Commplax is resurrected!
Thinking in JAX and JAX your codes.
cpr = eq.CPR() # create a single dimension carrier-phase-recovery module
cpr = jax.vmap(cpr) # vmap CPR to be N-dimension (e.g., 2 for dual polarizations)
cpr, output = cpr(input) # input.shape=(2,N) => output.shape=(2,N)# An example of 15-tap 3/4-spaced 8x8 MIMO with JIT
mimo = eq.MIMOCell(15, af=cma(lr=5e-5), dims=8, up=3, down=4)
mimo_updated, output = jax.jit(scan)(mimo, input)with sample mode, composition of feedback loop is easy
mimo = eq.MIMOCell(15, af=dd_lms, dims=2)
cpr = eq.CPR(dims=2, mode='feedback')
def cpr_mimo_loop(state, x):
cpr, mimo = state
y = cpr.apply(x) # remove carrier phase
mimo, y = mimo(y) # apply MIMO
cpr, y = cpr.update(y) # update carrier phase tracker
state = cpr, mimo
return state, y
(cpr_updated, mimo_updated), output = scan(cpr_mimo_loop, (cpr, mimo), input)PMD layer:
- M/K-spaced NxN MIMO
- Adaptive equalizers(CMA/LMS/RLS/Kalman...)
- Polyphase resampler
- Detector (e.g., Viterbi)
- Timing syncrhonizer
- Distribution matcher (e.g. CCDM)
- Probabilistic shaping
- Standards-compliant FEC (e.g., CFEC, oFEC)
Optical channel:
- fiber/WSS/ROSA/TOSA/...