Академический Документы
Профессиональный Документы
Культура Документы
11g
Group Members:
Lawrence Kaplan
Wilson Liang
Mark McSharry
Date: 6/11/2013
I Introduction
In our project, we followed the receiver architecture specified in the assignment as shown in Figure 1. In
addition, we have voltage regulators to tap out specific voltages needed for proper circuit performance, and
current sources used in the baseband amplifier stage. We used a cascaded LNA with input and output
matching networks, followed by a balun to split the RF signal into I and Q rails. We used a passive
differential input mixer with differential LO. We chose our LO to be 2.38Ghz to have an IF frequency of
20Mhz. After the down conversion, we used a passive LPF to filter out all the unwanted harmonics. After
filtering, the IF signal is amplified again and the I path is created by I = I+ -I- and the Q path is created by Q =
Q++Q-. The I and Q are then shifted to be 180o out of phase and added together to cancel out the image.
II Design
LNA
For our LNA design, we chose to utilize the common-source amplifier that we had previously utilized for the
homework shown in Figure 3. We chose this design to maximize the gain. We opted to include a commondrain stage at the output to help improve the stability of our circuit. To begin with, we used non-ideal
inductors at the gate and the source of the LNA to bias the input. We sized the source-degenerative inductor
to match the real part of the input impedance to 50; we then sized the gate inductor to the correct value
such that were able to nullify the imaginary component of the input impedance. Next, we added a matching
network to the output of our LNA, so as to match the output impedance to 50. To calculate the necessary
values of L and C, we found the real and imaginary parts of the output impedance and used Q to find the
correct inductor and capacitor sizes. Lastly, we added a blocking capacitor to the output of the LNA (after the
matching network) to block any DC signal from the LNA, as we had designed the mixer to be a passive
mixer with no DC bias (we did eventually change this misconception and opted to bias the passive mixer as
you will see later). Once we had completed the LNA design, we ran the following Cadence simulations: gain,
NF, IIP3, return loss, and stability for the amplifier by itself. Figures 4-6 show the LNA gain, noise figure,
and stability plotted over temperature, respectively. Figures 7-9 show the LNA IIP3 at -20C, 27C, 60C,
respectively. Lastly, Figure 10 shows the return loss over temperature.
Low-Pass Filter
For our LPF design, we used an RC filter shown in Figure 17. When simulating a one stage RC filter, the
attenuation levels were not rolling off fast enough, so we continued adding stages until we got the desired
response. After a few iterations, we found that 8 stages were enough to get the desired response. We believe
this filter should be improved upon since non-ideal resistors in the models will give us loss in the pass band.
At 2.38 GHz and R = 100, this corresponds to a capacitance of 633.146 F. The circuit layout can be seen in
Figure 18, while the output waves can be seen in Figure 19.
Voltage Regulator
The voltage regulator was used to generate an 800mV supply line. We were allotted two DC supplies, which
we used for VDD and a 600mV bias. We required a third supply at 800mV to bias several transistors.
Rather than build a mirror for each bias, we elected to build a single supply that could be tapped for all the
800mV biases. In the end this saved on power consumption, as the regulator was designed with three low
current paths from VDD to ground. Had we built a mirror for each bias we would have likely doubled this
power consumption, and in the process decreased our Figure of Merit. Also, because the regulator was
designed entirely from transistors it is not affected much by changes in temperature, as a resistive mirror
would be. To generate the appropriate output voltage, the values of the transistors were adjusted. The
second level, PMOS on the far right is the dominant controller of the V800m bias as seen in Figure 20.
due to the non-idealities of the load resistors as show. If the resistor value were to increase significantly, the
output bias could be driven to ground.
Figure 22:Gain and Bandwidth for the Baseband Amplifier terminated with 10k.
The IIP3 was also simulated at the previously mentioned temperatures. The curves can be seen in Figures
23-25. All three curves fit well, and have IIP3 values in the range of -6 to -11dBm. The simulation at -20C
has a non-linear component for input power between -10 and -20dBm.
LNA
Gain
(dB)
19.37
IIP3
(dBm)
-12
Mixer
-9
7.7
Baseband
Amp
12.9
-11
N/A
Total
23.27
-16.4
2.2
NF (dB)
1.1
IV Conclusion
Although our receiver did not work as a whole, we learned and appreciated the difficulty in designing a
receiver chain. The individual building blocks of our receiver show promising results. If we could get this
receiver working and tuned it properly, we believe that we can meet the specs based on our theoretical
calculations. Future improvements to this receiver includes an on chip local oscillator, new filter design that
requires less stages and better stop-band attenuation.