A Free Trial That Lets You Build Big!
Start building with 50+ products and up to 12 months usage for Elastic Compute Service
The soft decision does not make a decision. It only provides information on speculation and credibility, so that subsequent algorithms (such as the Viterbi algorithm) can be further processed and comprehensively determined based on other information.
For a single-bit decision, the difference in hardware and software judgments is represented by the number of BITs used for signal quantization in physical implementation. The number of bits quantified by hard decision is one bit. The decision result is not "0" or "1", and there is no room for maneuver. In Soft Decision, multiple bits are used to quantify the signal. One bit is the information of speculation, and the other bit provides the credibility information of the prediction.
Reasons for adopting soft decision
The signal processing structure and process of a G Receiver Based on Digital coherent receiving PM-QPSK are basically the same, no matter it adopts soft decision or hard decision, the difference is that soft judgments provide additional credibility information to the forward correction encoding (FEC) decoding unit, while hard judgments only provide single-bit judgment information and discard credibility information, the error correction advantage provided by the analog-to-analog converter and digital coherent reception is abandoned. The Reliability Information provided by the soft decision can further improve the FEC encoding gain.
The error correction encoding technology can jump out of the limitations of the transmission physical layer and compensate all physical transmission damages at the logic layer, especially the compensation for the influence of non-linear effects. The higher the encoding gain of the Error Correction Code, the lower the requirement on the incoming optical power and the lower the OSNR requirement of the receiver at the same transmission distance. On the other hand, the smaller the OSNR during optical signal transmission, the smaller the variation of the core optical power intensity, and the smaller the fluctuation of the core refractive index, the less obvious the influence of the nonlinear effect. Therefore, for G optical transmission systems that are subject to nonlinear effects, the 1dB error correction coding gain improves the system transmission performance much higher than the attenuation or dispersion limited optical transmission systems. According to the 60% G Test Results of China Mobile and China Telecom and the G benchmark, the transmission distance of G.655 optical fiber is 6 more than that of hard decision, and the transmission distance is increased.
Ways to implement soft decision
The realization of soft decision is benefited from the use of ADC. The G requirement is that the effective bits (ENOB) of the ADC must be greater than 6 bits. The input optical signal is converted to a four-way analog electrical signal by optical front-end. The four-way analog electrical signal is discretely quantified as a four-way digital signal by a analog-to-digital converter in terms of time and amplitude, the four channels (Ix, Qx, Iy, and Qy) estimated by digital signal processing are used for Channel balancing and carrier estimation.
The difference between the soft and hardware judgments is only the number of BITs output by the four-channel decision: the hard decision outputs one bit decision for each channel; the soft decision outputs only one bit of information for each channel, it also provides several bit credibility information. The soft decision provides the prediction information based on the judgment threshold and the credibility information based on several credibility thresholds. Figure 1 is a soft decision example of the I and Q components in a single polarization state of the digital coherent receiving PM-QPSK. In this example, a 3-bit soft decision is given for I and Q respectively. The decision threshold is used as the reference to provide 1-bit guess information, and the two-bit credibility information is provided as the reference to the three credibility thresholds.
Soft Decision-Based Error Correction Code
The error correction encoding algorithm is not necessarily related to the soft decision. However, some soft decision correction Encoding algorithms can provide credibility probability information to further improve the coding error correction capability and increase the coding gain. The third-generation Error Correction Code Based on Soft Decision and iterative algorithms has a code gain of more than 11 dB. The typical examples of this Code are Turbo and Low-QPS, compared with Turbo encoding, the non-checksum encoding method has better error correction characteristics and implementation complexity. Low Density Parity Check Code, which is a Low-Density Parity Check Code by Robert G. dr. Gallager proposed a linear grouping code with a sparse validation matrix in 1963. It not only has good performance in approaching Shannon limitation, but also has low decoding complexity and flexible structure, it is a hot topic in the field of channel coding in recent years. It has been widely used in fields such as deep space communication, optical fiber communication, satellite digital video and audio broadcast. It has become a strong competitor of the fourth generation communication system (4G), and its encoding scheme has been adopted by the next generation satellite digital video broadcast standard DVB-S2.
