You are on page 1of 22

VCO Design. Lab 3.

Prof. Jos Luis Gonzlez

Final Project.
Gilbert Cell Design.

Prof. Jos Luis Gonzlez.


Radio Frequency System-On-Chip
May, 30th of 2006

Cristina Domingo Sbat


Xavier Jimnez Gonzlez

1/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez

Gilbert Cell Design.


Contents.
Overview........................................................................................................................2
Introduction to mixers...................................................................................................2
Analytical study.............................................................................................................4
Gilbert Mixer Simulation..............................................................................................5
Conversion Gain........................................................................................................6
Port Isolation.............................................................................................................8
Power Dissipation, Large Signal Power Conversion Gain........................................9
S-Parameters............................................................................................................11
Noise Figure............................................................................................................13
1 dB Compression and IIP3.....................................................................................15
Effect of the Blocker on Gain and NF of Mixer......................................................17
Conclusions.................................................................................................................18
Proponed improvements..............................................................................................18

Overview.
This report tries to give a brief description of a mixer implemented by a Gilbert
Cell. After explaining some important concepts about mixers, an analytical analysis in
which transistors dimensions are calculated is presented. The main section of the work
is focused on the simulation of a Gilbert Cell. SpectreRF, of Cadence, is used for the
simulation in order to obtain the most significant parameters that characterize a mixer,
such as:
1.
2.
3.
4.
5.
6.

Power Consumption
RF to IF Conversion Gain
Noise and NF
Input and Output Impedance Matching
LO to RF and LO to IF Isolation
Linearity

Circuit design is implemented in a State-Of-Art of 0.35 m by an AMS CMOS


(c35b4) process.
Concluding this analysis a change on the original circuit is proposed in order to
improve the Noise Figure of the mixer.

Introduction to mixers
Mixers are used for frequency conversion and are critical components in modern
radio frequency (RF) systems. A mixer converts RF power at one frequency into power
2/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez
at another frequency to make signal processing easier and also inexpensive. A
fundamental reason for frequency conversion is to allow amplification of the received
signal at a frequency other than the RF, or the audio, frequency. A receiver may require
as much as 140 decibels (dB) of gain. It might not be possible to put more than 40 dB of
gain into the RF section without risking instability and potential oscillations. Likewise
the gain of the audio section might be limited to 60 dB because of parasitic feedback
paths, and microphonics. The additional gain needed for a sensitive receiver is normally
achieved in an intermediate frequency (IF) section of the receiver
The ideal mixer is a device which multiplies two input signals. If the inputs are
sinusoids, the ideal mixer output is the sum and difference frequencies given by

Typically, either the sum, or the difference, frequency is removed with a filter.

Circuit symbol for a mixer

However, a mixer is a non-linear device which provides an output that follows


the presented identity:
y ax bx 3

being a the mixers gain for the linear response and b the voltage gain for the third order
non-linearity. This non desired third order term will force the designer to consider some
aspects such as Intermodualtion products, Desenzitation, Blocking, 1dB Compression
Point or Interception Point, which will be introduced elsewhere in this work, when
designing transistors sizes and choosing design structures.
There are two types of mixer, passive and active. Generally
passive types (although have better IM3 performance) have higher
conversion losses and hence higher noise figures than active mixers.
Additionally, there are single balanced mixers and double balanced
mixers A Gilbert Cell is an active double-balanced mixer. Single
balanced mixers are much less complex, but have inferior
performance in terms of RF to IF and LO to IF rejection, compared to
double balanced mixers. Given below are the advantages and
disadvantages of double balanced mixers.

3/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez
Advantages:

Both LO and RF are balanced, providing both LO and RF


Rejection at the IF output.
All ports of the mixer are inherently isolated from each other.
Increased linearity compared to singly balanced.
Improved suppression of spurious products (all even order
products of the LO and/or the RF are suppressed).
High intercept points.
Less susceptable to supply voltage noise due to differential
topography.

Disadvantages:

Require a higher LO drive level.


Require two baluns (although mixer will usually be connected to
differential amplifiers).
Ports highly sensitive to reactive terminations.

Mixer operation
The RF signal is applied to the transistors M1 and M2 which perform a
voltage to current conversion. For correct operation these devices should not be

driven into saturation and therefore, signals considerably less than the 1dB
compression point should be used. MOSFets M3 to M6 form a multiplication function,
multiplying the linear RF signal current from M1 and M2 with the LO signal applied
across M3 to M6 which provide the switching function. M1 and M2 provide +/- RF

4/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez
current and M3 and M5 switch between them to provide the RF signal (or the
inverted RF signal) to the left hand load. M4 and M6 switch between them for the
right hand load. The two load resistors form a current to voltage transformation
giving differential output IF signals.

5/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez

Analytical study
When designing, it must be taken into account what are the requirements to
be fulfilled by the design. Likewise, it can be preferable to give more importance to
certain aspects, such as linearity or gain, in front of other characteristics or the
design which may be not that strict.
It is only commented in this section few hints of how a mixer can be
analytically design.
It can be desirable then, a differential input and output design or just a
single-ended design depending on whether previous and next blocks are balanced
or unbalanced ones. There is where BALUNs take place. However, a first step in
mixers design is to set the switching transistors to the same ratio W/L
(100u/0.35u). Other action that increases the linearity of the mixer lies on
introducing degeneration resistors or inductors to RF transistors sources (M1, M2).
Hence, the voltage gain of the mixer with source degeneration is given by:

Being gm the first stage transconductance defined as:

Nevertheless, gm can be also calculated as follows:

And defining the drain current through the transistor as:

The overdrive voltage (Vgs-VT) should be set at some value around 0.2 to 0.4V.
Depending on the current flowing through the LNA RF section we can determine
the optimum W/L ratio for the LO switching section.
Thus, for a desired high linearity performance of the mixer, this steps may be a
good reference when starting the design:
1.- Select a value for Rs. If we start with the LNA design , Rs will be realized by
an inductor Ls.
2.- Using the design equations of section 4(ii) decide on the IM3 value required
to calculate the
voltage overdrive (Vgs-VT). This is part of the LNA design process.
3.- Calculate gm and check for compliance of 0.4V.
4.- Determine LO switching W by selecting minimum L allowed and assuming
that the overdrive
voltage (Vgs-VT) will be between 0.2 and 0.4 V

6/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez

7/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez

Gilbert Mixer Simulation


The schematic that implements a Gilbert Mixer is presented on the foregoing
figure.

IF port

Supply
voltage
RF
port

Oscillation
part
LO
port

Bias current

This is the test bench for Gilbert Mixer:

8/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez
A Gilbert Cell is an active mixer, its to say, active components (transistors) are
used to achieve the frequency conversion. These types of mixers provide a certain
voltage/power gain to the output signal. Likewise we can define the voltage conversion
gain as the ratio of the RMS 1 voltages of the IF and RF signals and the power
conversion gain as the ratio of the power delivered to the load and the available RF
input power. Input and output equivalent resistances of the circuit must be taken into
account to achieve a certain power transfer. That means that the designer should have a
reliable idea of what output resistance value is expecting from a previous block (LNA
commonly) and what input resistance is presenting the next block in the RF chain (IF
Filter). However, a common design procedure relies on fixing these resistances to 50
Ohms just as we are going to assume for input resistance and output load. This
assumption leads us to the fact that so Voltage as Power Gains are the same when
presented in dB.
Note:

RF at 2.41 GHz------LO at 2.40GHz-----------IF at 10 MHz

Conversion Gain.
This is the ratio (in dB) between the IF signal (usually the
difference frequency between the RF and LO signals) and the RF
signal.
Conversion Gain can be obtained using a small signal analysis, like a PSS 2
together with a PAC3 or a PXF4 analysis.
But other gains such as VCG referred to RF frequency are typically obtained
when simulating.

Voltage versus frequency at RF and IF ports

Voltage Conversion Gain depends on the oscillation signal power as it is shown


here,
1

RMS: Root Mean Square


PSS: Periodic State S
3
PAC: Periodic Altern Current
4
PXF: Periodic XF
2

9/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez

VCG versus LO output power.

Thus, we notice a maximum and top power transfer when the local oscillator
power is set to be around 20 dBm. Consequently, subsequent steps in simulation are
done assuming this power provided by the local oscillator.
Voltage conversion gain also depends on the frequency of the IF signal (when f LO
is fixed and fRF is variable). So a simulation is done by fixing power in Local Oscillator
port (20 dBm) and power in the RF port (-50 dBm). The next graphic shows Voltage
Conversion Gain versus frequency at IF.

Voltage Conversion Gain.

10/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez
Consequently, we can state that the Gain of the mixer designed is
Power Gain (dB) = Voltage Gain (dB) = - 6 dB
Its value moves around -6 dB, what means that it is not a very good one because
active mixers like this should have a VCG greater than 0 dB.
We have done a similar simulation by using PXF. This method is done to use a
two-tone large-signal QPSS analysis which is more time-consuming. After simulating,
we check that the results are very similar:

Voltage Conversion Gain.

Port Isolation.
We can define the A-to-B port isolation as the ratio in dB between the A
frequency power level incident at the A port of the mixer and the A frequency power
emanating from the B port. It can be also referred as the A-B feedthrough (leakage).
Another important characteristic to measure in a mixer is the port-to-port
isolation.

11/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez
And this is the RF to IF isolation:

RF-to-IF port Isolation.

LO-to-RF and LO-to-IF Feedthrough:

LO-to-RF and LO-to-IF Isolation.

Power Dissipation, Large Signal Power Conversion Gain


Another remarkable characteristic of a mixer resides on how much power it is
able to transfer from the injected power at RF and LO to the desired output (main
harmonics of IF).
By using a QPSS analysis we have been able to identify this parameter in our
circuit..

12/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez
QPSS (Quasi Periodic Steady State Analysis) is an analysis that invokes a series
of PSS-like analyses over all the input frequencies, their harmonics and the
intermodulation products of the input frequencies.
QPSS allows arbitrary signal inputs, including sum of sinusoids which are not
periodic, so called quasi periodic extension of PSS. Similar to PAC (Periodic AC
analysis) it calculates the responses of the circuits that exhibit the frequency translation
like mixer, oscillator etc. Unlike PAC, PSS is not explicitly required before QPSS as it
simulates the moderate and large signal behaviour instead of small signal behaviour.
Likewise, this is the Large Signal Conversion Gain of our circuit:

Large Signal Conversion Gain

It must be taken into account that so QPSS as PSS provide the spectrum, not
scalar values. So if we want to give a value to the power consumption, a good
approach relies on the summation of the main fundamental harmonics presented
in the foregoing figure. We can state, thus, that our circuit accomplishes a good
behaviour referring to power consumption.

13/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez

Large Signal Conversion Gain (zoom at interest frequency, 10 MHz)

S-Parameters
Elementary circuit theory provides many methods for describing electronic
networks. Those methods, however, best describe DC and low-frequency circuits. They
fall short when the wavelengths of the signals of interest shrink to become comparable
to the physical dimensions of the circuit of interest. To characterize high-frequency
circuits, you can employ S-parameters (or scattering parameters) in place of the
impedance or admittance parameters that describe low-frequency circuits. Each
parameter is typically characterized by magnitude, decibel and phase. The expression in
decibel is 20log(Sij) because s-parameters are voltage ratios of the waves.
S11:
S21:
S12:
S22:

input reflection coefficient of 50 terminated output.


forward transmission coefficient of 50 terminated output.
reverse transmission coefficient of 50 terminated input.
output reflection coefficient of 50 terminated input.

Notice that:
PORT 1 RF port
PORT 2 IF port
PORT 3 LO port

14/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez
The advantage of s-parameters does not only lie in the complete description of
the device performance at microwave frequencies but also the ability to convert to other
parameters such as hybrid (H) or admittance (Y) parameters.
A QPSS analysis has been required as well to obtain the s-parameters.
These are the figures corresponding to the mixer S-parameters:

S21: Voltage Conversion Gain.

