manual: ECE1390 RF A2 - Mixer.pdf
repo: https://gitea.nodusk.me/jay/2025-meng-ece1390-rf
final report: ECE1390 RF A2 - Report.pdf

#🎯Todo

  • rewrite design procedure
  • write mixer note
  • write charge injection note
  • write clock feedthrough note

1 Specs

Metric Target Single-Balanced Double-Balanced
IIP3 > 5 dBm 5.9 dBm 5 dBm
Conversion Gain > -7 dB -6.9 dB -5.2 dB
Integrated Output Noise < 11 7.9 11.2
Mixer Power 0.66 mW 1.68 mW
Local Oscillator Power () -2 dBm -2 dBm
  • Direct Conversion Mixer for WiFi 802.11a (WLAN) Receiver
  • Center Frequency (): 5.4 GHz
  • IF: 20 MHz
  • Output Load: 110 fF
  • Noise Integration BW: 10kHz - 10MHz
  • Singled-Balanced and Double-Balanced topology
  • PDK: GF 22nm FDSOI
    • mosfets: slvtnfet_rf_b and slvtpfet_rf_b from cmos22fdsoi_rf

2 References

ECE1390 RF L6 - Mixer
RF Microelectronics

3 Research

source: RF Microelectronics - Chapter 6, p.337

A Mixer translates a signal to a different Frequency by multiplying it with a single tone. In an RF Receiver (RX), and RF signal is multiplied with a local local oscillator (LO) to produce a baseband or Intermediate Frequency (IF) signal. In an RF Transceiver (TX), an IF signal is multiplied with a LO to produce a RF signal. IF is the difference between Center Frequency () and LO frequency (), .

Pasted image 20251116023006.png#invert
RF Microelectronics, p.338

The mixer above uses a single-ended RF input and LO, leading to an inefficient LO, where the RF input is discarded for half the time. A more efficient topology is a single-balanced mixer, where two differential switches are used that commutate (toggle) the RF signal between two output paths. This topology leads to two main benefits; conversion gain is doubled compared to single-ended topology, and Charge Injection from the switches is eliminated. Though LO-IF Clock Feedthrough remains an issue. A double-balanced mixer solves this by connecting two single-balanced mixers such that they cancel clock feedthrough at IF port without affecting the IF signal.

Pasted image 20251116025217.png#invert
RF Microelectronics, p.348

Pasted image 20251116040309.png#invert
RF Microelectronics, p.349

The above are passive mixers, where mixing is done without any additional power (apart from and ). Conversion gain with the double-balanced passive mixer is -4dB. Active mixers integrate gain and mixing in a single stage with a Transconductance input. These also come in single-balanced and double-balanced forms, where the gain for both is given by

Pasted image 20251116124702.png#invert
RF Microelectronics, p.369

Pasted image 20251116133414.png#invert
RF Microelectronics, p.370

4 Design

single-balanced mixer:
single-balanced-mixer-sch.svg > invert

double-balanced mixer:
double-balanced-mixer-sch.svg > invert

In both topologies, the load resistor () is in parallel with the output capacitor (). This forms a low pass filter at the IF port whose 3 dB bandwidth must be designed to be larger than IF frequency (20 MHz).

 

To begin the design, some rough initial assumptions must be made for the overdrive voltage () for the mixer transistors, here 0.15V is a safe choice. Output voltage swing is limited to in single-balanced, and in the double-balanced. The can use the latter for both to simplify the designs. Maximum must be decided according to this voltage to ensure the transistors stay in saturation.

 

Assuming = 500 , the maximum is quite large, showing that there is more than enough voltage headroom for this design. To minimize Thermal Nosie, can be set lower. 100 is be used for now. The required to meet conversion gain specification (-7 dB) can now be determined by the following

To ensure a good balance between gain and linearity, a of 0.15 is selected for . Below the required and is calculated to get the desired .

 
 

The best performance was achieved with mixing switches set to a low . This doesn't align with methodology rules where a low results better gain but worse noise and linearity. #🎯Todo Need to revisit these simulations.

Both the single- and double-balanced mixers used the same input and switching devices, as well as the same . needed to be increased to achieve the desired gain and linearity.

Metric Single-Bal. Double-Bal.
Input Devices 10/20nm 10/20nm
Switching Devices 30/20nm 30/20nm
100 100
588 880
3/20nm 4.8/20nm
80/20nm
150 110
Bias filter , 20 k, 100 fF

5 Simulations

5.1 Testbench

testbench-sch.svg > invert

Pasted image 20251119155106.png

Pasted image 20251119155013.png

PORT_RF Setup

  • resistance: 50 Ohms
  • source type: sine
    • frequency name 1: FRF
    • frequency: frf Hz
    • amplitude 1 (dbm): prf
  • display small signal params: select
    • pac magnitude (Vpk): 1 V
    • ac magnitude (Vpk): 1 V

PORT_LO Setup

  • resistance: 50 Ohms
  • source type: sine
    • frequency name 1: FLO
    • frequency: flo Hz
    • amplitude 1 (dbm): plo

PORT_IF Setup

  • resistance: 50 Ohms
  • source type: dc

Variables

  • flo: 5.4G
  • frf: flo+20M
  • plo: 0
  • prf: -50
  • vlo_cm: 0.75
  • vrf_cm: 0.55
  • pacmagdb: prf

5.2 Conversion Gain

  • set PORT_RF source type to dc.
  • set PORT_RF pac magnitude (Vpk) to 1.

Analyses Setup

  • analysis: pss
    • fundamental tones: PORT_LO
    • beat frequency: auto calculate
    • output harmonics: number of harmonics -> 2
    • accuracy: conservative
    • run transient: yes
    • stop time: 0.3n
    • sweep: checked
      • variable name: plo
      • start: -20
      • stop: 40
      • sweep type: linear
      • number of steps: 20
  • analysis: pac
    • sweeptype: default
      • input frequency sweep range: Single-Point
      • freq: frf
    • maximum sideband: 2

Direct Plot Form

  • analysis: pac
    • function: Voltage
      • select: Net
      • sweep: variable
      • modifier: dB20
      • output harmonic: -1 20M
      • select output net (VIF) to plot

5.3 Noise

  • set PORT_RF source type to dc.
  • set PORT_RF pac magnitude (Vpk) to 1.

Analyses Setup

  • analysis: pss
    • fundamental tones: PORT_LO
    • beat frequency: auto calculate
    • output harmonics: number of harmonics -> 10
    • accuracy: conservative
    • run transient: yes
    • stop time: 0.3n
    • sweep: checked
      • variable name: plo
      • start: -20
      • stop: 40
      • sweep type: linear
      • number of steps: 20
  • analysis: pnoise
    • Sweeptype: absolute
    • start: 10k
    • stop: 10M
    • sweep type: linear
    • number of steps: 200
    • maximum sideband: 10
    • noise figure: selected
      • output: voltage
      • positive output node: /VIF
      • negative output node: /GND
      • input source: port
      • input port source: /PORT_RF
      • reference side-band: Enter in field -> -1

Direct Plot Form

  • analysis: pnoise
    • function: Output Noise
    • units: V/sqrt(Hz)
    • plot

5.4 IIP3

  • set PORT_RF source type to sine.
  • set PORT_RF pac magnitude (dBm) to pacmagdb.

Analyses Setup

  • analysis: qpss
    • fundamental tones:
      • 1 FLO flo 5.4G Large 3 PORT_LO
      • 2 FRF frf 5.42G Moderate 3 PORT_RF
    • accuracy: moderate
    • tstab: 0.5n
    • sweep: checked
      • variable name: prf
      • start: -70
      • stop: 15
      • sweep type: linear
      • number of steps: 20
  • analysis: qpac
    • sweeptype: default
      • input frequency sweep range: Single-Point
      • freq: frf+100k
    • maximum sideband: 2

Direct Plot Form

  • analysis: qpss
    • function: Compression Point
      • select: Port (fixed R(port))
      • gain compression (dB): 1
      • input power extrapolation point (dBm): -70
      • input referred 1dB compression
        • 1st order harmonic: 20M -1 1
        • select output port to plot
  • analysis: qpac
    • function: IPN Curves
      • select: Port (fixed R(port))
      • circuit input power: variable sweep
      • prf: -60
      • input referred IP3
        • order: 3rd
        • 1st order harmonic: 20.1MM -1 0
        • 3rd order harmonic: 19.9M 1 -2
        • select output port to plot

Results

single-balanced mixer:
single-balanced-mixer-sim.png

double-balanced mixer:
double-balanced-mixer-sim.png