DYNAMIC RANGE IMPROVEMENTS OF A BEACON RECEIVER USING DSP TECHNIQUES

Transcription

1 DYNAMIC RANGE IMPROVEMENTS OF A BEACON RECEIVER USING DSP TECHNIQUES Cornelis J. Kikkert James Cook University, Electrical and Computer Engineering, Townsville, Queensland, Australia, Keith.Kikkert@jcu.edu.au ABSTRACT This paper describes how DSP techniques can be used to improve the dynamic range of a satellite receiver for monitoring the beacons on communication satellites in order to gather statistics on the effects of rain on satellite communication. In many situations Radiometers have been used prior to the availability of satellite beacons. The use of DSP techniques allows both the beacon signal and the received noise to be measured at the same time in the one receiver, thus allowing the radiometer and beacon measurements to be correlated for developing improved radiometer models and enhance the accuracy of previous radiometer measurements. INTRODUCTION Attenuation of satellite signals due to rain is very significant for frequencies above 5 GHz [1, 2]. As the spectrum becomes more crowded, operators are forced to use higher and higher frequencies. To enable the higher frequency bands to be marketed, the probability of a communication channel not being available due to rain needs to be known for particular receiver locations and dish sizes. The most effective technique used to measure rain attenuation is to conduct an experiment, which monitors the received signal strength of a satellite beacon. Since the Uplink Power Control (UPC) beacons on the satellites are not modulated, they can be used to determine the attenuation due to rain. INTELSAT satellites have Ku band beacons at GHz and GHz. The Optus satellites have a Ku band beacon at GHz. With the introduction of Ka band services on newer satellites, Ka band beacons are now also becoming available and are used for rain attenuation experiments [3]. The Satellite Transmission Rain Attenuation Project (STRAP) team at James Cook University (JCU) has been involved in microwave propagation research for many years. As part of this research an analogue Ku band beacon receiver was developed (STRAP1) [4] and installed at the Bukit Timah Earth station in Singapore in This receiver used seven successive frequency shifting and amplification stages to place the beacon signal at the centre of the final Intermediate Frequency (IF) filter which is centred at khz. PLL Ku band LNB s have less than ±25 khz diurnal variation in frequency. This frequency variation is still many times the 100 Hz final IF filter bandwidth and a sophisticated frequency control system had to be used to keep the satellite beacon at the centre of the final IF filter. The original beacon receiver used analogue circuitry for its frequency control and beacon amplitude detection and is expensive to construct and align. A web search of commercial beacon receivers [5, 6] shows that most beacon receivers are still made using similar techniques and use a typical 50 khz final IF bandwidth, giving a 27 db lower dynamic range than the STRAP1 receiver with a 100 Hz bandwidth and a 37 db lower dynamic range than the 9.5 Hz bandwidth used in the STRAP4 DSP based receiver described in this paper. DIGITAL RECEIVER PRINCIPLES The JCU STRAP team have designed a series of Satellite Beacon Receivers [7, 8, 9, 10], which use Digital Down- Conversion and Signal Processing Techniques to detect the satellite beacon signal. The beacon signal at the 5.5 MHz IF stage used in the beacon receiver is digitised using a 10-bit Analogue to Digital Converter (ADC). A sampling frequency of 20 MSPS satisfies the Nyquist rate and avoids any harmonic aliases. A Direct Digital Down-Converter (DDC) IC, is used to produce an IF at DC with both In-phase (I) and Quadrature (Q) components. The required frequency shift can simply be changed, by controlling the numerical oscillator in the DDC IC. Since the I and Q data are centred at DC, a decimating filter can be used to restrict the bandwidth to be analysed to the required resolution.

