synthetic diagnostic is a forward model of a real instrument applied to the output of a simulation, producing the signal that detector would actually record — so that a code's prediction and a measurement can be compared as the same kind of object.
A simulation outputs physical fields: density, temperature, emissivity, fluctuation amplitude, defined everywhere in the 📝plasma and at every instant. A detector outputs something far narrower — a count rate or a voltage, integrated along a line of sight, blurred by finite spatial and temporal resolution, band-limited in wavenumber and frequency, and mixed with background and noise. The two are not the same object and cannot be laid against each other directly. The older way of closing the gap was inversion: process the measurement backwards into a physical quantity and compare there. But inversion is mathematically ill-posed and carries assumptions of its own, so any disagreement could be blamed on it. A synthetic diagnostic runs the comparison in the well-posed direction instead, pushing the model through the instrument rather than the measurement back through the model. That reversal is what made turbulence-code validation quantitative: raw fluctuation levels from a 📝gyrokinetics simulation mean little beside what a beam-emission or correlation-radiometry system reports until the instrument's own filtering has been applied to them.
Applied to a predicted plasma rather than to a reproduction of an existing one, a synthetic diagnostic does something stronger: it states in advance what a named instrument will read. This is the mechanism by which a projection is made falsifiable before the machine is built — not "the plasma will be this hot," but "this spectrometer, at this port, through this collection optic, will record this line shape at this rate." Much of the literature behind 📝SPARC Early Campaign Diagnostics is written that way, the 2024 Review of Scientific Instruments companion papers sizing and siting instruments against synthetic signals computed from predicted 📝SPARC plasmas — X-ray crystal spectrometry, 📝neutron diagnostics flux monitors, bolometry, 📝runaway electrons synchrotron emission. Deciding that an instrument will see what it needs to see is a design step; publishing what it will see is a commitment.
The term also enforces the distinction it depends on. A synthetic signal is model output end to end: it inherits every assumption of the 📝integrated modeling chain that produced it and none of the authority of a measurement. Only the real detector produces a datum. Keeping the two labelled correctly is what allows the eventual comparison to settle anything, because a gap between a synthetic spectrum and a recorded one is then attributable to the physics model rather than to the mismatch between what was computed and what could ever have been seen.