S13: Power transfer from LO port to RF port

S23: LO-IF feedthrough

15/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez

S31: RF-LO feedthrough

This graphic shows Noise Figure in the range of frequency from 0 to 10 MHz.

Noise Figure
Noise figure is defined as the ratio of SNR at the IF port to the SNR of the RF
port.
Typically, the signal present at the image frequency is not desired. The mixer
translates both RF and the image signals at the same IF. So for a noiseless mixer the
output SNR is half the input SNR i.e. NFSSB of a noiseless mixer is 3dB.

16/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez

However, in some applications (direct conversion receivers) the signal present at


the image frequency contains useful information, and hence the NF DSB is measured and
calculated.
Noise Figure SSB and DSB:

Noise Figure SSB and DSB

We define output noise as the amount of undesired signal power found at the
output signal frequency.

17/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez
Output Noise:

Output Noise

1 dB Compression and IIP3


In small signal conditions output power increases linearly with increase in the
input signal power. When circuits shift toward large signal operation this relation is no
longer linear. The 1dB compression point is a measure of this nonlinearity. This is
power where the output of the fundamental crosses the line that represents the output
power extrapolated from small signal conditions minus 1dB.
The recommended approach to calculate the 1dB CP and IIP3 is to apply large
LO and one medium RF tone and perform the QPSS analysis. Then the second tone as a
small tone close to the RF signal frequency is applied to perform afterwards the QPAC
analysis. The power of the 2nd small RF signal tone, pacmag, has to be small enough
that IMP1 and IMP3 are in their asymptotic ranges.

18/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez
1dB Compression point and IIP3:

1dB Compression point

IIP3 using QPSS and QPAC:

IIP3

19/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez
Effect of the Blocker on Gain and NF of Mixer
In band and out of band blockers are specified for all standards (GSM, DECT
etc). These blocker desensitize the receiver i.e. the gain and NF of the receiver for
desired signal is drastically degraded. All communication standards include the blocking
requirements for both mobile terminals and base stations. The requirement defines
several in-band and out-of-band blockers.
NF in presence of Blocking Signal:

NF in presence of Blocking Signal

Voltage Conversion Gain in presence of Blocking Signal:

Voltage Conversion Gain in presence of Blocking Signal

20/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez

Conclusions

Down conversion of receiving signal is achieved. The mixer provides the desired
IF frequency at 10 MHz.
Not as good as desired Interception Point as well as 1dB Compression Point are
obtained. We observe not a very good linear behaviour.
When applying a Blocking signal at the input, a good mixer response can be
noticed in terms of NF but not in VCG.
Good power consumption behaviour is accomplished.
Mixer sensibility is considerably lower than expected. This leads to a masking of
the RF signal (medium power signal, -50 dBm) by the LO signal (large power
signal, 20 dBm).
Feed through (Isolation, leakage) is far from desirable. This affects as well to
mixer sensibility
A very poor gain is a remarkable aspect to consider. This leads to dreadful
values of NF as well as SNR.

Proposed improvements
As a result of such a bad noise result, we have referenced our proposed
improvments to a paper titled Noise in RF-CMOS Mixers: A simple Physical Model
written by Hooman Darabi and Asad A. Abidi, who are fellows of the IEEE. In this
paper, they refer to three kind of noise with effect on the output signal and state what
factors take influence on them:
-

low frequency noise: inversely dependant on LO slope and periode

square-like waves and large periodes prevents this kind of


noise.

High frequency noise: it does not depend on transistor size but on LO


amplitude and the bias current.

Increasing gate area of transistors which provide bias current


could be reasonable.

Transconductance noise: inversely dependant on LO amplitude.

They also explain that increasing the gate area of the


switch transistors lowers flicker noise as well as lowering
the OD5 voltage; however, it degrades mixer bandwidth.

Lowering LO amplitude minimizes this kind of noise.

OD Voltage: Over Drive Voltage.

21/22

VCO Design. Lab 3.


Prof. Jos Luis Gonzlez
Important note: when OD Voltage assumes similar values as LO
amplitude both noise effects contribute comparably to the output noise. Hence, a
tradeoff between linearity and noise takes place in active mixers.

22/22

You might also like