Professional Documents
Culture Documents
Cornelis J. Kikkert, Craig Blackburn Electrical and Computer Engineering James Cook University Townsville, Qld, Australia, 4811. E-mail: Keith.Kikkert@jcu.edu.au, Craig.Blackburn@jcu.edu.au.
Abstract
This paper describes the realisation of a differential BPSK demodulator using a high speed ADC, an EPLD and an EPROM. By incorporating both I and Q data in the demodulation process, a significant improvement in performance is obtained. Computer simulation shows the Bit Error Rate (BER) performance versus Received Carrier to Noise Ratio (CNR) is virtually identical to the theoretical performance of a differential phase shift keyed (DPSK) detector. This paper also describes the realisation of the special PLL required, to recover the data clock. This PLL uses an EPLD, a DAC, a conventional loop filter and a conventional VCO.
Communication texts [3] suggest the use of differential phase detection, based on the block diagram of figure 1. However these texts do not mention the strict synchronisation requirements. If the input waveform is a carrier: X(t) = Sin(ct) The delayed signal will be: Y(t) = Sin (ct + ) Eqn. 1b where is the phase shift of the delayed waveform with respect to the present waveform. The output from the multiplier will then be:
Sin( c t ) Sin( c t + )dt = Sin 2 ( c t ) Cos( ) Sin( c t ) Cos( c t ) Sin( ) 1 1 = [1 Cos(2 c t )] Cos( ) Sin(2 c t ) Sin( ) 2 2
1. Introduction
The transmission of data from the GMS weather satellite is a BPSK signal, at a carrier frequency of 1.6871 GHz with a 660 kbit/sec data rate. Because of the >200 dB free space loss of the transmitted signal, the received CNR is only a few dB above the minimum value required to demodulate the data with acceptable error rates. To receive these signals, a low cost BPSK demodulator is required, which can accurately demodulate BPSK signals at these low carrier to noise ratios. The reception of Binary Phase Shift Keyed (BPSK) signals has traditionally been difficult. The carrier has either a 0 degree or a 180 degree phase shift depending on the data and the carrier amplitude is zero during the data transitions. As a result, a conventional Phase Locked Loop (PLL) demodulator cannot be used. Differential detection of BPSK signals is easy in theory, but difficult in practice. A Costas Loop [1] or a Phase Locked Loop (PLL) with special phase detectors[2], can be used to recover the carrier. That carrier is then used to synchronously demodulate the received signal. From theory, synchronous demodulation of BPSK data gives the best BER performance. In practice a Costas Loop will only perform accurately at a high Carrier to Noise Ratio (CNR). At a low CNR the loop will tend to loose lock. The BPSK demodulator described here remains locked under noisy conditions. The new BPSK demodulator uses digital technology, resulting in a lower production cost and simpler hardware than the Costas Loop.
Eqn. 1a
Eqn. 2
For a typical BPSK system, there are many carrier cycles per data period. For ideal operation is 0 or 180 and then the second term can be ignored, leaving a DC component plus a large ripple.
2.1 Realisation
The signal at the multiplier output in Figure 1 and shown in equation 2 must be used as the input for the data clock recovery circuitry. This large ripple causes problems, with the stability of the clock recovery hardware. Using In-Phase (I) and Quadrature (Q) components as shown in Figure 2, removes most of the ripple. The Analogue to Digital Converter (ADC) digitises the input carrier waveform of equation 1a. The digital delay line then delays this waveform by one data bit. The signals at the tapping points 1 and 16 then correspond to those that are multiplied in figure 1. The delay line is clocked
1.00 0.90
to PLL
1 2 16 17
0.80 0.70
Input
Output
ADC
0.20 0.10
with the sampling clock that is also used for the ADC. The sampling frequency is chosen such that one sampling period delay corresponds to 90 phase shift of the carrier waveform. If a sine waveform is present at tapping point 1, then a cosine waveform is present at tapping point 2. The signals at tapping points 1 and 16 will thus be InPhase (I) components and those at tapping points 2 and 17 will thus be Quadrature (Q) components. At the adder in figure 2, the following waveform will thus be present: X(t) = Sin(ct)Sin(ct+)+Cos((ct+)Cos(ct++) Eqn. 4. Where is the phase shift of the carrier over one data bit period, including the BPSK modulation and is the variation from the 90 phase shift. In practice will be zero, and will be either 0 or 180 depending on the BPSK data. The resulting waveform at the adder will thus be either: X(t) = Sin2(ct) + Cos2(ct) = 1 or X(t) =- Sin2(ct) - Cos2(ct) = -1 Eqn. 5.
