Title: Impact on Constellation Due to Imperfect Mixing
1Impact on Constellation Due to Imperfect Mixing
2Key Receiver Design Issues AGC (1)
- Intermediate Frequency (IF) filter sets noise
bandwidth of the Receiver - Implementation impacted by cost, signal loss, and
adjacent channel rejection
3Key Receiver Design Issues AGC (2)
- Automatic Gain Control (AGC)
- Placement for minimal noise (after IF for
constant noise figure) - Large dynamic range to match the A/D dynamic
range - Response time of AGC loop is critical for min.
distortion and maximum dynamic range
4Digital AGC
To Software Receiver
A/D Converter
Input Signal
Amp
Gain Control
Energy Detector
Slew
Gain Factor Mapping
D/A Converter
Inactive
Mode Selector
?
Tracking
-
Reference Level
5AGC Modes
Reference Level
Slew Mode
Slew Mode
High amplitude level
Low amplitude level
Input Signal Level
Tracking Mode
Tracking Mode
AGC Inactive Zone
6Key Transmitter RF Design Issues
- Power Efficiency
- Modulation Accuracy and Linearity
- Spurious Signal Reduction
- SNR of Transmitted Signal
- Power Control Performance
- Output Power Level
7Transmitter Component Issues Ocsillator Mixer
- Transmit IF VCO
- noise floor
- power consumption
- phase noise provides significant modulation to
narrowband signals - Up-Converter
- Linearity to reduce spurious products
- Noise floor
- Power consumption
8Key Transmitter Component Issues Modulation
- Modulator
- balance between IQ required to keep distortion
(sidebands) down - Noise figure
- Power consumption
- Variable Gain Amplifier
- Linearity and fidelity
- Noise figure
9Transmitter Component Issues Transmit Filters
- Transmit Filters
- Isolation of transmitter noise from PA leaking
into the receiver (supplement duplexer) - low loss required
10Transmitter Component Issues Power Amplifier
(1)
- Power Amplifier (very critical)
- Cost - especially for base stations
- Noise floor
- Spurious response (source of interference)
11Transmitter Component Issues Power Amplifier
(2)
- Packing to handle heat
- Low distortion traded for power efficiency traded
for bandwidth - (in practice only about 25 of the battery is
effectively used during the talk time)
12General Performance Metrics
- Noise Characterization and Figure
- Spurious Free Dynamic Range
- Blocking Dynamic Range
- Intermod
- Power Consumption
13Noise Characterization (1)
- Noise is introduced into resistive components due
to thermal actions. - where k is Boltzmans constant (1.38.10-23 J/K),
T is the temperature in Kelvin, R is component
resistance (in ohms), and B is the bandwidth in
Hz.
14Noise Characterization (2)
- Antenna is the first and the base line source of
noise for which other noise sources are compared. - Thermal noise and quantization noise introduced
by the A/D
15Noise Figure
- Noise Figure (NF) measure the amount of noise an
element (or elements) adds to a signal. - NF SNRin/SNRout
- where SNRin is the input SNRout and is the device
output SNR. - Active Components The manufacturer of a device
usually supplies a noise figures for equals the
loss of the passive components.
16Using the Noise Figure (1/2)
- It is possible to provide an equivalent system
wide noise figure NFtotal that relates the noise
back to the antenna. - (equation 1)
- Here NFi represents the noise figure at the ith
stage and Gi represents the gain at the ith stage
(units are linear).
17Using Noise Figure (2/2)
- Given a component with a noisy input having noise
power Pi-1 (dBm), gain Gi (dB) and noise figure
NFi (dB) the output noise power Pi (dBm) is given
by - Pi (dBm) Pi-1 (dbm) NFi (dB) G (dB)
-
- Units are linear unless proceeded by (dB) or
(dBm).
18Example NF Calculations (1/2)
NF3 2 dB G310 dB
NF22 dB G2-2dB
NF46 dB
To Next IF Chain
Cable, G1-3dB
From Anttenna
BPF
X
LNA
LO
The total noise figure equals 5.975 .
19Example Noise Calculations (2/2)
Does ordering of the components yield optimal NF?
NF3 2 dB G310 dB
NF22 dB G2-2dB
NF46 dB
To Next IF Chain
Cable, G1-3dB
From Anttenna
BPF
X
LNA
- the total noise figure equals 3.6.
- In the system, the LNA has the biggest impact on
the noise figure (because of its high gain) - In general, it best to have higher gain
components (like the LNA) located as early as
possible in the RF chain.
