Radio frequency circuits design techniques

First lecture is an introduction to the field of RF, we have not gone through any technical matter (see intro.pdf and intro_EE-426_ICT.pdf).

What is important to remember is the complexity and coexistence of many different standards. Coexistence and spectrum resource allocation with related regulations is a very important point. It's a wonder that up to 7 wireless standards coexist into our phones (NFC, BT/BLE, WiFi, GPS, 4/5G, UWB, basic SAT-COM). There are obviously much more (look at the FCC spectrum allocation between 3kHz to 300GHz)!

ICT has started its revolution 30y ago with www, then 25y 2G/BT/WiFi, 20y camera in our phones, 15y smartphone (app era), 10y BT low energy to prepare earbuds, smartwatch and wearables. ICT is the combo of computing and wireless enabled by integrated circuits and their continuous scaling (1E6 density improvements over 50y) and the cornerstone of the ongoing digitalization revolution (our data to the cloud). LLMs such as ChatGPT are likely another major breakthrough in ICT but all of the above have already completely revolutionized our lives. 

The Friis equation and the min sensitivity one presented at the end are key to remember as they dictate major architectural choices for any comm app. We will revise this at the beginning of next course!

 

During the first hour, we discussed typical numbers like 0 and 30dBm on 50 Ohm, sensitivity around -100dBm, propagation equation and a generic conceptual transceiver architecture with main building blocks and corresponding signal levels across the receiver chain. This is now summarized in the "2nd_lecture_concepts.ppt" file linked as the 1st file below.

We have gone through the first chapter of Enz course (notes included in this week pdf LC_resonant_circuits). In addition you have additional material (lecture notes Chapter 1 Passives, previous week and  Chapter 2, resonant circuits this week if you want to read it for yourself)

You should be familiar with Q definition, transformation between series/parallel equivalent, LC resonant parallel (used in load or shunt elements) and series (typically used along the signal path; an LNA uses both circuits), calculate the Q of a resonant circuit with lossy C and L. L and C do not change much when going from series to parallel or vice versa while RP=Q^2*Rs for Q high enough (Q=10, 1% error, Q=3, 10%).
Understand the link between Q and BW (1/Q). The LNA gain for example with be prop to the Q of the load but its BW will be 1/Q. At 2.4GHz, we have 80MHz BW, this is 3% hence Q should not exceed 30. When taking into account component spread, our resonance (1/LC) will get shifted and we miss the desired band. High-Q desired but be aware of the limits !

With what we learned today you should be capable of designing a simple VCO, calculate its consumption and phase noise. As well you have now almost the skills to design an LNA! We will run examples.

Exercises: Friis and range (in this week's folder, beware of the typo with respect to the exponent which should be negative) and series 1 (Q-factor) located in next week folder.  Corrections in this week's folder for Friis/range

Check out the summarizing ppt file located in previous week as first linked material to get a summary of last week's discussions on propagation, range and indicative voltages present in a transceiver. 

This week we have covered impedance matching. Lecture notes are to be found in last week's folder (LC_resonant_circuits, 2nd part on Z match, slides) or Chapter 3 (textbook) which you find below.

You may practice any Z-matching exercises that you find in both lecture notes. Or pick your own, for example at 2.4GHz, with Zin=50 Ohm and Zout= 1kOhm. Calculate the matching network with all types -inverted L, PI/T after you fix the Q (e.g. 10 or try 30 since BW=80MHz @ 2.4GHz), and with the widest possible BW with 2 cascaded inverted-L network. Implement and compare the transfer function of those circuits in LT Spice (see below) after modifying the example provided below which explains how to simulate a RLC series or parallel circuit. Try HP/LP version or mixed ones, with and without component reductions (e.g. the L/C in slide 28!) 
There are additional exercises below (exercises chapter 3 part 1 (inverted-L), exercises chapter 3 part 2 (pi-T))

You may compare the results with an ideal L and a real one. Search for Coilcraft or Murata on the web to find the equivalent models (look for high-Q series and choose the component size e.g. 01005, 201, 402, 805 or larger depending on the L value). 
RF Inductors | Coilcraft

First LAB was conducted during the exercises (see LABS tab after the per week headers). It's a Jupyter notebook prepared by Filippo using Python to model a BPSK transceiver that should be handed back! 

LECTURE

1st part

Smith Chart and Impedance matching (Chapter 3 pp 23-36).
You should be able to place any Z or Y on the chart, draw cst Q lines and implement L, pi, T or LL matching and calculate the value of the components from the chart

2nd part

Filter design Lecture notes Chapter 4 without the SAW filter section. 

You should know how to determine the filter order, use filter tables (no need to calculate analytically the components) and do the HP/BP transformation, modify the source, load impedance, simulate and compare different implementations!

 

EXERCISES

Cadence intro done see LAB section at the end of the semester plan. LT Spice as alternative but no support (easier more intuitive and a loads of example in the WEB).
It is expected you are able to simulate things with one or the other SW. Simulate resonant RLC and matching networks. 

Reuse examples from the lecture notes (calculated ones) and resolve on Smith chart!

See under next week: exercices_chapter3_part3 and solve the problems 

Check the URL for the online Smith chart simulator, you may add your series, shunt elements, adjust values to implement your matching, add cst Q line (see at the bottom). Watch out, the matching drawing goes from left to right (thus we are looking from the right towards the left). In the lecture note we start from the load (right) and add components towards the source (left). You should thus mirror things horizontally with the online tool (Z black box is that of the load) and you should end up on the Zs* location. Add spread to your components to see how it impacts the final result.   

