A 50-300-MHz low power and high linear active RF tracking filter for digital TV tuner ICs

Abstract

A low power and highly linear CMOS active tracking bandpass filter is presented to overcome a local oscillator harmonic mixing problem for Digital TV tuner ICs. A transconductor linearization technique based on a method of dynamic source degenerated differential pair is adopted to improve the linearity performance. The newly proposed low power high quality factor (Q) biquad and the linearized transconductor with negative resistance load (NRL) enables a low power and high Q RF tracking filter design. The total chip area is 0.25 mm X 0.9 mm. The fabricated tracking filter based on the 0.13 um CMOS process shows 48~300 MHz tracking range with 10~50 MHz bandwidth, more than 38 dB 3rd order harmonic rejection, 6 dB unwanted signal rejection@N+6 channel offset, and a maximum IIP3 of 6 dBm at 5 dB gain while drawing 6.4 mA from a 1.2 V supply.

Full text

Page 1

A 50-300-MHz Low power and High Linear Active
RF Tracking Filter for Digital TV Tuner ICs

Y. Sun, C. J. Jeong, I. Y. Lee, J. S. Lee, S. G. Lee

Korea Advanced Institute of Science and Technology

Abstract-A low power and highly linear CMOS active tracking
bandpass filter is presented to overcome a local oscillator
harmonic mixing problem for Digital TV tuner ICs. A
transconductor linearization technique based on a method of
dynamic source degenerated differential pair is adopted to
improve the linearity performance. The newly proposed low
power high quality factor (Q) biquad and the linearized
transconductor with negative resistance load (NRL) enables a low
power and high Q RF tracking filter design. The total chip area is
0.25 mm X 0.9 mm. The fabricated tracking filter based on the
0.13 um CMOS process shows 48~300 MHz tracking range with
10~50 MHz bandwidth, more than 38 dB 3rd order harmonic
rejection, 6 dB unwanted signal rejection@N+6 channel offset,
and a maximum IIP3 of 6 dBm at 5 dB gain while drawing 6.4
mA from a 1.2 V supply.

I. INTRODUCTION

Today’s digital TV (DTV) standards, such as advanced TV
systems committee-terrestrial (ATSC-T), Open Cable, digital
video broadcasting terrestrial (DVB-T), and DVB-C cover a
wide range of operating frequencies, from 48 to 860MHz. The
tuner, as a key element of DTV, converts received RF signals
into digital signals, which can be further processed for the
sound and picture. To develop a broadband DTV tuner, there
are many challenging technical issues including harmonic
mixing, image rejection, dynamic range, and linearity [1].
Unlike narrowband receivers, with broadband tuners, when
receiving lower-band channels (48~288 MHz), the harmonics
of the LO signal down convert the higher-band channels
(288~860 MHz). Therefore, harmonic rejection is an
important feature of DTV tuners to secure the down
conversion of the lower-band channel signals. Moreover, TV
broadcasting environments with multiple strong interferers are
another challenge in terms of selectivity and linearity
requirements. Those strong adjacent channel signals can
saturate the tuner RF front-end or degrade receiver sensitivity
by the down conversion of the LO phase noise into the desired
channel. For decades, the problems of linearity, harmonic
rejection and wide range signal strength have been resolved by
the adoption of an external tracking filter in front of the LNA.
Such external tracking filters are implemented with coils and
varactors, which are bulky, require high tuning voltage and
manual tuning, and furthermore, the tuner has to be shielded
from outside electromagnetic signals, i.e., CAN type tuners.
Therefore, the bulky size, poor temperature characteristics,
and manufacturing difficulty have been the long lasting
problems of CAN type tuner [2]. Lately, motivated by the
need for a smaller size tuner, industries have started to develop
more integrated tuner solution, i.e., a silicon tuner, to
overcome the deficiencies of the CAN type tuner. A silicon
tuner can offer the advantages such of smaller size, ease of
manufacture, and superior thermal stability. With a silicon
tuner, the issues that have been resolved by the off-chip
tracking filter still remains, but this time, these issues need to
be implemented on-chip. The on-chip tracking filter can be
implemented as passive or active. Fig.1 shows a typical silicon
tuner architecture in which the tracking filter is located after
the LNA but before the mixer.
With most of the previously reported silicon tuners, RF
tracking filters are based on the LC type, which still requires
bulky inductors, capacitors and varactors with high tuning
voltage [3], [4]. An active RF tracking filter can be an
alternative solution that can provides the advantages of small
size, low cost, and on-chip integration. However, the key
issues with using an active RF tracking filter for a DTV tuner
are poor linearity, high noise figure (NF), and large power
consumption. The previously reported on-chip active RF
tracking filters cover a tuning range of up to 300MHz with
large bandwidth (low Q), but consume a large amount of
power [5] [6], and most of them are based on low pass filter
type (LPF) which means that the filter provides only the
harmonic rejection not the channel selection. This paper
presents the very low power integrated RF tracking bandpass
filter with high linearity, wide frequency tuning range, and
good harmonic and unwanted signal rejection ratio. Section II
describes the design details of the proposed Gm-C type RF
tracking bandpass filter and Section III discusses the
measurement results. Section IV concludes.

