SATRE modem Two Way Satellite Time and Frequency Transfer receiver implemented as a GNU Radio processing block. The SATRE is a proprietary hardware sold by TimeTech but based on the MITREX microwave two-way satellite time transfer modem developed at Univ. of Stuttgart in the 1980s: most information concerning the CDMA and transmitted codes are collected from the open litterature describing the latter (see bibliography/ directory). TimeTech has added a digital communication layer that is not understood yet, since the current investigations focuses on reception for time transfer only.
Two large (~100 MB) files have been recorded using either a B210 or an X310 Ettus Research SDR receiver, the former set to a 50 dB gain on the AD9361 front end tuned for frequency transposition of the 70 MHz intermediate frequency output of the TX and RX monitor of a SATRE modem, the latter fitted with a two BasicRX frontends (no gain), the RX monitor channel being fitted with a single Agilent MSA886 monolithic amplifier. Both datasets store short integer (16-bit) data, interleaved complex, interleaved TX and RX channels. Notice that due to a connection mistake, RX and TX have been swapped between the two datasets. The sampling rate is 5 MS/s, so that 100 MB files will hold 100(MB)/5(MS/s)/2(complex)/2(short int)/2(channels)=2.5 s worth of data.
A detailed description of the SATRE, and its predecessor the MITREX modem, encoding is found in Appendix 2 of G. De Jong & al, "Results of the calibration of the delays of Earth stations for TWSTFT using the VSL satellite simulator method", 27th PTTI (1995) at https://apps.dtic.mil/sti/pdfs/ADA518490.pdf
- Start by checking that a signal has been recorded and the coarse frequency offset by squaring the BPSK modulated signal for getting rid of the spectrum spreading by the modulation and collapsing the energy in each carrier
- Check the code length by autocorrelating the signal and verifying that the code length is 10000 bits at a rate of 2.5 Mchips/s or a Pulse Repetition Interval of 4 ms
- The
identify_delay.mscript is the most important for identifying delays between received stations. Here we start by correlating the frequency shifted signal with the pseudo-random sequence. Which code is associated with which frequency shift has been identified by exaustive search as described in the./reverse_codesubdirectory, although the list of frequency offsets for each station can be found in A. Kanj, "Etude et développement de la méthode TWSTFT phase pour des comparaisons hautes performances d’étalons primaires de fréquence" [in French] (2013) at https://tel.archives-ouvertes.fr/tel-00831596/document on page 54 (table 2.2)
The frequency offset only needs to be corrected by less than the inverse of the code duration, or
4 ms at a rate of 2.5 Mchips/s and a code length of 10000 chips. Hence, an initial coarse frequency
offset was found by squaring the signal with a resolution of 5 MS/s/N=131072=38 Hz which is much less
than the required 250 Hz, and after an initial correlation sequence identification the phase of the
correlation is unwrapped for fine frequency offset identification to be used as long as this offset
is not larger than 1 Hz. The frequency_residue variable in identify_delay.m checks that the
frequency offset does not diverge during processing.
- Following fine frequency correction, each correlation peak every 4 ms is fitted with a parabola for oversampling and fine tuning the correlation delay with an improved time resolution equal to the measurement signal to noise ratio
We display on top-left the correlation delay after oversampling -- exhibiting the sharp jump by half a chip period every beginning of a second -- and on bottom the station correlation delay subtracted from the Paris Observatory clock correlation delay. Top right is the phase after unwrapping the squared signal phase exhibiting the BPSK digital communication signal, and bottom right is the sentences reshaped with the sentence length of 250 bits to try and find some pattern in the bit sequence.
- The same result is achived for ranging using
identify_range.m, this time comparing the transmitted signal and the received signal (assumes one has access to the SATRE emitter for recording the TX Monitor signal)
The standard deviation stdval on the time delay resulting from analyzing this 2.5-s long sequence is 5 to 10 ns
depending on the recorded signal signal to noise ratio.
All satellite link information are documented in the files stored by BIPM. Looking at OP records, the downlink frequency for communications within Europe is 10953.9500 MHz while downlink from the US uplink is 11497.0600 MHz. The latter is verified in the following measurement taken from a TV satellite reception parabola dish with the LNB local oscillator set to 9.75 GHz.
From the USNO and NIST files, it can be assumed that the USA downlink frequency is 11747.7400 MHz although this link has not been verified experimentally (lacking a parabola dish pointed towards Telstar11N located in the USA).
The SATRE modulation is BPSK at 2.5 Mchips/s so that sampling at 5 MHz is sufficient for collecting all the information transfered over the communication channel. However at 5 MHz, the sampling period is only 200 ns and the time resolution of the correlation output is insufficient. A classical solution is oversampling the correlation peak with a parabolic fit to identify finely the time of flight. Running this algorithm on the data sampled at 5 MHz leads to periodic fluctuations whenever the correlation peak switches from one sampling interval to the next:
Quite surprisingly, interpolating the data (ie not adding any information since interpolation involves zero-padding
the Fourier transform of the signal and the code product prior to calculating the inverse Fourier transform
to recover the correlation in the time domain
The SATRE manual mentions communicating "via HiRate data all necessary information to enable calculation of the predicted arrival of time (PAT) at the satellite by remote modems receiving the ranging signal" and goes on describing the time of flight compensation formula.
The only information about HiRate is that Besancon Observatory identifier is 727 as displayed on the SATRE modem screen.
Having identified that HiRate is encoded at 250 bps using BPSK, it is quite natural to assume differential encoding to avoid π phase rotation of the Costas loop/atan(Q/I) output. The differential encoding of 0d727=0b01011010111 is 100111001111 (start with d(0)=1 and iterate with m(k)={0, 1, 0, 1, 1, 0, 1, 0, 1, 1, 1}, defining d(k+1)=1 if d(k)==m(k) and d(k+1)=0 otherwise.
After removal of the differential BPSK encoding, all header/footer bits look similar and the HiRate index code repeats
every 500 bits or two sentences as shown on the cross-correlation of the output variable from hirate_digital_mode/digi15.m
with the binary encoding of 0d727.
Bit map after removal of the differential encoding:
Comparing the digital signals collected simultaneously and decoded from station 09 and 15 combined as an RGB image exhibits the bits common to all modems and thos individual to each transmission:
