December 2021 Eclipse Festival of Frequency Measurment IQ recording by station EA4GPZ

This dataset contains the IQ recordings done by the station EA4GPZ (Daniel Estévez), in Madrid (Spain), during the HamSci December 2021 Eclipse Festival of Frequency Measurment. The dataset contains 200 sps IQ data for WWV at 5, 10 and 15 MHz, and RWM at 4.996, 9.996 and 14.996 MHz. Each of the six frequencies is in a separate file. Additionally, the full recording span of 10 days is divided into 3 segments, since recording was briefly interrupted twice for technical reasons.

These recordings have been done with a Hermes-Lite2 beta2 DDC/DUC HF SDR transceiver connected to a random wire antenna, which is an end-fed wire of approximately 17 metres length set up as a sloper. The Hermes-Lite2 used a 10 MHz external reference coming from a Vectron MD-011 GPSDO. The generation of the sample clock and its divisions are exact factors with respect to the 10 MHz reference, so the sample rate of these files is exactly 200 sps locked to the GPSDO.

The LO of the DDC is not exactly set to the nominal frequency, so the following mHz scale frequency errors are present:

• 4996 and 5000 kHz: 1/3*2^(-32)*76.8e6 Hz (approximately 5.96 mHz).
• 9996 and 10000 kHz: -1/3*2^(-32)*76.8e6 Hz (approximately -5.96 mHz).
• 14996 and 15000 kHz: no frequency error

The sign convention of these is that the frequency errors listed above should be added to the frequencies measured from the 200 sps IQ data in order to get the correct frequency. For more information about these frequency errors, see this post by the author.

The timestamps of the files are only accurate to a fraction of a second, since they are generated on the computer performing the recording and are not synced in any manner to the GPSDO. It should be possible to assess the accuracy of the timestamps by using the time signals in the recordings.

The recordings were made using GNU Radio 3.8 and gr-hermeslite2. Three FPGA receivers on the Hermes-Lite2 were tuned to 5, 10 and 15 MHz using an IQ sample rate of 48 ksps on each receiver. The IQ data was first decimated by a factor of 4 with a low-pass FIR filter. For the WWV channels, the output of this filter was decimated by a factor of 60 with another low-pass FIR filter. For the RWM channels, the output of the downsample by 4 filter was multiplied by a complex exponential with a frequency of 4 kHz (generated as a periodic vector to avoid any rounding off errors in its frequency) and then decimated by a factor of 60 with a low-pass FIR filter. Data was recorded to 16 bit integer raw IQ files and then converted to SigMF format.

The Hermes-Lite2 was set to an LNA gain of 18 dB throughout the recording, and a digital gain factor of 500 was used in the GNU Radio flowgraph to better adjust the signal amplitude to the range offered by the 16 bit IQ output files. The amplitude scale of this receiver system has not been calibrated to an absolute reference.

Warning: Between the third and second recording segment (i.e., starting at 2021-12-08T12:57:25) there was a power outage that rebooted the Hermes-Lite2 and the GPSDO. Due to operator mistake, the external reference of the Hermes-Lite2 was not enabled after the power outage. Therfore, in the thrid segment, the Hermes-Lite2 is running with its onboard TCXO. This segment should be used carefully for frequency mesaurement. In principle it could be possible to correct the frequency error, because the GPSDO 10 MHz signal leaks into the recording of WWV at 10 MHz.