II. ACTIVE RF TRACKING FILTER DESIGN

A. Filter Type
On-chip active filters can be implemented as active RC,
MOSFET-C, Gm-C, or switched capacitor filter types [7]. It is
well known that the active RC, MOSFET-C, and switched
capacitor filters can provide high linearity due to negative
feedback. However, their operating frequency bands are
limited due to the limited gain-bandwidth product of the
operational amplifier. High frequency operation is possible at
the cost of huge power consumption. The Gm-C filters are
known to show superior high frequency performance while
consuming a small amount of power, and to show good
tenability, but linearity is poor due to the open loop operation
nature. However, there have been many efforts to improve the
linearity of the Gm-C filter to a level comparable to that of
active RC filters. Therefore, the Gm-C type filter has been
adopted in the proposed RF tracking filter design.

B. Proposed High Quality Factor (Q) Biquad

978-1-4244-5759-5/10/$26.00 ©2010 IEEE

Page 2

Fig. 1. A Typical block diagram of RF front-end for silicon DTV tuner.

The biquad based filter architecture is chosen to build the
tracking filter considering the tenability and the simplicity in
analysis and control. Fig. 2 shows the conventional
biquadratic structure of a typical Gm-C bandpass filter.
Transconductors Gm1-4 along with C1 and C2 determine the
filter’s center frequency w0 and Q. The transfer function of the
tracking filter is
sC G
V
Vssw
+
where the center frequency and Q are

G G
wQ
CCG

21
2
22
100
/
m
Qw
= −
+
, (1)
34341
0
1222
1
mmmm
m
G G C
C
==
，
, (2)

and bandwidth is

021
/Q/
m
BGC
ω==
, (3)

In order to provide good harmonic rejection and unwanted
signal rejection, the Q factor of the biquad should be high.
From (2), it is possible to achieve high Q by adopting large
value Gm3 and Gm4. However, this method consumes large
power, and, moreover, if we want to keep the same center
frequency, the capacitor value should also increase, further
increasing the chip size. Another method to achieve high Q is
to reduce the value of Gm2. To achieve accurate analysis, the
parasitic capacitance and finite output resistance of the
transconductor are included, as shown follows
(32 ) 3
oi
sCs CCC
sCs CCCg
→+++
111
222
3
()
oeff
+
o
oioeffo
g
=
sCg
sCg
→+++=+
, (4)

where Ci, Co are effective input and output capacitors which
include the parasitic capacitance, and go is the output
conductance of the Gm cell. Including these effects, the more
accurate transfer function is shown in (5).
(
[
effeffm effeffo
V s CCs G CCg
++
, (5)
where the updated center frequency and Q are
3
mmomo
G Gg G g
w
CC
21
2
22
1122212342
)
3]3
effom
g
effommomo
sCg G
C
V
G GgG g
+
= −
++++

2
342
0
12
2
34212
2212
(3
+
)
3
effeff
mmomo
C
eff
g
eff
m effeffo effo
G G
G C
gG g C
g
+
C
Q
C
++
=
++
=
, (6)
Equation (6) proves that even with an ideal transconductor
(go=0), Q could not go to infinite. In order to develop a biquad
very suitable for RF tracking filter, the biquad should be low


Fig. 2. Conventional biquadratic structure of a typical gm-c band pass filter.