LTSpice: folder with Z-matching added, copy and practice, compare L, pi and T and low-Q 2L matching networks frequency characteristics, implement some of the course example

1st hour: 

We saw BP filter last week. Today most RF filters are acoustic either SAW or BAW due to better manufacturing tolerances, sharp cut-off characteristics and reduced board space usage. BP_filter_summary gives you some high level view to compare BP and HP/LP filter response.

RF_filter_BAW_epfl presents some research work done in the past at EPFL/CSEM exemplifying co-design of MEMS & IC RF front-end.

The Qorvo whitepaper is to give you some hints on BAW acoustic filters. Currently the market is split 50/50 between SAW (surface acoustic wave) and BAW (bulk). No coexistence of radio standards without acoustic filters!

2nd hour:
MOS regime of operation, IC factor, small signal equivalent circuit with various conductances (G, S, D), noise. See lecture notes: MOS_Fundamentals from E. Vittoz. With this material you should be able soon to design your own active circuits and we will go through examples.

The 2nd lab to run on Cadence or LT-Spice is available after the weekly plan. It's typical of exam questions with calculation using EQs given in the lecture notes. Then you should simulate your results to verify your calculations.

There is also below the lab example done last year (also with some Cadence tip)
Exercices: implement examples on Cadence, passband, low-pass, high-pass, Chebyshev, Butterworth, with given frequency and input output resistance. Hints: make a parametric model so that you may quickly vary things. Check that your filter is robust against component variations and could be implemented with discrete components (search e.g. for murata or coilcraft RF inductors). 

1st hour: lecture notes "Noise made simple", NF, F definitions, equivalence to SNRo/SNRi , Friis equation for cumulated noise calculation, derivation of sensitivity equation: given SNRo and NF impact. 

2nd hour: F calculation for 2 resistors see word file, done in class! Redo it 1) after adding an impedance transformation network and changing RL to e.g. 1kOhm, what do you conclude?; 2) add a simple common source MOS transistor loaded with a resonant RLC network of your choice and calculate how F is affected when including the transistor thermal noise! Here we keep the 50Ohm physical resistor and make the assumption that any reactive component due to the MOS T is resonated out with an inductor.

2nd hour: also started the derivation of common source LNA input impedance to replace the 50Ohm physical impedance added above!

FALL BREAK

Recap on KCL (Kirchhoff's Current Law) with example on LNA output node, + when entering devices be it VDD or GND (no difference in AC), be careful with MOS to define proper sign (e.g. + on D and - on S when dVG but + on S and - on D when dVS. I_in used for Z_in calc is pos since on other side of = sign.   Condition for input match of common source, L-degenerated LNA design discussed on the board; 1 degree of freedom thus possible to add L1 in series with LNA to give an extra DOF. The Q of this matching network is very close to 1 with typ components thus wideband match but no filtering!
Went through Intermodulation and distortion lecture notes: didactical, concentrate on summary slides and system level rather than equations. Understand the difference between H3 and IM3 meas in a band limited (BP filter) system. Understand the cascaded IM3 calculation and difference with NF (which blocks degrade most NF and IP3). Now you know why 1dB of gain should be combined with 1dB of filtering to ensure coexistence of various standards e.g. on a phone!   
Exercises: Calculate the cumulated NF of the RX sketched below with and without the various filters. 
Exercises: finish the design of the LNA including MOS sizing and run it on Cadence, plot Re(Zin), Im(Zin) and verify that it goes to 50 Ohm and 0 Ohm respectively. At first pick a low frequency e.g. 100MHz so that components are big and parasitis negligible. Then repeat e.g. at 2.4GHz. Look at passives on the web (Coilcraft RF high-Q series) to get realistic component values. Safe to use >0.5pF, >10nH! See how adding L1 in series gives you more freedom. Understand what happen when a parasitic cap placed between input and ground affect the circuit (Smith Chart). You should know how to change your design slightly to get back to 50Ohm input!
In Cadence, we will show you next week how to simulate the NF so that you could compare it with calculations. 
Check the ISPEC ppt file. It explains how to do your sizing for IC=1 (always a good start), knowing gm or current
You may also practice the current conveyor LNA exercice to verify if you master the design technique.

We covered Transceiver architectures (see lecture notes).
The cumulated NF corrections are available below, practice the cumulated IP3 exercise on the same architecture with and without the filters (corrections also available directly). 
There is a nice paper on why IM2 and IM3 are important in 0-IF RX. Read it to improve your understanding!
A Cadence IP3 tutorial is available. Analyze your LNA or a single transistor!
You should now be able to analyze any few nodes circuit with KCL including NF calc! Another LNA example is provided. 

We have covered mixers (see attached lecture notes)

-you may search by yourselves about passive mixers and 25% duty cycle driven mixers

-implement simple and double-balanced mixers in Cadence of LT-Spice  

We have studied how to design VCOs
We have also covered the design of crystal oscillators

 

We have studied the implementation of VCOs using vco_ppt_lecture_notes. You also have alternative material from Enz/Dehollain for your own curiosity.

Exercises: derive the phase noise analytical expression as a function of the tank parameters (power of noise / signal power); hint: we consider that noise is split equally between phase and amplitude noise (1/2 factor)

We have covered PA design, see attached lecture notes!

Discussions on PLL system and blocks design, see attached slides!