NIRCam Detector Performance
NIRCam detectors have been optimized to balance a variety of performance metrics such as read noise and dark current. Unless otherwise specified, the following values were measured during ground testing in a flight-like configuration. Because the instrument's operating temperature won't be known until on-orbit commissioning, some of these values are subject to change.
Correlated double sampling (CDS) read noise was measured in pairs of frames differenced to subtract bias structure and kTC (thermal) noise. The read noise measured in this difference of two images equals the read noise of a single image multiplied by sqrt(2).
Dark current is generally negligible in the short wavelength detectors (~2 e–/ks) and more significant in the long wavelength detectors (~35 e–/ks).
"Effective noise," including both read noise and dark current, was measured for 1,000 s integrations. The combination of ASIC-commanded parameters and tuning yields gains in electrons per analog-to-digital units (e–/ADU) near 2.
Tables 1 and 2 provide measurements obtained in cryogenic vacuum ground testing (CV3 in 2015/2016) followed by more detailed descriptions of the various parameters and measurement methods.
Table 1. Average NIRCam detector properties measured in ground testing
|Parameter||Short wavelength |
|Long wavelength |
|CDS read noise|
(correlated double sampling)
|16.2 ± 1.3 e–||13.5 ± 0.2 e–|
|Dark current in 1,000 s||1.9 ± 1.1 e–/ks||27 ± 5 e–/ks|
|Effective noise in 1,000 s |
(93 groups of full-frame RAPID)
|5.9 ± 0.5 e–||8.6 ± 0.1 e–|
2.05 ± 0.4
|1.82 ± 0.4|
|Well capacity (saturation level|
with superbias subtracted)
|105,750 ± 2,264 e–||83,300 ± 1,200 e–|
|Quantum efficiency (QE)||70% @ 0.6 µm|
80% @ 1.0 µm
90% @ 2.0 µm
|80% @ 3 µm|
90% @ 4 µm
60% @ 5 µm
|Interpixel capacitance (IPC)||0.53% ± 0.04%||0.59% ± 0.04%|
|Post-pixel coupling (PPC)||0.08% ± 0.02%||0.19% ± 0.03%|
|Persistence (for same exposure time|
as previous nearly saturated image)
Values are averaged across all detectors in each channel. Uncertainties are either standard deviations or uncertainties of individual measurements, whichever is larger. Gain uncertainties are estimated to be ~20%. Quantum efficiency as a function of wavelength is shown in Figure 1.
Table 2. Individual NIRCam detector properties measured in CV3
|Detector||CDS read noise |
|Effective noise in |
1,000 s (e–)
|Saturation raw (ADU)|
|A1||16.16 ± 0.15||0.3 ± 1.6||5.46 ± 0.07||2.08||60,300 ± 1,760|
|A2||14.88 ± 0.28||2.7 ± 1.9||6.56 ± 0.09||2.02||63,100 ± 600|
|A3||14.62 ± 0.15||2.8 ± 1.6||5.57 ± 0.06||2.17||59,300 ± 220|
|A4||14.99 ± 0.18||3.3 ± 1.8||5.46 ± 0.08||2.02||62,000 ± 1,330|
|B1||16.65 ± 0.08||2.2 ± 1.5||5.43 ± 0.04||2.01||62,300 ± 440|
|B2||18.06 ± 0.21||0.8 ± 1.5||6.18 ± 0.10||2.14||61,000 ± 360|
|B3||16.76 ± 0.13||1.7 ± 1.4||5.61 ± 0.08||1.94||62,500 ± 860|
|B4||17.70 ± 0.40||1.1 ± 1.8||6.54 ± 0.10||2.03||61,000 ± 520|
|A5||13.33 ± 0.20||33.5 ± 3.1||8.55 ± 0.08||1.84||59,200 ± 960|
|B5||13.61 ± 0.20||35.0 ± 2.5||8.57 ± 0.05||1.80||58,500 ± 470|
Table note: the short wavelength detectors are numbered 1–4 in each module (A and B). The long wavelength detectors, listed at the end (shaded in yellow), are number "5" in each module.
The gain values reported in Tables 1 and 2 are from per-pixel gain maps derived in the NIRCam detector laboratory at the University of Arizona using the photon transfer curve (PTC) method. These gains should be considered preliminary, with uncertainties on the order of 20%. Gains have not been corrected for interpixel capacitance.
The noise of a given pixel readout is dominated by contributions from thermal kTC noise, pixel read noise, and 1/f noise. Since kTC noise is simply a random pedestal offset that occurs during the pixel reset, this noise source is effectively removed via correlated double sampling (CDS) or fitting a slope to an integration ramp. Integrations with only a single group will be unable to remove this reset noise, incurring approximately 35 e– worth of kTC noise.
