things have moved forward, KrakenSDR is now fully controllable by the new (unpublished) ThorSDR code.
Licencing may change to the MIT license, but this is still under consideration. SoapySDR support is also under consideration,
Changes are being made and the ThorSDR code seems quite stable at the moment. Also, we are considering to add soapysdr support in the future. Note that this driver will also support other reference noise switching calibration systems. e.g. KrakenSDR
***10.8.2022 PROJECT DEVELOPEMENT HALTED, WE'LL NEED TO FIND SOMEONE TO CONTINUE THE WORK WITH OUR RECEIVER PROTO.
Perhaps this should be someday converted into a GNU Radio OOT module as a signal source, but I do not have the time until I finish my PhD.
Mostly the same requirements as in my previous coherentsdr proof-of-concept. One added dependency, GNU Readline, for the shell. Some unworking features: adding and deleting receivers during runtime, application does not always exit cleanly.
Matlab client included. A system object interfacing to a .c MEX implementation (matlab c++ interfacing seems to be too slow). This is not built automatically by cmake, instead it has to be compiled manually (instructions in the folder).
Required for compiling:
:zmq: - the zero message queue: sudo apt-get install libzmq3-dev
:fftw3f: - fastest fourier transform in the west: sudo apt-get install fftw3-dev
:volk: - vector optimized library of kernels: sudo apt-get install volk
:librtlsdr: -Tejeez/Keenerds experimental librtlsdr fork: git://github.com/tejeez/rtl-sdr
+GNU Readline
Requires common 28.8 MHz clock & reference signal (noise) for synchronization. Some examples in the electronics folder, but this is still missing the coupler module PCB files.
Currently may return a receive matrix where data on some channels are from a previous sample buffer. This seems to happen under heavy CPU load, at least on limited platforms (testing on RockPI 4 with 21 signal channels). This can be noticed by observing discontinuous channel sequence number (readcnt). Occasionally, all channels skip a buffer in unison, i.e. 8192 sample gap in reception. Perhaps I may need to add some buffering to the packetizer singleton. Under construction.
UPDATE 2.2.2021: We have concentrated on experiments with the receiver and presented a paper at EUSIPCO2020. "Phase-coherent multichannel SDR - Sparse arraybeamforming" can be found at: https://aaltodoc.aalto.fi/handle/123456789/102361
Please cite this as: Laakso , M , Rajamäki , R , Wichman , R & Koivunen , V 2020 , Phase-coherent multichannel SDR - Sparse array beamforming . in 28th European Signal Processing Conference, EUSIPCO 2020 - Proceedings . , 9287664 , European Signal Processing Conference , EURASIP , pp. 1856-1860 , European Signal Processing Conference , Amsterdam , Netherlands , 24/08/2020 . https://doi.org/10.23919/Eusipco47968.2020.9287664
The next paper, in which we utilize deep neural networks on data captured with coherent-rtlsdr to do near-field localization (with surprisingly accurate results), was presented at VTC2021 Helsinki. The article is titled "Near-field localization using machine learning: An empirical study" and it can be accessed at: https://aaltodoc.aalto.fi/handle/123456789/102361
Please cite as: Laakso , M & Wichman , R 2021 , Near-field localization using machine learning: An empirical study . in 2021 IEEE 93rd Vehicular Technology Conference, VTC 2021-Spring - Proceedings . , 9449002 , IEEE Vehicular Technology Conference , vol. 2021-April
Added the reference noise controller firmware, shamelessly edited from libopencm3 examples. This uses the 2$ STM32F103C8T6 (a.k.a bluepill) board to control 2 GPIOS for switching operating voltage for the noise amplifiers & case fan (yes, last summer I had problems with the amplifiers overheating while taking measurements during a sunny day). DFU upgradeable with dfu-util once flashed (hack, but works).