AXIS High-Speed X-ray Camera

• AXIS High-Speed Camera is a photon-counting X-ray CCD instrument that achieves arcsecond-resolution imaging with rapid readouts (>5 fps and up to ~14 fps) to minimize photon pile-up.
• It employs a 2×2 mosaic of 1440×1440 back-illuminated MIT/LL CCID100 detectors with 24 μm pixels to deliver high throughput and maintain spectral resolution (<150 eV FWHM at 6 keV).
• The system integrates custom ASICs, digital waveform capture, and FPGA-based event recognition and transient alert modules to enable versatile, time-domain X-ray observations.

The AXIS High-Speed Camera is the sole focal-plane instrument of the Advanced X-ray Imaging Satellite (AXIS). It is a fast, photon-counting X-ray CCD camera intended to preserve AXIS’s large collecting area, low background, and near-arcsecond imaging while also supporting rapid transient response. In the AXIS overview, the camera is described as producing position-, energy-, and time-tagged photon detections that can be formed into images, spectra, and light curves, and later Phase A design papers refine the concept around MIT Lincoln Laboratory CCDs, Stanford ASIC readout, digitally sampled front-end electronics, and FPGA-based onboard event recognition and transient alerting (Reynolds et al., 2023).

1. Mission function and observatory context

AXIS is framed as a Probe-class successor to the high-angular-resolution X-ray imaging tradition of Chandra, but with a much larger high-resolution field and much higher throughput. The camera exists because AXIS couples fine imaging to substantially higher throughput than earlier soft X-ray imaging spectrometers, so the focal plane must operate much faster than Chandra ACIS and Suzaku XIS in order to limit photon pile-up while retaining CCD-class spectral performance (Miller et al., 2023).

In the 2023 overview, the mission is characterized as delivering low-background, arcsecond-resolution imaging over 0.3–10 keV across a 24 arcmin diameter field of view, with the camera read out at $>5$ frames/s so that spacecraft jitter does not irreducibly blur reconstructed images (Reynolds et al., 2023). The 2025 Phase A update keeps the camera in the same instrumental role but states the formal detector requirements as frame rate $\ge 5$ fps, goal $\ge 20$ fps, pixel size $24\,\mu\mathrm{m}$, read noise $\le 3\,e^-$ RMS, and focal plane temperature $-110 \pm 0.1^\circ\mathrm{C}$, with mission spectroscopy requirements of $\le 70$ eV FWHM at 0.5 keV, $\le 100$ eV FWHM at 1 keV, and $\le 150$ eV FWHM at 6 keV (Miller et al., 19 Aug 2025).

The concept also evolved at observatory level. The 2023 overview described a low-inclination low-Earth orbit chosen to support rapid communications, low detector background, and slow detector degradation (Reynolds et al., 2023). By the 2025 design update, the camera is being described in the context of an L2 halo orbit, which drives changes in thermal design and motivates a commandable shutter for protection against solar energetic particles (Miller et al., 19 Aug 2025). This shift is significant because the camera architecture is tightly coupled to radiation environment, background control, and thermal stability.

2. Focal-plane architecture and detector design

The core architecture is a $2\times 2$ mosaic of back-illuminated X-ray CCDs. The overview paper states that AXIS uses a Focal Plane Assembly (FPA) built around a $\ge 5$0 array of back-illuminated MIT/Lincoln Laboratory X-ray CCDs, each with $\ge 5$1 pixels and $\ge 5$2 square pixels (Reynolds et al., 2023). With the AXIS optics, those pixels were described in 2023 as 0.5 arcsec square pixels, intentionally oversampling the PSF by a factor of 2–3; the 2025 design update quotes 0.55 arcsec per pixel for the CCID100 implementation (Reynolds et al., 2023).

The same update identifies the flight detector as the MIT Lincoln Laboratory CCID100, a back-illuminated frame-store CCD with $\ge 5$3 active pixels, $\ge 5$4 pixels, 100 $\ge 5$5 thickness, 16 p-channel JFET outputs, MBE passivation, and a thin aluminum layer directly deposited on each CCD for optical blocking (Miller et al., 19 Aug 2025). The frame-store region is shielded by an aluminum cover so that an exposure can be transferred rapidly into storage before serial readout. The focal plane consists of four independent detectors, each with its own ASIC and electronics chain, and the 2025 design explicitly notes that failure of one detector does not affect the other three (Miller et al., 19 Aug 2025).

The detector technology is presented as an evolution of established MIT/LL X-ray CCD lineage rather than a wholly new sensor class. The 2023 camera design paper emphasizes two enabling changes relative to earlier MIT/LL devices: a single-layer polysilicon gate structure, which allows fast, low-power clocking, and a two-stage pJFET output amplifier, which provides similar noise at about 10× faster readout rate than older on-chip MOSFET designs (Miller et al., 2023). The backside is passivated by molecular beam epitaxy (MBE) depositing a 5–10 nm heavily doped silicon layer, intended to preserve soft-X-ray charge collection near the entrance window (Miller et al., 2023).