The SIDECAR ASICs1 generate significant 1/f noise during detector operations and signal digitization. The magnitude of this noise is highly sensitive to the ASIC temperature. Luckily, the majority component of the 1/f noise is common to all four detector amplifiers. Efficient use of the side reference pixels can effectively reduce the measured CDS noise by 1–3 e– depending on the magnitude of 1/f noise. Subarray observations that exclude the column reference pixels will be unable to track the 1/f contributions.
Pixel read noise occurs in the HgCdTe detector layer and is due to random thermal motion of electric charge in the pixel interconnects (Johnson noise). CDS read noise values were measured for multiple pairs of frames that were differenced to remove the array's static bias structure along with kTC contributions. The read noise for a single frame will be lower by a factor of sqrt(2). Quoted values include reference pixel correction, which reduces the 1/f noise component.
1 System for Image Digitization, Enhancement, Control, and Retrieval Application-Specific Integrated Circuit
All infrared detectors exhibit some level of dark (leakage) current. By virtue of the relatively low operating temperatures (~40 K) and light-tight environment, NIRCam's detectors show extremely low dark current values. Short wavelength detectors are consistent with 0.001 e–/s, while long wavelength detectors show moderately higher values (0.035 e–/s).
The "effective noise" values were measured using a number of 1,000 s dark integrations (full frame RAPID 1 with 93 groups). Linear fits were simultaneously applied to each pixel in a given integration, and the standard deviation of the slope parameter determines the effective noise. The results include contributions from both read noise and dark current. In general, the effective noise should follow the MULTIACCUM noise equation (Robberto 2010, Equation 1); however, a consistent excess of ~10% above the expected value was found in the measurements.
1Bold italics style indicates words that are also parameters or buttons in software tools (like the APT and ETC). Similarly, a bold style represents menu items and panels.
Quantum efficiency (QE)
Polynomial fits to quantum efficiency versus wavelength are shown for the average short wavelength detector and individually for each long wavelength detector (one in each module). (From filters version 4.0: April 2016.)
Interpixel capacitance (IPC)
Interpixel capacitance (IPC) is a type of capacitive coupling between neighboring pixels (Moore et al. 2006). This phenomenon is commonly confused with charge diffusion, which involves migration of charge carriers to adjacent pixel cells. Instead, IPC arises from the interaction of electric fields that alter the measured voltages.
In its simplest form, this effect can be parameterized as a convolution kernel described as a 3 × 3 matrix with values totaling unity. Measurements for NIRCam detectors show values consistent with an assumption of symmetric coupling: the kernel will have values of zero in the corners, some value α in the left, right, top, and bottom positions, and the remaining 1−4α in the central pixel. NIRCam arrays show α ranges from 0.005 to 0.006. In other words, approximately 2%–2.5% of the flux measured from a given pixel gets equally redistributed between the four adjacent neighbors.
Uncorrected IPC will attenuate the spatially measured Poisson noise, which results in overestimates of the conversion gain and reported quantum efficiencies.
Post-pixel coupling (PPC)
As the detector clocking scheme addresses and reads each pixel, the analog output signal requires a settling time to transition between the pixel values. If the settling time is longer than the time between analog-to-digital conversion (ADC) pulses, then the measured signal for a given pixel will have a value that has not fully transitioned to the real analog signal. This effect has been dubbed post-pixel coupling (PPC). PPC is measured to be ~0.1% and ~0.2% or less in the short and long wavelength detectors, respectively.
Near-IR detectors commonly exhibit latent images persisting between integration ramps. After array reset, pixels previously subjected to illumination show an anomalous charge accumulation rate that is initially high but decreases to dark current levels over time. The amount of accumulated persistence directly correlates with the stimulating flux as well as the pixel dwell time. Persistence is a deterministic phenomenon that can be corrected and removed from the science images. NIRCam detectors show the latent emission to be approximately 0.1% of total illumination (Leisenring et al. 2016).
NIRCam persistence studies were performed under lab conditions that differ from the final flight configuration. Due to optimized tuning runs, the final bias settings will be different compared to those used in the NIRCam detector lab. This difference will have an effect on the size of the full depletion region, subjecting flight observations to a slightly altered trap population than has been characterized.
Leisenring, J., Rieke, M., Misselt, K., et al., 2016, SPIE 99152N
Characterizing persistence in JWST NIRCam flight detectors
Moore, A.C., Ninkov, Z., Forrest, W.J., 2006, Optical Engineering, 45, 7
Quantum efficiency overestimation and deterministic cross talk resulting from interpixel capacitance
Robberto, M., 2010, JWST-STScI-002100, SM-12
NIRCAM Optimal Readout II: General Case (Including Photon Noise)