LO
20Calculating Sensitivity (1)
- Sensitivity of the receiver to achieve a minimal
signal-to-noise ratio SNRmin is defined as - S dBm Noise floor dBm SNRmin dB
- where
- Noise floor dBm 10 log (kTB) NFtotal dB
- 10 log(kT) dB NFtotal dB 10 log(B) dB
- and B is the end of system bandwidth and NF
is the overall system noise figure.
21Calculating Sensitivity (2)
- For room temperature, the sensitivity becomes
- S dBm -174 dBm/Hz NF dB 10 log(B) SNRmin
- A good conservative practice keeps the noise
floor due to analog components lower than the
noise introduced by the A/D converter.
22Distortion Characterization 1 dB Compression
Point (1)
- Devices that exhibit cubic characteristic, the
third order distortion power grows at a rate of
3x the rate of the desired signal. - Eventually the device begins to saturate and when
the actual output power level differs by 1 dB
with the ideal output value, the 1 dB compression
point P1dB is reached.
23Distortion Characterization 1 dB Compression
Point (2)
- Amplitude compression tends to block the
detection of lower level signals in the presence
of stronger signals and the blocking dynamic
range (BDR)quantifies this effect. - BDR P1dB - MDS
- The MDS level occurs when the input -signal is
equal to the noise floor.
24RF Distortion - BDR
Output Power G ? Input Power
Output Power(dB) Input Power(dB) GdB
MDS Minimum Detectable Signal
P1dB,in Input 1 dB compression point P1dB,out
Output 1 dB compression point
BDR Blocking Dynamic Range BDR P1dB,in - MDS
25Spurious Free Dynamic Range (SFDR) Definition
- The difference between the input levels for the
MDS and the onset of third-order distortion (when
the third order distortion equals the noise
floor) defines SFDR.
26SFDR Measurement (1)
- The on-set of third-order distortion can be
determined using the two-tone test, where two
closely spaced tones of equal amplitude form the
input to the system and the amplitude is
increased until the third-order cross-product
produces a signal equal to the noise floor.
27Spurious Free Dynamic Range (2)
- This test mimics real world situations where
adjacent channel interference can cause
significant intermodulation distortion. - Typical dynamic range values extend from 60dB to
90 dB.
283rd Order Intercept (IIP3)
- IIP3 is found by extrapolating the fundamental
and third-order intermod. product lines until
they intersect. - The output power at this point is called the
third-order intercept point (OIP3). - SFDR can be found from the two linear equations
for the harmonic and third-order intermodulation
product - SFDR 2/3 IIP3 MDS
29RF Distortion - Intermod
Predicts Susceptibility to Adjacent Channel /
Nearby Interference
IIP3 3rd Order Input Intercept Point OIP3 3rd
Order Output Intercept Point SFDR Spurious Free
Dynamic Range SFDR 2/3 (IIP3 MDS)
IIM3 Intermod due to 3rd Order IIM3 3?PI - 2
?IIP3 (dBm)
30System Level DistortionCharacterization
- The effects of non-linear distortion are
cumulative. An overall IIP (either IIP2 or
IIP3), IIPtotal can be computed using the
following approximation. - where IIPi represents, in mW the Intermod
Intercept Point (IIP) for stage i. - Like parallel resistors, the overall total is
limited by the lowest value and the non-linearity
at the later stages becomes more critical since
its impact is magnified by the gain of all of the
previous stages.
31RF Distortion Intermod for Cascaded Devices
IIP3 2 dB G310 dB
IIP22 dB G2-2dB
IIP46 dB
Cable, G1-3dB
From Antenna
To Next IF Chain
LNA
BPF
X
LO
Dominated by Worst IIPi
32A/D Distortion Characterization
- Composite RF and A/D noise and distortion is
needed to quantify the overall receiver
performance. - A conservative design approach is to choose an
A/D converter that introduces insignificant noise
contribution compared to the overall RF chain.
33Example of A/D Impact
- For instance, given an input noise at the antenna
of 99 dBm, and a conversion gain of 25 dB and a
noise figure of 10 dB, the input noise to the A/D
is Ptotal (-99dBm 25dB 10dB) -64 dBm. - The percentage of noise power actually delivered
to the A/D load from the RF front end can then be
calculated. This noise voltage due to the analog
components can be compared to the noise figure of
the A/D converter. - A more precise analysis can determine the
overall noise voltage by summing the effective
voltage due to quantization with the voltage due
to the analog components, VA/D,total Vquant
VA/D,analog where Vquant iA/D RA/D.
34Example of A/D Impact (2)
Ranalog
iA/D Pquant / RA/D
V A/D,total
RA/D
i analog P analog,total / (R analog RA/D )
-
i analog effective current from analog noise
(RMS) iA/D effective current due to A/D
quantization noise P analog,total noise power
presented by the analog front end Pquant
quantization noise power Ranalog equivalent
analog resistance in series with A/D
converter RA/D resistance of the A/D converter
35RF Distortion Cascaded SFDR
Cascaded SFDR Recipe
- Determine Input Noise Power
- Calculate System Gain
- Calculate NFTotal
- Calculate Output Noise Power
- Calculate MDS
- Calculate IIPTotal
- Calculate SFDR
36Using SDR to Change the Cake Equation
- Software Radios have the added benefit of using
both software and hardware which changes
traditional tradeoffs - Examine two problems addressable by software
radio - PA nonlinearity vs efficiency
- RF flexibility vs performance
37Significance of the PA
- Quality determines capacity
- Output power defines coverage
- Impacts size of BTS
- Dominates infrastructure costs
- Major contributor to BTS operating costs
- Dominates power consumption
38Transmitter Component Issues Power Amplifier
(1)
- Power Amplifier (very critical)
- Cost - especially for base stations
- Noise floor
- Spurious response (source of interference)
39Transmitter Component Issues Power Amplifier
(2)
- Packing to handle heat
- Low distortion traded for power efficiency traded
for bandwidth - (in practice only about 25 of the battery is
effectively used during the talk time)
40Handling Multiple Channels Todays Realizations
SCPA Single Carrier Power Amplifier MCPA
Multi Carrier Power Amplifier
Antenna
Radio
Low power combiner
Radio
Band pass filter diplexer
MCPA
Radio
MCPA based BTS GSM, GPRS EDGE
Radio
41Realizing Multiple Channels with SDR and a Single
PA
Antenna
Wideband digital radio
Band pass filter diplexer
MCPA
Advantages over SCPA and Multiple Radio MCPA Most
cost effective with multiple carriers More
flexible More efficient Saves space Disadvantage
No redundancy and demanding PA specs
42Summary of Cost Drivers in TX Design
- Signal Peak-to-Average Power Ratio (PAR)
- Signal Peak-to-Minimum Power Ratio (PMR)
- Transmitter Power Control Dynamic Range (PCDR)
- Signal Bandwidth
- Transmitter Duplex Mode half or full
- Bandwidth Confinement Requirements (transmit
mask) - Adjacent Channel Power (ACP, ACPR, ACLR)
Earl McCuner, SDR Radio Subsystems using Polar
Modulation, SDR Technical Conference, Nov. 11,
2002, pp. 23-27.
43TX Requirements for Common Standards
Earl McCuner, SDR Radio Subsystems using Polar
Modulation, SDR Technical Conference, Nov. 11,
2002
44Non-Linearity and Power Amps
- Linearity
- Class A Best
- Class AB, B Mid-range
- Class C Worst
- Efficiency
- Class A Worst
- Class AB, B Mid-range
- Class C Best
45Simple Power Amplifier Model(no memory effect)
46Memory Effects In PA (1/2)
- High power, wideband amplifier characteristics
exhibit hysteresis-like effects - Frequency-dependent electrical memory effects at
high frequencies - Thermal memory effects at low frequencies
- Linearization scheme must cancel the dynamic
behavior of the PA
47Predistortion with Memory
Tapped Delay Line PD (TDL PD)
Complex gain polynomial
Hammerstein PD
Filter
48Memory Effects in PA (2/2)
hysteresis loops
- 4-carrier W-CDMA input, PAPR 13.7 dB
- Class B power amplifier (30W approx.)
