Direct Conversion Mixer for WiFi 802.11a

#class/ECE1388 #📦project #analog #rf #mixer

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

Assets

Manual: ECE1390 RF A2 - Mixer.pdf
Report: ECE1390 RF A2 - Report.pdf
Repository: https://gitea.nodusk.me/jay/2025-meng-ece1390-rf

References

Specifications

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 $μ V_{r m s}$	7.9 $μ V_{r m s}$	11.2 $μ V_{r m s}$
Mixer Power		0.66 mW	1.68 mW
Local Oscillator Power ( $P_{L O}$ )		-2 dBm	-2 dBm

Direct Conversion Mixer for WiFi 802.11a (WLAN) Receiver
Center Frequency ( $f_{c}$ ): 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

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 ( $f_{R F}$ ) and LO frequency ( $f_{L O}$ ), $I F = | f_{R F} - f_{L O} |$ .

!400
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.

!400
RF Microelectronics, p.348

!200
RF Microelectronics, p.349

The above are passive mixers, where mixing is done without any additional power (apart from $P_{R F}$ and $P_{L O}$ ). Conversion gain with the double-balanced passive mixer is $2 / π \approx$ -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

A_{v} = \frac{2}{π} g_{m_{1}} R_{D}

!200
RF Microelectronics, p.369

!400
RF Microelectronics, p.370

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 ( $R_{d}$ ) is in parallel with the output capacitor ( $C_{l o a d}$ ). 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).

f_{- 3 d B} = \frac{1}{2 π R_{D} C_{l o a d}}

fif = 20 MHz
Cload = 110 fF

Rdmax = 1/ (fif * 2 * pi * Cload)

To begin the design, some rough initial assumptions must be made for the overdrive voltage ( $V_{o v}$ ) for the mixer transistors, here 0.15V is a safe choice. Output voltage swing is limited to $V_{D D} - 2 V_{o v}$ in single-balanced, and $\approx V_{D D} - 3 V_{o v}$ in the double-balanced. The can use the latter for both to simplify the designs. Maximum $R_{d}$ must be decided according to this voltage to ensure the transistors stay in saturation.

R_{d, m a x} = \frac{2 V_{I F, p p}}{I_{b i a s}}

Vov = 0.2 V
Vdd = 0.9 V
Id1 = 500 uA

Vpp = Vdd - (3 * Vov)
Rd = 2 * (Vpp / Id1)

Assuming $I_{b i a s}$ = 500 $μ A$ , the maximum $R_{d}$ is quite large, showing that there is more than enough voltage headroom for this design. To minimize Thermal Nosie, $R_{d}$ can be set lower. 100 $Ω$ is be used for now. The required $g_{m}$ to meet conversion gain specification (-7 dB) can now be determined by the following

A_{v} = \frac{2}{π} g_{m_{1}} R_{D}

Rd = 100 ohm
gain = -7
Av = 10^(gain/20) V/V
gm1 = Av / (Rd * 2/pi)

To ensure a good balance between gain and linearity, a $V_{o v}$ of 0.15 is selected for $M_{1}$ . Below the required $I_{D 1}$ and $W_{1}$ is calculated to get the desired $V_{o v}$ .

gm1 = 7 mS
vov = 0.15 V
id = vov*gm1/2


l = 20 nm
kn = 480 uA/(V^2)
W =  (l * 2 * id) / (kn * vov^2)

The best performance was achieved with mixing switches set to a low $V_{o v}$ . This doesn't align with $\frac{g_{m}}{I_{D}}$ methodology rules where a low $V_{o v}$ 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 $P_{L O}$ . $I_{D}$ needed to be increased to achieve the desired gain and linearity.

Metric	Single-Bal.	Double-Bal.
Input Devices $M_{1, 4}$	10 $μ m$ /20nm	10 $μ m$ /20nm
Switching Devices $M_{2, 3, 5, 6}$	30 $μ m$ /20nm	30 $μ m$ /20nm
$R_{D}$	100 $Ω$	100 $Ω$
$I_{D 1}$	588 $μ A$	880 $μ A$
$M_{b}$	3 $μ m$ /20nm	4.8 $μ m$ /20nm
$M_{t}$		80 $μ m$ /20nm
$I_{r e f}$	150 $μ A$	110 $μ A$
Bias filter $R_{b}$ , $C_{p}$	20 k $Ω$ , 100 fF

Simulations

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

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

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

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:
!400

double-balanced mixer:
!400