A GMSK VHF-uplink/UHF-downlink transceiver for the CubeSat missions

Functional and thermal performance characteristics of a very high frequency/ultra high frequency (VHF/UHF) transceiver based on Gaussian minimum shift keying (GMSK) modulation are presented. The transceiver has been designed for CubeSats telemetry and commanding needs or low rate data download. The design is validated at 27 dBm, 30 dBm and 33 dBm transmitting powers over −20 ∘C\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^\circ \hbox {C}$$\end{document} to +51 ∘C\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^\circ \hbox {C}$$\end{document}. Under these thermal conditions, the transmitter spurious dynamic response shows little if any change and the average sensitivity of receiver at the 12 dB signal noise and distortion (SINAD) is −116.7 dBm at 140 MHz and −116.78 dBm at 149.98 MHz. The transmitter and receiver frequencies are stable and the current consumption as well the output RF levels are steady. The design has been verified against a simulation model which allows system tradeoff analysis. The measurements demonstrate the transceiver made with commercial grade parts has dependable performance at the low earth altitudes and orbital heating conditions.


Introduction
The very high frequency/ultra high frequency (VHF/UHF) is widely recognized as a model of authority for excellence in performance in space communications. The technology has distinguished itself over several decades as reliable in the high-end missions [1,2]. The technology also rules as the standard communications fundament in the missions flying the CubeSat class of satellites [3,4]. Emerged from the terrestrial mobile radio applications, Gaussian minimum shift keying (GMSK) [5,6] is a type of fast frequency shift keying (FSK). It, too, has settled as a robust, spectrum compact, continuous phase but constant envelope modulation scheme, suited for noisy environments and requiring nonlinear Class C amplification. In this paper, we examine, through functional and temperature measurements, the several performance metrics of a GMSK-based VHF/UHF communication subsystem that has been successfully flying the ZA-CUBE-1 mission in full duplex mode and its simplex variant will be flying the ZA-CUBE-2.

VHF/UHF transceiver
VHF/UHF transceiver ( Fig. 1) is compatible with the Cube-Sat nanosatellite standard and CubeSat Kit PC/104 form factor. The frequency of RX/TX operation is software selectable within the band and is adjustable in channel bandwidth steps. Additionally, a Consultative Committee for Space Data Systems (CCSDS) 1 / 2 rate convolutional encoder with constraint length K = 7 (171 octal, 133 octal) can be enabled to encode the data. The commercial band transceiver has the features listed in Table 1. The AFSK does not operate in full duplex mode exclusively. A variant of the transceiver operates in the amateur band.

Signal design for GMSK
In GMSK modulation, the non-return-to-zero (NRZ) binary pulse is shaped by a pre-modulation, Gaussian, lowpass filter which has an impulse response with no zero crossings. The filtered baseband GMSK pulse model [7] is given as where T is the symbol period and B is the 3 dB bandwidth. When the filter is not applied, bandwidth-symbol time product BT→ ∞ i.e., the modulation is MSK and the Gaussian shape of the pulse reduces to a rectangular shape: g(t) = 1 2T for 0 ≤ t ≤ T . The MSK power spectrum rises above the GMSK, loses content in the sidelobes and incurs penalty on the link quality through increase in the occupied channel bandwidth (OCBW) and adjacent channel power ratio (ACPR). The GMSK spectrum, however, is more compact �� and improves these metrics, but at the cost of higher intersymbol interference (ISI), SNR and per bit SNR ( E b N 0 ) at the receiver. Figure 2 illustrates that the spectral density of GMSK attenuates sooner than MSK. The impulse response of the Gaussian filter in GMSK is f 3 dB = BT ⋅ bit rate . For BT =0.3 the cutoff frequency ranges from 360 to 2880 Hz. When BT is small, the pulse peaks and spreads over less number of bit periods and ISI increases. When BT is large, the pulse attenuates but spreads over more number of bit periods, resulting in better ISI. For efficient transmissions, in a GMSK system, it is imperative to design the pulse for a nominal BT to suppress the out-of-band spectrum leakage. For example, the European Telecommunications Standards Institute (ETSI) Global System for Mobile communications (GSM) specification imposes BT = 0.3. Figure 3 shows the simulation model of the transmitter section. The RF power is adjustable in 27 dBm, 30 dBm and 33 dBm levels.

