S. Brand, Philips Semiconductors, PCALE QAM Demodulation 0 QAM Demodulation
o Application area o What is QAM? o What are QAM Demodulation Functions? o General block diagram of QAM demodulator o Explanation of the main function (Nyquist shaping, Clock & Carrier Recovery, AGC, Adaptive Equaliser) o Performance o Conclusion
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 1
Example Application Area
“Wireless Cable” Digital TV using Microwave Transmission
QAM Modulation Set-top Box
Multiplexing Radio Channel Compression
· Compression = bit rate reduction · Multiplexing = assembly of multiple programs · Modulation = conversion to transmission format · Set-top Box = Integrated Receiver Decoder (IRD), provides a subscriber access to a wide range of programs
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 2
What is QAM? o Amplitude Modulation of 2E 2E o Two Orthogonal Carriers o o xi (t) = ------ai cos(wct) + ------bi sin (wct) Ts Ts
a =+3 a =+1 a =-7 +7 i-1 i i+1 Q +5 +3 7 +1 -1 I 5 -3 -5 Ts 3 -7 ÖEo bi-1=-5 bi=+5 bi+1=-1 1 I +7 ÖEo +5 Tc -1 +3 +1 -1 Q -3 -3 -5 -5 -7 -7 time -7 -5 -3 -1 1 3 5 7 64QAM in time domain 64QAM Constellation diagram
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 3
M-ary QAM { { I Q
Satellite Cable b5b4b3b2b1b0 b1b0 Noise power
signal power
S/N > 3 dB for M=4 S/N > 21 dB for M=64 S/N > 27 dB for M=256
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 4
What to do to recover the information?
Functions Result Automatic Gain Control Optimal position of constellation diagram in reception window Quadrature I & Q base band signals down conversion (Half) Nyquist Filtering Pulse shaping Clock Recovery Sampling reference for A/D Converter Carrier Recovery Carrier frequency reference Adaptive Equaliser Compensate for channel distortion Demapping Representation of received data in bits
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 5
System Block Diagram
2 QAM DEMODULATOR IF fs I C fine AGC
f f f I
ÖN g n i
1,0,-1,0 Complex p
Tuner BPF LPF ADC p Equaliser a 0,-1,0,1 m
Q e 4fs ÖN d n
o VCO VCXO i t c
e AGC clock loop carrier n DTO
n detect detect filter detect o
y Digital C r
y e r e l e v Analogue b v o a c o DAC DAC DAC e c C e R
R r
e k C i r c G r o a l A C C
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 6
Automatic Gain Control * 2 loops AGC n Filtering & IF down ADC I conversion Equalisation Q * Coarse AGC to prevent ADC from overloading coarse fine AGC AGC * After Nyquist filtering and Equalisation ‘small’ QAM remains. * Fine AGC to position contella- tion diagram to decision window Q Q Q
I I I
Tuner output Equaliser output Fine AGC output
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 7
(Half) Nyquist Filtering * Pulse Shaping required to realise ÖN I ISI=0 in limited BW 1,0,-1,0 ADC * ISI=0 when zero crossings occur at 0,-1,0,1 multiples of Ts=1/fs 4fs ÖN Q * Achieved with Nyquist Criterion (DVB: a = 15%) Ts=1/fs S Sn Sn+3 n+6 * Cascade of Transmitter & Receiver Sn+1 Ts BW=¥ fulfil Nyquist Criterion (Half Nyquist each ) time 0 freq S Sn+5 n+2 Sn+4 (1+a)fs * Digital implementation (T = 9 T ) fs delay symbol
BW=8MHz * This delay is in the loops and thus influences the demodulator archi- time 0 freq tecture
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 8
Clock Recovery
nd 1,0,-1,0 * Recovery with 2 order PLL
ADC ÖN * Clock Detector
0,-1,0,1 - Energy Maximization algorithm - After Half Nyquist Filter to achieve clock vcxo DAC det. ISI=0 at detector input
* Half Nyquist Filter in loop is allowed - Received clock has crystal accuracy (100 ppm at 7 Msym/s)) - Loop BW may be small I, Q signal - Delay in loop is allowed (no instability) time * Quadarture Demodulation Ts - fclock = 4 fsymbol - Simple with j-n (n=0,1,2,3,...) Recovered clock
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 9
Carrier Recovery nd delay * Recovery with 2 order PLL
IF * Carrier Detector LPF ADC ÖN equaliser - Decision directed
4fs - After equaliser - PD (lock) and PFD (unlock) vco vcxo * PFD for large acquisition range (100 kHz) * PD for stable behaviour once in lock carrier DAC det. * Half Nyquist & Equaliser in loop - Large delay causes problems for distur- bances like: * phase noise * microphonics (mechanical vibrations) I or Q time Tcarrier * Alternative solution required Recovered carrier
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 10
Carrier Phase Disturbances (1) Additive White * AWGN Disturbance Gaussian Noise - Random distribution - Mainly inserted in the cable channel
* Result - Enlarged constellation points
* PLL Properties - Average the noise - Loop BW small Implementation Loss - Low IL Cable Tuner QAM demod
s(t) + r(t)
n(t) BW
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 11