0.00 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% Frequency Error wrt Data Frequency
of 660 kHz, or 165 kHz of the 18.48 MHz BPSK carrier for a 3 dB drop in output. This frequency stability requirement is quite feasible, even for down conversion from a 1.6871 GHz satellite signal.
to PLL
Decision Logic
1 2 16 17
Q I
to PLL
depending on the value of the BPSK data. Note that the ripple at 2c has been removed completely. The resulting waveforms are shown in Figure 3. By calculating the variation of and resulting from the BPSK carrier frequency and the sampling frequency being not correctly related, the tolerance of this BPSK demodulator can be plotted as shown in figure 4. Using both the I and Q tapping points, doubles the output signal and results in a minimal ripple. For an IF frequency shift of 25% of the data rate, the ripple is less than 10% and the data output is more than 70% of the ideal output voltage. For hardware realisation of the BPSK demodulator, one can thus have a frequency shift of 25%
1.2 0.8
By adding two more multipliers and an adder to the block diagram of Figure 2, the technique can be extended to demodulate QPSK signals as well, as shown in Figure 5. A phase locked loop (PLL) is required to recover the 660 kHz data clock. The 10.56 MHz sampling frequency for the ADC and delay line is 16 times the 660 kHz data clock and is produced by this same PLL.
2.2 Hardware
An 8 bit ADC is used as this has been demonstrated by computer simulation to give a satisfactory performance. The delay line is thus 17 bit long and 8 bit wide. This requires 136 flip-flops and is programmed into a Lattice ispLSI1032 EPLD. The multipliers shown in figure 2 need to be able to multiply two 8 bit numbers inside the 94.7 ns period of the 10.56 MHz sampling clock. This is achieved by using a 512 kbyte EPROM as a look-up table. The 16 EPROM data inputs are formed by the two 8 bit inputs to be multiplied. Since the multiplied I data is the same as the multiplied Q data one clock pulse later, only one EPROM is required if the EPLD is used to store and delay the multiplied value for one clock period.
I +Q
Amplitude
-0.8 -1.2
Time
Figure 3. Demodulator Waveforms.
A second ispLSI1032 EPLD is used to add the I and Q multiplier outputs and to perform the integrate and dump data detection using an adder for the integrator. The added I and Q multiplier output is the raw digital data output required for the data synchronisation PLL. This signal is turned into an analogue signal using a Digital to Analogue converter (DAC), the output of which is then filtered to provide some smoothing. This filtered raw demodulated data is then used as reference input to a PLL, to recover the 660 kHz clock and hence the 10.56 MHz sampling clock. The whole block diagram shown in figure 2 is thus realised using two low cost EPLDs, one EPROM a low cost ADC and a low cost DAC. Figure 6 shows the resulting hardware. The left board is the BPSK demodulator and the right board is the Phase Locked loop described later. Figure 7 shows the binary input data, the BPSK data and the demodulated output obtained with this hardware. The top trace is the binary data, the second trace is the differentially encoded signal, which is then modulated with a carrier. The bottom trace is the demodulated output. The system performs very well, even under noisy signal conditions.
1.E+00
Baseband Simulated Baseband Theoretical IF Simulated
1.E-01
1.E-02 BER
1.E-03
1.E-04
1.E-05
1.E-06 0 2 4 6 Eb/No 8 10 12
BPSK Data
Data out
This will occur 25% of the time on average, which is sufficient to keep the PLL phase locked. Since the analogue Phase Detector Output is produced by a DAC, no drift occurs in-between Phase Detector Output updates.
The binary data has a time between zero crossings that varies in multiples of 1.515 s. Since the zero crossings do not occur every data bit, a conventional phase detector cannot be used. One of the authors [4] has developed an analogue phase detector, which is suitable for this application. This phase detector will however only lock if the free running frequency of the PLL is very close to the 660 kHz data rate. That cannot always be guaranteed. To ensure reliable operation of the receiver system, frequency detection is required to ensure that the PLL always locks. Figure 9 shows the waveforms used in the Phase/Frequency detector developed for the Data Clock recovery. The Clock is the 660 kHz data clock, which is derived from a 84.48 MHz voltage controlled oscillator (VCO) by dividing by 128. The data signal is the raw binary data, obtained by digital to analogue conversion and filtering of the adder output of Figure 2 (to PLL). At a transition of the data, a Phase and Period Counter is set to zero. The 84.48 MHz VCO increments that counter and its output ramps up as shown. When the falling edge of the Clock occurs, the content of this counter is latched and transferred to the Phase Detector Output. This Phase Detector Output is converted to an analogue voltage, the filtered value of which controls the VCO. If more than one ones or zeros are being transmitted in succession, there will not be a second transition 1.515 s after the first one and the counter simply keeps ramping up until the maximum count of 192 is obtained, where the counter limits. If this happens, the Phase & Period Counter contents are not transferred to the Phase Detector Output. The next data transition will then reset the Phase and Period Counter back to zero. The phase detector will thus only produce a phase output when two transitions in the data are produced 1.515 s apart.