2 Fig. 1 shows the block diagram of the indoor unit of the beacon receiver, with emphasis placed on the DSP control. A commercial LNB is used at the focal point of the antenna dish in order to convert the Ku or Ka band beacon signal into the MHz frequency region. A Ku band PLL LNB has been used to provide the results in this paper. A Xtal multiplied UHF Local Oscillator (LO) is used in order to minimise the phase noise. The UHF LO and the cavity filter are specially designed for the frequency of the beacon on the satellite. The VHF LO is controlled by the DSP to ensure that the beacon signal is placed at the centre of the final IF filter bandwidth. Fig. 2 shows the hardware for the indoor unit corresponding to the block diagram of Fig. 1 in operation. Xtal Multiplied UHF LO UHF IF BW 5MHz 135 MHz Cavity Filter and Amplifier Input MHz, -80 dbm typ MHz VHF PLL LO Coarse Frequency Control 5.5 MHz IF BW 150 khz 5.5 MHz ADC 20 MSPS Cos(5.5 MHz t) Sin(5.5 MHz t) Decimating Filter Decimating Filter Fine Frequency Control Q Data, 156, 39 or 9.8 ksps Decimation Control I Data, 156, 39 or 9.8 ksps Fine Frequency Control DSP Processing Engine Digital Data to Logger Fig. 1 Block Diagram of the Beacon receiver. Fig. 2 The indoor unit corresponding to Fig.1. The I and Q data resulting from the decimation are sent to the DSP board to determine the satellite beacon signal amplitude and frequency. In a DSP based system the detection of a satellite beacon signal is best done using an FFT. The output from the FFT is then analysed to determine the exact frequency and amplitude of the satellite beacon signal. During each measurement, which occurs at 9.5 times per second, 1024 I and Q values are sent serially to the Digital Signal Processor. A fourth order Blackman-Harris window is applied to the data and a 1024 point Radix-4 FFT is

3 performed. A peak detection algorithm is applied and the magnitude and frequency of the largest signal, which is the satellite beacon, is determined. The decimation ratio is varied by the DSP engine, depending on the operating conditions. For normal measurements, decimating by 2048 is used and this achieves a 33 db increase in SNR compared with no decimation and thus gives a significant improvement in the dynamic range of the 10-bit Analogue to Digital Converter. At full resolution, each bin of the FFT is 9.5 Hz wide. Typically 95% of the energy of a satellite beacon transmitter is contained in less than a 100 Hz bandwidth. The beacon signal energy must thus be evaluated by summing the signal power over several bins. The number of bins used depends on the FM noise of the beacon and the FM noise of the receiver. RADIOMETER OPERATION In this receiver, the background noise power can be evaluated by summing the noise over a number of FFT bins, away from the carrier. Since the noise figure of the Beacon Receiver is very low, the indicated noise is substantially due to the transmission medium. The noise level indication can be used like a radiometer, making this beacon receiver unique and allowing both the reduction in signal level and the increase in thermal noise during a rain fade to be measured along exactly the same path and at the same frequency. By considering the different noise sources that contribute to the measured noise and how these change as the atmospheric medium temperature changes through the presence of rain drops, a relation between attenuation and medium temperature, known as the radiometer equation, can be established [2, 9] as follows: Attenuation = Log [ Ht m H ( t (1 H ) t t ) g + t rec Nrec ] kbg m c 10 (1) In this expression the following variables and constants are used: H - Antenna Factor (0.9) t m - Medium Temperature (290 K). t c - Cosmic Noise (10 K). t g - Ground Temperature (310 K). t rec - Receiver Temperature (180 K, 2 db Nf) N rec - Measured noise at receiver output k - Boltzmann s constant 1.38 x10-23 B - Bandwidth (Hz) G - Receiver Gain The measured noise power rises as the rain fade increases. A satellite beacon receiver directly measures the rain attenuation and a radiometer only provides an indirect measurement. A satellite beacon receiver is thus more desirable for gathering rain fade statistics. The accuracy of the Radiometer derived attenuation becomes subject to large errors [10] for attenuations greater than 15 db, thus limiting the fade depth that can be measured. A radiometer however does not require a beacon signal and can thus be used prior to the introduction of satellite beacons. As a consequence there have been many radiometer experiments carried out and comparatively few satellite beacon measurements, this combined