Splice Optimization Recipe Explorer

How a starting recipe is picked

1. Each fiber is described by its fiber definition in the database: class (SMF, PM, rare-earth doped…), sub-type (Panda, Bowtie, ZBL), doping, cladding OD, core size, MFD at its recorded wavelength and NA. When a definition has no NA, it is estimated from core size and MFD with the Marcuse relation. MFD is rescaled to your operating wavelength.
2. Fiber-to-fiber distance. OD and MFD differences are measured on a log scale (10% = one unit), core and estimated NA at 20% per unit. A different type counts two units, a different sub-type or doping counts one or two. Unknown values count one unit.
3. Pairs are compared both ways round. When a stored recipe fits better mirrored, its left/right θ-alignment settings are swapped to match your orientation. A recipe for the same two fiber names (vendor prefixes ignored) counts as the same pair.
4. The nearest recipes are blended with inverse-square distance weights (k, below). Arc and rearc times are blended on a log scale because they span 500 ms to 50 s. Alignment settings are a weighted vote.
5. Only parameters the fibers actually predict use the blend. A leave-one-out test on the library decides this. Parameters where similar fibers predict no better than the library-typical value (the median) start from that typical value, and their sweep uses the library's middle 50% range.
6. Physics checks are added on top: MFD-mismatch loss floor, cladding mismatch (fiber holders, arc toward the larger fiber), PM needs a P-model splicer, and SM → MM direction.

Arc powers are in bits relative to the arc-calibrated standard, so they only carry over between splicers that have been arc-calibrated. By default only FSM-100-series recipes (and records with no splicer) are used. Every change you make on the Database tab is used by the finder straight away. Other splicers can be included from the filter row on the Find tab.