Basic Concepts in RF Design

ICE669 (H. J. Yoo)

 Nonlinear systems
• Linear system • Example
– output can be expressed as a linear combination
of inputs 


 = 
 ω


  
 =   ω 

– otherwise, nonlinear  
   


 >

• nonzero initial conditions or finite offsets
make system nonlinear
• Time invariant system
– time invariant: time shift in input results in the
same time shift in output


→  

 
− τ
→  
− τ

• Differential or balanced system

– system with odd-symmetry
– no even order harmonics

– (b) nonlinear time-variant system

– (c) linear time-variant system
• a linear system can generate frequency
components that do not exist in the input
signal
Razavi98

ICE669 (H. J. Yoo)

 Gain Compression
• Memoryless, nonlinear system • Gain Compression
– For a nonlinear system that can be approximated – Small signal gain varies with input level
 

≈ α


+ α    

+ α    

– Most circuits, output is compressive or saturating
→ occurs if α3 < 0
– Small signal gain is a decreasing function of A
• Harmonics – 1-dB compression point: input signal level that
  
 =   ω
  causes small-signal gain to drop by 1 dB
 
 = α
  ω
+ α      ω
+ α      ω

α   α  
= α
  ω
+  +  ω
 +   ω
+  ω

 
α   α    α  α 
=  + α
 +    ω
+   ω
+   ω

     

– Even order harmonics result from αj with even j

– Amplitude of nth harmonic consists of a term 

 α
+ α  
−  = 
 α
− 
proportional to An and other terms proportional to 
higher power of A
α

– Neglecting latter for small A, nth harmonic grow in 
−  =

proportion to An α

– Measure of maximum applicable input signal

– Around -20 to -25 dBm (63.2 to 35.6 mVpp in 50 Ω
Razavi98
system) in typical front-end RF amplifiers

ICE669 (H. J. Yoo)

 Desensitization and Cross Modulation
• Desensitization and Blocking • Cross Modulation
– Desensitization – Phenomenon that occurs when a weak and a
• Problem when process weak, desired signal strong interferer pass through a nonlinear system
with a strong interferer – Amplitude of interferer modulates the amplitude of
• Weak signal can experience smaller gain as weak signal
large signal tends to reduce average gain – From (1), if amplitude of interferer is modulated by


 = 
 ω

+   ω 
  +   ω 
  ω 


 
 = α

 + α    
 + α    

 
   
 = α

+ α  

   
= α

+ α  
 + α  
   ω
+ !
       
 
<<   ×  + +  ω 
+   ω 
  ω
+ !
   
  
 
 = α
+ α    
 ω
+ !
   – Desired signal at the output contains amplitude
• Gain becomes a decreasing function of A2 modulation at ωm and 2ωm
– For sufficiently large A2, gain drops to zero → – Common case of cross modulation
“blocked” • Amplifiers that must simultaneously process
– In RF design, “blocking signal” usually refers to many independent signal channel — e.g.,
interferers that desensitize a circuit though gain CATV transmitters
does not fall to zero • Full duplex CDMA systems — request for
– Many RF receivers require to withstand blocking super-linear LNA
signals 60 to 70 dB higher than the wanted signal

Razavi98

ICE669 (H. J. Yoo)

 Intermodulation
• Intermodulation • Third-order IM (TOI)
– when two signals with different frequencies are – if the difference between ω1 and ω2 is small, the
applied to a nonlinear system component at 2ω1 - ω2 and 2ω2 - ω1 appear in the
– the output in general exhibits some components vicinity of ω1 and ω2
that are not harmonics of the input frequencies – in typical two-tone test, A1 = A2

 = 
 ω

+   ω 

 
 = α

 + α    
 + α    


   
ω = ω
 ω  α

+ α  
 + α  
   ω


   
   
+ α
 + α   + α   
   ω 

   
= ω
± ω  α  
  ω
+ ω  
– Corruption of a signal due to intermodulation
+ α  
  ω
− ω  
between two interferers
 • a weak signal with two strong interferers
= ω
± ω   α  
    ω
+ ω  



+ α  
   ω
− ω  



= ω  ± ω
 α   
  ω  + ω




+ α   
 ω  − ω



Razavi98

ICE669 (H. J. Yoo)

 Third-order Intercept Point (IP3)
• Third-order intercept point (IP3) • Calculation of IP3 without extrapolation
– IIP3 and OIP3


 =   ω

+   ω 

     
 
 = α
+ α       ω

+ α
+ α       ω 

     
 
+ α     ω
− ω  
+ α      ω  − ω

+ ⋅ ⋅ ⋅
 

• Example
 
 α
>> α      α
  = α     
 
 = 
 
   = 

≈ − 
  
 = 


 
 α

∴   =
 α

• Relationship between A1-dB and AIP3


− 
 
 

 
 
= ≈ − 
  ! = 
≈ "
 

 Razavi98

ICE669 (H. J. Yoo)

 IP3 of Cascaded Nonlinear Stages
• Overall IIP3 of cascaded stages – As α1 increases, the overall IP3 decreases
• with higher gain in the first stage, the
second stage senses larger input levels,
thereby producing much greater IM3 product

– Since each stage has a narrow passband, out-of-

 

 = α

 + α   
 + α   

  band signals are heavily attenuated
#  
   
 = β


 + β
 
 + β  
 
   α

≈  + 
[
  
 = β
α

 + α    
 + α    
 ] 
   
  

[
+ β  α

 + α    
 + α    
 ] 

+ β [α 
 + α 
]
 – for three or more stages

   
 + α   
= α
β

 + α  β
+ α
α  β  + α
 β     
 + !   α
 α
 β


≈  +  +  + ⋅⋅⋅
 α
β
   
    
∴   =
 α  β
+ α
α  β  + α
 β   

≈ + +
 + ⋅⋅⋅
 
  
– for worst-case estimate,
(Powers & gains in real value, not in dB)
 α  β
+ α
α  β  + α
β 


=

   α
β
– For gain greater than 1, the IP3 of latter stage
 α  β  α 
becomes increasingly critical
= + + 



 
 β
  
Razavi98

ICE669 (H. J. Yoo)

 Adjacent Channel Power Rejection (ACPR)
• ACPR • Problem
– Adjacent channel power rejection – Lead to higher inband noise power
– Signal power leakage to the adjacent channel →
degrade adjacent channel (used by others) S/N
• Origin ratio
– Intermodulation

• CDMA ACPR
– Measure spectral power of the channel (1.23 MHz
Nonlinear
bandwidth)
System
– Measure upper and lower band edge (
885 KHz)

 

  of the next adjacent channel power (30 KHz

 

 bandwidth)


 
 
 [] =
 
 
 

– Alternative method


 [ ]
 ⋅ %

! 
$
=





PA
 

ICE669 (H. J. Yoo)

 Random Process (I)
• Time average vs ensemble average • Probability density function (PDF)

   = ())'   <  <  + 

– Gaussian (or normal) distribution

 −   −  
   = *(
σ π σ 
 
−   −  
  
<  <    =
σ π 
∫ *(
σ 


  − 
 +  
   = ∫
π 
*(


∫
+∞
< 
 >= & 


 = ∫ 
  
 →∞  −   −∞

 +   +∞
  
 = ∫   
  
<   
 >= &
 →∞  ∫−  
  

−∞ – AWGN: additive white Gaussian noise

– Q1: Is the time average measured today equal to – (ex) binary signal corrupted by noise
that measured tomorrow?
A: Not necessarily. If system is “stationary”, its
statistical properties are invariant to a time shift.
– Q2: Is the time average of a stationary process
equal to ensemble average?
A: Not always. But in most cases they are equal.
Razavi98

ICE669 (H. J. Yoo)

 Random Process (II)
 
• Power spectral density (PSD)      = &    
 →∞ 
– a Fourier transform of a random process is also a 
random process      = ∫ 
 *(−  π



– how much power the signal carries in a unit
bandwidth around f → characterized by power – Since Sx(f) is an even function of f for real x(t),
spectral density −  


∫− 

   
 + ∫    
 = ∫     

 

– in general, PDF and PSD have no relationship

• thermal noise: Gaussian PDF, white PSD
• flicker noise: Gaussian PDF, 1/f PSD

• Random signals in linear systems

– if a signal of Sx(f) is applied to a linear, time-
invariant system with transfer function H(s), then

    =     !   