Gbit/s Error Correction Coding Technology
Beacon fire technology 100g dwdm adopts digital coherent receiving PM-QPSK modulation technology, and Its Error Correction Code adopts the combination of 7% hard decision and 13% soft decision, respectively placed in framer and ASIC, 2 shows. Among them, 7% of the hard decision error codes are the second-level chaincodes defined by G.975.1, and 13% of the soft decision error codes are low-density parity codes (Low-Density Parity codes). These combinations actually constitute three-level chaincodes.
This three-level chain connection code is used because of the strong large error correction capability of the Low-Density byoc encoding, which can reduce the error code of 2.5e-2 to less than 1e-5, however, due to loops and deadlocks in the decoding process, low-level error codes cannot be reduced to less than 1e-12. The third-level link hard-judgment Error Correction Code defined by G.975.1 is used externally to eliminate the influence of the "error code flat layer" of the low-level checksum codes. This configuration, on the one hand, utilizes the error correction capability of the Low-density encoding (low-density encoding) for large numbers of codes, and uses external hard-coded error codes to eliminate the impact of "code leveling layer, on the other hand, the high gain of mature commercial 7% hard-coded Error Correction Codes can minimize the complexity, power consumption, and latency of the low-cost-effectiveness of the low-cost code. For example, a combination of 992,956 hard-coded 3% (17%, 7936) and RS () encoding can get a net coding gain of 9.7dB by eliminating an error code flat layer) + G.975.1 uses the 7% hard-coded error code to eliminate the 13% low-level bitwise code layer. The net encoding gain of 11.5dB is obtained, and the length of the low-level bitwise code is less than half.
The G third-level link Error Correction Code of beacon fire technology uses the net coding gain of 11.5dB and 2E-2 with low coding complexity, processing latency, and power consumption, this greatly improves the reliability, stability, and robustness of the system.
G soft decision current network application maturity problem and future development from the above analysis can be seen: Based on Digital coherent receiving PM-QPSK modulation g receiver, whether it uses soft decision or hard decision, the signal processing structure and process are basically the same. The difference is that the soft decision provides additional credibility information for the forward correction encoding (FEC) decoding unit, hard decisions only provide single-bit decision information, discard credibility information, and discard the error correction advantages provided by analog-to-digital converters and digital coherent receivers. (That is to say, from a purely technical perspective, the maturity of soft judgments over hard judgments is actually a pseudo problem .)
The Reliability Information provided by the soft decision can further improve the FEC encoding gain. According to the telecom and Mobile G Testing Results and G benchmark, soft decision has many advantages over hard decision: ① the maximum error rate before soft decision correction is as high as 2.5E-2, the maximum error code before a hard decision is 6E-3; ② receiver OSNR margin: the row Standard specifies that the soft decision is 13dB (BOL), and the hard decision is 14.5dB (BOL); ③ transmission distance: the G.652 soft decision contains four more segments than the hard decision, and the G.655 soft decision contains six more segments than the hard decision.
From the industrial chain perspective, the soft decision industry chain is more mature than the hard decision chain. The chip vendors that are engaged in DSP in the industry all adopt soft decision technology without exception, while hard decision is only developed by some manufacturers.
From the perspective of cost, the internal structure of hardware and software judgments is almost identical, with only the difference being the DSP processing chip. The current soft decision price is slightly higher than the hard decision price, because the DSP has invested too much in the early stage, and the current production is not large, resulting in a high price. However, once the cost is increased, the cost allocation decreases rapidly. In the long run, because the soft decision industry chain is more mature, the overall price of the soft decision can be lower than the hard decision.
Start building with 50+ products and up to 12 months usage for Elastic Compute Service