A 32-wavelength architecture carrying 400 Gb/s per wavelength has a gross arithmetic capacity of 12.8 Tb/s. Apply an illustrative 20% combined overhead and you have 10.24 Tb/s of ideal net data. That calculation is valid. It is also insufficient to claim a working link.
Where the model becomes a system
The real design decision starts with per-wavelength received optical power, not the aggregate number on the slide. Diffraction, aperture size, beam-pointing errors, terminal optical losses, source linewidth, Doppler dynamics, receiver noise, and error-correction overhead determine what each channel can actually carry.
Channel multiplication is valuable because it can reuse an optical aperture and packaging. But capacity is not free: every additional lane adds power dissipation, control complexity and manufacturing variables. More wavelengths can worsen the economics unless optical integration and DSP are designed together.
The experiment that matters
We would rather see a repeatable two-channel bench result with measured BER, calibrated attenuation and a thermal sweep than an elaborate simulated 10 Tb/s dashboard. Our current LinkLab tool screens architecture assumptions; it cannot certify receiver feasibility, FEC performance, or flight availability.
SpaceWDM's thesis is that terrestrial DWDM concepts are worth adapting to orbital systems. Its feasibility will be earned one measured constraint at a time.