!    = !  =  π 

Razavi98

ICE669 (H. J. Yoo)

 Noise
• Thermal noise • this noise is negligible at low frequency, but
– Brownian random motion of thermally agitated it can dominate at RF
charge carriers 
– generated in every physical resistors ' = &δ ∆ 
 =
%  
• modeled by a voltage source in series • δ~4/3 in long device
(Thevenin representation) or current source
– both drain and gate noise share a common origin,
in parallel (Norton representation)
they are correlated
' &∆
' = &∆  =  =

 
 = & ∆
 
• Shot noise
– purely reactive elements generate no thermal
– Gaussian white process associated with the
noise
transfer of charge across an energy barrier
– p-n junction, gate-channel
• Thermal noise in MOSFET
– drain current noise  =   ∆


 = γ
  ∆ • Flicker noise in MOSFET
– random trapping of charge at oxide interface
• γ~1 at zero VDS in long device, 2/3 in – modeled as a voltage source in series with gate
saturation, 2~3 for short-channel NMOS
– gate noise " %  "
 =

 
∆ ≈ ⋅ ω  ⋅  ⋅ ∆
• Thermal agitation of channel charge cause  #$  
fluctuation of channel potential. This couples
capacitively with gate terminal, leading gate – total flicker noise is integration from f1 to f2 → “DC”
noise infinities are not a practical problem Razavi98

ICE669 (H. J. Yoo)

 Input-Referred Noise
• Input-Referred Noise • Noise addition
– Noise voltages do not combine in linear:
superposition do not apply

• (Ex) MOS equivalent noise model

– require both Vn and In for adequate representation

– must take into account correlation between Vn
and In
– determine Vn by shorting input, and In by opening
input

• Correlation
– Noise voltages produced independently, 
%  ' =  

 %    (
 =  


• No relationship between instantaneous

 
voltage values 
+  = &  %  
 
 = 
+  +  
 "& "&
' =   = 
%   %  (

• C = 1: full correlation, C = 0: uncorrelated
– since Vn and In represent same noise mechanism,
they are correlated
Razavi98

ICE669 (H. J. Yoo)

 Noise Figure
• Definition • Noise figure calculation


+


+ +
 
)* =



+ + +
+

+ 
) )
+ )  )
=  = = + 
)
)
)


)
 ) )

= =

=

)    )  )

) ()
+ )  ) )
+ )  '  + ' +      '  + ' +     

= = =  )  (- ,   )* = =
)
)
)

'  ' 

• Noise Figure vs Noise Factor ' +     

= +
– Noise factor = SNRin/SNRout ' 
– Noise figure = 10log10(noise factor) – no correlation between VRS and Vn (or In), but
– Noise figure (or noise factor) measures the SNR correlation between Vn and In
degradation as a signal pass through a system – NF is typically specified for 1-Hz BW
• if a system has no noise, then NF = 1 = 0dB
• if input signal contains no noise, then SNRin ' +      ' 
)* =  +  )* =  

= ∞ and NF = ∞ (even though the system & 

 & 
has a finite internal noise) : this does not
occur in real world – function of source impedance RS
• in RF system, 50-Ω input and output
resistance are used
Razavi98

ICE669 (H. J. Yoo)

 Noise Figure Analysis
• Ex: resistors • Ex: feedback amplifier


'
= & ( !!   )α =
 +  
& ( !!   ) 
∴ )* = 
= +
   
  ⋅ &
  +   
– NF is minimized by maximizing RP
• condition for minimum noise figure does not
match with that for maximum power transfer


(cf) '


=

'
  
=  ⋅ '
= ⋅ '
 = '
 '

 +    +   ( +  ) )* =
 
&
  
∴ '


= '
 = α '
 
α =      ( + %     )
( +  )
 +  =  + %    +  + % 
 
     % 

Razavi98

ICE669 (H. J. Yoo)

 Noise Figure of Cascaded Stages (I)
• Cascaded noisy stages

'  '
α   '
 
= α
 '  +  
  + '
  )*

= =
  


 =α '[ +   + '  ]




 
 




⋅ &  α
 
α  α  ⋅ &

= α [ ' +   + '  ]
  α  
'

+ α      

+ '   
' 
 
   
=
 
 
 

 


α
 
α  ⋅ &
' 
= α   '
 
'

    

+ '   



   = +

α
= α  = α  =  α
 ⋅ & α
 
⋅ & 
 + 



+ 
  
 +  
'  +   
  + '
      + '   
  

= α

α   α  = +  ⋅   


&  
&

 
= α

.