beacon receiver and radiometer in one instrument is a valuable tool for developing models for correlating the radiometer and satellite beacon measurements at the same frequency and along the same transmission path. Path and Frequency coincident measurements have not been possible, as the beacon signal would normally interfere with the radiometer. This receiver is however able separate the beacon signal and the received noise, by using an FFT. FREQUENCY CONTROL In order to measure the satellite beacon signal, it must be firstly be located such that the down converted signal is at DC, so that the decimating filter does not remove the wanted signal. Fig. 1 shows the how the DSP is used to control the

4 VHF LO, the digital down conversion and the decimation filter. Two forms of frequency control are used. A coarse frequency control changes the divider ratios of the PLL used for the VHF LO. This allows the LO to be swept over a 2 MHz frequency range around MHz. For this coarse frequency sweep, the decimation ratio is set to 128, resulting in a 78 khz bandwidth, and the Numerically Controlled Oscillator (NCO) of the DDS, which is shown in Fig. 1 to generate the 5.5 MHz Sin and Cos signals, is set to exactly 5.5 MHz. The maximum amplitude of any signal in that bandwidth is determined by the DSP. The VHF LO frequency is then changed by 50 khz, and the maximum signal is recorded again. This process is repeated until the whole 2 MHz frequency range around the expected beacon frequency is covered. The division ratios are then set such that the largest signal in the scanned frequency band is set to within ±25 khz of the centre of the 5.5 MHz IF filter. The fine frequency control adjusts the frequency of the NCO in the DDS to ensure that the beacon frequency is within centre third of the final 5 khz decimation filter bandwidth. To prevent the receiver locking onto a spurious signal, this is done in two steps. The first step consists of an NCO sweep. The decimation ratio is set to 512 resulting in a 20 khz bandwidth. The NCO is then swept in ten 20 khz steps, covering the whole bandwidth of the 5.5 MHz IF filter. The peak signals from each of these measurements are determined and the NCO is then adjusted to make the beacon signal frequency 0 Hz at the input to the decimating filters. This NCO sweep ensures that the receiver does not lock onto a sideband of a modulated beacon. For the second step, the decimation ratio is set to 2048 resulting in a 4.9 khz bandwidth. The frequency that was set as the result of the NCO sweep is now fine-tuned to place the beacon at exactly 0 Hz at the input to the decimating filters. This frequency control works well and a satellite beacon can be located anywhere inside a 2 MHz bandwidth and the receiver can lock onto this signal within two seconds after being switched on. If the DSP encounters a modulated beacon and a CW beacon inside the 2 MHz bandwidth, then the receiver locks onto the CW beacon, thus preventing the receiver from locking onto the telemetry beacons that are located 750 MHz away from the UPC beacon on the Optus satellites. During normal operation, the NCO frequency is adjusted after each FFT, to ensure that the beacon signal is always at 0Hz during decimation. When the beacon drifts out of the centre third of the 5.5 MHz IF filter, the division ratios of the VHF LO PLL are adjusted to move the beacon to the centre of the 5.5 MHz IF filter after which an NCO sweep is done to ensure that the beacon is again at 0 Hz at the input to the decimating filters after this adjustment. The beacon is thus tracked, even if the Ka or Ku band LNC drifts with temperature. The LO of a PLL Ku band LNB typically varies less than ±25 KHz over the operating temperature range, however the LO of a Dielectric Resonator Oscillator (DRO) Ka band LNB can drift ±1 MHz with temperature. These Ka band LNB s need to be temperature controlled to ensure that the change of LO with temperature is less than the 2 MHz VHF LO frequency sweep. If the receiver looses lock, such as can occur during a very deep fade, an NCO sweep is initiated after a few seconds, to ensure that the beacon will be found once it emerges above the noise. If after one minute the beacon is still not found, then a VHF LO sweep is initiated. Under normal conditions, the receiver initiates a NCO sweep every ten minutes if no fade is in progress. This ensures that the beacon cannot remain locked onto a spurious interference signal. Fig. 3a STRAP 2 Received Beacon, (DC LO control). Fig. 3b STRAP 3 Receiver Beacon, (PLL LO control). The UPC Ku band beacon on typical satellite has 95% of the energy in a 100Hz bandwidth. The phase noise produced by each of the local oscillators in the receiver will spread this energy over a wider band, so that the performance of the beacon receiver is critically dependent phase noise of these oscillators. A PLL LNB used to convert a Ku band beacon signal at GHz for the Optus satellite has a phase noise of 75 dbc at 1kHz away from the carrier, and does not