Analysis setup and the simulation model
In addition to the bandlimited nonlinear distortions present in the system, the channel is additive white Gaussian noise (AWGN), having zero mean noise with infinite   To avoid receiver complexity in digital hardware to keep up with high, impractical sampling rates, the MSK demodulator is generally implemented as equivalent in-phase quadrature (IQ)-type orthogonal quadrature detector [8] in which the IQ signal variation is little compared to the phase variation of the carrier. Regardless of the occupancy of bandwidth, the average bit error probability (BEP) of coherent detectors whether BPSK, QPSK, OQPSK or precoded MSK type is equivalent [9]. This ideal situation occurs when the local oscillator phase and frequency are completely synchronized to the received carrier. The BEP is given as An estimate of the bit error rate in AWGN transmission for maximum likelihood quantized and unquantized detection and several modulator structures is shown in Fig. 4. The thermal noise power spectral density at a temperature N 0 is given as [10] where h is Planck's constant, k is Boltzmann's constant and T is the temperature, and is the frequency band under consideration. Note that h kT ≪ 1 . The thermal noise power varies according to Eq. (4) in the thermal cycle from 3.02e−8 to 4.93e−8 W.
The free space path loss (FSPL) is taken at 600 km altitude.
where d the distance between transmitter and receiver and f the signal frequency are in Hz and m, respectively, and C is speed of light in vacuum. The periodogram of the output signal is shown in Fig. 5. The second harmonic occurs at −40 dBc above which and up to the fundamental is the clear signal space known as the spurious free dynamic range (SFDR). The model has been setup in a way that the largest spur happens to be the second harmonic. This has been the case according to the empirical data as we shall see in Sect. 3.
Various nonlinearities contributed by the power amplifier (PA) are modeled which distort the signal and add offsets in the carrier frequency, phase and the IQ gain. The PA is modeled using a truncated power series polynomial P having peak voltage amplitude A. The model suits a weak nonlinear device giving off only static, memoryless, nonlinearity to an instantaneous input. This is so as the PA on the transceiver has been organized to operate in the linear range of amplitude-amplitude modulation (AM-AM conversion).
The polynomial shown in Eq. (6) describes the spectral contributions in the amplifier output. where coefficients a k of the Taylor series expansion are real constants. On a real amplifier, these coefficients can be determined in laboratory on AM-AM power sweep test by employing curve fitting techniques on the measured trace on a signal analyzer.
Considering the input signal to be narrowband and assuming the amplifier transfer characteristics are frequency independent [11], the amplifier output is digitally corrected by gauging the input as a linear ramp. To start the approximation using the distortion data, a third degree polynomial is numerically fitted, with the least square method, to an input of linear ramp. The amplifier output P(x) comprising of harmonic powers in amplitude x, and the coefficients given in Eq. (7) are computed. Note the dominance of linear gain, followed by the bias and the second harmonic, and small dynamics of the nonlinearity.
The corrected response is in Fig. 6 showing that the harmonics now have been significantly rarified.
The resulting dynamic range through the correction and improved spectral purity is then determined by estimating the SINAD, SNR and total harmonic distortion (THD) metrics against a sweep of input attenuation. The optimum metrics thus obtained are plotted in Fig. 7.
The use of attenuator models has bearing on the load being driven and tradeoff between the load and the tolerable attenuation level becomes necessary. As shown in Fig. 7, it is expected that the power loss to an ideal load would occur around 15 dB at which the dynamic figures of merit tend to be the most favorable.
The strongest TX signal and expected upper ACPR is shown in Fig. 8. The ACPR is about 38 dBc down at full power. Therefore, the strength of spectral growth may be expected to be insignificant, chiefly due to the linear amplification. At the ground station, the received signal is modeled for sensitivity level of − 129 dBm. Figure 9 shows the sensitivity.

Thermal analysis
Thermal finite element analysis (FEA) of the transceiver board shows the temperature field depicted in Fig. 10 and at least 70 • C below the PA die surface of the board. In addition to the thermal plane in the PCB stack-up, the thermal gradient is readily counterbalanced using an enclosure which will also act as the RF shield for protection for the radiated immunity of the RF front end.
An infra-red (IR) temperature gun is used to infer the temperature of the hot spots on the actual board at UHF frequency under nominal RF test conditions. Figure 11 shows this imagery.