Carrier Phase Disturbances (2) Phase Noise/ * Phase Noise & Microphonics -No random distribution Microphonics Q -Mainly inserted in the tuner by LC oscillators which are sen- sative for mechanical vibra- tions (Microphonics) * Result -Rotation of constellation dia- I gram. * PLL Properties Implementation Loss -Follow the phase disturbance -Loop BW large Cable Tuner QAM -Low IL demod * PLL properties for AWGN and s(t) X r(t) BW phase noise are in contradic- tion
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 12
Phase noise versus AWGN * Loop BW trade of between: a. Ability to follow phase noise ]
B b. Ability to average AWGN d [
s s * Rule of thumb: o L 1 n
o BW = ------f i 1000
t symbol a
t AWGN n
e AWGN+Phase noise * Simulations show this is m e
l approximately correct p m I * Optimum depends on S/N and amount of phase noise
* Problem: Optimum loop BW instable due to large delay in the loop (Half Nyquist + Equal- iser). OPTIMUM Loop BW [kHz]
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 13
Double Loop Carrier Recovery * Introduction of second loop with large delay (relatively) small delay IF carrier LPF ADC equaliser ÖN det. * Outer loop
4fs -Adjust (static) frequency offset inner loop vco vcxo dto Loop -Small loop BW due to large delay filter -PD/PFD
outer loop DAC * Inner Loop - Optimum loop BW as trade off between phase noise & AWGN - Large Loop BW due to small delay - PD only I or Q time * Conclusion: optimum loop BW can Tcarrier be selected and causes no instability Recovered carrier
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 14
Equalisation
* Nyquist Criterion specifies a frequency domain condition on the received pulses to achieve ISI=0
* Generally this is NOT satisfied unless the channel is equalised
* Equalise the channel = compensate for channel distortion
* Unfortunately, any equalisation enhances noise from the channel
* Tradeoff between: Accurately minimising ISI
Minimising the noise
* Different types of Equaliser
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 15
Multipath Distortion Multipath * Multipath distortion causes ISI Reflection * Each original point consists of M new points in the shape of constella- tion diagram F * Amplitude, delay and phase of the echo determine shape/size of the small constellation diagrams A * Varying channel requires Adaptive Equaliser
Cable Tuner QAM demod
s(t) + r(t)
Amplitude, delay, phase
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 16
Equaliser Structure
Linear Equaliser (LE) Decision Feedback Equaliser (DFE)
Symbol Symbol Decision Decision
Iout Qout
Coefficients Coefficients Coefficients
Iin Complex Iout Iin FFE + DFE FIR filter Qin Qout Qin
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 17
Equaliser Structure
Linear Equaliser (LE) Decision Feedback Equaliser (DFE)
H(z) G(z) H(z) G(z) in out in out + + in out in out + +
- A Z-t -A Z-t A Z-t Z-t A 2 (-) Residual ISI (A ,2t) (+) No residual ISI –t H (z) = 1 + Az 1 G(z) = ------–t –t G(z) = 1 – Az 1 + Az
2 –2t H(z) G(z) = 1 – A z H(z) G(z) = 1 (+) ‘Fast’ acquisition (-) ‘Slow’ acquisition (-) ‘High’ noise amplification (+) ‘Low’ noise amplification zeroes poles noise noise
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 18
Equaliser Adaptation Algorithm
Zero Forcing (ZF) Mean Square Error (MSE)
(+) Complete elimination of ISI (+) Minimize sum of ISI and noise (-) Penalty = Noise amplification (+) Less noise amplification by (-) Allowing residual ISI
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 19
Adaptive Equaliser ZF (Zero Forcing) MSE (Mean Square Error) Suited for QAM with M£64 Suited for QAM with M£64 (-) residual ISI does not allow higher M (-) residual ISI does not allow higher M E
L (+) Fast acquisition (+) Fast acquisition (+) High stability (+) High stability
No suitable solution Required for QAM with M>64 (-) Because of complete elimination (+) Stablity guaranteed when of ISI system is instable when zero zero in spectrum in spectrum Equaliser E Equaliser F channel D channel (-) ‘Slow’ acquisition
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 20
Measurement Results Implementation Loss
R (2) 0.6 dB
E (3) 1.1 dB B (4) 1.6 dB (5) 1.9 dB (6) 2.0 dB
(1) Theory (2) AWGN (single car loop) 1 2 3 4 5 6 (3) AWGN (double car loop) (4) 1 ray echo (5) 2 ray echo (6) 3 ray echo
S/N
Wireless Communications S. Brand, Philips Semiconductors, PCALE QAM Demodulation 21
Conclusion
Single Chip QAM Demodulator with low Implemenation Loss - Double Loop AGC for optimum usage of A/D Converter - Delay in half Nyquist filter and equaliser require double carrier recovery loop structure to achieve high performance on phase noise & microphonics - Adaptive equaliser * LE/ZF or LE/MSE preferred for QAM withM £64 * DFE/MSE required for QAM with M>64
Wireless Communications