Later calibration papers add further device-level details. The CCID100 is described there as fully depleted, 100 $\ge 5$6 thick, with 16 output ports using 2-stage pJFETs, an on-chip Al optical blocking filter, and radiation-tolerance features including a trough and charge injection (Grant et al., 19 Aug 2025). A plausible implication is that the mature AXIS camera concept should be understood as a large-format, multi-output, frame-store CCD spectro-imager optimized simultaneously for angular resolution, count-rate handling, and broad-band X-ray redistribution calibration.

3. Readout electronics, waveform sampling, and onboard autonomy

The detector is coupled to a custom Stanford ASIC, the MCRC (Multi-Channel Readout Chip). In the 2023 camera design, MCRC v1.0 is fabricated in a 350 nm process and provides 8 channels per chip. Each channel includes selectable gain, low-noise analog amplification, CCD output biasing via integrated current sources, and differential buffering for transmission to ADCs. Reported ASIC characteristics include achievable pixel rate 5 Mpix/s per channel, voltage gain 6.2 V/V or 12.1 V/V, input-referred noise 1.63 $\ge 5$7 RMS, channel-to-channel crosstalk $\ge 5$8 dB, power consumption $\ge 5$9–35 mW per channel, and radiation tolerance $\ge 20$0 krad (Miller et al., 2023). The 2025 update retains the same readout concept but describes the packaged implementation as an 8-channel ASIC of size 4.2 mm $\ge 20$1 2.9 mm, with dual 8-channel MCRC-V1 ASICs used for the 16-channel CCID100 (Miller et al., 19 Aug 2025).

The front-end architecture is explicitly digital-first. In the 2023 concept, the Front End Electronics (FEE) digitize the CCD video waveform at 40 Msamples/s and apply digital processing to preserve low noise at high speed (Miller et al., 2023). In the 2025 design update, each Camera Control (CC) board uses a Microchip PolarFire FPGA and Microchip ADCs for 50 Msamples/s digital waveform capture, then processes the waveform into a raw image (Miller et al., 19 Aug 2025). The FEE has been reorganized so that one CC board handles one CCD + ASIC pair, with four identical CC boards split across two identical boxes, partly because the analog flex cables must be $\ge 20$2 cm (Miller et al., 19 Aug 2025).

Downstream processing occurs in the Back End Electronics (BEE). The 2023 camera paper describes FPGA-based Event Recognition Processor (ERP) boards that perform bias correction, bad-pixel masking, local-maximum detection, $\ge 20$3 neighborhood extraction, multiplicity or grade determination, event filtering, and telemetry packaging (Miller et al., 2023). The AXIS overview states that the ERP greatly reduce[s] the telemetry stream, making a fast, large-format CCD system operationally practical (Reynolds et al., 2023). The 2025 update gives the same subsystem a mission-level role: it performs high-speed identification of candidate X-ray events, reduces telemetry by several orders of magnitude, and monitors for transients (Miller et al., 19 Aug 2025).

The Transient Alert Module (TAM) is the camera’s main autonomy element for time-domain astronomy. It operates continuously during science operations and triggers on thresholds identifying either new sources or sources that have varied dramatically from their baseline within the AXIS field. The resulting alert packet contains source location, flux, rise time, and spectral hardness; source localization is stated to be accurate to $\ge 20$4 arcsec, with delivery through a low-latency commercial L-band service and community dissemination in $\ge 20$5 minutes from initial detection (Reynolds et al., 2023). The same operational framework supports external target-of-opportunity response through Transient Broker Networks, with AXIS able to be on source in $\ge 20$6 hours from receipt of an alert at the science operations center and quick-look science data available within $\ge 20$7 hours (Reynolds et al., 2023).

4. Readout cadence, operating modes, and measured performance

Fast readout is not treated as an optional enhancement. AXIS is a photon-counting instrument that reconstructs images on the ground by registering and stacking individual events, and the overview explicitly states that low-noise reaction wheels and fast CCD readout ($\ge 20$8 frames per second) are required so that spacecraft jitter does not blur the reconstructed image. A metrology system tracks relative mirror–focal-plane motion on the same cadence as the CCD readout, embedding frame rate directly into the observatory’s imaging architecture (Reynolds et al., 2023).