=  
)* − 
)*

= )*
+


Razavi98

ICE669 (H. J. Yoo)

 Noise Figure of Cascaded Stages (II)
• Alternative derivation – For  stages, Friis equation,
– NF can be represented by available power gain
)* −  )* − 
)* = )*
+ +!+


!    −

 +


+
 
 =
 +  +
 
• NF of each stage is calculated with respect
• Available output power: power that circuit to the output impedance of previous stage
would deliver to a conjugate-matched load
• Available source power: power that source – the noise contributed by each stage decreases as
would deliver to a conjugate-matched circuit the gain preceding the stage increases
– first few stages are the most critical
– for cascaded noisy stages, – if a stage exhibits attenuation (loss), then NF of
the following circuit is amplified
α
 
'
 '

 =   =

 

 

∴  = α
 




    + '   
)* = )*
+   ⋅   


α

& 


    

+ '   
= )*
+ ⋅
α
 
 &


)* 

− 
= )*
+

Razavi98

ICE669 (H. J. Yoo)

 Noise Figure of Lossy Circuits
• Lossy passive circuits • Cascade of lossy filter and LNA
– passive devices attenuate desired signal, and
contribute noise

– Thevenin equivalent circuit for signal, )* 

 = $ = −




 +  +
  '     '  )* − 
$= = 
 =
 
)*

= )* 

 +
 +


+
  '  
 '  $−

= $ +  )* −  $ = $ ⋅ )*
  ' 
' 
= &
  = = $ + )* 
 

( + 
) 
'
   + 

 
&
– With overall NF estimation, filter, attenuator, etc,
'
(  + 
)
∴ )* = = that comes in front of active components, can be
 ⋅ &   '  

treated as a part of active components
  ⋅ &
 '
   + 

'
 

= =$
' 
– Passive reciprocal network: NF is same as loss
(not apply to active circuit that has insertion loss)
Razavi98

ICE669 (H. J. Yoo)

 Sensitivity
• Definition – If input is conjugate matched, then
– minimum signal level that a system can detect
& 
with acceptable signal-to-noise ratio at the output  = = & = −2&!01
 


– since =  × 
− 
3& , &(

)
 
 
)* = = ∴ 
   = %−2&!01 + )* + 
  $ + ) 
)
)

/ 
: input signal power per unit BW



+ +
/ : source resistance noise per unit BW
/ : function of bandwidth

 =  ⋅  ⋅ 
/ +(sensitivity) degrades with higher data
rates (BW increases)
– for flat channel,

 

=  ⋅ )* ⋅ )
⋅  Pin,min


   =   
+ )* 
+ )  
+ 
 
 SNRout

/ : minimum total input power over 

Input (referred) noise floor = output noise floor - gain

Razavi98

ICE669 (H. J. Yoo)

 Dynamic Range
• Definition

   %&$
, =

   %&$
4 
: maximum input level that the circuit can
tolerate
4 : minimum input level at which circuit
provides a reasonable signal quality
– differently quantified in different applications

• Spurious-free dynamic range (in RF) 4 

is obtained when = * (input noise floor)
4 : sensitivity    + *

   =
4 
: maximum input level in a two-tone test for 
which IM3 < noise-floor
 * = −2& + )* + 
 


−  
– since  = F + ),
  = 
 +
    + *

= 
 +  
=  
 +  *, = −  * + )  


 −  
 
 −  
    − * 
  = 
 + = = − ) 
  
   +  

∴ 
 = • Ex

4 )*5 5,&5

601)5
→*,5
Razavi98

ICE669 (H. J. Yoo)

 Impedance Transformation
• Impedance Transformation • Transform a resistance to a higher value
– transformer can transform impedance – capacitive divider
– high frequency transformer exhibits loss,
capacitive coupling between primary and
secondary, and even unwanted resonances

• RC circuits
  
– inductive divider 

≈  +    
 


– if Q is relatively high and band is narrow, then two  $ 



≈  +
  
circuits are equivalent when ZS = ZP  $ 
 
+(  =  + (  =   !! 
 
 +    
= • Transform a resistance to a lower value
  +   

∴    =     +    −     =

ω   


   ≈    ≡   >>      ≈
  (ω  )



 ≈   ≈ -  
 (ω )

Razavi98

ICE669 (H. J. Yoo)