KerberosSDR Developments in 2020
Posted: Thu Oct 03, 2019 8:33 am
Thanks for your support with KerberosSDR! The people who've purchased units are helping make this open source code happen.
Posted here is a rough road map of what we're planning for the next stage of KerberosSDR development. Please feel free to post any suggestions for features. We won't be giving a timeframe for these developments, but we are expecting most of them to be completed by the end of 2020.
[*] DAQ Double buffering - We are currently working on revised DAQ code that will allow for arbitrary buffer sizes to be set. The buffer size defines the number of samples the code first collects, then gives to the DSP code to work on. Different buffer sizes may be optimal for different signal types. This is necessary for future developments.
[*] Networked DFing - With several networked KerberosSDR units spread out around a city it will be possible to instantly direction find any signal.
[*] Intermittent Signal DFing Improvements - Right now it is difficult to use DFing with very short intermittent signals. Adjustments to the code will allow us to work on sample chunks that consist of mostly noise.
[*]Passive Radar Blips On a Map (PR + DF) - This will be a combination of the passive radar and direction finding techniques. The idea is that it should be possible to direction find the passive radar blips if we use all 4 on board RTL-SDRs and 4 directional antennas. This would allow us to plot the estimated location (via the distance + bearing) of an object being tracked by passive radar on something like Google Maps. Several improvements to the PR code such as target tracking are required for this to be completed.
[*] DAQ Unit Separation - Processing performance could be significantly improved if the DAQ subsystem and DSP processing code is separated onto two different pieces of computing hardware. Presently if the DSP code is set to run too fast, the DAQ code will lose sample sync due to the CPU + OS on a single device not being able to keep up with the task switching requirements. In the long run it would be better to dedicate a Pi3/4 or other cheap SBC to the DAQ and another computing unit to the DSP. This will allow people to easily read the coherent data stream in other programs too like GNU Radio via an Ethernet network stream. For mobile DFing we will keep the option to use just one SBC.
[*] Android DFing App Bug Fixes + Optimizations - There are a few bugs related to entering out of bounds settings in the app. The app also has room for speed optimization.
[*] Proper confidence values for DFing - Currently confidence is roughly calculated. There exists a potential method for calculating if probabilisticly and this will be tested.
[*] Beamforming - Once the DAQ system upgrades are completed, beamforming and listening to actual signals should be possible. Beamforming will allow you to electrically optimize an antenna array for receiving in one particular direction.
[*] Direct support for two combined KerberosSDR units - Combining two KerberosSDR's gives you 8 antenna inputs. This can improve DFing accuracy, and improve passive radar coverage.
[*] DSP Performance Improvements - There is room for DSP code speed improvements. Ideally we'd like to get passive radar updating significantly faster.
[*] Notching - Calibrating with the antennas connected results in better results, UNLESS there are signals in the spectrum. So calibrating with the antennas can be difficult/impossible in the real world. Adding notches in the DSP may help with this.
[*] Hardware: Switches and an improved calibration PCB - It should be investigated if switches degrade the phase calibration or not. With switches we could eliminate the need to disconnect the antennas when recalibrating.
[*] Improved clutter filters for passive radar - The clutter filter removes passive radar blips from stationary objects. Improved algorithms are faster and result in a cleaner PR display.
[*] A more user friendly UI - Wizards and calculators should be added to make setting up the KerberosSDR settings easier.
Posted here is a rough road map of what we're planning for the next stage of KerberosSDR development. Please feel free to post any suggestions for features. We won't be giving a timeframe for these developments, but we are expecting most of them to be completed by the end of 2020.
[*] DAQ Double buffering - We are currently working on revised DAQ code that will allow for arbitrary buffer sizes to be set. The buffer size defines the number of samples the code first collects, then gives to the DSP code to work on. Different buffer sizes may be optimal for different signal types. This is necessary for future developments.
[*] Networked DFing - With several networked KerberosSDR units spread out around a city it will be possible to instantly direction find any signal.
[*] Intermittent Signal DFing Improvements - Right now it is difficult to use DFing with very short intermittent signals. Adjustments to the code will allow us to work on sample chunks that consist of mostly noise.
[*]Passive Radar Blips On a Map (PR + DF) - This will be a combination of the passive radar and direction finding techniques. The idea is that it should be possible to direction find the passive radar blips if we use all 4 on board RTL-SDRs and 4 directional antennas. This would allow us to plot the estimated location (via the distance + bearing) of an object being tracked by passive radar on something like Google Maps. Several improvements to the PR code such as target tracking are required for this to be completed.
[*] DAQ Unit Separation - Processing performance could be significantly improved if the DAQ subsystem and DSP processing code is separated onto two different pieces of computing hardware. Presently if the DSP code is set to run too fast, the DAQ code will lose sample sync due to the CPU + OS on a single device not being able to keep up with the task switching requirements. In the long run it would be better to dedicate a Pi3/4 or other cheap SBC to the DAQ and another computing unit to the DSP. This will allow people to easily read the coherent data stream in other programs too like GNU Radio via an Ethernet network stream. For mobile DFing we will keep the option to use just one SBC.
[*] Android DFing App Bug Fixes + Optimizations - There are a few bugs related to entering out of bounds settings in the app. The app also has room for speed optimization.
[*] Proper confidence values for DFing - Currently confidence is roughly calculated. There exists a potential method for calculating if probabilisticly and this will be tested.
[*] Beamforming - Once the DAQ system upgrades are completed, beamforming and listening to actual signals should be possible. Beamforming will allow you to electrically optimize an antenna array for receiving in one particular direction.
[*] Direct support for two combined KerberosSDR units - Combining two KerberosSDR's gives you 8 antenna inputs. This can improve DFing accuracy, and improve passive radar coverage.
[*] DSP Performance Improvements - There is room for DSP code speed improvements. Ideally we'd like to get passive radar updating significantly faster.
[*] Notching - Calibrating with the antennas connected results in better results, UNLESS there are signals in the spectrum. So calibrating with the antennas can be difficult/impossible in the real world. Adding notches in the DSP may help with this.
[*] Hardware: Switches and an improved calibration PCB - It should be investigated if switches degrade the phase calibration or not. With switches we could eliminate the need to disconnect the antennas when recalibrating.
[*] Improved clutter filters for passive radar - The clutter filter removes passive radar blips from stationary objects. Improved algorithms are faster and result in a cleaner PR display.
[*] A more user friendly UI - Wizards and calculators should be added to make setting up the KerberosSDR settings easier.