ECE 6361:Mixer Design Review

2.4 GHz Single Balanced Mixer

Jan Brosi Vasileios Iliopoulos

Project Objectives I

IF Frequency 140MHz

IF Bandwidth 5 MHz

RF Frequency 2400-2485 MHz

LO Frequency 2260-2345 MHz

LO Power 8 dBm (max)

Conversion Loss >-9.5 dB

RF Power -10 dBm (max)

LO-RF Rejection <-20dB

LO-IF Rejection <-30 dB

IF Input power 0 dBm (max)

Spurious Signals at RF port

Relative to desired RF with 8 dBm LO power

1000-2120 MHz <-30dBc

2325-2400 MHz <-40 dBc

2400-2485 MHz <-50 dBc

2485-2900 MHz <-40 dBc

Project Objectives II

Board:4-layer PPE (1.7x2.6 inches)Schottky diodes: dual series connected

Physical Construction I

27mm

Reduced-size rat-race balun with coupled lines on different layers

Radial stubs for matching at RF/LO frequency

Lumped elements for matching at IF and for DC-return

Physical Construction II

Board parameters: εr=2.85 ±0.5 tanδ=0.002-0.003 core thickness=1mm resin thickness=50μm copper thickness=12μm

Diodes: Zetex ZC2812ECT IS=9.5nA, RS=16.3Ohm N=1.27, Cj0=1.1pF

Lumped Elements Toko 2021 series chip inductors Panasonic chip capacitors

Simulation with Agilent ADS using a combination of Multilayer & Microstrip models

Diode Model: ADS P-N diode model with parameters from manufacturer’s data sheet

Inclusion of parasitic elements (bends, steps, T- junctions, vias, pads, parasitics of lumped elements).

Balun design with S-parameter simulation and gradient method optimization for best performance

Mixer harmonic balance simulation with 8 orders and gradient method optimization

Model Description & Simulation

Simulation Results

RFfreq=2.485GHzdBm(V_RF)=-8.614

LOfreq=2.345GHzdBm(V_RF)=-29.697

Spur2freq=2.065GHzdBm(V_RF)=-60.577

Spur1freq=2.765GHzdBm(V_RF)=-50.357

RFfreq=2.485GHzdBm(V_RF)=-8.614

LOfreq=2.345GHzdBm(V_RF)=-29.697

Spur2freq=2.065GHzdBm(V_RF)=-60.577

Spur1freq=2.765GHzdBm(V_RF)=-50.357

2.41.4 3.0

-70

-60

-50

-40

-30

-20

-10

0

-80

10

freq, GHz

dBm

(V_R

F)

RF

LO

Spur2Spur1

Simulated circuit meets all specs

Measurement of conversion loss in down-conversion using the frequency offset function of HP89441 Network analyzer

The LO port is fed by E4432 Signal Generator with 8.5dBm input power (assuming 0.5 dB cable loss)

Use of a low-pass filter at the input port of the Network analyzer to ensure better phase-locking

The filter response was calibrated out Accuracy of the measurement:

Signal generator: ± 0.5dB Network analyzer: ± 0.2dBTotal accuracy: ± 0.7dB

Conversion Loss Measurement

LO Frequency Max. Conv. Loss over IF BW

Uncertainty Range

Specs met

2.26 GHz 8.4 dB 7.7-9.1 dB Yes

2.3025 GHz 8.8 dB 8.1-9.5 dB Yes

2.345 GHz 9.5 dB 8.8-10.2 dB No

Conversion Loss Measurement Results

Conversion loss increases with increasing LO-frequency

Measurement of spurious signals and LO-RF isolation for up-conversion using HP8994E Spectrum Analyzer.

Measurement of LO-IF isolation for down-conversion Sweep over the LO frequency range Accuracy of the spurious signals measurement

Relative accuracy of spectrum analyzer: ±0.5dB

Accuracy of the LO-RF isolation measurement Absolute accuracy of spectrum analyzer: ±1dB Signal generator: ±0.5dB

Total accuracy: ± 1.5dB

Spurious & Isolation Measurement

Spurious & Isolation Measurement Graph

LO RF

LO+2IF

LO+3IF

LO+4IF

LO-IF

LO-2IF

LO-3IF

SignalWorst value in

frequency rangeUncertainty range Spec

SpuriousLO-3·IF

-29.7dBc -29.2 to -30.3dBc -30dBc

SpuriousLO-2·IF

-20.3dBc -19.8 to -20.8dBc -30dBc

SpuriousLO +2·IF

-27dBc -26.5 to -27.5dBc -40dBc

SpuriousLO + 3·IF

-30.9dBc -30.4 to -31.4dBc -40dBc

SpuriousLO+4·IF

-41.7dBc -41.2 to -42.3dBc -40dBc

Spurious Measurement Results

SignalWorst value in

frequency rangeUncertainty range Spec

IsolationLO-RF

-12dB -10.5 to –13.5dBc -20dB

IsolationLO-IF

-21.5dB -20 to -23dBc -30dB

Isolation Measurement Results

Circuit assembled as designed has bad performance and does not meet any of the specs

Slight change of inductor & capacitor values and interchange of their position increased mixer performance

The mixer seems to be shifted to lower frequencies, it works much better at fLO=1.9 GHz

Thus it was tried to reduce the size of the mixer Coupled lines were shortened by use of new vias Line edges were smoothened with copper material to

reduce length

Resulting circuit had best performance and was used for measurements

Changes made to the Circuit

Comparison to Simulations

The measurements don’t match with the simulations

Trying to fit ADS-model to measurements by change of dielectric constant, distance between layers, coupled line offset, length and width of lines, stub length and width, vias and pads, losses and diode model.

This should account for production tolerances and previously not considered effects

Changes didn’t have a significant effect on the simulated mixer performance

Indication that ADS multilayer-model is not accuarate

A single-balanced mixer using a reduced-size rat-race balun with mulilayer coupled-lines was designed, simulated, fabricated and measured

Great differences between model and measurements of the produced mixer occurred

Performance could be increased by change of lumped elements and reducing the size of lines

Specifications for Conversion loss are almost met, for Isolation and some spurious signals not

Re-simulation was not successful

Summary & Conclusions I

Summary & Conclusions II

The ADS multilayer-model does not seem to work well in our case, another software or model should be tried

Board parameters like layer thickness and dielectric constant should be reevaluated

The mixer works much better at lower frequencies, the balun seems to be downshifted

For the next production cycle, the line lengths should be decreased to account for that