RF / microwave circuit design

Academic projectIn progress

RF Frequency-Conversion Chain

Academic project - chain design in Keysight ADS, roughly 2.5 GHz to roughly 5.2 GHz

Identification

Type
Academic project, ENSEEIHT / INP Toulouse, SYSCom track
Tool
Keysight ADS
Input
Around 2.5 GHz
Output
Around 5.2 GHz
Function
Amplification, filtering, mixing, LO generation
Reference impedance
50 ohm throughout

Target specification

Conversion gain
Around 18 dB
Noise figure
Less than or equal to 6 dB
IIP3
Greater than 20 dBm
Rejection
Around 50 dBc
Stability
Unconditional over the band of interest
Status
Targets, not results. No simulated or measured figure is published yet.
01

Academic project

Academic project, coursework inside the SYSCom track at ENSEEIHT. The specification is the target the design is held to; no simulated or measured figure is published yet.

02

The design problem

Take a signal at around 2.5 GHz and deliver it at around 5.2 GHz, through a chain that amplifies, filters, mixes and generates its own local oscillator, while meeting gain, noise, linearity and rejection targets at the same time. The targets are the design: each one is easy alone and they pull against each other.

  • Gain against noiseNoise figure is set mostly by the first stage, so the front end wants gain. Gain early costs linearity later.
  • Linearity against gainIIP3 above 20 dBm constrains how hard the chain may be driven, and every stage of gain moves the whole chain closer to compression.
  • Rejection against complexityAround 50 dBc on the image and the spurious products is a filter problem, and every filter costs insertion loss that the gain budget has to absorb.
  • StabilityAn amplifier that oscillates has no specification at all. Stability is checked over the band, not only at the design point.
03

Chain and functions

  • AmplificationLow-noise amplification at the input, where the noise figure is decided, and gain stages placed so that the linearity budget survives.
  • FilteringBand selection before the mixer, image rejection around it, and output filtering of the conversion products.
  • MixingThe frequency-conversion stage itself, with its conversion loss, isolation and the spurious products it generates.
  • LO generationThe local oscillator that drives the mixer, and its own purity, drive level and leakage.
  • Matching50 ohm matching at every interface, because a mismatch inside the chain is not just a loss, it changes the response of the stages around it.
04

Analysis to be carried out

Which analysis is run, and what each one settles.

Analyses and what they decide
AnalysisSettles
S-parameterMatching, gain, isolation and the stability factors across the band
NoiseChain noise figure, and which stage actually sets it
Harmonic balanceSteady-state nonlinear behaviour: conversion, compression, harmonics and mixing products
Two-tone nonlinearIIP3 and the third-order intermodulation products
Budget analysisGain, noise and linearity accumulated stage by stage along the chain
\[F_{\text{tot}} = F_1 + \frac{F_2-1}{G_1} + \frac{F_3-1}{G_1G_2} + \cdots\]
\[\frac{1}{\mathrm{IIP3}_{\text{tot}}} \approx \frac{1}{\mathrm{IIP3}_1} + \frac{G_1}{\mathrm{IIP3}_2} + \frac{G_1G_2}{\mathrm{IIP3}_3} + \cdots\]

The two cascade relations that make the chain a single design problem rather than five independent ones. Friis pushes noise responsibility to the front, the linearity cascade pushes it to the back, and the gain distribution is what settles the argument between them.