My question is as follows:

  • To implement successive interference cancellation (SIC) process in downlink non-orthogonal multiple access (NOMA), should receivers know all power allocated to each symbol?

Consider a single-cell network consisting of one base station (BS) and $N$ users. Let $x_i$ be the symbol for user $i$, $h_i$ be the complex-valued channel gain between user $i$ and the BS, $p_i$ be the power allocated to symbol $x_i$. Then, the transmitted signal is given by $$ x = \sum_{i=1}^N \sqrt{p_i} x_i. $$ The corresponding received signal at user $k$ is given by $$ y = h_k \sum_{i=1}^N \sqrt{p_i} x_i + n_k, $$ where $n_k\sim\mathcal{CN}(0,\sigma^2)$ is the additive white Gaussian noise of user $k$.

Assuming that $\lvert h_1 \rvert^2 \le \lvert h_2 \rvert^2 \le \cdots \le \lvert h_N \rvert^2$, user $k$ decodes its own symbol $x_k$ according to the following process:

  1. Decode $x_1$ by treating $\sum_{i=2}^N h_k\sqrt{p_i}x_i + n_k$ as noise.
  2. Subtract $\hat{x}_1 \triangleq h_k\sqrt{p_1}x_1$ from $y$.
  3. Decode $x_2$ by treating $\sum_{i=3}^N h_k\sqrt{p_i}x_i + n_k$ as noise.
  4. Subtract $\hat{x}_2 \triangleq h_k\sqrt{p_2}x_2$ from $y-\hat{x}_1$.
  5. Repeat to (i) decode $x_l$ by treating $\sum_{i=l+1}^N h_k\sqrt{p_i}x_i+n_k$ as noise, and (ii) subtract $\hat{x_l} \triangleq h_k\sqrt{p_l}x_l$ from $y-\sum_{i=1}^{l-1}\hat{x}_i$ until $x_k$ is decoded.

In my opinion, to decode its own symbol $x_k$, user $k$ should know $p_l$ for any other user $l$ such that $\lvert h_l \rvert^2 < \lvert h_k \rvert^2$. Is this opinion true? If this is true, the BS should inform the user about the power allocated to each symbol. Nevertheless, is NOMA useful?


Yes, one would require to know the power of other users. There are two options, the network can either signal the UEs this information in a control frame or the UEs need to estimate the power of the streams allocated for other users (knowing its own power allocated and some smart signal processing).

In any case interference cancellation techniques are suspetible to the "effectiveness" of the cancellation. Usually the SNR degardes rapdily with decrease in estimation accuracy or error in knowledge of the parameters of the other users, such as power

| improve this answer | |
  • $\begingroup$ Thanks to your answer, I got it. May I request any references related to the smart signal processing? $\endgroup$ – Danny_Kim Jul 8 at 3:20
  • 1
    $\begingroup$ Mostly ideas related to energy detection after segregating the stream of a user. Energy detection for random signals is explained in Stephen m Kay Detection theory book for ex. $\endgroup$ – Dsp guy sam Jul 8 at 6:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.