# Modelling transmitter and receiver impairments in communication system

A paper I read studied the effect of transmitter impairments and receiver impairments by modeling both as additive white Gaussian noise. The noise for transmitter impairmenets are added before the antenna before transmission. The noise modelling receiver impairments is added as an additional white Gaussian noise to the thermal white Gaussian noise. A conclusion was made as follows

The main reason for this is that the noise added at the transmitter (modeling Tx EVM) experiences the same frequency- and time-selective fading as the desired signal. In contrast, the noise added at the receiver (modeling Rx EVM) does not experience frequency-selective fading (only time-selective fading).

Does the quoted conclusion above make sense?

It makes sense to me. If the ideal transmitted signal is $s(t)$, the signal that is actually transmitted is $s(t)$ plus noise and distortion. This is the signal that suffers attenuation and fading, and to which an extra heaping of noise and distortion is added in the receiver. Modeling transmitter imperfections is not something that is included in most introductory textbooks, but it's very important to implement a system with good performance. You may see for example "RF System Design of Transceivers for Wireless Communications" by Qizheng Gu for more.
• After reading the paper, I think that the explanation is this. In the paragraph above section 3, the variance of the noise used to model Rx EVM is given as $\epsilon_{Rx}P_n$, where $P_n$ is the received instantaneous power. $P_n$ varies with time according to the multipath channel dynamics. That's why the variance of the EVM noise depends on the channel's time-selective fading properties.