Phase Detector Output 660 kBps Data
8 Bit
3.3 Stability
This phase/ frequency detector has some interesting stability problems. If we consider the phase detector only, the gain of the analogue circuit and the corner frequency
16 Bit
Adder
15 Bit
Latch
DAC
Loop Filter
VCO
Threshold Detection
Divide by 16
Divide by 8
of the RC filter used at the output of the DAC are carefully controlled to ensure the correct damping factor and natural frequency of the PLL. Since digital hardware of 8 bit precision in some parts and 16 bit precision in other parts is used, a slightly conservative damping factor needs to be used in the design, to result in a stable system. The PLL locks easily with the phase detector only. As expected, the lock-in range is however limited.. When the frequency detector is used by itself, the resulting Frequency Locked Loop works very well and quickly adjusts the frequency of the VCO to within 660kHz of the required frequency, when the frequency control stops providing further frequency correction. This is a coarse frequency control. Using nonlinear techniques, which are beyond the scope of this paper, the phase detector provides a fine frequency correction,which ensures that the PLL acquires lock. The EPLDs can simply be reporgrammed on-board to change the phase and frequency detector to different configurations during development. The use of EPLDs is thus very convenient for the development of new systems like this. When the phase detector and frequency detector are combined, there can be an interaction between them, which together with the finite precision used in the digital hardware, can cause instability. This instability manifests itself as small amplitude limit cycles. These limit cycles can be removed using careful design of the phase/ frequency detector parameters and careful selection of the loop filter parameters. In order to be able to select the correct phase/frequency detector parameters and the correct loop filter, the PLL including all its nonlinear and quantisation effects needs to be simulated using MATLAB or an Excel spreadsheet. Figure 11 shows the transient phase error response obtained from this Excel simulation of the Phase/Frequency detector. The transient includes a 5 MHz initial frequency error, to permit the operation of both the frequency and phase detector to be demonstrated and to show that the PLL will lock regardless of the initial conditions. Three operating regions can clearly be seen. The first part up to 0.35 ms consists of region where the frequency error is large and
180 150 120 90 Phase Error (degrees) 60 30 0 -30 -60 -90 -120 -150 -180 0.0 0.1 0.2 0.3 0.4 0.5 Time (ms) 0.6 0.7 0.8 0.9 1.0
frequency corrections are made every 1.515 s measurement interval shown in Figure 9. The second region is from 0.35 ms to 0.88 ms where the phase detector is controlling the frequency counter to ensure that the PLL acquires lock. During this time the operate at the exact frequency. The third region from 0.88 ms onwards shows the PLL in lock. For a 5 MHz initial frequency error, the PLL locks within 1 ms. Figure 12 shows the eye diagram and the recovered data clock. It can be seen that the PLL locks accurately. The recovered data clock is used to drive the timing for the integrate and dump filter in the BPSK phase detector, thus ensuring that the transmitted data is recovered with the maximum likelihood and the minimum error.
4. Conclusion
A BPSK detector and its associated Data Clock recovery circuitry is realised using low cost digital hardware. The BPSK detector performs well and locks reliably to the data, even in high noise conditions. A novel type of Phase/ Frequency detector is presented, which ensures a reliable clock recovery of random binary data using predominantly digital hardware. This recovered data clock permits the recovery of the received data using an integrate and dump filter, thus ensuring the optimum detection of the BPSK data.
5. References
[1] Costas, J. P. Synchronous communications proceedings of the IRE, Vol 47 pp 2058-2068, 1959. [2] Piper B. and Kikkert C. J. The hardware design for the reception of GMS SVISSR Weather Satellite Signals. Workshop on Applications of Radio Science, Canberra, June 1995. [3] Roden, M. R. Digital Communication System Design, Book, Prentice Hall, 1988. [4] Kikkert C. J. Appendix 1 of D. A. Pucknell, Fundamentals of Digital Logic Design: with VLSI Circuit Applications, Book, Prentice Hall, 1990.