Fig. 3. Proposed low power and high Q biquad used for RF tracking filter.

power, high Q and high linear. With an emphasis on these
aspects, we propose a new simple low power and high Q
biquad for the RF tracking bandpass filter. The proposed low
power and high Q Gm-C biquad is shown in Fig. 3. By
removing the Gm2 stage and adopting the negative resistance
load (NRL) in the Gm-cell design, we both remove Gm2 and
reduce go value in (6). Therefore, a high Q and low power
biquad for RF tracking filter design is achieved. The
quantitative analysis is provided in this paper. The sC1’ and
sC2’ of proposed biquad are
(2)
oi
sCs CCC
→+++
′
→+++
where the center frequency and Q are

(
2
m
mmo
G G
G Gg
wQ
CCC
′′
From (8), also note that, C1eff’ and C2eff’ are smaller than C1eff
and C2eff respectively. We can find that the Q factor of the
proposed biquad is higher than the conventional one. With an
ideal transconductor, the Q factor of the proposed biquad
architecture can be infinite. Moreover, the center frequency is
also slightly higher without increase of the Gm value. As
shown in Fig. 3, there is only one internal node in the biquad
architecture, which means that the first and last stages can
share the common-mode stability and gain enhancement
circuit, thus further simplifying the architecture and also
reducing the power. The proposed biquad is very suitable for
RF tracking filter design in the aspect of low power and high
Q.
Cascaded sections of the two biquads are used to realize a
fourth order Butterworth RF tracking filter. Center frequency
tuning is achieved with the integrated programmable capacitor
bank, C1, C2, which can also compensates for the variation of
process, supply voltage, and temperature. In contrast with
integrated varactors, an integrated capacitor bank enables a
higher tuning range and better linearity performance. The
111
222
22
()
oeff
′
o
oioeffo
g sCg
sC s CCCg sCg
′′
=+
=+
, (7)
2
2
343
′
4
34
0
1212
2)
2
m
′
o effeff
effeffeffo effo
gCC
gCg
+
+
==
+
，
. (8)

Page 3

proposed RF tracking filter shows the potential to the whole
frequency band tracking filter design.

B. Gm-cell Design
The Gm-cell is the main building block in Gm-C type filter
Fig. 4 shows the schematic of the Gm-cell used in this work,
in which the dynamic source degeneration technique is
adopted to improve the linearity [8]. The four cross-connected
transistors, M3~6, operate in linear mode and provide
dynamically varying degeneration resistance to the differential
pair, M1-2, leading to nearly constant transconductance to the
input signal amplitude variation. As vin increases from zero,
the channel resistance of one of the two degeneration
transistors (M3 or M4, M5 or M6) is reduced, so that the
transconductance stays constant. In order to achieve the
optimized linearity, the size ratio between M1, 2 and M3-6, is
defined as a= (W/L)M1, 2/(W/L)M3-6. Fig. 5 shows the simulated
transconductance (gm) and the gm” of the Gm-cell as a
function of the differential input voltage for various transistor
size ratios a. From Fig. 5, a=2 shows the flattest gm and the
smallest gm” which means that this value gives the best
linearity.
In order to achieve high Q biquad which proved foregoing,
the negative resistance load (NRL) is adopted in our
transconductor design, formed by M7 to M10, as shown in Fig.
4. This NRL is very suitable for low supply voltage design,
since it does not need stack transistors which would introduce
extra internal nodes. Moreover, our design can maintain high
frequency performance with no generation of parasitic poles.
The operation of this load is such that M9 and M10 introduce
positive feedback between the nodes Vo+ and Vo- and a
negative resistance is produced that will be used to
compensate for the parasitic output resistance of the two
differential output nodes.
The common mode feedback circuit, formed by Mcf1-4, is
used to stabilize the DC operating point and bias the circuit
properly. The circuit has enough phase margins and good
linearity after adopting a source degeneration resistor Rc.
Other advantages to the Gm-cell are shown in Fig. 4. The
Gm-cell exhibits good high-frequency performance as no
additional internal nodes are created in the circuit. In addition,
the input impedance of the Gm-cell is highly capacitive, and
can easily be incorporated into the integrating capacitors of the
filter.
CMFB



Fig. 4. Schematic of Gm-cell with negative resistance load used in the
proposed RF tracking filter.


Differential input voltage [V]
-.4-.2 0.0 .2.4
Tranconductance [mA/V]
0.0
.2
.4
.6
a=1.68
a=3
a=2.2
a=2
a=1
Differential input voltage [V]
-.4-.2 0.0.2.4
Gm" [mA/V3]
-40
-20
0
20
40
a=1.68
a-3
a=2.2
a=2
a=1



Fig. 5. Simulated (a) gm and (b) g”m of the Gm-cell as a function of input
voltage for various size ratios.

IV. MEASUREMENT RESULT

The proposed RF tracking filter is fabricated using the 0.13
um CMOS process. For the measurement, a buffer is added at
the output and the filter dissipate 6.4 mA excluding the buffer
from a 1.2 V supply. Fig. 6 shows the chip micro-photograph
which includes the output buffer for measurement with a size
of 0.25 mm x 0.9 mm. Fig. 7 shows the measured frequency
characteristic of the filter for various center frequencies. As
can be seen in Fig. 7, the proposed RF tracking filter shows
tuning range of 48~300MHz. With proposed high Q biquad,
the 3-dB bandwidth ranges from 8 to 50MHz. Measured 3rd-
order harmonic rejection ratio (HRR) and unwanted signal
rejection ratio (URR) @ N+6 channel offset versus center
frequency is shown in Fig. 8. More than 35 dB HRR and 6 dB
URR are achieved. A two-tone test with 10 MHz separation is
applied to obtain the IIP3. Fig. 9 shows the measured IIP3
versus the center frequency for the input signals with 10MHz
offset at -23dBm power. As can be seen in Fig. 10, the filter
IIP3 varies from 1.9 to 6 dBm. Fig. 11 shows the measured NF
over the 48~300 MHz center frequency at the 50ohm source
impedance.
To compare the performance of the proposed filter with
that of other reported filters, a figure-of-merit (FOM), defined
in [15], is adopted. Table I compares the performance of the
proposed filter with that of other reported filters. This FOM
does some injustice to filters with high quality factor, to band
pass filters, and to filters with good blocking performance; this
design has all these properties, so the comparison is
conservative. However, as can be seen in Table I, the proposed
on-chip active RF tracking filter still shows good FOM with
lowest power consumption.



Fig. 6. Chip micro-photograph of the proposed RF tracking filter

Page 4

Frequency (Hz)

100x106
200x106
300x106
400x106
500x106
600x106
Magnitude (dB)
-50
-40
-30
-20
-10
0
10

Fig. 7. Measured frequency response of the RF tracking filter.
Center Frequency (Hz)

050x106
100x106
150x106
200x106
250x106
300x106
HRR / URR (dB)
0
10
20
30
40
50
URR
HRR

Fig. 8. Measured third order HRR and unwanted signal rejection ratio @ N+6
channel offset versus channel frequency with different chips testing.
Frequency (MHz)

150x106
160x106
170x106
180x106
Output power (dBm)
-90
-80
-70
-60
-50
-40
-30
-20
-10

Fig. 9. Measured two tone test spectrum with 160 MHZ and 170 MHz input,
at 165MHz center frequency, input power is -23dBm.
Center frequency (Hz)

50x106
100x106
150x106
200x106
250x106
300x106
IIP3 (dBm)
1
2
3
4
5
6
7

Fig. 10. Measured linearity performance over the frequency range.
Center frequency (Hz)

050x106
100x106
150x106
200x106
250x106
300x106
NF (dB)
19.0
19.5
20.0
20.5
21.0
21.5

Fig. 11. Measured NF over the frequency range.


