flow Chart

Designs and Flowcharts

 Systems Flowchart

Systems Flowchart

Risk Assessment

Risk Management Part 1

Risk Management Part 1

Risk Management Part 2

Risk Management Part 2

Link to Risk Assessment Working Document (Excel Sheet)

Our three highest risks are the error of TDoA, interfering signals at the 437 MHz range, and the budget. The uncertainty in our TDoA calculations needs to be estimated as soon as possible, because this uncertainty will affect the accuracy of the orbit determination. If the TDoA uncertainty is too high, then the OD accuracy may fall out of our engineering requirements. In the next phase, we will have to figure out the uncertainties of TDoA and OD because it will decide the required distances between ground station and SDR sampling resolution required.

Next, we need to make a proof of concept antenna at 437 MHz to see if we can actually receive a satellite signal with an omnidirectional antenna. It is possible that other signals could drown out satellite signals and make them unrecognizable.

Lastly, our budget will affect the quality the SDRs and other related equipment that we can buy. This will affect how accurate of an OD calculation we can get. Our most common risks are technical and resource related. Although technical risks have a low likelihood the severity of the risk is very high. The most concerning resource related risks are in reference to our current budget, this will require thorough planning to avoid unnecessary spending.

Design Review Materials

Systems Design Review Documents

Presentation Pre-Read Sheet

SDR Presentation

SDR Presentation Notes

SDR Post Presentation Action Items

Plans for next phase

Team Vision

Individual Three Week Plans:

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

https://creativecommons.org/licenses/by-nc-sa/4.0/


Home | Planning & Execution | Imagine RIT

Problem Definition | Systems Design | Preliminary Detailed Design | Detailed Design

Build & Test Prep | Subsystem Build & Test | Integrated System Build & Test | Customer Handoff & Final Project Documentation