Having little exposure to Industry, academic researchers wonder if they are designing the transmitter for the right metrics? Are transmitter specs all about output power and linearity? And besides what metrics are used by industry to characterize linearity? If you have such questions, you have landed at correct place. Take a look at the transmitter specs used by industry.
You knew it. Output power matters.
Almost all the transmitters you see today are single side band, that means you want to transmit either upper side band (\(f_{lo}+f_{bb}\)) or lower side band (\(f_{lo}-f_{bb}\)) but not both. Residual side band, measured in dBc, is a measure of how much other sideband is rejected. Say, \(f_{lo}+f_{bb}\) is your desired sideband and you have spec of -60dBc residual sideband band, that means \(f_{lo}-f_{bb}\) should be 60dB lower than \(f_{lo}+f_{bb}\). To ge to such levels though, IQ calibration is required.
Something that always leaks. LO leakage, measured in dBc, means how much your LO signal is lower compared to your desired signal. LO signal is present at TX output either by direct leakage or through different mechanisms that produce LO leakage like up-conversion of DC offset or downconversion of \(2f_{lo}\) components.
This metric is not readily available in textbooks to learn about yet it is a very important one because it is only \(4f_{bb}\) away from your desired signal, hence called 4FMOD. Say, \(f_{lo}+f_{bb}\) is your desired signal, then anything at \(f_{lo}-3f_{bb}\) frequency would be called CIM3 or P4FMOD. Similarly, if \(f_{lo}-f_{bb}\) is your desired band, then P4FMOD would be at \(f_{lo}+3f_{bb}\). It helps to remember that P4FMOD always falls on the opposite side of LO (meaning if your signal is on upper side of LO, P4FMOD would be at lower side).
It is generated by two ways:
We have dedicated a separate article to 4FMODs.
Similar concept as CIM3. It is generated when \(HD_{5}\) of baseband signal upconverts to \(f_{lo}+5f_{bb}\). Since this is again \(4f_{bb}\) away from main signal, it is also called as 4FMOD although a secondary one because it comes from weak (\(5^{th}\) order) non-linearity.
Gain control is crucial part of any TX, and it is a big deal because you are interested in seeing same performance (linearity & noise) over the gain control range which is very hard, and also precise gain steps are required over parts and temperature.
RF or mm-wave, TX output is required to be matched to interface with next block (usually a front-end filter or PA). A 2:1 VSWR is considered good.
A lot of simulations are done to check isolations between different EMs on-chip (inductors, baluns, TR lines etc.). You can dead optimize your TX for distortion and noise, only to see poor isolation between two blocks giving you spurs that you just cannot get rid of, and you start failing your emissions or ACLR. One particular study that is done before tapeout is isolation between VCO inductors and TX. Any TX signal that leaks to VCO will pull it, degrading its phase noise or generating spurs that eventually end up on your TX output.
Things couple in three main ways:
Spurious emissions refer to different spurs (DAC, CLK, PLL, LO) modulated or un-modulated making there way to TX output. TX output is scanned across wide frequency range and any spur that is above a threshold (determined by emission mask) is either planned to be filtered by front-end filter if it is far away from signal, if not then starts a long debug process to figure out what caused it and how to fix it in next tapeout.
Transient overshoot and settling times are checked from ON to OFF, OFF to ON and between switching gain states. How long does it take for EVM or ACLR to stabilize during transients, that is also looked at.
Author: RFInsights
Date Published: 09 Oct 2022
Last Edit: 02 Feb 2023