Silicon Photonics / Co-Packaged Optics Link-Budget Simulator

Model whether a co-packaged optical link closes — then run it: the simulation executes on the ChipFoundryServices distributed compute pool. As AI training clusters outgrow the reach and energy of copper SerDes, the interconnect moves onto the package as silicon photonics. An integrated laser feeds a ring/Mach–Zehnder modulator; light couples off the die through a grating/edge coupler into fibre, travels to a remote package, couples back in and lands on a Ge photodetector + TIA. Two things gate the design. The optical budget: launch power minus modulator, two couplers, on-chip waveguide, fibre and WDM mux/demux losses gives the received power, which must sit above the receiver sensitivity (the minimum optical power to hit the target BER — itself worse at higher data rate and under PAM4) with margin to spare. And energy-per-bit: wall-plug laser power over the WDM comb, plus the driver + TIA energy, divided by the aggregate bit rate, has to beat the pJ/bit budget. The link ships only if it clears both gates. Reduced-order educational model. See also the SerDes / die-to-die, chiplet / CoWoS, HBM bandwidth, network-on-chip, package warpage, interconnect RC and power-delivery network simulators and the compute-pool status.

Co-packaged optics Aggressive DWDM Long reach 2 km
Optical path — the dB loss budget
Receiver & data
WDM comb & energy
Link margin vs fibre reach — margin falls as the fibre gets longer (more dB of attenuation). It crosses the dashed red zero line where the received power drops to the receiver sensitivity: that is the maximum reach. The marker (●) is your operating point — above the line the link closes, below it the link fails
Left: link margin vs line rate — sensitivity worsens with data rate (more noise bandwidth), so margin falls; it crosses the dashed red zero line at the rate ceiling (marker ● = your rate). Right: the optical loss stack — each element as a share of the total insertion loss; the amber bar is the dominant loss to attack first
Developer API — same simulation over HTTP (load-balanced across the pool): 
curl -X POST https://www.chipfoundryservices.com/edge/photonics \
  -H "Content-Type: application/json" \
  -d '{"laser_power_dbm":10,"modulator_loss_db":4,"coupling_loss_db":1.5,
       "waveguide_loss_db_cm":1.5,"waveguide_length_cm":2,"fiber_loss_db_km":0.2,
       "fiber_length_m":100,"wdm_loss_db":3,"data_rate_gbps":100,"responsivity_a_w":0.9,
       "tia_noise_pa_rthz":20,"modulation_format":"PAM4","num_wavelengths":8,
       "laser_wall_plug_pct":15,"driver_energy_pj_bit":1,"energy_budget_pj_bit":5}'
Returns JSON with outputs (total_insertion_loss_db, modulator_loss_db, coupling_loss_total_db, waveguide_loss_db, fiber_loss_db, wdm_loss_db, dominant_loss, received_power_dbm, q_factor, symbol_rate_gbaud, receiver_sensitivity_dbm, link_margin_db, meets_link, aggregate_bandwidth_gbps, aggregate_bandwidth_tbps, laser_electrical_mw, driver_electrical_mw, total_electrical_mw, laser_power_fraction_pct, energy_per_bit_pj, meets_energy, meets_budget, verdict), the full profile (48-point margin_vs_reach sweep [fiber_length_m, link_margin_db] and 48-point margin_vs_rate sweep [data_rate_gbps, link_margin_db]), the serving node, and compute_ms. Endpoint aliases /edge/siliconphotonics, /edge/copackagedoptics, /edge/cpo, /edge/opticalio, /edge/opticallink, /edge/wdm.