Sequence: Modulation!

and demodulation!
!
Episode 9 !
"Going further"!
!
I/Q demodulator!
By Marie-Laure Boucheret!
!

Professor!
at Institut National Polytechnique de Toulouse!
INP / ENSEEIHT!

Week 4 44!
In episodes 4 and 5, we saw the I/Q modulator.! 00101101! I bits! 01100100! Q bits!
Its function is to transform one!
(PSK modulation) or two (QAM modulation)!
Bit-to-symbol mapping! Bit-to-symbol mapping!
bitstreams forming a constellation into a!
continuous signal to be transmitted over the !
channel.! Shaping! Shaping!

I channel bits!
Modulated! Frequency translation!
I/Q!
Modulator! signal!
Q channel bits!

Figure 1: I/Q modulator (QAM)! Figure 3: Detailed I/Q modulator (QAM)!

I and Q channel bits! The functions implemented in the modulator are:!

I/Q! Modulated! bit-to-symbol mapping, signal shaping,!
modulator! signal! and frequency translation.!
Only the bit-to-symbol mapping function differs for!
Figure 2: I/Q modulator (PSK)! PSK modulation (only 1 bitstream at the input of!
the bit-to-symbol mapping).!
Let's move on to the I/Q demodulator,! Noisy!
introduced in episode 6. The input signal of this! signal!
demodulator is the noisy output signal of the!
channel. The output is the noisy constellation. ! Translation!
To baseband!
I/Q!
demod.!
Filtering!

Noisy! Noisy! Sampling!

signal! constellation!
Noisy!
Figure 4: I/Q demodulator! constellation!

The functions implemented in the I/Q demodulator Figure 5: Detailed I/Q demodulator!
are:!
- Translation to baseband! We will now go over the various functions.!
- Reception filtering! !
- Sampling at symbol time! Note: basic knowledge of filtering is required!
for what follows!
Let's start with the translation to baseband.!
It is the dual operation of the translation to! I’(t)!
Signal on carrier!
the carrier frequency as described below.! frequency! 2cos(2πf0t)!
I(t)!
oscillator!
Signal on carrier! s(t)! -2sin(2πf0t)!
cos(2πf0t)! frequency!
oscillator! Q’(t)!
-sin(2πf0t)! s(t)! Figure 7: translation to baseband!
yyy

Q(t)! Let's determine the signals I’(t) and Q’(t).!

s (t ) = I (t ) cos( 2πf 0t ) − Q(t ) sin( 2πf 0t )
Figure 6: translation to carrier frequency!
I ' (t ) = 2s (t ) cos( 2πf 0t )
2
f0 is the carrier frequency! I ' (t ) = 2 I (t ) cos (2πf 0t ) − 2Q(t ) cos( 2πf 0t ) sin( 2πf 0t )
2
2 cos ( x) = 1 + cos( 2 x) 2 cos( x) sin( x) = sin( 2 x)
The translation to baseband is described in figure I ' (t ) = I (t ) + {I (t ) cos( 4πf 0t ) − Q(t ) sin( 4πf 0t )}
7. In order to simplify the calculations,!
we assume noiseless transmission.!
In the same way:!
We will use a low-pass filter on each channel!
s (t ) = I (t ) cos( 2πf 0t ) − Q(t ) sin( 2πf 0t ) to eliminate signals centered on frequency 2f0.!
Q' (t ) = −2s (t ) sin( 2πf 0t ) !
2
A low-pass filter has a zero center frequency.!
Q' (t ) = −2 I (t ) cos( 2πf 0t ) sin( 2πf 0t ) + 2Q(t ) sin (2πf 0t ) Its frequency response is chosen so that signals!
2
2 sin ( x) = 1 − cos( 2 x) 2 cos( x) sin( x) = sin( 2 x) I(t) and Q(t) are not distorted, and that signals!
centered on 2 f0 are cut off.!
Q' (t ) = Q(t ) − {I (t ) sin( 4πf 0t ) + Q(t ) cos( 4πf 0t )}
!
We therefore obtain the following I’(t) and Q’(t) ! !
signals:! !
I ' (t ) = I (t ) + {I (t ) cos( 4πf 0t ) − Q(t ) sin( 4πf 0t )} !

