A survey night is a purchased input. Telescope time, instrument time, and staff attention all carry a price, and the efficiency of the spectrograph sitting at the focus sets how much science comes back per night. Throughput is the fraction of photons that survive the optical train and register on the detector. When that number drops, the cost per measured target rises, and the survey either shrinks or slips.

A Night at the Telescope Has a Price

Observatories allocate time in nights or hours, and those allocations map onto real budgets: electricity, cryogens, staff, travel, and the depreciation of a facility built over decades. A spectrograph that delivers 30 percent end-to-end throughput turns a large fraction of incoming starlight into heat rather than data. The night still costs the same.

That asymmetry is the quiet problem in survey economics. Capital budgets are visible and celebrated. Operations budgets are recurring and easy to trim. A team that wins a night on a 2-meter class telescope has won an asset measured in hours, and every optical surface between the sky and the detector taxes that asset.

Computing budgets sit on the other side of the ledger. Pipelines, archives, and reprocessing are often funded separately from observing, which means a throughput gain that produces more data can land as an unfunded cost. The site has examined related questions in how funding panels weigh a null result, where the accounting of a result matters as much as the result.

Throughput Is a Multiplier on Every Exposure

Optical losses compound. A mirror coating that reflects 90 percent, a slit that passes 70 percent, a grating that diffracts 60 percent into the wanted order, and a detector that converts 80 percent of arriving photons: multiply those and the ceiling is far below unity. Each element is a research project in its own right.

Echelle spectrographs, which spread light into many orders at once, are especially sensitive to this arithmetic. They demand stable optics and high efficiency across a wide band, and any misalignment or contamination shows up as a wavelength-dependent loss. The SOPHIE echelle spectrograph at Haute-Provence Observatory, operational for the community since 2006, was built for asteroseismology and radial-velocity planet detection, and its design reflects that trade.

Detector quantum efficiency sets the hard ceiling. Silicon devices can convert a large majority of photons in some bands, but that peak is narrow, and the red end is harder. A spectrograph optimized for the blue will underperform for a survey targeting cool stars. The mismatch is not a failure of engineering; it is a budgeting decision made years earlier.

The Worked Example of a Radial-Velocity Survey

Consider a radial-velocity survey looking for planets around faint stars. The signal is a periodic shift in spectral lines, and detecting it requires high signal-to-noise per resolution element. Signal-to-noise scales with the square root of the number of detected photons, so throughput enters the exposure time directly.

If throughput falls by a quarter, the exposure must run roughly 1.8 times longer to reach the same signal-to-noise. That is arithmetic, not opinion. On a survey of hundreds of targets, the extra time compounds into weeks, and weeks of a shared facility are not recoverable within the same season.

Seasonal visibility makes the slip worse. A target field is observable for part of the year, so lost nights push measurements into the next window. Radial-velocity surveys also need many epochs per star to fold a orbit, and gaps in the cadence degrade the period recovery. A throughput shortfall becomes a scheduling problem, then a statistical one.

This is where model-organism habits from other fields mislead. A clean lab measurement of grating efficiency does not predict performance in a cold, moving instrument on a telescope. The site has followed similar inference gaps in cold-atom work across two labs, where the apparatus, not the theory, set the disagreement.

Funding Structures Reward New Instruments

Capital grants favor hardware that can be named, photographed, and opened. A new spectrograph is a deliverable; a throughput upgrade to an existing one is maintenance. Review panels, composed of people who built instruments themselves, tend to reward the former. The result is a fleet of instruments with aging optics and no line item to fix them.

Operations budgets are harder to defend because they do not generate announcements. A coating re-aluminization, a detector replacement, or a recalibration campaign competes against a new camera that would attract a press release. Publication pressure pushes the same direction: discovery claims travel further than efficiency measurements.

The objection is real. New instruments open new wavelength ranges and new science, and a community that only maintains will eventually fall behind. The trade is not between maintenance and progress; it is between a marginal throughput gain on an existing facility and a genuinely new capability. Reviewers rarely see both numbers side by side.

Integral Field Units Change the Cost Equation

Integral field spectrographs collect spatially resolved spectra from a two-dimensional region in a single exposure, using lenslet arrays or image slicers to reformat the field into a slit. The technique spans the optical and infrared, roughly 0.32 to 24 micrometers, and it is used beyond astronomy in biomedical imaging and Earth remote sensing.

For a survey, an integral field unit can replace a raster scan, trading many pointings for one. That reduces overhead, which is often the dominant cost on faint targets. But the reformatting optics add surfaces, and the light is divided among many spatial elements, so per-pixel throughput drops. Multiplexing and efficiency pull against each other.

Data volume shifts the cost to pipelines. A single integral field exposure can produce tens of thousands of spectra, and the calibration, sky subtraction, and cube reconstruction are not trivial. A survey that budgets for telescope time but not for pipeline development has moved the expense to a place where it is harder to see and harder to fund.

The Maintenance Burden and Its Trade-offs

Keeping a spectrograph near its design throughput is an ongoing campaign. Mirror coatings degrade from oxidation and dust, and the rate depends on the environment: a dry high-altitude site might see reflectivity fall by a few percent per year, while a humid coastal site could lose more. Re-aluminization requires removing the mirror from the instrument, a process that takes the spectrograph offline for days to weeks and carries its own risk of damage. The trade-off is between lost observing time now and degraded throughput later.

Detector performance also drifts. Charge transfer efficiency in CCDs can decline after years of radiation exposure, and the effect is worst for faint sources because the signal is spread over many pixels. A detector that once met specifications may now lose a measurable fraction of the signal in the readout, and the loss is not uniform across the chip. Replacing a detector is a major intervention, often costing as much as a new instrument and requiring a new calibration pipeline.

These maintenance costs are predictable but rarely budgeted. A survey that assumes the instrument will perform as it did at commissioning is building on sand. The honest approach is to model a gradual decline and to plan for interventions, treating them as part of the survey's cost rather than as an unwelcome surprise.

What a Survey Team Can Do Now

Audit throughput end-to-end before writing a proposal. Measure the full train with a calibrated source, including the detector, rather than quoting component specifications from vendor sheets.

Model exposure-time cost per science target, and express the survey in nights and in currency. A proposal that states the cost per measured star is harder to cut than one that states a target count.

Budget pipeline and archive work as infrastructure, with staff lines and a maintenance schedule, not as a postdoc's side project. Calibration products age, and a survey without a maintained pipeline is a data set without a future.

Publish throughput curves alongside survey data, so the next team inherits a measured number rather than a nominal one. The site has made a related case for carrying methods across fields in a piece on turbulence models and galaxy formation, where the borrowed machinery was only as good as its calibration.

Treat operations as a first-class cost line in every proposal. A facility that cannot afford to keep its optics clean will not deliver the survey that its telescope time implies.