The baseline cadence was stated in 2023 as 5 fps, with a “current best estimate” of 7 fps for a full frame using eight CCD outputs and ASIC channels per detector running at 2 MHz and a parallel transfer speed of 1 MHz; that operating point was associated with an out-of-time fraction of 0.5% (Miller et al., 2023). By 2025, the CCID100 design is described with 16 outputs, 2 MHz serial operation, 1 MHz parallel transfer, and an expected frame rate of approximately 14 fps, comfortably above the baseline 5 fps requirement though still below the $\ge 20$9 fps mission goal (Miller et al., 19 Aug 2025). Across these papers, the progression is from an 8-output baseline sufficient for 7 fps to a 16-output implementation expected to deliver $24\,\mu\mathrm{m}$0 fps.

The camera can also be operated in brighter-source or faster-timing modes. The 2023 overview states that the team is confident of meeting a goal of 20 fps using some combination of increasing the number of outputs, increasing the output rate to 5 MHz, reading only a small sub-array of the aimpoint detector, or using continuous-clocking mode (Reynolds et al., 2023). The same paper notes that 5 MHz operation may come at the cost of increased noise and reduced soft X-ray response, so it is framed as a science-dependent option. Continuous clocking is the most explicit high-time-resolution mode: it eliminate[s] all spatial information along CCD columns but improves time resolution to the sub-ms regime (Reynolds et al., 2023).

Prototype measurements show that the speed goal is being pursued without abandoning X-ray spectroscopy. In the 2023 camera paper, a back-illuminated CCID-89 prototype at 2 MHz and about 220 kHz parallel transfer yielded noise <2.5 $24\,\mu\mathrm{m}$1 RMS on six of eight outputs at $24\,\mu\mathrm{m}$2C, and a representative Mn K$24\,\mu\mathrm{m}$3 line width of 137 eV FWHM at 5.9 keV, meeting the AXIS $24\,\mu\mathrm{m}$4 eV FWHM at 6 keV requirement (Miller et al., 2023). The 2025 update reports that on a back-illuminated CCID89, all eight output nodes met the $24\,\mu\mathrm{m}$5 RMS requirement at 1 MHz, and that all 1 MHz spectral FWHM measurements across the AXIS energy range met baseline requirements (Miller et al., 19 Aug 2025).

A complementary Stanford prototype program couples CCDs directly to the MCRC ASIC. The 2025 readout paper reports CCID-93 + MCRC performance of 2.31 $24\,\mu\mathrm{m}$6 read noise and 121 eV FWHM at 5.9 keV at 173 K and 2 MPixels/s after bias optimization, with operation demonstrated up to 5 MPixels/s, where the same chain produced 3.84 $24\,\mu\mathrm{m}$7 and 125.8 eV FWHM at 5.9 keV (Stueber et al., 19 Aug 2025). That paper also introduces a 4D bias-optimization method over RGH, RGL, OG, and RD, using 400 combinations and about 45 minutes per scan, which is especially relevant because the full CCID100 has 16 output channels that may require per-node tuning (Stueber et al., 19 Aug 2025).

5. Calibration philosophy and technology maturation

AXIS calibration is explicitly inherited from the methodology developed for Chandra/ACIS and Suzaku/XIS, but adapted to the fast, many-output CCID100 architecture. The calibration plan is centered on pre-integration CCD-level characterization rather than post-assembly full-system recalibration, and its purpose is to generate the response information needed for RMFs and ARFs (Grant et al., 19 Aug 2025). In standard X-ray terminology, the Redistribution Matrix File (RMF) encodes spectral redistribution, including the Gaussian core and off-peak structures, while the Ancillary Response File (ARF) encodes effective area as a function of energy and source position.

The screening campaign uses a dedicated vacuum chamber capable of stable operation down to $24\,\mu\mathrm{m}$8, with an Archon controller used in a mode intended to mimic the flight front end (Grant et al., 19 Aug 2025). The initial high-energy source is $24\,\mu\mathrm{m}$9Fe, providing Mn-K$\le 3\,e^-$0 at 5.9 keV and Mn-K$\le 3\,e^-$1 at 6.4 keV; low-energy screening uses $\le 3\,e^-$2Po + Teflon, producing C-K at 277 eV and F-K at 677 eV (Grant et al., 19 Aug 2025). The notional full calibration set extends across the AXIS band with C-K: 277 eV, O-K: 525 eV, Al-K: 1.5 keV, Mn-K: 5.9 keV, and Cu-K: 8.0 keV, using an In-Focus Monochromator (IFM) for 175 eV to 1.5 keV and fluorescence targets plus commercial X-ray sources at higher energies (Grant et al., 19 Aug 2025).