V. CONCLUSION


was proposed and implemented. The low power and high Q
factor with good linearity performance are based on the
proposed high Q biquad with optimized Gm-cell design. The
In this paper, a low power on-chip active RF tracking filter
TABLE 1
PERFORMANCE SUMMARY AND COMPARISONS
Ref
Tech.
Filter type
HRR
URR
Freq range
OIP3
N
NF
Power
FOM
(FOM*)
This work
130nm
BPF
40
6~30
50~300
11
4
20
7.6
120
(11926)
[6]
180nm
LPF
30
No
50-300
16.9
3
14
72
152
(3675)
[9]
250nm
LPF
-

80-200
18.8
7
-
210
-
(1264)
[10]
250nm
BPF
-

30-120
10.1
8
-
120
-
(327)
Units
nm
-
dB

MHz
dBm
-
dB
mW
-
NF listed in this table is average values for fair comparison.
3/(
MAX
FOMOIPfTR N
=⋅⋅⋅
FOM* (without NF)

proposed RF tracking filter, implemented with the 0.13 um
CMOS process requires
Measurement results show a 50~300 MHz frequency tuning
range with narrow bandwidth (8~50 MHz). The average NF,
the maximum OIP3, and the power consumption (without
buffer) are 20 dB, 11 dBm, and 7.6 mW (from a 1.2 V
supply), respectively which shows the lowest power
consumption among all published tracking filters.

ACKNOWLEDGMENTS

This work was supported by the National Research
Foundation of Korea (NRF) grant funded by the Korea
government (MEST) (No. R0A-2007-000-10050-0).

REFERENCES

[1] Mark Dawkins, Alison Payne Burdett, and Nick Cowley, “A single-chip
tuner for DVB-T”, IEEE journal of solid-state circuits, vol.38, no. 8,
August 2003.
[2] Alvin Wong, “Silicon TV Tuners Will Replace Can Tuners as
Transistors Replaced the Vacuum Tube”,
[3] J.R.Tourret et al., “SiP tuner with integrated LC tracking filter for both
cable and terrestrial TV reception,” IEEE Journal of Solid-State Circuits,
42, (12), pp. 2809-2821, Dec. 2007.
[4] Olivier Jamin, Vincent Rambeau, Frederic Mercier, Insaf Meliane, “On-
chip auto-calibrated RF Tracking Filter for Cable Silicon Tuner,” in
Solid-State Circuits Conference, ESSCIRC 2008. 34th European,
Sept.2008 pp.158-161.
[5] Gupta, S. Lerstaveesin, D. Kang, and B-S. Song, “A 48-to-860MHz
CMOS direct-conversion TV tuner,” ISSCC 2007, pp. 206-207.
[6] Kuduck Kwon, H.T Kim, and Kwyro Lee, “A 50–300-MHz Highly
Linear and Low-Noise CMOS Filter Adopting Multiple Gated
Transistors for Digital TV Tuner ICs,” in IEEE Trans. Microwave
Theory and Techniques, Feb. 2009, Volume: 57, Issue: 2, pp.306-313.
[7] R. Schaumann and M. E. Van Valkenburg, Design of Analog Filters.
Oxford, U.K.: Oxford Univ. Press, 2001
[8] V. Prodanov, G. Palaskas, J. Glas, and V. Boccuzzi, “A CMOS AGC-
less IF strip for Bluetooth,” Proc. IEEE 27th European Solid-State
Circuits Conf, pp. 488-491, Sep. 2001.
[9] S.Dosho, T. Morie, and H. Fujiyama, “A 200-MHz seventh-order
equiripple continuous-time filter by design of nonlinearity suppression in
0.25-um CMOS process,” IEEE J. Solid-State Circuits, vol. 37, no. 5, pp.
559–565, May 2002.
[10] G. Bollati, S. Marchese, M. Demicheli, and R. Castello, “An eighthorder
CMOS low-pass filter with 30-120 MHz tuning range and programmable
boost,” IEEE J. Solid-State Circuits, vol. 36, no. 7, pp. 1056–1066, Jul.
2001.
1)
F Pdc
−
[6]
no off-chip components.