Q' (t ) = Q(t ) − {I (t ) sin( 4πf 0t ) + Q(t ) cos( 4πf 0t )} f!

2f0!
Figure 8!
I(t) and Q(t) are baseband signals, their center! In figure 8, the frequency response of a possible!
frequency is therefore zero.!
The terms in parentheses are signals on carrier!
low-pass filter is shown in red. !
frequency, centered on frequency 2f0. !
Note that baseband signals (here I(t) and Q(t)) ! LPF! I(t)!
are in the passband of the filter and are therefore n
ot distorted. ! 2cos(2πf0t)!
Signals centered on 2 f0 are in the stopband of the oscillator!
filter (i.e. the frequency response of the filter is ! s(t)!
zero at these frequencies). Therefore, only the ! -2sin(2πf0t)!
baseband signals come out of the filters.! LPF! Q(t)!
!
The diagram of the translation to baseband! Figure 9: Translation to baseband !
including the low-pass filter is given in figure 9.!
In this diagram, the low-pass filters are !
represented by the function "LPF" (low-pass filter).! We have shown that we can recover the!
in-phase I(t) and quadrature Q(t)!
components of the signal from the signal s(t)!
on carrier frequency.!
Let's focus now on reception filtering. ! In the diagram from figure 10, we replaced!
! the low-pass filters by the matched filters !
We saw that, for each channel, a low-pass filter is! MF. The outputs of the matched filters !
required to eliminate the component of the signal! are denoted by IMF(t) and QMF(t).!
centered on 2f0.!
!
It is shown that a second low-pass filter is!
required to reduce the bit error rate (BER).!
This is the filter that gives the maximum !
MF! IMF(t)!
signal-to-noise ratio (SNR) at the device output.!
It is called matched filter. ! 2cos(2πf0t)!
! oscillator!
For satellite communications, where only direct !
path between the transmitter and the receiver is ! -2sin(2πf0t)!
considered, the matched filter is the same as that ! MF! QMF(t)!
used in the transmitter (shaping). !
It is shown, by using the properties of the filters,!
that the low-pass filter from figure 8 may be ! Figure 10!
removed, since the matched filter also eliminates!
signals at frequency 2f0.!
!
Example of matched filter output signals! Example of matched filter output signals!
(NRZ filtering): case without noise! (NRZ filtering): case with noise!
Bits transmitted: 1 0 1 0 0 1 1 0 1 0 (channel I) !
Figure 13:!
High SNR: slightly!
Figure 11:!
disturbed signal!
Shaping output!
(transmission)!

Figure 14:!
Low SNR: strongly!
Figure 12:! disturbed signal!
Matched filter!
output (reception)!
We now have to implement the sampling function !
from figure 5.! Figure 15:!
! Signal without !
The signal is sampled at frequency Rs! noise!
(symbol rate), i.e. at one sample per symbol.!
!
Consider the outputs of matched filters from!
figures 12 and 14 (I channel). The samples !
obtained are shown in blue in figures 15 and 16.!
!
Note that, without noise, the values of the !
samples are +1 and -1.! Figure 16:!
With noise (figure 16), the signs of the samples! Noisy signal!
are the same, which means that the bits will be!
correctly estimated.!
!
With even more noise,we see samples leading to
erroneous decisions (samples circled in green in !
figure 17).!
!
 In the same way, the output of the matched filter!
of the Q channel is sampled at the symbol rate.!
!
Figure 15! We thus obtain a noisy constellation at the output !
of the I/Q demodulator.!
!
The full I/Q demodulator is shown in figure 18!
!

MF! Sampler! noisy I!

Figure 17! 2cos(2πf0t)!
oscillator!
s(t)! -2sin(2πf0t)!
MF! Sampler! noisy Q!

Figure 18: I/Q demodulator !