The calibration products are intended to characterize gain, offset, read noise, spectral resolution, line shape, charge transfer inefficiency (CTI), cosmetic defects, dark current, quantum efficiency (QE), and behavior in off-nominal modes (Grant et al., 19 Aug 2025). The paper states that good flight devices are expected to have CTI values of a few $\le 3\,e^-$3 or lower and “close to zero,” and that devices are screened against read noise $\le 3\,e^-$4 RMS, spectral resolution $\le 3\,e^-$5 eV FWHM at 6 keV, and $\le 3\,e^-$6 eV at 0.5 keV (Grant et al., 19 Aug 2025). Rather than fitting only Gaussian widths, the plan is to accumulate long exposures—using the ACIS heuristic of at least 10,000 events in each X-ray line and each region of interest—and fit a physics-based detector response model including fluorescence peaks, escape peaks, and low-energy shoulder or hump structures (Grant et al., 19 Aug 2025).

A notable innovation concerns QE transfer. Heritage Chandra and Suzaku campaigns used an absolutely calibrated CCD reference detector at the PTB beamline at BESSY, but the AXIS packaged CCD+ASIC geometry makes the older two-CCD chamber arrangement impractical. The new plan is to use an sCMOS reference detector from the Sony STARVIS family, absolutely calibrated at BESSY-II, and then mounted on a translation stage within the AXIS calibration chamber so that the same beam can be measured sequentially by the reference detector and the test CCD (Grant et al., 19 Aug 2025). The CCD QE model is described as a “slab and stop” model, with standard absorption and transmission curves and fitted layer thicknesses (Grant et al., 19 Aug 2025).

The development program is correspondingly staged. Prototype devices such as CCID93, CCID94, and CCID89 are used to understand speed, noise, soft response, and ASIC coupling; the first lot of prototype CCID100 detectors had completed fabrication by the 2025 update and was about to enter X-ray testing (Miller et al., 19 Aug 2025). After full camera assembly, however, the calibration plan states that no further large-scale ground calibration activities are planned, beyond aliveness and performance verification with a $\le 3\,e^-$7Fe source in the camera door (Grant et al., 19 Aug 2025). This places unusual weight on the pre-integration CCD-level campaign.

6. Scientific role, heritage, and interpretive issues

The scientific drivers behind the camera are explicitly time-domain and high-throughput. The AXIS overview links fast detector operation to supernova shock breakouts, gravitational-wave counterparts from binary neutron star mergers, Galactic Plane time-domain surveys for ultra-short-period mass-transferring double white dwarfs, and studies of variability, flares, and long-term activity in exoplanet host stars and stellar clusters (Reynolds et al., 2023). The onboard TAM and the fast target-of-opportunity framework are direct consequences of those science cases rather than auxiliary mission features.

In heritage terms, AXIS positions the camera between two earlier X-ray traditions. Relative to Chandra, the novelty is much faster CCD operation while preserving high-resolution imaging spectroscopy and expanding the high-resolution field; the AXIS overview characterizes the observatory as delivering almost two orders of magnitude greater high-resolution survey grasp than Chandra (Reynolds et al., 2023). Relative to Swift, the relevant comparison is time-domain capability rather than detector similarity: AXIS is described as having $\le 3\,e^-$8 the sensitivity of Swift and as extending Swift’s transient legacy into a much more sensitive imaging regime (Reynolds et al., 2023).

Several recurring misconceptions are addressed implicitly by the design papers. First, the AXIS High-Speed Camera is not a conventional optical high-speed area camera. It is a photon-counting X-ray imaging spectrometer, and its “high speed” refers to fast frame-store CCD operation, pile-up control, event-based telemetry, and transient response rather than video-style full-frame imaging. Second, the concept is not portrayed as a speculative detector replacement for CCD astronomy. The overview characterizes it as enabled by large-format, small-pixel, high readout rate CCD detectors with good spectral resolution, building on “a long line of successful space instruments spanning the last three decades,” while locating the principal innovations in the ASICs, digital waveform capture, event processor, and transient alert module (Reynolds et al., 2023).

A third interpretive issue concerns contamination, radiation environment, and soft response. The AXIS overview states that contamination control is critical to preserving soft-X-ray response and describes a warm contamination blocking filter at $\le 3\,e^-$9 protecting CCDs cooled to $-110 \pm 0.1^\circ\mathrm{C}$0, explicitly as a lesson learned from Chandra/ACIS (Reynolds et al., 2023). The later L2 design update retains contamination control but reworks the thermal architecture around a $-110 \pm 0.1^\circ\mathrm{C}$1 passive interface, trim heaters at $-110 \pm 0.1^\circ\mathrm{C}$2, a freestanding aluminum + polyimide contamination blocking filter on 95% open stainless-steel mesh, and a 4 mm aluminum shutter for radiation protection and non-X-ray background measurements (Miller et al., 19 Aug 2025). This suggests that the high-speed camera is best understood not as an isolated detector package, but as a tightly integrated sensor, electronics, thermal, contamination-control, and autonomous-operations system whose defining purpose is to keep CCD-grade X-ray imaging spectroscopy viable in a much higher-throughput observatory regime.