Maybe you remember, at March 22, Cees Bassa and Tammo Jan Dijkema made a planar scan with the Dwingeloo 25m radiotelescope, and they detected some transmissions of Sprites, and they recorded this in a big file. You can download this file from their site,
https://charon.camras.nl/public/satellites/The file size is 2.7 Gbyte, and the name is: KICKSAT_PI9CAM_2019-03-22T19:55:09_437.35MHz_500ksps_complex.raw
Based on this recording, Daniel Estevez now published his analyzis of this recorded data:
"Detecting the Sprites from KickSat-2"
https://destevez.net/2019/04/detecting-the-sprites-from-kicksat-2/#more-6350It is a long paper with many interesting details.
Here I emphesised only some sentences:
- It is interesting note from him, at the time, when was the recording, "At that moment, the Sprites were up to 5 minutes ahead KickSat-2, due to their much higher drag to mass ratio. They all probably reentered a few days after this." this is corresponds to the predictions, published here.
- He extracted not complete-transmissions from some sprites, he noted "shorter transmissions probably indicate that the transmission failed due to lack of power. As soon as the transmission started, the bus voltage dropped and the transmission stopped. It is also interesting to note that despite the large number of sprites launched, only 4 of them have been successfully detected. This seems to indicate that the failure ratio of these Sprites is very high."
-" Only one of them was functioning normally, transmitting complete messages. The remaining Sprites showed transmissions shorter than a complete message. Since more than 100 Sprites were launched by KickSat-2, this seems to indicate a high failure rate"
We can add the note, this analysis based only on this 11 minutes long recording. If we have another recorded data, maybe this result will extended.
- An interesting aspect of this analysis, that is it very "computationally expensive", need a high-end configuration: "the correlation against the complete recording amounts to more than 2.8 thousand million FFTs of 2048 samples. The algorithm took nearly 58 hours to run in my i7-2620M CPU, using dask for parallelization. This means that my machine is doing 13800 FFTs per second.".
t.janos