I'm simulating the optical PMD-QPSK using optisystem,and I'm measuring the effect of Nonlinearity on BER,and the problem that I'm using a DSP for compensating the Dispersion,Polarization multiplexing and carrier Phase Estimation.
The problem that the effect of nonlinearity doesn't appear because it produce a phase error which is corrected or changed by the DSP component.
When I tried to disable the carrier phase estimation in the DSP the system also produce error even if there is nonlinearity.
How can I measure the nonlinearity effect on BER with this problem
What symbol rate and link dispersion are you simulating?
In the low chromatic dispersion limit, each symbol will experience a similar non-linear phase shift, and the phase recovery signal processing will compensate for this - as would occur in a real-world implementation.
You need sufficient chromatic dispersion to distort the signal waveform, so that peak power in the fibre varies for different symbol sequences. This will result in a non-linear Kerr phase shift which depends on the bit pattern of adjacent symbols, and a distortion of your QPSK constellation which is not corrected by a simple phase recovery algorithm.
Thanks Mr.Robinson for your response and the notes that you declared.
Actually, I'm using 112Gb/S PM-QPSK with symbol rate 28 Gbaud/s and the dispersion D = 19 ps/(nm-km). Dispersion slope = 0.
Do you think this value of dispersion is low?
I can't predict net dispersion from the dispersion coefficient alone.
Are you simulating a single span system, so that the power decreases monotonically with propagation distance, or an amplified system with one or more intermediate optical amplifiers along the link?
What matters is the net dispersion, and hence the waveform shape, at locations in the fibre where the optical intensity is high. It follows that the non-linear phase shift will be sensitive to fibre attenuation coefficient and other optical losses in the system.
At 28 Gbaud you should see some waveform distortion after propagating a few kilometres. What are your optical power, fibre effective area, propagation distance and fibre attenuation coefficient?
Are you able to display the waveform in your simulation and monitor changes as you increase optical power in the fibre?
I'm changing the transmitted power from -5dBm to 15dBm to show the effect of nonlinearity on BER.Also I'm using 1000 Km fiber divided in 10 spans with 10 EDFA,Noise figure of the ampilfier = 6dB.the effective area of the fiber is 80 um^2 and gamma = 1.77 1/(watt.km)
Yes,I'm displaying signal spectrum every step in SSFM of the fiber and the spectrum change but without large distortion.
I get a large error for all the data or fixed small error due to noise.
where the carrier phase estimation round the constellation to the nearest corner.
I can't get an error that proportional to value of the transmitted power.It is On/Off results
Linear chromatic dispersion will not change the spectral intensity, so this is not a useful measure of time-domain waveform distortion. However, I would expect to see significant spectral broadening after 10 spans at +15 dBm due to Kerr effect distortion.
You don't specify fibre attenuation coefficient. At 0.21 dB/km, the effective non-linear interaction length (Leff) is around 20 km. At -5 dBm launch power, the average non-linear phase shift over each 100 km span is approximately 0.011 radian. The net distortion over 1000 km will be small, with little spectral broadening.
At +15 dBm, the average non-linear phase shift increases to approximately 1.1 radian per span. There is sufficient dispersion and distributed non-linearity for the non-linear distortion to be noise-like. The magnitude should be sufficient to degrade your BER significantly.
Are you able to display the QPSK constellations after phase recovery? I would expect the scatter of the constellation points to reduce initially as the signal power rises from -5 dBm, due to an improvement in linear signal to noise ratio. At higher powers non-linear distortion should dominate, and the scatter of constellation samples should increase again.
If you can't get sensible results from your 10-span system, try a single span simulation which may give results which are easier to understand. Is the scatter of constellation points at low powers consistent with your optical signal to noise ratio?
How is BER measured? How many symbols do you propagate? Is your BER a simple count of errors at the receiver, or do you add additional noise at the receiver and measure the change in BER with OSNR?
But +15dBm is a large value and in some papers ,I found that the nonlinearity appears at 5 dBm.
I count the bits with errors at the receiver by comparing Tx bits with the received one.and I'm sending about 131,000 bits each bits with 4 samples..
I'll try to test the system wit single span and using large powers as you mentioned and i'll reply with the results.
- Badges
- Science topic
- Similar topics
- Computer Communications (Networks)
- Simulators