A 2.4 ghz cmos lna input matching design using resistive feedback topology in...
ECE265B_FinalReport_finished all_v1.0
1. Tunable Narrowband Receiver (1GHz-1.5GHz) with Active Mixer 1
Tunable Narrowband Receiver (1GHz -1.5GHz)
with Active Mixer
Jihang Lu, Haoran Pu, Fanyu Yang, and Nian Jiang
Abstract – Based on IBM 0.18m-CMOS
technology, an Implementation of a narrowband
receiver tunable from 1 GHz to 1.5 GHz is
reported in this paper. To achieve high
performance on NF (noise figure) and linearity,
the receiver consists of a differential source
degenerate LNA (low noise amplifier), a negative
body-biasing active mixer, and a differential TIA
(transimpedance amplifier) with CMFB (common
mode feedback). In LNA, APD (active post-
distortion) is introduced to improve IIP3.
Feedback is used in TIA to improve linearity.
Based on the simulation results, the overall
performance of the proposed receiver achieves <3
dB NF and > 8 dBm IIP3 across the frequency
band from 1GHz to 1.5GHz.
Index Terms – tunable narrow band receiver,
LNA, active mixer, TIA, APD, linearity
improvement, noise figure, resistive feedback,
CMFB.
I. Introduction
For a narrowband receiver, high linearity is critical,
as this property is essential to alleviate severe gain
compression, cross modulation and harmonic
generation [1]. This paper proposes a tunable
narrowband receiver with high performance on
linearity. The paper begins with the project target
performance and background information. Section III
to Section V describes the major components of the
receiver. The overall performance is presented at the
end of this paper.
II. Background and Project Target
High linearity receivers are significant for
communication system due to the requirement of
signal-to-interference-plus-noise ratio (SINR). Since,
in general, TIA tends to have good linearity, the LNA
and mixer of the receiver have significant impact on
overall linearity of the system. For LNA, several
linearity improvement methods has been proposed. [2]
introduces an idea of 3rd
order distortion cancellation
by using 2 different biases. In [3], the theory of active
post distortion (APD) is used to improve IIP3. For
mixer, several topologies in [4], [5], and [6] has been
used to improve the linearity of active mixer.
In previous work on tunable receiver, [7] has
proposed such a receiver with parameters shown
below:
Frequency
Range
(MHz)
Maximum
Gain (dB)
NF at
Max
gain
(dB)
IIP3 at
Max
gain
(dBm)
IIP3
(dBm)
Power
Consump
-tion
(mW)
470-862 >80 7.9 -8 +2 120
In this paper, we target to design a receiver with
following specs:
Gain
(dB)
NF
(dB)
LO Output
Power
(dBm)
IIP3
(dBm)
IIP2
(dBm)
Power
Consumption
(mW)
> 45 < 3 ~ 0 > +6 > +45 < 40
III. LNA
A. Topology
Since the receiver is a tunable narrowband receiver
and has high requirement on noise figure and
linearity, inductor source degenerated LNA is used.
Capacitors in series with switches control operation
frequency from 1 GHz to 1.5 GHz, 100 MHz per
stage. In addition, several techniques are introduced
to improve noise figure and linearity.
B. Band-Tuning
Tunable narrowband LNA has basic source
degenerated LNA structure (Figure 1), but its input
matching frequency and output tank peak frequency
are changeable from 1G Hz to 1.5G Hz. In the
tunable range, conversion gain S21 is larger than 15
dB, reflection coefficient S11 is less than -10 dB, NF
is less than 3 dB, out of band IIP2 is larger than +45
dBm and IIP3 is larger than +6 dBm. Five switches
control five capacitors in parallel with of the
common source NMOS to adjust input matching
frequency changing from 1 GHz to 1.5 GHz, 100
MHz per stage. Another five switches control the
resonance frequency of the load tank. The LNA is
designed sensitive to the added capacitors. In other
words, the added capacitors can change input
matching frequency and load tank frequency easily
2. Tunable Narrowband Receiver (1GHz-1.5GHz) with Active Mixer 2
with small capacitance values. Small capacitor is
important because large capacitor can introduce large
parasitic resistor and capacitor, and require large area.
