In this section, we analyze the exact BER expression for both the conventional NOMA and the proposed NOMA. In this paper, it is assumed that the data for user 1 and user 2 are modulated by QPSK. Each user’s symbol is detected as one of , assuming that the priori probabilities of each symbol are equal, and each symbol is modulated according to QPSK as {}.
  4.1. Exact BER of Conventional NOMA
In the conventional NOMA system, both user 1 and user 2 decode the received signal to detect 
. Additionally, user 1 detects and removes 
 to decode its own signal, 
. Therefore, the error probability for 
 at user 
k is first calculated. In QPSK modulation, since each symbol consists of 2 bits, the probability that user 
k fails to decode 
 can be expressed as the average of the error probabilities for the first and second bits of 
, as follows:
        where 
 denotes the event that user 
k incorrectly detects 
 and 
 denotes the event that user 
k incorrectly detects the 
j-th bit of 
 (
) [
8]. 
Figure 4 shows the constellation diagram of the signals received by each user in a conventional NOMA system. Each point represents the combination of signals 
 and 
 under SC. It is denoted as 
, and the received signals at each point are expressed by (5) as 
.
In 
Figure 4, an error occurs if any point on the constellation diagram deviates from its appropriate decision region. It is evident that the first bit of the QPSK symbol is determined by the quadrature part, and the second bit is determined by the in-phase part. For instance, if the actual first bit of 
 is 1, in order to detect the first bit correctly, points 
 and 
 should be located below the boundary of the decision region. The second bit can also be approached in the same way as before. That is, since user 
k mistakenly decides bit 1 as 0 when point 
 or 
 is greater than 0, 
 and 
 can be expressed as follows:
        where 
 denotes the event that user 
k incorrectly detects the 
j-th bit of 
 due to the position of point 
i (
). By substituting (
5) and the QPSK symbols, (
13) can be expressed as follows:
        where 
 and 
 denote the in-phase and quadrature components of the noise, respectively, and 
 is the symbol energy, 
 [
9]. By substituting (
14) into (
12), the error probability of user 2 is defined as follows:
        where 
Q(·) denotes the Gaussian 
Q function. The simplified expression of (
15) is as follows:
        where 
 is the SNR for the 
i signal points given in 
Figure 4. For 
, 
 can be expressed as follows:
Then, the average BER at user 2 can be obtained as follows:
        where 
 is the probability density function (PDF) of SNR. In this paper, assuming a Rayleigh fading channel, the PDF of the SNR is defined as follows:
        where 
 denotes the average of 
.
Using ([
10], Equations (13.3)–(7)) and (
19), (
18) is given as follows:
Next, user 1 decodes 
 after performing SIC to remove 
. The error probability of user 1 depends on whether 
 is detected correctly or incorrectly. Therefore, the error probability of user 1 is defined as follows:
        where 
 denotes the complement of 
. 
Figure 5a shows the constellation of 
 after user 1 correctly detects 
 and performs SIC. The expressions for each signal are derived from the signal expressions in 
Figure 4 with the terms containing 
 removed. Therefore, the signal at each point is expressed as 
. For the correct detection of the bits in 
, the points C or C′ on the in-phase and quadrature axes should not fall outside their respective decision regions.
User 1 incorrectly detects the first bit of 
 if either 
 or 
, and incorrectly detects the second bit of 
 if either 
 or 
. Hence, the bit error probability of user 1, under the condition that no error occurs when detecting 
 symbols, is given as follows:
        where 
 and 
 are the probabilities that user 1 correctly detects the 
j-th bit of 
 as given in (
13). If the fundamental theorem of conditional probability, 
, is applied, (
22) can be redefined as follows:
For the purpose of simplifying (
23), we define the SNR for the C or C′ signal point given in 
Figure 5a as follows:
Then, the error probability of 
, given that user 1 correctly detects 
, is expressed as follows:
Figure 5b,c show the constellations of 
 after user 1 incorrectly detects 
 and performs SIC. These represent the cases where the first bit of 
 is in error and where the second bit is in error, respectively. The expressions at points D and E, contrary to the case of 
, represent the in-phase and quadrature components of the signal, respectively, where terms containing 
 or 
 are not removed, but instead added to the signals in 
Figure 4: 
. In 
Figure 5b, user 1 incorrectly detects the first bit of 
 if either 
 or 
, and incorrectly detects the second bit of 
 if either 
 or 
. Similarly, in 
Figure 5c, user 1 incorrectly detects the first bit of 
 if either 
 or 
, and incorrectly detects the second bit of 
 if either 
 or 
. Hence, the bit error probability of user 1, under the condition that error occurs when detecting 
 symbols, is given as follows:
        where 
 and 
 are the probabilities that user 1 incorrectly detects the 
j-th bit of 
 as given in (
13). Similar to (
23), (
26) can be redefined as follows:
 For the purpose of simplifying (
27), we define the SNR for the D and E signal point given in 
Figure 5b,c as follows:
Then, the error probability of 
, given that user 1 incorrectly detects 
, is expressed as follows:
By substituting (
25) and (
29) into (
21), the error probability of user 1 is obtained as follows:
Then, the average BER at user 1 can be obtained as follows:
Using ([
10], Equations (13.3)–(7)) and (
19), (
31) is given as follows:
  4.2. Exact BER of Proposed Transmitter and Receiver
In the proposed method, the transmitter sends two types of data: one is the information for user 2, and the other are the redundancy data used to transform the former data into the information for user 1. In this method, user 2 only requires the former data, while user 1 requires both the former and the latter data. Therefore, we proceed by separating the former and latter data to calculate the error probability. In this section, we assume 
. First, the constellation diagram of the former data is represented as shown in 
Figure 6.
The received signals at each point are expressed as 
. User 
k incorrectly detects the first bit of 
 if 
, and incorrectly detects the second bit of 
 if 
. Thus, the error probability for 
 at user 
k is expressed as follows:
To simplify (
33), SNR at points 
 can be expressed as follows:
By substituting (
34) into (
33), the error probability of user 2 is defined as follows:
Then, the average BER at user 2 for the proposed receiver can be obtained as follows:
Using ([
10], Equations (13.3)–(7)) and (
19), (
36) is given as follows:
Next, for user 1, the error probabilities of both 
 and 
 should be considered. The error probability of 
 is the probability of incorrectly detecting 
, and the error probability of 
 is the probability of incorrectly detecting 
. Examining the error probability of user 1, the XOR bit operation results in 1 when the two bits are different, and 0 when they are the same. Due to this characteristic, if both 
 and 
 are incorrectly detected, the final 
 obtained in 
Section 3.2 would be detected without error. Therefore, the error for user 1 occurs only when either 
 or 
 has an error. The average error probability of user 1 in the proposed receiver is expressed as follows:
        where 
 denotes the event that user 1 incorrectly detects 
. Due to (
33) and (
34), the error probability of 
 for user 1 is given as follows:
Then, the average error probability of 
 at user 1 is given as follows:
Since 
 is a combination of 
 and 
 combined in series, its average error probability for user 1 can be represented as the average value of the error probabilities for the 
 and 
 of user 1 in conventional NOMA, as follows:
By substituting (
12) and (
32), (
41) is obtained as follows:
By substituting (
40) and (
42) into (
38), the average error probability of user 2 is given as follows: