DAC and ADC distortions analysis in optical DMT and PAM systems

Fig 5. 18 Modulation bit allocation curve of optical DMT system by experiment and simulation

Fig 5. 19 Constellations of optical DMT system by experiment and simulation

5.4. DAC and ADC distortions analysis in optical DMT and PAM

77

locked to the same value, and the transmitter signal power of DMT modulation is 2.76dB lower than the multi-level PAM because the swing of DAC output is fixed for all the cases. The baud rate for PAM is swept down from 32G symbols per second to 4G symbols per second until the specified FEC BER limit 5e-5 is met. By doing this, one can compare the distortion tolerance between DMT and PAM systems.

Fig 5. 20 Performance of optical DMT and PAM4 as white noise changes

Fig 5. 21 Performance of optical DMT and PAM8 as white noise changes

Fig 5. 22 Performance of optical DMT and PAM16 as white noise changes

Fig 5. 23 Capacity trend of PAM4/8/16 and DMT as white noise changes

Simulation with different white noise powers was evaluated to investigate the impact of white noise on the performance of optical DMT and PAM systems. The simulated results with different white noise powers comparing DMT and PAM4 are shown in Fig 5. 20. Seen from the figure, PAM4 system has more capacity than DMT. It reveals that at the large white noise and far less than 100Gbps case, PAM4 has the advantage to DMT because of 2.73dB larger signal power when the DAC swing is fixed. And capacity of DMT system linearly decreased from 55.52Gbps to 17.82Gbps, while capacity of PAM4 linearly decreased from 64Gbps to 20Gbps. The similar decrease slope shows that DMT and PAM4 systems have similar tolerance to white noise.

Results of DMT and PAM8 are shown in Fig 5. 21. Obviously DMT has better performance than PAM8. Capacity of DMT linearly decreases from 107.63Gbps to 34.67Gbps, while capacity of PAM8 also linearly decreases from 93Gbps to 22.5Gbps. And the capacity loss due to the enhancement of white noise is almost same for PAM8 and DMT. This similar capacity decrease trend of the two confirms the similar tolerance to white noise in DMT and PAM8 system. And the similar phenomenon could be observed from the comparison between DMT and PAM16 as shown in Fig 5. 22.

Capacity comparison of 4/8/16-level PAM and DMT are summarized and shown in Fig 5. 23.

PAM4 has more capacity than DMT because of higher signal power. But this advantage is no longer valid when PAM level is increased to 8 and 16 with the increased sensitivity to noise. On the other hand, the similar trends of capacity decrease with white noise increase of DMT and PAM with different levels confirms that the impact from white noise is same for both DMT and PAM modulation. It can be expected because the considered distortion is the white noise.

79

Fig 5. 24 Performances of DMT and PAM4 as colored noise changes

Fig 5. 25 Performances of DMT and PAM8 as colored noise changes

Fig 5. 26 Performances of DMT and PAM16 as colored noise changes

Fig 5. 27 Capacity trend of PAM4/8/16 and DMT as colored noise changes

Colored noise power are changed to evaluate the impact on DMT and PAM system in the simulation platform. Results of DMT and PAM4 are shown in Fig 5. 24. Comparing from the figure, PAM4 has more capacity than DMT. And similar with white noise case, it indicates that at large colored noise and far less than 100Gbps case, PAM4 has the advantage to DMT because of 2.73dB more signal power. Fig 5. 24 shows that with the increase of colored noise, the capacity of PAM4 drops more sharply than DMT. And this implies that PAM4 is less tolerant to colored noise than DMT. And the similar trend could be observed from the comparison between DMT and PAM8 as shown in Fig 5. 25. DMT and PAM16 results are shown in Fig 5. 26. DMT has much better performance than PAM16. And system capacity decreases 42.4% to DMT and 71.4% to PAM16 for the same colored noise change. Much more capacity decrease in PAM16 system reveals that PAM16 is less tolerant than DMT to the colored noise.

Capacity of PAM 4/8/16 and DMT are compared in Fig 5. 27. DMT has less capacity than PAM4 because of lower signal power and better performance than PAM8 and PAM16. From the different trends of capacity drop with the increase of colored noise, it is concluded that unlike the white noise case, DMT is more tolerant than multi-level PAM to colored noise especially when PAM level is high. And this is caused by the noise enhancement nature of the linear equalization in PAM systems.

Fig 5. 28 Performance of DMT and PAM4 as narrowband interference changes

81

Fig 5. 29 Performance of DMT and PAM8 as narrowband interference changes

Fig 5. 30 Performance of DMT and PAM16 as narrowband interference changes

Fig 5. 31 Capacity trend of PAM4/8/16 and DMT as narrow-band interference changes

Simulation results of DMT and PAM4 under different power of narrow-band interference are shown in Fig 5. 28. It appears that DMT has better performance than PAM4. Fast roll-off can be clearly seen in PAM4 curve compared with DMT, which mean PAM4 has worse tolerance to narrow-band interference than DMT. And only 20.5% capacity decrease of DMT from 78.91Gbps to

62.73Gbps is observed while to PAM4 the decrease is 81.3% and from 64Gbps to 12Gbps. The great capacity decrease to PAM4 indicates that narrow-band inference does greater harm to PAM4 than to DMT.

Results of DMT and PAM8 are shown in Fig 5. 29. Still fast roll-off is observed in capacity curve of PAM compared with DMT. When capacity of DMT decreases 18.5%, capacity of PAM8 quickly decreases 74.1% and from 87Gbps to 22.5Gbps. Serious capacity loss to PAM8 reveals that narrow-band inference can greatly reduce performance of PAM8. It indicated that DMT has much more tolerance to narrow-band inference than PAM8.

Results of DMT and PAM16 are shown in Fig 5. 30. Similar phenomenon is observed as to PAM4/8. Capacity curve of PAM16 falls more quickly than DMT. With capacity decrease of 7.8% to DMT from 96.79Gbps to 89.21Gbps, PAM16 would lose 66.7% capacity from 42Gbps to 14Gbps.

And narrow-band interference would cause great reduction to PAM16 performance. So DMT is quite more tolerant than PAM16 to narrow-band inference.

Capacity comparison is summarized and shown in Fig 5. 31. DMT performs much better than all PAMs with narrow-band interfered channel. With different narrow-band interference, great performance decrease is observed to PAM4/8/16 while DMT is robust. Comparing with white or colored noise and referring to DMT, narrow-band interference reduces more capacity to PAM 4/8/16. The reason is that DMT is multi-carrier modulation while PAM is single-carrier. The narrow-band interference only degrades some sub-carrier of DMT, but it degrades the whole signal of single carrier PAM. DMT has the natural advantage to combat frequency-selected channel distortions. Narrow-band interference would greatly impact all PAM symbols especially for the high-level ones like PAM16.

From the comparisons of the impacts from white noise, colored noise and narrow-band interferences, it is concluded that firstly DMT and PAM show the similar tolerance to white noise, and secondly PAM is a little more sensitive to colored noise especially when PAM level is high.

Notably, colored noise in our evaluation as shown in Figure 8 has a slowly roll-off shape of 10dB attenuation in 20GHz bandwidth. More sensitive impact could be expected to PAM with more sharply roll-off colored distortions in real-world scenarios. Thirdly, DMT system has much better tolerance to narrow-band interferences than high-level PAM system.