Transmitter characterization
The performance of radio is evaluated with a test setup that is shown in Fig. 12. The radio transmission is analyzed over the thermal cycle using HP8920 RF communications testset. The FM reference applied is ANSI/TIA/EIA-RS-316-B, RS-152-C. The test frequency range is 25 to 1000 MHz and 25 to 866 MHz.
An automation software [12] loads the radio configuration using an Arduino device over the I 2 C interface to the radio FPGA and also performs the electrical and functional tests in the thermal cycle. The configuration includes: up and

DC and RF power characterization
The health monitoring of the transceiver board is illustrated in Fig. 13. The board telemetry is updated once every second. The switched mode power supply (SMPS) and PA temperature sensors are located close to the respective components they monitor, namely the SMPS and PA ICs. The temperature is monitored during transmission. The SMPS and PA temperature registers contain the temperature   Table 2 lists the power consumption when measured in the thermal cycle. On the average, at 0.5 W, 1 W and 2 W transmit levels and the receiver also running, the drift in power consumption is 3%, 0.8% and 1.6%, respectively, while the receiver consumption swings as much as 10% in the cycle. Both the receiver and transmitter tend to consume less power in the cooler end of the thermal cycle.
The output at different transmit levels is listed in Table 3. Overall, the maximum variation in the RF power is no more than 1.1 dBm in the thermal cycle. From both tables, it is clear that the nominal performance of the transceiver may be expected in 1 W mode.
Within the thermal cycle, the relationship of the DC and RF power at different transmitting levels can be seen in Figs. 14 and 15.

FM deviation
The FM stability of the output modulation is measured over temperature as the maximum difference between the modulated frequency and the carrier central frequency. Table 4 shows the average test results in thermal cycle. The strong stability is likely to compact the OCBW and spectral leakage according to the Carson's rule [13] as

Spectral purity and spurs
The spectrum is searched for the presence of harmonics, harmonic products, nonharmonics or any other spurious  component under a single tone test. The center frequency is tuned in 1 MHz steps from 400 kHz to 1000 MHz on the testset. Furthermore, to isolate susceptibility to any interferer in the vicinity, the source of the emissions is verified to be the transceiver itself. This is done by unkeying the transmitter. Table 5 lists the levels of the dominant harmonics found in the spectrum at the maximum transmit level and within the thermal cycle. These levels are fairly consistent and low enough throughout the cycle. In addition, they confirm the amplifier being weakly nonlinear. The second-order spurious response, generally, slightly worsens at the hot end of the thermal cycle. This observation is true for all power levels as indicated in Fig. 16.
At different power modes, the effects of variation in RF power, carrier frequency and the spurious response in the thermal cycle are plotted in unison in Fig. 17. The level of second harmonic gives insight to the DC component in the carrier as their amplitudes are equal, caused by the square law term ( a 2 A 2 ∕2 ) in the power series (Eq. 6). The DC component is at least − 7.7 dBm or 92 mV.

Forward and reverse power
The forward and reverse power registers can be read via the I 2 C telemetry interface (Fig. 13). These registers are used to calculate the forward and reverse RF transmit power, as well as the return loss, and thus, indicate any mismatch in impedance between the transmitter and the antenna. That is, the transmitter may be calibrated for possible antenna voltage standing wave ratio (VSWR) conditions. The registers contain 12-bit values of the PA's forward and reverse power which are converted into voltages as The coupled forward or reverse power is obtained as The actual forward or reverse power is obtained by adding 32.5 dBm to P coupled fwd|rvs . The spread of forward and reverse power is illustrated through the ovals in Fig. 18 at the respective transmitting levels.

Return loss
The return loss is the difference of the actual forward and the actual reverse power, i.e., The losses measured at several attenuation levels of the signal using directional couplers are displayed in Fig. 19. The return losses are 13.86 dB, 14.18 dB and 14.10 dB at the 27 dBm, 30 dBm and 33 dBm, respectively. Recalling the simulation on selecting the best attenuator had shown a 15 dB attenuator (Fig. 7). The discrepancy may be improved by carefully choosing better coefficients of the polynomial.

Amplifier nonlinearity
Although GMSK is insensitive to phase distortion and nonlinearities, it is still prudent to operate the PA in small-signal (linear) region for radiated emissions control, most notably, for better ACP and reducing the spectral spread measured as the OCBW. There is a large, linear, dynamic operating range available in the PA which is utilized for the aforementioned purposes.
A basic block diagram of the amplifier circuit is in Fig. 20. PWR_REF returns a voltage proportional to the DC component of the drive current in the output stage and PWR_SENSE contributes to the RF components as well the DC. The difference of the current components yields the RF-only current and the resultant RF power. The logarithmic difference and the RF power maintains a linear relationship up to 33.5 dBm. This is the upper limit of linearity to which we operate at maximum 2 W mode. Beyond 34.5 dBm, which is the beginning of the 1 dB compression point and with the employed power control scheme, the amplifier may saturate, gain flattening may be seen or even constriction may appear, unless a different control scheme e.g., hybrid couplers and input/output matching networks are used in a balance amplifier configuration. (11) Return Loss = −10 log 10 P reflected P fwd

Amplifier efficiency
The price of linearity is paid in efficiency. Figure 21 shows the amplifier efficiencies tabulated at different transmit levels.
A maximum efficiency of 33.5% is observed at the coldest end of the thermal cycle in 2 W mode. Even at the hottest temperature and in the least power mode, the amplifier is at least 24.5% efficient.

