Per-span optical fiber nonlinearity compensation using integrated photonic computing
A method for per-span optical fiber nonlinearity compensation comprises determining values of fiber parameters characterizing one or more target optical fibers in one or more respective spans of a link, and applying selected weight values to one or more photonic computing chips (PCCs), each PCC integrated in a different respective span of the link, wherein selection of the weight values is based on the values of the fiber parameters and a mapping associated with each PCC. The method further comprises transmitting an optical signal through the link, wherein each integrated PCC emulates an inverse of a nonlinear transfer function of the target optical fiber in the respective span, thereby reducing nonlinearity contributed by the one or more target optical fibers to the optical signal.
1. A photonic computing chip (PCC) comprising:
photonic circuit elements configured to apply optical signal processing to an input optical signal to generate an output optical signal, wherein the input optical signal comprises nonlinearity contributed by a target optical fiber; and
at least one electronic circuit element configured to control the optical signal processing based on values of fiber parameters characterizing the target optical fiber and a mapping associated with the PCC,
wherein the optical signal processing comprises operations emulating an inverse of a nonlinear transfer function of the target optical fiber, such that the PCC reduces the nonlinearity in the output optical signal relative to the nonlinearity in the input optical signal.
2. The PCC as claimed in claim 1 , wherein the operations are based on machine learning.
3. The PCC as claimed in claim 1 , wherein the PCC comprises
a first element configured to generate, from the input optical signal, P weighted input signals characterized by P respective first weights, wherein P is a positive integer and P≥2; and
a second element configured to generate the output optical signal of the PCC from a plurality of weighted output signals characterized by a respective plurality of second weights,
wherein controlling the optical signal processing comprises controlling respective values of the first and second weights.
4. The PCC as claimed in claim 3 , wherein the PCC comprises
P third elements, each configured to divide a respective one of the P weighted input signals into N channelized input signals corresponding to N respective channels of an optical spectrum of the respective weighted input signal, wherein N is a positive integer and N≥2;
for each third element, N fourth elements configured to apply N respective nonlinear operations to the N channelized input signals, thereby generating N channelized compensated signals;
P fifth elements, each configured to combine the N channelized compensated signals generated for a respective one of the P third elements.
5. The PCC as claimed in claim 4 , wherein each fourth element comprises an electro-optic (EO) modulator and a photodiode (PD) configured to tap the respective channelized input signal, and wherein controlling the optical signal processing comprises controlling the EO modulator based on an output of the PD and a weight matrix.
6. The PCC as claimed in claim 4 , wherein the third element comprises an arrayed waveguide grating (AWG) serving as a demultiplexer, and wherein the fifth element comprises an AWG serving as a multiplexer.
7. The PCC as claimed in claim 4 , wherein the third element comprises a tunable demultiplexer, and wherein the fifth element comprises a tunable multiplexer.
8. The PCC as claimed in claim 1 , wherein the PCC comprises
at least one splitting element configured to separate an optical signal into a plurality of orthogonal input signals;
for each orthogonal input signal, a duplicate version of at least a portion of the photonic circuit elements and the at least one electronic circuit element, the duplicate version being configured to process the orthogonal input signal to generate an orthogonal output signal, thereby resulting in a plurality of orthogonal output signals corresponding to the plurality of orthogonal input signals; and
at least one combining element configured to combine the plurality of orthogonal output signals.
9. The PCC as claimed in claim 1 , wherein the fiber parameters comprise one or more of a zero-dispersion wavelength λ 0 of the target fiber, a dispersion slope S of the target fiber, a nonlinear coefficient γ of the target fiber, a length L of the target fiber, and a loss coefficient α of the target fiber.
10. The PCC as claimed in claim 1 , wherein the operations are based on the Regular Perturbation Method (RPM).
11. The PCC as claimed in claim 10 , wherein the PCC comprises
a broadband optical splitter configured to generate, from the input optical signal, P+1 input signals, wherein P is a positive integer and P≥2;
a delay element configured to apply a delay to one of the P+1 input signals, thereby generating a delayed input signal;
for each of the remaining input signals, a set of RPM elements characterized by RPM weights and configured to optically process the respective input signal based on the RPM, thereby generating a respective compensated signal, for a total of P compensated signals; and
a broadband optical combiner configured to combine the delayed input signal and the P compensated signals,
wherein controlling the optical signal processing comprises controlling respective values of the RPM weights.
12. The PCC as claimed in claim 10 , wherein the PCC comprises
a first element configured to divide the input optical signal into N channelized input signals corresponding to N respective channels of an optical spectrum of the input optical signal, wherein N is a positive integer and N≥2;
N second elements, each configured to generate, from a respective one of the N channelized input signals, P input signals, wherein P is a positive integer and P≥2;
for each second element, a set of RPM elements characterized by RPM weights and configured to optically process the P input signals based on the RPM, thereby generating P compensated signals;
N third elements, each configured to combine the P compensated signals generated for a respective one of the N second elements, thereby generating a respective channelized compensated signal; and
a fourth element configured to combine the N channelized compensated signals,
wherein controlling the optical signal processing comprises controlling respective values of the RPM weights.
13. The PCC as claimed in claim 1 , wherein the PCC is positioned between a first amplification stage and a second amplification stage within an erbium-doped fiber amplifier (EDFA).
14. A method comprising:
determining values of fiber parameters characterizing one or more target optical fibers in one or more respective spans of a link;
applying selected weight values to one or more photonic computing chips (PCCs), each PCC integrated in a different respective span of the link, wherein selection of the weight values is based on the values of the fiber parameters and a mapping associated with each PCC; and
transmitting an optical signal through the link, wherein each integrated PCC emulates an inverse of a nonlinear transfer function of the target optical fiber in the respective span, thereby reducing nonlinearity contributed by the one or more target optical fibers to the optical signal.
15. The method as claimed in claim 14 , wherein determining the values comprises measuring the values from the target optical fiber or obtaining the values from provisioning.
16. The method as claimed in claim 14 , wherein the fiber parameters comprise one or more of a zero-dispersion wavelength λ 0 of each target optical fiber, a dispersion slope S of each target optical fiber, a nonlinear coefficient γ of each target optical fiber, a length L of each target optical fiber, and a loss coefficient α of each target optical fiber.
17. The method as claimed in claim 14 , wherein the PCC is designed in accordance with a nonlinear compensation model based on machine learning.
18. The method as claimed in claim 14 , wherein the PCC is designed in accordance with a nonlinear compensation model based on analytical equations.
19. The method as claimed in claim 14 , wherein the mapping associated with each PCC comprises a look-up table (LUT) or an artificial neural network (ANN).
20. The method as claimed in claim 14 , further comprising
generating the mapping associated with each PCC using a simulated version of the respective PCC and a plurality of simulated optical fibers; and
fine tuning the mapping associated with each PCC using the respective PCC and a plurality of manufactured optical fibers.