C. NF Reduction
NF of the source degenerated LNA can be roughly
calculated as:
NF is reduced by reducing the size of the source
degenerated inductor and matching inductor .
The two inductors have the minimum line spacing in
order to have large inductance and small resistance at
the same time. In addition, is reduced as much as
possible to reduce noise figure and power
consumption, but the gain should be larger than +15
dB for all the stages.
D. Linearity Improvement
To improve linearity, the theory of active post
distortion is introduced to improve IIP3. The
structure used (Figure 2) is based on modified super-
position method. The circuit cancels third order non-
linearity by adding two current together to reduce
third order distortion term.
Fig. 1 (left). LNA Topology
Fig. 2 (right). Schematic Diagram of Active Post Distortion [3]
Under weakly nonlinear small signal condition,
higher order nonlinearities can be neglected.
Assuming only and contribute to
nonlinearities, gate voltage and drain current of
can be expressed respectively as:
Where and are the ratio of trans-conductance
as shown in Figure 2. Then we can find first and third
order terms for the output current:
The first equation shows, by adding Active Post-
Distortion, gain is reduced. In the second equation,
the first term is the third order nonlinearity
contribution from the original third order distortion
and the second term is the third order nonlinearity
contribution from the original second order distortion.
In our case, set and . Gain is
suppressed a little but total third order distortion due
to the original third order distortion is significantly
reduced. Since the second order distortion is
relatively small when compared to first order term,
the second term in the second equation can be
ignored [3].
Since a very good IIP2 (+45 dBm) is required, the
LNA need second order distortion reduction. We use
differential input differential output LNA to eliminate
second order distortion. As shown below, third order
distortion is suppressed as well by using differential
LNA. However, power consumption is doubled.
E. LNA Performance
Final result of LNA across frequency is shown as
below in Table 1.
TABLE 1.
SIMULATION RESULT OF LNA
Frequ-
ency
(GHz)
Gain
S21
(dB)
S11
(dB)
Noise
Figure
(dB)
Out of
Band
IIP3
(dBm)
Out of
Band
IIP2
(dBm)
Power
Consu-
mption
(mW)
1.0 +15.8 -12.6 2.67 +8.60 +44.1 28.88
1.1 +17.0 -14.5 2.49 +9.46 +45.3 28.88
1.2 +17.9 -16.9 2.38 +10.7 +47.4 28.88
1.3 +18.6 -19.5 2.33 +15.8 +47.5 28.88
1.4 +19.0 -22.1 2.32 +12.7 +55.9 28.88
1.5 +19.3 -28.1 2.34 +10.6 +50.1 28.88
F. Future Improvement
The circuit has best performance at 1.4GHz, while
its spec degrades as frequency moves away from
1.4GHz. The degradation is largely due to input
matching which affects all other specifications.
Instead of switching capacitors, wideband input
matching can solve the degradation problem.
3. Tunable Narrowband Receiver (1GHz-1.5GHz) with Active Mixer 3
Besides, variability over process and temperature
should be considered since the bias point of M2 and
M3 are insensitive.
IV. Active Mixer
A. Topology Selection
Active mixer is the bottleneck for linearity of the
whole system. With double balanced topology, IIP2
is cancelled with the differential output, and the main
issue is IIP3. We test several topologies with IIP3
target of 10 dBm. Eventually the topology we use is
negative body-biasing.
B. Source Degenerated Topology
In [5] a topology with modifications to the original
double balanced active mixer structure is presented.
It uses PMOS transistors as load to avoid load
resistance limit if an actual resistor were used. The
tradeoff is that flicker noise of the load transistors
shows up at baseband. We add inductors between
sources of switching pairs and source of RF input
pair to tune out large parasitic capacitance introduced
by large switches. Most importantly as presented in
the paper we use very large source degeneration
NMOS transistors
to maximize linearity. However,
instead of an IIP3 of +10 dBm as shown in the paper,
when simulating with PSS shooting we can only get -
3.5 dBm. Obviously this is too far away from our
target.