49Distortion Effects
Non-Linearity gtgtgt Spectral Regrowth
Low PA Efficiency reduces battery life
50Why is this a Problem? (1/2)
- Modern comm. systems use non-constant envelope
modulation - QAM
- Non-Constant envelope signals require linear
amplifiers - Changes in amplitude cause spurious emissions
51Why is this a Problem ? (2/2)
- Linear amplifiers are power hungry
- Can lead to short battery life
- Amplifiers are the most expensive part of the
base station system
52Non-Linearity and Power Amps
Tradeoff
- Linearity
- Class A Best
- Class AB, B Mid-range
- Class C Worst
- Efficiency
- Class A Worst
- Class AB, B Mid-range
- Class C Best
Linearity for Efficiency
53Techniques to Change the Amplifier Nonlinearity
Tradeoff
- Backoff
- Cartesian Feedback
- Feedforward
- Analog Predistortion
- Digital Predistortion
- Linear amplification using Nonlinear Components
(LINC) - Envelope Elimination Restoration (EER) (also
known as polar amplifier) - Coding
54Backoff
- Non-linear region is at higher power levels
- Operate amplifier at a fraction of its rated
maximum power - Backoff level depends on amplifier
Non-linear Region
Linear Region
55Backoff Pros and Cons
- Benefits
- Simple
- Drawbacks
- Wastes amplifier capacity/efficiency
- Requires amplifier with power limit significantly
higher than operating point - Very expensive solution
56Predistortion
- Wideband power linear amplifiers are the most
costly part of a base station. - Predistortion can alleviate power amplifier
distortion - especially for non constant envelop
signals
57Predistortion Concept
- Input is distorted before feeding to the Power
Amplifier (PA) - The predistortion function generates anti-phase
Inter-Modulation Distortion components to those
generated in PA
z is the complex input to the predistorter and G0
is the linear gain of the overall system
58Benefits of Predistortion
- A trade-off exists between power efficiency and
linearity of the amplifier -- more favorable
trade - Reduced sidelobe regeneration
- 20dB possible
- Secondary benefits include compensation of
carrier leak and linearity problems of the mixers
and I/Q mismatch in the RF chain, possibly due to
the mixers.
59Conceptual Approach to Predistortion -- Analog
Domain
Analog Domain
Digital Domain
Gain F(Sr)
Transmitted signal So KSrCos(?ct)
Sr ideal modulated signal
Sp
D/A
Transmitter IF
Gain G(Sp2)
Note F(Sr) G(Sp) K, K constant gain
Predistortion function
Non-linear power amplifier stage
60Conceptual Approach to Predistortion -- Digital
Domain
Transmitted signal So KSrCos(?ct)
Digital Domain
Analog Domain
Gain F(Vm)
Gain G(Sp)
Vm ideal modulated signal
Sp
D/A
Transmitter IF
Predistortion function
Non-linear power amplifier stage
61Example Predistortion Technique
- AM/AM and AM/PM distortion compensated
- Large calibration table needed and must be
updated - Linearity of feedback loop is an issue.
62Digital Predistortion
- W-CDMA input, PAPR 7.8 dB
- 3rd order polynomial Predistorter (PD)
63Implementation Issues
- Linearity and fidelity of the correction loop
- Demodulator distortion in down-converters affects
performance - Capabilities of the A/D stage
- Must over-sample to capture harmonics
- Convergence vs. Stability
- Large time constants for aging and thermal
effects of the PA - Slow convergence is acceptable
64Flexibility Tradeoff
- SDR Necessitates a Flexible Front-end in terms of
Center Frequency and Bandwidth - Modern Signaling Requires High Performance
Components, Indicates Specific Component Design
65New Amplifier Topologies (1)
- Not really new (Re-emerging)
- See Chapter 8 of RF Power Amplifiers for
Wireless Communications by Steve C. Cripps - Linear Amplification with Nonlinear Components
- LINC for short
- Uses two nonlinear amplifiers and combines power
66New Amplifier Topologies (2)
- Envelope Elimination and Restoration (EER)
- The signal is split amplitude and phase
- Use nonlinear amplifier for phase and restore
envelope by modulating supply voltage
67New Amplifier Topologies (3)
- Both EER and LINC show improved efficiency and
potential for improved linearity - These amplifiers made practical by digital radio
design - Commercial products based on both methods have
been introduced
68LINC (1)
- First proposed by Chireix in 1935
S1(t)
S(t)
S2(t)
69LINC (2)
- S(t) is decomposed into 2 constant envelope
signals S1(t) and S2(t) - Two nonlinear power amplifiers are used
- Design of combiner is interesting
- At low envelope levels, this is inefficient
- Four port combiner has been used, but this
wastes half the output power. However this
still approached 50 efficiency and improves
linearity.
70EER (1)
- First introduced by Kahn in 1952
Video Power Conditioner
Envelope Detector
May be a digital input
Splitter
Limiter
Power Amplifier
71EER (2)
- Signal is split in to envelope and phase
components - Envelope component is used to modulate PA supply
voltage - Constant envelope phase component is used to
drive the nonlinear power amplifier - Ideally 100 efficient
72Why is this Related to SW Radio?