5 cause a significant degradation to the total noise performance. The phase noise of the Xtal multiplied UHF oscillator is much lower than that. For the first version of our DSP based beacon receiver STRAP2 [4], the VHF LO was not phase locked and was DC voltage controlled by the DSP. Fig. 3a shows the spectrum of the Optus satellite beacon, received in Townsville using this receiver. An analysis of the data file for Fig. 3a shows that a 380 Hz measuring bandwidth covers more than 95% of the beacon signal power. To ensure that all the signal power is measured the power contained in 41 FFT bins around the peak value is added as the signal power. This is a 380 Hz bandwidth is significantly wider than that of the satellite and is the reason for phase locking the VHF LO in the STRAP3 receiver [10]. The same analysis of the spectrum for the STRAP3 receiver, shown in Fig. 3b shows that 95% of the energy is contained within an 80 Hz bandwidth, so that only 11 FFT bins need to be added for the signal power. Since the noise level is measured using the same number of FFT bins, a 6 db reduction in noise, corresponding to a 6 db increase in dynamic range is obtained. Further improvements in the dynamic range can be obtained by further improvements of the phase noise of the VHF LO. Fref Div M Phase Detect Loop Filter VCO Div N Output Fig. 4 PLL Block diagram. The block diagram of a PLL is shown in Fig. 4. The STRAP3 beacon receiver [10], uses a PLL VHF LO where M is fixed to produce a 25 khz phase-detector input and N is varied to obtain the required output frequency with 25 khz increments. Comparing Fig. 3a and Fig. 3b, it can be seen that the noise performance is significantly improved. There are however still some 50 Hz sidebands visible, which cause errors in the determination of both the beacon signal power and the radiometer noise. The output frequency of a PLL is: F = out N M F ref The PLL is control system and it s behaviour is determined by the natural frequency of the characteristic function. The natural frequency is set by the loop filter components, as well as the phase detector and VCO characteristics. As a result the term loop-filter bandwidth is often used instead of natural frequency of the characteristic function. The natural frequency must be much less than one tenth of the phase detector input frequency for the system to be stable. The frequency control system must be able to shift the VHF VCO in smaller than 25 khz steps, so that the natural frequency must be much less than 2.5 khz for a PLL where M is fixed. The phase noise of a PLL below the natural frequency is primarily determined by the reference oscillator, F ref. The phase noise above the natural frequency is primarily determined by the phase noise of the VCO. Since the reference oscillator is normally a crystal oscillator, the close-in phase noise is lower than the phase noise further away from the carrier, where the phase noise is due to the VCO alone. This effect can be seen in Fig. 3b, where there is a dip in the spectrum very close to the carrier. The only way to decrease the 50 Hz sidebands present in Fig 2b is to increase the natural frequency, such that it is much greater than all the 50 Hz harmonics. To achieve this a natural frequency greater than 5 khz is required, yet we still need to increment the LO in smaller than 25 khz steps. In the STRAP4 receiver, this is achieved by varying both N and M, such that the frequency increment of the LO is between 18 khz and 28 khz and the input to the phase detector is between 100 khz and 300 khz. The required 190 M and N values for a 2 MHz frequency sweep are stored in a lookup table in the DSP. Fig. 5 shows the resulting spectra. If the loop-filter bandwidth is not quire wide enough, then the phase noise will rise away from the carrier, as is shown in Fig. 5a. Increasing the loop-filter bandwidth further as shown in Fig. 5b, results in a flat noise spectrum, which can then be used for Radiometer measurements. The beacon signal for the STRAP4 receiver is now contained inside 5 FFT bins, resulting in a 9 db increase in dynamic range compared with the STRAP2 receiver of Fig. 3a. In addition the noise spectrum away from the carrier is much more suitable for radiometer measurements.

6 To ensure that the noise floor is due to sky noise and not due to hardware noise, a signal generator with the same amplitude and the same 1.45 GHz frequency as the beacon signal was applied to the input of the indoor unit. The measured noise floor was 20 db lower than the noise floor for the satellite beacon, showing that the receiver can be used as an accurate Radiometer. The spectra in Fig. 3 and Fig. 5 contain 512 bins where the signal or noise is present without attenuation. In the STRAP2 receiver 41 bins were used for the signal determination and a total of 82 bins sufficiently far away from the beacon were used for the noise calculation. In the STRAP4 receiver the 471 bins, being all the 512 useable bins minus 41 bins centred on the beacon signal are used for the noise calculation. Using more FFT bins for the noise calculation reduces the standard deviation of the measured noise. Fig 5a. PLL LO with wide loop-filter bandwidth. Fig 5b. PLL LO with wider loop-filter bandwidth (STRAP4). DIGITAL FILTERING Video filtering can be applied to the spectrum in order to decrease the variations in signal and noise at successive samples and thereby improving the precision of the measured data. Increasing the effective data rate, allows more filtering to be applied. Time Time data from DDS after decimation Fig. 6 Sliding FFT to increase the data rate. This can be achieved by using a sliding time window where a new is done for every 256 new data samples, so that 4 FFT s are applied to every data sample, as shown in Fig. 6. The data is however sufficiently independent to enable the noise to be filtered, without compromising the accuracy of either the signal or the noise. The DSP has more than enough capability to handle this 4-times increase in the number of FFT s required to be done. This filtering reduces the maximum noise by a factor of 3 db, resulting in a 3 db increase in lock-in range. For the 20 MSPS ADC sampling rate, decimation by 2048 and using a 1K FFT with a sliding time window, 38 FFT updates are produced per second. The rain attenuation data can have variations at a 1 Hz rate that need to be measured. As a result a 1.5 Hz first order video filter can be applied to all the bins in the spectrum, without loss of data. These filtered FFT bins are then used to determine the beacon signal power and the noise power. From actual beacon data measured using the STRAP4 receiver, it was determined that by simply adding the power of 5 FFT bins around the peak value, the signal power is obtained with an accuracy better than 0.08 db. It should be noted that if a Ka band DRO LNB is used, the number of bins will need to be increased to allow for the increased phase noise of that LNB, however that is the only change that needs to be made to the DSP code.

7 Since 4 FFT s are performed on every sample, 38 Signal and Noise power measurements are obtained each second. To further reduce the random variations in the signal and noise data, a 6 th order Butterworth Digital filter with a cut off of 0.5 Hz is applied to the noise values. Since the beacon signal does not have as much statistical variation, a simple first order filter with 0.5 Hz is applied to the signal values. This results in a 5-sample (0.13 second) delay between the signal and the noise outputs, which can be ignored in most applications. Finally, the data rate is reduced by decimation by 4 to have the 9.5 samples per second data rate of the STRAP2 beacon receiver. By comparing Fig 7a and 7b, the improvements due to the digital filtering can clearly be seen. The Radiometer output can be used without any additional digital filtering being required during post logger processing. Fig. 7b also shows that there is a good agreement between the measured beacon signal attenuation and the attenuation calculated from the radiometer. For deeper fades the temperature used in the radiometer equation (1) needs to be adjusted [10] to partially track atmospheric temperature changes that occur during a heavy downpour. Radiometer power Meas. Signal Noise Signal and Noise power (db) Samples (approx per second) Fig. 7a STRAP2 Signal and Noise data Fig. 7b STRAP4 Signal and Noise data for a 10 db fade. By producing a signal from a signal generator to be the same as the clear sky satellite beacon and adding noise to be equivalent to the sky noise, the beacon receiver signal accuracy and the lock in range can