RF Mixers: EECS 242

EECS 242:
RF Mixers
UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

Mixers
The Mixer is a critical component in communication

circuits. It translates information content to a new
frequency.
Information
Mixer
PSD

Why use a mixer (transmit side)?
1)  Translate information to a frequency appropriate
for transmission
Example: Antennas smaller and more efficient at
high frequencies
2)  Spectrum sharing: Move information into separate
channels in order to share spectrum and allow
simultaneous use
3)  Interference resiliance
Geographic map
1,2 of cell sites
3,4
1,2

Why use mixer in the receiver?
RF band
Q of filter
Desired
channel
Bandpass filter at ωo
requires a high-Q for
narrowband signals
Ch.1 2 3 4 5
Δf ~ 200 kHz (GSM)
High Q

Mixers in Receivers (cont)
High Q ⇒ Insertion Loss
Filter center frequency must change to select a given
channel ⇒ tunable filter difficult to implement
Mixing has big advantage! Translate information down
to a fixed (intermediate frequency) or IF.
1 GHz ⇒ 10 MHz: 100x decrease in Q required
Don’t need a tunable filter
High Q channel filter
IF
Issue: Mixer has
high noise factor
Superheterodyne receiver architecture

Mixers Specifications
  Conversion Gain: Ratio of voltage (power) at output

frequency to input voltage (power) at input frequency
  Downconversion: RF power / IF power
  Up-conversion: IF power / RF power
  Noise Figure
  DSB versus SSB
  Linearity
  Image Rejection
  LO Feedthrough
  Input
  Output
  RF Feedthrough

Mixer Implementation
We know that any non-linear circuit acts like a mixer
Two tones f(x)

Non-linear
2nd order IM
ω1 , ω2

Squarer Example
x x2 y
DC & second harmonic Desired mixing
Product component:
What we would prefer:
LO
IF
RF
A true quadrant multiplier with good dynamic range is difficult to fabricate

LTV Mixer
LTI No new frequencies
LTV New tones in output
Example: Suppose the resistance of an element is modulated

harmonically

Time Varying Systems
In general, any periodically time varying system can achieve
frequency translation
consider n=1 plus n=-1

Desired Mixing Product
Output contains desired signal (plus a lot of other signals)

→ filter out undesired components

Convolution in Frequency
Ideal multiplier mixer:
p(t)
periodic input
y(t)
input x(t)

Convolution in Frequency (cont)
X(f)
X(f) peaks at fRF
f
fRF
Translated spectrum peaks:

Y(f)
n=1 n=2 n=3 f
Input spectrum is translated into multiple “sidebands” or

“image” frequencies
⇒ Also, the output at a particular frequency originates from
multiple input frequency bands
How Low can you LO?
Take the simplest mixer:
output IF IF LO1 RF LO2

x(t)
Low side injection High side injection
Side note:
Which LO frequency to pick? LO1 or LO2?
Channel spacing
No. of channels
Tuning range: ⇒ fLO larger implies smaller tuning

range

Image Problem
Back to the original problem:
RF LO IMAGE RF LO IMAGE
Question: Why filter before mixer in spectrum analyzer?

Answer: Image rejection Image reject filter
IF
Channel selection
Image reject filter
LNA
LO
Receiver architecture is getting complicated…

Origin of Image Problem
If we could multiply by a complex exponential, then image

problem goes away…
IF frequency
High side injection
(Low side injection) Image Freq.

Review of Linear Systems and PSD
Average response of LTI system:

Average Value Property
“DC gain”

Output RMS Statistics
Recall the definition for the autocorrelation function

Autocorrelation Function
is a real and even function of ω

since is a real and even function of τ

Autocorrelation Function (2)

Average Power in X(t)
Consider x(t) as a voltage waveform with total average
power . Let’s measure the power in x(t) in the band
0<ω<ω1.
Ideal LPF
+ +
- -
The average power in the frequency range 0<ω<ω1 is now
W/radian
W/Hz

Average Power in X(t) (2)
Generalize: To measure the power in any frequency range

apply an ideal bandpass filter with passband ω1< ω<ω2
The interpretation of φxx as the “power spectral

density” (PSD) is clear

Spectrum Analyzer
A spectrum analyzer measures the PSD of a signal
Poor man’s spectrum analyzer:
Wide
dynamic Sharp filter
range mixer
vertical
Phase Sweep
noise VCO CRT
generation horiz.
Linear wide
tuning range

EECS 242:
Current Commutating
Active Mixers

Balanced Mixer
  An unbalanced mixer has a transfer function: Has “DC”
  which contains both RF, LO, and IF

  For a single balanced mixer, the LO signal is
“balanced” (bipolar) so we have No “DC”
Has “DC”
  As a result, the output contacts LO but no RF component

  For a double balanced mixer, the LO and RF are balanced
so there is no LO or RF leakage

Noise in an Ideal Mixers
Consider the simplest ideal multiplying mixer:
RF IF Noise
LO
IF RF LO IM
•  What’s the noise figure for the conversion process?
•  Input noise power due to source is kTB where B is
the bandwidth of the input signal
•  Input signal has power Ps at either the lower or
upper sideband