- Both LINC and EER require standard I/Q signal
translated into different format - LINC needs two phase modulated signals
- EER requires polar form signal
- PA can use direct baseband digital signal rather
than RF analog signal - The dividing line between the radio and the PA is
blurred
73Power Supply Issues Battery Life
Battery technology energy density doubles every
35 years.
Powers, Proc. of IEEE, April 95
74Other Considerationswith Batteries
- Environment (NiCa is terrible)
- Shelf life (length of time a charge is
maintained) - Cell Voltage
- Cycle Life (how often can it be recharged)
75Example Power Budget for a Typical Handset
- An Actual DECT Receiver Chip
- LNA power consumption 40 mW
- Mixer/downconverter section 50 mW
- A/D converter and BPFs 100 mW
- Total 200 mW
Rudell, et al., Proc. 1997 ISSCC
76Micro ElectroMechanical (MEM) Systems
- RF MEMS is a unique technology that offers a
significant impact on RF flexibility, performance
and cost
A Near Perfect Switch
77HRL MEMS Circuit
DC Bias
RF Signal
78Example of MEMS in SDR
MEMS Reconfigurable Antenna
MEMS Switchable Impedance Matching Circuit
ADC
LO
Software Control
79MEMS Applications and Future (1)
- Switches should be the first widespread
application - - Very low insertion loss 0.1 to 0.2 dB
- - Good isolation that depends on switch
configuration - - Good RF power-handling (gt 1 W)
80MEMS Applications and Future (2)
- Switches could enable a new class of RFICs
- - Integrated RF systems (e.g., multiband
radios-on-a-chip) - - New system architecture (e.g., reconfigurable
apertures) - - New RF functionality (e.g., quasi-optical beam
steering)
81MEMS Designs for RF Front Ends
- Design flexible filters using two-value
switchable capacitors - Tunable capacitors
- Two distinct capacitor values Con and Coff
- Two value capacitors arranged in parallel to form
digitally tunable capacitors
82MEMS Designs for RF Front Ends
- Inductors
- Fixed or variable
- High Q inductors for filters
- Tunable filters
- Use MEMs filter banks to create tunable RF filters
83MEMS Designs for RF Front Ends
E-tennas Reconfigurable Antenna
- Tunable antenna with narrow fixed bandwidth
- Patch antenna connected by RF switches
84Role of Superconductors In Software Radios (1/3)
- Extremely fast ADCs and DACs
- Based on superconducting quantum interference
device (SQUID) - Enable Digital-RF processing (TRF)
- Sampling rates now 20-40 GHz and 15 bits
- Enable more precise predistortion with DAC to RF
low feedback time for predistorter
Deepnarayan Gupta,etl, Benefits of Superconductor
Digital-RF Transceiver Technology to Future
Wireless Systems, SDR Technical Conference, Nov.
2002, pp 221-226. Superconductor Digital-RF
Transceiver Components, SDR Technical Conference,
Nov. 2002, pp227-232.
85Role of Superconductors In Software Radios (2/3)
- Sensitive enough to eliminate the LNA and lower
noise floor resulting in - Lower power
- Lower interference
- Greater range
- Greater capacity
86Role of Superconductors In Software Radios (3/3)
- High Purity Clock Sources to Reduce Jitter
- Extremely Fast Decimation and Matched Filters
87Transmitter Design for SDR
DIGITAL PREDISTORTER
DAC
Software can be used to control sampling rate
and resolution for different signaling standards
I Q can be predistorted by software to
compensate for nonlinearity of power amplifier
Software based power management is possible
(i.e., periodically turn the PA off, or adjust
bias to lower power consumption
Biasing can be dynamically adjusted by software
to reduce distortion
Flexibility in gain and bandwidth needed for
multimode operation
Mixer
Tuning controlled by software
BPF
PA
LO
88Receiver Design for SDR
LNA
BPF
AGC
Software based power management is possible
(i.e., periodically turn the LNA off, or adjust
bias to lower power consumption
Flexibility in gain and bandwidth needed for
multimode operation
Receiver noise and distortion can be minimized by
software controlled gain and attenuation
Sampling rate, resolution, interference rejection
controlled by Software
Biasing can be dynamically adjusted by software
to reduce distortion
Bandwidth and center frequency controlled by
software
Mixer
ADC
LPF
Tuning controlled by software
LO
89Summary
- RF design for multimode radios can be very tricky
- Best design must balance the performance of the
RF, A/D, and back-end DSP - Numerous tradeoffs must be made
- Software radio techniques can be used to
compensate for imperfection in RF components and
change the nature of the tradeoffs