Detector Technologies#

ZSpec uses quantizing CCDs (qCCDs) in its three visible channels and linear-mode avalanche photodiode arrays (LmAPDs) in its three infrared channels. ZImager uses three ORCA-Quest 2 qCMOS cameras for simultaneous multiband imaging.

Detector baseline#

Subsystem	Preferred detector	Active format and pixel size	Fallback / status
ZSpec VIS: B, G, R	STA5900 qCCD with tapered ALD anti-reflection coating	2,048×4,096; 15 µm	STA5500 qCCD, then conventional STA or Teledyne-e2v CCD
ZSpec NIR: YJ, H, K	Leonardo LmAPD; Ike Pono and IBEX are candidates	2,048×2,048; 15 µm target	H2RG/H4RG-family HgCdTe array
ZImager: three optical bands	Hamamatsu ORCA-Quest 2 qCMOS	4,096×2,304; 4.6 µm	Commercially available; cameras in use

Each ZSpec channel has one detector; no detector mosaics are required. The VIS qCCD frame store is additional to the active image area.

ZSpec VIS: qCCDs#

Skipper readout samples the same charge packet nondestructively. For independent samples, read noise falls approximately as 1/√N, approaching single-electron charge resolution. ZShooter’s preferred architecture combines differential signal/reference sampling, many parallel outputs, and a frame store that allows readout while the next exposure integrates.

Device	Role	Readout performance
STA5500	Available qCCD fallback; 4,096×2,048 image area and matching frame store	Approximately 1 e⁻ read noise in 13 s
STA5900	Preferred faster device; bulk red-sensitive and epitaxial blue-optimized versions	Target approximately 1 e⁻ in 2.8 s; charge quantization in approximately 120 s

Low read noise supports short sub-exposures for cosmic-ray rejection and changing conditions, while retaining high native spectral resolution. OH-contaminated samples can be down-weighted before binning to the science resolution. Integer-electron peaks provide calibration of gain, zero signal, linearity, and serial clock-induced charge, with potential brighter-fatter diagnostics. On-chip binning remains available when the observation can trade native sampling for sensitivity.

Tapered AR coatings#

The fixed echellogram maps a narrow wavelength range to each detector location. A spatially varying coating can match that local range, targeting less than 1% reflectance and reducing ghosts, fringing, and scattered light.

VIS channel	Coating passband	Full channel span
B	310–430 nm	120 nm
G	430–650 nm	220 nm
R	650–980 nm	330 nm

The initial coating study uses a graded Al₂O₃ layer bounded by HfO₂, deposited by atomic layer deposition (ALD). The thickness gradient and local bandwidth must be matched to the current 2k×4k echellograms; process calibration and repeatability remain development work.

Conventional CCDs preserve wavelength coverage. Their read-noise and QE penalties are included in the fallback comparison below.

ZSpec NIR: LmAPDs#

HgCdTe LmAPDs multiply photoelectrons before the readout integrated circuit (ROIC) adds noise. This reduces effective read noise in short integrations and between OH lines. Leonardo’s Ike Pono and IBEX are candidate implementations for the three NIR channels.

The current ZShooter model and design targets are:

Parameter	LmAPD baseline	HxRG comparison
Format	One 2k×2k, 15 µm array per YJ/H/K channel	H2RG/H4RG-family options
Effective read noise	<1 e⁻ RMS in a 60 s sampling-up-the-ramp exposure	H2RG comparison: 6 e⁻
QE	Approximately 80% effective QE, including the avalanche excess-noise penalty	Approximately 95% nominal QE
Dark current	Target <3.6 e⁻ pixel⁻¹ h⁻¹, excluding glow	H2RG comparison: 23 e⁻ pixel⁻¹ h⁻¹

Avalanche multiplication introduces an excess-noise factor of approximately 1.2 in the project model. Its S/N penalty is represented by the reduced effective QE. In K, sky and thermal shot noise reduce the advantage of low read noise; the higher-QE H2RG is slightly deeper in the one-hour comparison below.

Earlier LmAPD testing reported glow of approximately 0.012 e⁻ pixel⁻¹ frame⁻¹ with a modified pixel structure (Huber et al. 2024). This is a measurement from the development program, not a specification for either current candidate array.

Leonardo update — detector team contribution needed. Leonardo has made significant progress; both Ike Pono and IBEX may meet ZSpec’s needs, and quotation requests are being accepted. Add the latest device results and compare the two options: available format and ROIC, read noise versus gain and cadence, QE/excess noise, dark current and glow, persistence, and uniformity. Record packaging, procurement/NRE terms, and delivery estimates. Update the model assumptions where the new results support a change.

ZImager: qCMOS#

The ORCA-Quest 2 provides 4,096×2,304 pixels at 4.6 µm, with 0.30 e⁻ RMS read noise in ultra-quiet scan mode and improved UV QE relative to the original ORCA-Quest. Three cameras support acquisition, spectral flux normalization, and simultaneous multiband time-series photometry. Windowed readout supports high-cadence observations; the 0.30 e⁻ figure applies to the ultra-quiet mode, not every readout rate.

The ZImager design gives the optical layout, detector scale, and simulator parameters. Camera mode, subarray size, cadence, and read noise must be qualified together for the timing modes, including exposure synchronization, timestamp accuracy, and sustained data rates.

Detector fallback comparison#

The saved dark-time calculation uses one hour on source, a 0.70″ slit, 0.65″ natural seeing, airmass 1.1, and R≈1,000. OH emission is included, with independent S/N optimization for each detector choice. Values are medians over complete matched spectral bins.

Channel	Conventional detector	Limiting magnitude relative to the qCCD/LmAPD baseline
B	CCD	1.06 mag shallower
G	CCD	0.46 mag shallower
R	CCD	0.34 mag shallower
YJ	H2RG	0.53 mag shallower
H	H2RG	0.25 mag shallower
K	H2RG	0.06 mag deeper

The calculation includes an equal-duration independent sky reference and the different detector QE curves. It bins after readout; it does not include CCD on-chip binning. The Performance page gives the baseline limiting magnitudes and a separate detector break-even calculation that does include on-chip binning.

Outstanding detector work

qCCD readout: verify STA5900 noise versus readout time, clock-induced charge, frame-transfer overhead, and availability; retain STA5500 and conventional CCD alternatives.
Tapered coatings: derive thickness gradients and instantaneous bandwidths from the current B/G/R echellograms; verify reflectance, calibration, and fabrication repeatability.
LmAPDs: complete the Ike Pono/IBEX comparison above, including detector/controller packaging and the operating gain, temperature, and sampling cadence for YJ, H, and K.
qCMOS: qualify readout modes, synchronization, timing, photometric stability, and sustained data rates for the selected subarrays.

Technical references