Receiver characterization
The receiver current varies 11% during the thermal cycle. The measurements are shown in Table 6 which improve at low thermal noise in the cold end of the cycle.

Minimum discernible signal (MDS) and spurious-free dynamic range (SFDR)
The receiver front-end noise figure (NF) is measured to be less than 1.5 dB. This yields an equivalent noise floor in a 12.5 kHz channel according to The noise floor sits at −132 dBm to −131 dBm at the extremes from −20 to +50 • C . The noise floor is the minimum discernible signal level. Upon evaluation of several boards, the SFDR is specified at − 120 dBm. The extreme nonlinearity of the receiver is then estimated as the input intercept point of third order (IIP 3 ), shown in Eq. (13), which includes not only the harmonics and their variants but also the intermodulation (IM) product differences of any interferer with the carrier.  The IIP 3 turns up around − 49 dBm. This figure is quite low, for example, compared to GSM-900 service. The reason is the absence of interferes such as the co-located subscribers inducing IM distortion in the terrestrial mobile radio services.

RSSI
The receive signal strength indicator (RSSI) telemetry (Fig. 13) is a 12-bit value returned over two bytes. The RSSI provides an indication of the current received signal strength. This telemetry should typically be captured when a signal is being received from the ground. Table 7 provides an indication of the RF level given as the output voltage. The higher the received voltage the better the RF signal being received. Note that this is a noisy telemetry channel unless a carrier signal is present.
The conversion is according to Eq. 14.

12 dB SINAD
The FM receiver's signal noise and distortion (SINAD) ratio at the output of the receiver is defined as Effectively, the SINAD is SNR without the harmonics. Typical specifications, rooted in the law enforcement agencies and emergency vehicle audio communications intelligibility, call for 12 dB [14]. The level corresponds to 4:1 at which the contribution of noise and distortion is 25% to the signal amplitude. The receiver input is fed with a VHF frequency signal that is modulated with a 1 kHz tone that introduces spurs and distortions in the output. At the output, the SINAD meter of HP8920 measures the power of the modulating tone and the noise and distortion. The power of the modulating tone is filtered out and the noise and distortion power is measured in the channel bin. The ratio of the two power measurements is the SINAD. A meter averaging function is used for 10 readings to measure a consistent signal quality.

12 dB SINAD sensitivity
The minimum level of the modulated radio carrier input that is required to yield 12 dB SINAD ratio at the receiver at specific power output is the receiver sensitivity. During the SINAD test, the amplitude of the receiver input is adjusted such that the SINAD reads 12 dB. This level of the receiver input signal is the 12 dB SINAD sensitivity.  The sensitivity is measured during the full thermal cycle at three VHF frequencies. On the whole, the level turns out to be − 117 dBm and is rather insensitive to the variation in temperature. This is illustrated in Fig. 22. The largest variation is found at the highest frequency of the channel. The test reference applied is EIA RS-204-D.
Although not exclusively measured, an estimate of the noise and distortion content at the sensitivity level, and its variation in the thermal cycle may be established through Fig. 23. The typical contribution of noise and distortion is around − 146 dBm. That is, about 30 dBc below the sensitivity.

Receiver Eb∕N 0
Using the 12 dB SINAD (instead of SNR), we estimate worst case Eb∕N 0 at the receiver sensitivity and when the bit rate is maximum. It turns out to be 15.6 dB according to Eq. 16 and will improve if determined against the SNR.

Frequency stability
Receive center frequency stability was checked by measuring the frequency where the best sensitivity is obtained and the center frequency did not shift with temperature either.

Conclusion
The averaged results on the communication performance of several VHF/UHF transceiver units for the CubeSats have been presented. The analysis began by first developing a simulation model to reason out the expected performance. The results are in accordance with the simulations and the analysis laid down. If the amplifier linearity is maintained, (16) Eb∕N 0 = SINAD f bit rate excellent radio performance may be expected. Appropriate thermal balance mechanism would avoid thermal memory effect problems and distortion during power conversion. To sum up, the work demonstrates robust performance of GMSK modulation and the transceiver impairments are small. The performance data may assist in system engineering activities such as the antenna design, spectrum analysis for the filing of bandwidth license, revising electric power budget and selecting the frequency for the highest quality line-of-sight link budget.