be determined by reducing the signal generator output until the receiver looses lock. These measurements show that the beacon amplitude determined by the receiver is well within the 1 db accuracy of the attenuators used. For the STRAP2 receiver the clear sky signal to noise ratio with the Optus B3 satellite received at Townsville, Queensland, Australia, was 31 db and the lock in range was 36 db. The STRAP4 beacon receiver has a 45 db SNR and a 50 db lock in range, a significant improvement over the STRAP2 receiver. Since the noise power is referenced to the power in 5 FFT bins and the receiver can detect and stay locked to a signal when the signal in one bin is 2 db larger than the adjacent bins, the receiver can remain locked when the beacon signal power is less than the displayed noise power. Since the clear sky beacon signal at the antenna is close to 120 dbm, and the receiver has a lock in range of 50 db, the STRAP4 receiver can acquire and remain locked onto beacon signals of close to -180 dbm. DATA LOGGER A data logger was developed using LabView to record the data from the beacon receiver and interface with the temperature sensors and rain-gauge. The data logger has a graphical display showing the recorded signal and noise over a 12-hour period. Fig. 8 is a photograph of the screen of the logging computer, during a day when there was no rain but some clouds. The Optus B3 satellite has a diurnal signal strength variation as it moves around in it s orbit, resulting in a 3 db variation in signal strength. The noise power varies 1.5 db due to temperature variations. By taking long-term differences between the signal and the noise, this temperature variation can be compensated for. It can be seen that the noise variations during a 10 minute interval are kept within 0.3 db, making the radiometer noise variations comparable to conventional radiometers. The beacon signal to noise ratio is shown as 44 db during the peak signal period. In addition the beacon signal can be increased by 3 db by using circular rather than linear polarisation for the feed structure of our antenna.

8 Fig. 7. Photograph of Data Logger screen, showing a 12 hour data plot. 6. CONCLUSION The JCU-STRAP4 satellite beacon receiver uses DSP technology: To control the frequency of a beacon receiver, to accurately measure of the attenuation of the satellite beacon due to rain and to measure the transmission medium noise temperature at the same time. The receiver has a typical lock in range of 50 db This was not previously possible. The use of digital technology results in a larger dynamic range than is obtained by conventional beacon receivers. The receiver acquires the satellite beacon signals, at the antenna, with levels as low as -180 dbm in close to one second. 7. ACKNOWLEDGEMENT The author acknowledges the work done by Brian Bowthorpe in writing the initial DSP code for the beacon receiver and Roy Snijders for implementing the digital filtering and the updated PLL control of the VHF LO into that DSP code. 8. REFERENCES [1] Report Attenuation by gases CCIR 15th Plenary Assembly, Geneva, Vol5, [2] Allnutt, J. E., Satellite-to-Ground Radiowave Propagation, Peter Peregrinus Ltd., 1989, ISBN [3] R. Acosta, S. Johnson, W. Feliciano, R. Pollard, L. A. Gonzalez, NASA s New Radio Wave Propagation Experiment, 8 th Ka Band Utilisation conference, Baverno Italy, September [4] Kikkert C. J. The Design of a 12 GHz Narrowband Low Noise Receiver, 1992 Asia-Pacific Microwave Conference, Adelaide, pp [5] Sri Sai Communications Satellite Beacon Tracking Receiver [6] Novela Satcoms, B350 Series Satellite Beacon Receivers [7] Kikkert C. J., Bowthorpe B. and Allen G. Satellite Beacon Receiver Improvement Using Digital Signal Processing The Fourth International Symposium on Signal Processing and its Applications (ISSPA96), Gold Coast, August 1996, pp [8] Kikkert C. J., Bowthorpe B. and Ong Jin Teong, A DSP Based Satellite Beacon Receiver and Radiometer, 1998 Asia Pacific Microwave Conference, APMC98, Yokohama, Japan, 8-11 December 1998, pp [9] Bowthorpe B. Microwave Propagation Impairements in Tropical Rain, PhD Thesis, James Cook University, [10] C. J. Kikkert, B. J. Bowthorpe and J. T. Ong, Improvements To A DSP Based Satellite Beacon Receiver And Radiometer, ICICS99, Second International Conference on Information, Communications & Signal Processing, December 7-10, 1999, Singapore.