C. Negative Body Biasing Topology
The topology we eventually used is as presented in
[4]. It is a double balanced active mixer that utilizes
negative body-biasing to cancel third-order
transconductance. The NMOS input pairs for RF+
and RF- signals each provide adequate gm of 19mS
with moderate current consumption of 4.72mA. Each
input transistor is separated into two transistors to
allow negative body biasing at one of the transistor.
Negative bias voltage applied to body of one of the
transistor (TN81 as in Figure3) can increase its
threshold voltage and shift its curve right by the
same amount. Combining the new curve with the
original curve (from TN4 as in Figure 3), a flat region
can be created where is minimized.
Fig.3. Schematic of negative body biasing
The correct gate and body bias is found using
parametric sweep. is determined to be -1.16 V
and is determined to be 690 mV to give IIP3 of
7.9 dBm. As later simulation turns out, IIP3 reduces
as LO frequencies move down. This implies that
curves are frequency dependent, and the optimum
bias points shifts as frequency shifts. Luckily, the
degradation of IIP3 is compensated by gain
degradation of LNA as frequency reduces, and the
system IIP3 is largely maintained.
D. Flicker Noise Reduction
When the active mixer is tested with the receiver
chain, flicker noise shows up at input pairs of TIA.
This flicker noise comes from large current through
switching pairs at zero crossings of LO. To alleviate
this problem, a cross-coupled PMOS pair is added.
Cross-coupled pair themselves introduces flicker
noise, so is increased about 5 times until NF of
whole system reaches the specification.
E. Suggested Improvements
As mentioned above, the negative body biasing is
only optimized for a single LO frequency. To
maintain linearity over various frequencies, some
other topologies may be more suitable. However, as
shown in [6], tunable receiver generally has less than
0 dBm IIP3. New topologies are needed for active
mixers that are both highly linear and tunable.
Also IIP3 results vary a lot with simulation method
being used. PSS (including shooting and harmonic
balance) and HB give widely spread results. The
large difference makes it hard to decide the best
topology.
F. Final Results of Active Mixer
The final results of active mixer are shown below
in Table 2.
4. Tunable Narrowband Receiver (1GHz-1.5GHz) with Active Mixer 4
TABLE 2.
SIMULATION RESULT OF ACTIVE MIXER
Gain
(dB)
NF (dB) Power
(mW)
IIP3
(dBm)
IIP2
(dBm)
12.7 ±2 16.8 ±2 8 3.65 ±4 57 ±7
V. TIA
A. Topology Selection
For the last stage of receiver, which is the TIA,
two-stage differential OP-AMP with common mode
feedback (CMFB) is used. The simplified circuitry is
shown in Figure 4.
Fig.4. Circuitry of Differential OP-AMP with CMFB
The amplifier is designed with a differential first
stage and a common source second stage. CMFB is
designed using a low gain amplifier, which is added
to keep the common mode output low. The TIA is
biased at ½ VDD (0.85V).
Analysis and design procedure of the differential
OP-AMP will be briefly discussed in next two
sections.
B. Linearity Boosting
To improve linearity, input impedance needs to be
minimized without degenerating bandwidth and
phase margin. Thus, open loop gain of the OP-AMP
should not be too small. To improve the open loop
gain, for the first stage, boosting of the PMOS
common source amplifier is required, while of the
NMOS below also has to be kept large enough. For
the second stage, of the NMOS common source
amplifier should be large. The open loop gain of the
OP-AMP presented can be boosted up to 60dB in
simulation (Figure 5).
Fig.5. Open Loop Gain of the OP-AMP
Moreover, RC feedback is added between input and
output to construct a TIA, which has good linearity at
baseband.
C. Stability Optimization
In order to maintain stability at baseband, enough
phase margin (larger than 50 degrees) is critical.
Compensation RC network needs to be carefully
designed to achieve this goal. Therefore, capacitance
should be large so as to guarantee proper phase
margin. Moreover, the product of resistance and
capacitance may not be overly large to avoid
corrupting bandwidth. Phase margin of the TIA
proposed in this report is 59 degrees, which ensures
the stability of the receiver system.