Noise in Ideal Mixers
•  At the IF frequency, we have the down-converted
signal G⋅Ps and down-converted noise from two
sidebands, LO - IF and LO + IF
IF RF LO
For ideal mixer, G=G’=G’’
For a real mixer, noise from multiple sidebands can fold

into IF frequency & degrade NF

Noise in CMOS Current Commutating Mixer
(After Terrovitis, JSSC)
I1 I2
M1 M2
LO Assume is is small relative to IB and
perform Taylor series expansion
RF M3
vx
-vx
+1
All current M2 Both on

through M1

Noise in Current Commutating Mixers
M1 M2
i1 i2
is
Note that with good device matching

Expand p1(t) into a Fourier series:
Only odd coefficients of p1,n non-zero

Single Balanced Mixer
RL IF Assume LO signal strong so that

current (RF) is alternatively sent
Switching to either M2 or M3. This is
+ Pair equivalent to multiplying iRF by ±1.
LO
- RF current
Transconductance
RF M1 stage (gain)
Period waveform with period = TLO

Current Commutating Mixer (2)
g(t) = square wave =
Let
gain
LO-RF isolation good, but LO signal appears in

output (just a diff pair amp).
Strong LO might desensitize (limit) IF stage (even
after filtering).

Double Balanced Mixer
ID1 ID2
•  LO signal is rejected up
to matching constraints
•  Differential output
+ 1 2 3 4 + removes even order
LO LO non-linearities
- -
•  Linearity is improved:
Half of signal is
processed by each side
Transconductance •  Noise higher than
single balanced mixer
since no cancellation
occurs

Common Gate Input Stage

Gilbert Micromixer
  The LNA output is often single-ended. A good balanced RF

signal is required to minimize the feedthrough to the output. LC
bridge circuits can be used, but the bandwidth is limited. A
transformer is a good choice for this, but bulky and bandwidth is
still limited.
  A broadband single-ended to differential conversion stage is
used to generate highly balanced signals. Gm stage is Class AB.
Active and Passive Balun

Bleeding the Switching Core
  Large currents are good for the gm stage (noise,

conversion gain), but require large devices in the switching
core  hard to switch due to capacitance or requires a
large LO (large Vgs-Vt)
  A current source can be used to feed the Gm stage with
extra current.
Current Re-Use Gm Stage

Single, Dual, and Back Gate

Rudell CMOS Mixer
  Gain programmed using current through M16 (set by

resistance of triode region devices M9/M10)
  Common mode feedback to set output point
  Cascode improves isolation (LO to RF)
Passive Mixers/Sampling

Sub-Sampling Mixers

Triode Region Mixer

Improved Linearity
LO, HIGH M2 || M3
Cascode To improve M1, apply
Amp local series feedback
M1
RF
Provide input matching and

feedback RF
Zs
⇒ No DC headroom sacrificed

Recap: CMOS Mixer Operation
I1 I2
M1 M2
LO
RF M3
Periodic
Fourier Series expansion

References
  Noise in current-commutating CMOS mixers
Terrovitis, M.T.; Meyer, R.G.;
Solid-State Circuits, IEEE Journal of
Volume 34, Issue 6, June 1999 Page(s):772 - 783
  Intermodulation distortion in current-commutating
CMOS mixers Terrovitis, M.T.; Meyer, R.G.;
Solid-State Circuits, IEEE Journal of
Volume 35, Issue 10, Oct. 2000 Page(s):1461 – 1473
  A systematic approach to the analysis of noise in mixers
Hull, C.D.; Meyer, R.G.;
Circuits and Systems I: Fundamental Theory and
Applications, IEEE Transactions on [see also Circuits and
Systems I: Regular Papers, IEEE Transactions on]
Volume 40, Issue 12, Dec. 1993 Page(s):909 - 919

RF Mixers: EECS 242

Uploaded by

Copyright:

Available Formats

RF Mixers: EECS 242

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

RF Mixers: EECS 242

Uploaded by

Copyright:

Available Formats

EECS 242:

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

The Mixer is a critical component in communication

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

Δf ~ 200 kHz (GSM)

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

Superheterodyne receiver architecture

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

 Conversion Gain: Ratio of voltage (power) at output

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

Two tones f(x)

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

DC & second harmonic Desired mixing

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

LTI No new frequencies

LTV New tones in output

Example: Suppose the resistance of an element is modulated

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

consider n=1 plus n=-1

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

Output contains desired signal (plus a lot of other signals)

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

X(f) peaks at fRF

Translated spectrum peaks:

n=1 n=2 n=3 f

Input spectrum is translated into multiple “sidebands” or

Take the simplest mixer:

output IF IF LO1 RF LO2

Tuning range: ⇒ fLO larger implies smaller tuning

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

Question: Why filter before mixer in spectrum analyzer?

Image reject filter

Receiver architecture is getting complicated…

If we could multiply by a complex exponential, then image

(Low side injection) Image Freq.

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

Recall the definition for the autocorrelation function

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

is a real and even function of ω

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

The average power in the frequency range 0<ω<ω1 is now

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

Generalize: To measure the power in any frequency range

The interpretation of φxx as the “power spectral

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

Poor man’s spectrum analyzer:

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

 which contains both RF, LO, and IF

 As a result, the output contacts LO but no RF component

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

Consider the simplest ideal multiplying mixer:

UC Berkeley EECS 242 Copyright © Prof. Ali M Niknejad

For a real mixer, noise from multiple sidebands can fold

  Conversion Gain: Ratio of voltage (power) at output

  which contains both RF, LO, and IF

  As a result, the output contacts LO but no RF component

  The LNA output is often single-ended. A good balanced RF

  Large currents are good for the gm stage (noise,

  Gain programmed using current through M16 (set by