Mega Channel Spectrometer
Status: Unfinished
Introduction
When implementing spectrometers in ROACH2 hardware with over 100,000 channels, using standard biplex FFT architecture becomes unfeasible due to the large amount of computational resources required. An alternative structure exists that make possible the implementation of over a million channel spectrometer in ROACH2, which consist in cascading two FFT blocks together (the output of one FFT is the input of the other). Specialized blocks called 'corner turner' are use between the FFTs with the purpose of input the data of the last FFT in the correct order (mathematically, they perform a matrix transposition). The disadvantages of this method are that memory requirements increase, and channel leakage is present in non-consecutive channels (different from the standard leakage from FFT blocks).
Documentation
Mega channel spectrometer docs:
Paper the Cooley-Tukey (i.e. corner turner) method: https://ieeexplore.ieee.org/document/1458140
SETI page about a similar concept (simpler, yet more flawed) to increase the channels of an spectrometer: https://casper.ssl.berkeley.edu/wiki/SETI_Spectrometer
Mailing list discussion about SETI spectrometer: https://www.mail-archive.com/casper@lists.berkeley.edu/msg05384.html
Edgardo Huaracán report comparing Cooley-Tukey and SETI spectrometer: https://drive.google.com/open?id=1RVbaDImfxhs6qMvRBe_foY6WvXj8A11S
ROACH2 QDR docs (necessary for the mega-channel spectrometer implementation):
QDR block documentation by CASPER: https://casper.ssl.berkeley.edu/wiki/Qdr
Mailing list discussion about using ROACH2 QDRs: https://www.mail-archive.com/casper@lists.berkeley.edu/msg05774.html
Chenwei report on QDR: below
Felipe Vega report on QDR: missing?