After several rounds of optimization based on
system requirements, size of the transistors are
adjusted to ensure enough gain, bandwidth and phase
margin. The performance parameter of the final
design of TIA is listed in Table 2.
TABLE 3.
SIMULATION RESULT OF TIA
Power Gain Noise Figure
(Integrated, 1Hz-
5MHz)
Phase Margin
TIA 76dB 3.08 dB 59 degree
IIP3 IIP2
TIA 23 dBm 83 dBm
VI. Result Summary
Simulations are conducted in Cadence. Simulation
results for 1.5 GHz are shown in Figure 6.
5. Tunable Narrowband Receiver (1GHz-1.5GHz) with Active Mixer 5
Fig.6. Simulation results at 1.5 GHz
The overall system’s simulation result across
frequency is shown in this part.
The general performance is summarized in Table 4
.
Table 4.
General Performance of the Receiver across Frequency
Frequency
(GHz)
Power
(mW)
Gain
(dB)
Integrated NF(1Hz-
5MHz) (dB)
1
41.5
51.79 2.93
1.1 52.84 2.69
1.2 53.2 2.59
1.3 53.59 2.52
1.4 53.98 2.52
1.5 53.7 2.98
All specs except the power consumptions have met
the project target. Power consumption is 1.5mW
higher comparing with requirement.
As shown in Table 5, IIP3 of the receiver > 8dBm
across the frequency. IIP2 is ranged from 52dBm to
95dBm over frequency 1GHz and 1.5GHz, with the
best performance achieved at 1.2GHz.
Table 5.
Linearity Result of the Receiver across the Frequency
Frequency (GHz) IIP3 (dBm) IIP2 (dBm)
1 8.44 89.07
1.1 8.46 68.57
1.2 8.69 95.29
1.3 8.8 77.352
1.4 8.6 62.18
1.5 8.58 52.68
VII. Conclusion
A tunable narrow band receiver working in the
frequency 1GHz to 1.5GHz is presented in this
project report. The performance of receiver has met
the project requirements with power usage slightly
higher comparing with the target. The linearity of this
system is greatly improved with the implementation
of active post distortion theory and third-order
transconductance cancellation on LNA and active
mixer respectively.
REFERENCES
[1] T. W. Kim, B. Kim, and K. Lee, "Highly linear
receiver front-end adopting MOSFET Trans-
conductance Linearization by multiple Gated
transistors," IEEE J. Solid-State Circuits, vol. 39,
no. 1, pp. 223–229, Jan. 2004.
[2] Vladimir Aparin and Lawrence E. Larson.
“Modified derivative superposition method for
linearizing fet low noise amplifiers.” Mircow.
Theory and Techniques, 53(2): 571–581, Feb.
2005.
[3] K. Namsso, A. Vladimir, B. Kenneth, P. Charles.
“A Cellular-Band CDMA 0.25um CMOS LNA
Linearized using Active Post-Distortion.”
Qualcomm, ESSCIRC, France, 2005.
[4] K. Liang, C. Lin, H. Chang, and Y. Chan. “A
New Linearization Technique for CMOS RF
Mixer Using Third-Order Transconductance
Cancellation.” IEEE Microw. Wireless Compon
Lett., vol. 18, no. 5, pp. 350-352, May 2008.
[5] M.Brandolini, P. Rossi, D. Sanzogni, and F.
Svelto. “A +78 dBm IIP2 CMOS Direct Down-
conversion Mixer for Fully Integrated UMTS
Receivers.” IEEE J. Solid-State Circuits, vol. 41,
no. 3, pp. 552-559, Mar. 2006.
[6] C. Wu, H. Hsieh, L. Lai, and L. Lu, “A 3–5 GHz
Frequency-Tunable Receiver Frontend for
Multiband Applications.” IEEE Microw.
Wireless Compon Lett., vol. 18, no. 9, Sept. 2008.
[7] Kulkarni, R., et al. "UHF Receiver Front-End:
Implementation and Analog Baseband Design
Considerations." IEEE Transactions on Very
Large Scale Integration
Systems 20.20(2012):197-210.