I
p
ðt
1
Þ¼adðt
1
Þsin cðpdfTÞcos 2pdft
1
þ
T
2
þ du
0
Q
p
ðt
1
Þ¼adðt
1
Þsin cðpdfTÞsin 2pdft
1
þ
T
2
þ du
0
ð4Þ
where t
1
is the start of the integration time and T is the total
integration time.
Figure 1 depicts the scheme of a general FLL. The
code and carrier correlators wipe off the PRN code and
the Doppler frequency to get i
p
ðtÞ and q
p
ðtÞ. The inte-
gration and dump module integrates the input signals,
i
p
ðtÞ and q
p
ðtÞ, to get the coherent integration results,
I
p
ðtÞ and Q
p
ðtÞ, respectively. The frequency discriminator
estimates the Doppler frequency residual, d
^
f , from a
series of I
p
ðtÞ and Q
p
ðtÞ. A low-pass loop filter is fol-
lowed to eliminate the high frequency noise in the
estimation. The filtered estimation is then used to control
the NCO, which generates the orthogonal local carrier
signals, i
LO
ðtÞ and q
LO
ðtÞ. This design has a limitation
when using a phase-differential frequency discriminator.
The limitation is that non-coherent integration cannot be
applied because it does not reserve phase information.
Hence, it is not possible to increase the sensitivity
through non-coherent integration.
FFT-based frequency discriminator
The complex form of (4) is:
r
p
ðtÞ¼I
p
ðtÞþjQ
p
ðtÞ
¼ adðtÞsin cðpdfTÞexp j 2pdftþ
T
2
þ du
0
ð5Þ
where adðtÞsincðpdfTÞ is the amplitude of the complex
r
p
ðtÞ which will attenuate with the increase in dfT.Ina
stable tracking state, the frequency residual dfT is rela-
tively small, so the amplitude is approximately adðtÞ.
Hence, r
p
ðtÞ can be expressed as:
r
p
ðtÞ¼I
p
ðtÞþjQ
p
ðtÞ
¼ adðtÞexp j 2pdftþ
T
2
þ du
0
ð6Þ
The phase of expðjð2pdf ðt þ
T
2
Þþdu
0
ÞÞ will rotate with
t, where df determines the rotation speed. The navigation
data bit dðtÞ can be {1, -1} arbitrarily. This will cause a
180° phase reversal in r
p
ðtÞ if the integration is done over
data bits transitions. An assisted GNSS (AGNSS) receiver
can overcome this problem by using aiding information to
predict the data bit values provided it has accurate timing
to within 0.5 ms, eliminate their effect, and overcome the
data bit length limitation (Djuknic and Richton 2001). The
integration time limitation in the GPS L1 C/A signal is
20 ms if the locations of the data bits transitions are
known. However, it is difficult to determine the data bit
transitions in a stand-alone GNSS receiver. Therefore, a
complex squaring operation can be applied to circumvent
the effects of data bits, without the need for determining
the data bits transitions or values. In this operation, the
noise is squared, and so it is not averaged out by summa-
tion. This nonlinear process induces the so-called squaring
loss (van Diggelen 2001). The complex squaring result of
(6) is as follows:
S
p
ðtÞ¼ðI
p
ðtÞþjQ
p
ðtÞÞ
2
¼ I
2
p
ðtÞQ
2
p
ðtÞþj2I
p
ðtÞQ
p
ðtÞ
¼ a
2
exp j22pdftþ
T
2
þ d/
0
ð7Þ
The effect of navigation data bits on S
p
ðtÞ does not exist,
and the frequency and phase of S
p
ðtÞ are doubled. To track
the frequency, the frequency discriminator should estimate
the Doppler frequency residual, d
^
f , from the series S
p
ðtÞ
and then use it to update the NCO.
The proposed frequency discriminator uses the FFT
method to estimate the frequency. Equation (7) shows that
S
p
ðtÞ is a single tone complex signal, where the residual
frequency is its frequency parameter. The frequency of a
single tone complex signal is estimated using FFT. The
adopted FFT approach can be described by:
Carrier
Correlators
Frequency
Discriminator
Code
Correlators
Loop
Filter
NCO
,
()
IF I
t
,
()
IF Q
st
()
LO
it
()
LO
qt
Integration
Dump
()
p
it
()
p
qt
()
P
It
()
P
Qt
ˆ
f
δ
Fig. 1 Scheme of general FLL
GPS Solut (2016) 20:225–237 227
123