NIRISS Detector Performance

Various fundamental properties of the JWST NIRISS detector that affect the overall science performance were measured in ground testing in late 2015 and early 2016.

On this page

See also: NIRISS Detector Readout, NIRISS Detector Readout Patterns

The NIRISS detector is a 2048 × 2048 HAWAII-2RG device with a wavelength cut-off around 5 μm, in the same "family" of detectors as the 2 NIRSpec detectors and the 2 guider detectors. It is also similar to the NIRCam long wavelength devices.

The NIRISS detector and the application-specific integrated circuit (ASIC) have been tested in flight-like configuration during the Integrated Science Instrument Model (ISIM) cryo-vaccum test 3 (CV3), which took place between November 2015 and February 2016. The specific ASIC settings were selected to minimize the channel-to-channel cross talk observed in full frame readout. The detector properties reported here are based upon analysis of the CV3 test data, with one exception. The measurements given here were generally taken at a detector temperature of about 38 K, which is close to the expected on-orbit temperature.

Various detector properties are observed to be different in the epoxy voids, which cover about 20% of the total detector area. Where possible, values are given for the void regions and for the main area of the detector.

Summary table

Words in bold are GUI menus/
panels or data software packages; 
bold italics are buttons in GUI
tools or package parameters.

Table 1. Values for a detector temperature of 38 K, except for the total noise which was determined at 44 K

QuantityValueValue in the void (if applicable)
Read noise11.36 ADU or 18.32 e
Total noise8.7 e– for a 100 group NISRAPID integration
Gain1.61 e/ADU
Inter-pixel capacitance0.54% to each row/column adjacent pixel0.29% to each row/column adjacent pixel
Dark current0.0257 ± 0.0021 ADU/s0.0125 ± 0.0015 ADU/s
Full well capacity95,000 to 105,000 e
Pedestal drift22 e or 13.6 ADU

Further details are given below.

Read noise

Measurements of the read noise were derived from taking correlated double sample (CDS) images from long dark integrations and finding the average dispersion about zero. This measurement includes the read noise as well as some contribution from the 1/f noise and a small contribution from the dark current signal, although the measured value is expected to be dominated by uncorrelated 1/f noise and the read noise. The average value of the CDS read noise was found to be 11.36 ADU or 18.32 e. These values are for raw integrations before the reference pixel correction. The read noise per frame is therefore about 8.03 ADU or 12.95 e. For a single frame exposure the kTC noise will also be present. The observed dispersion in the first frames of raw dark exposures corresponds to a total read noise and kTC noise of 40.2 ADU or about 65 e. This dispersion is higher than expected; the kTC noise is expected to be of order 30 e.

In full frame readout, each channel shows offsets from frame to frame, called the pedestal drift. The pedestal drift for NIRISS is about 22 e or 13.6 ADU. This is mostly removed by the reference pixel correction.

Total noise

Figure 1 shows the total noise as a function of the number of groups in an integration, for up to 100 groups in NISRAPID readout (a total exposure time of 1,062.9 s). This analysis was done using dark ramps taken at 44 K rather than at 38 K. Analysis of the CV3 38 K dark ramps suggests that the total noise is slightly higher than the values shown in Figure 1. The derived total noise for a 100 group NISRAPID integration, or a 25 frame NIS integration, is 8.7 e. The "read noise" is determined by removing the dark current component from the total noise, but this value includes read noise and 1/f noise. Hence the "read noise" values so obtained are somewhat larger than the read noise value determined from the CDS frame analysis given above.

Figure 1. Total noise for NIRISS detector as a function of integration time

The derived total noise of the NIRISS detector at 44 K from CV3 data. The total noise includes the read noise, 1/f noise, and the dark current component. The curve labeled "read noise" is the total noise minus the estimated dark current contribution, and so it includes both read noise and the 1/f noise. The dashed curve is the theoretical noise function from Robberto (2009).


The NIRISS detector gain was measured with the photon transfer method in a large set of integrations with close to uniform illumination. The average gain value was found to be 1.61 e/ADU, with roughly a 10% uncertainty. The gain value has been corrected for the inter-pixel capacitance (IPC) effects. As the IPC is smaller in the voids, the raw measured gain before correction is slightly larger in the voids than outside the voids. However the net gain after the correction appears to be the same inside and outside of the voids.

Dark current

The average dark current measured at 38 K was 0.0125 ± 0.0015 ADU/s in the voids and 0.0257 ± 0.0021 ADU/s for the rest of the detector. In electrons, the corresponding rates are 0.0202 ± 0.0024 e/s and 0.0414 ± 0.0034 e/s respectively. Figure 2 shows the dark current across the NIRISS detector.

Figure 2. Dark current across the NIRISS detector

The areas of low dark current are the epoxy voids. There are many small voids in channel 1 and a single large void, in channels 2 and 3, on the right.

Inter-pixel capacitance

The inter-pixel capacitance (IPC) was measured from isolated hot pixels or cosmic ray hits in the CV3 dark ramps. Averaging over many such pixels, the IPC effect appears to be symmetric from the central pixel to the 4 adjacent pixels along the row or column. The average magnitude of the coupling is 0.29% (per pixel, so 4 times this overall) in the voids and 0.54% outside the voids.

The post-pixel coupling adds about 0.15% to the signal observed in the pixel next along in the read direction. Thus the "raw" IPC values show an asymmetry in the read direction that differs between channels 1 and 3 and channels 2 and 4 of the detector.

Detector full well capacity

The detector full well capacity was measured in the detector testing at Teledyne Imaging Sensors prior to delivery. The general full well depth is about 105,000 e. With the gain setting selected for flight the full well capacity is not reached for some pixels because they pass the A/D conversion limit of 65,535 ADU before the full well depth. About half of the detector pixels are subject to A/D saturation and the other half reach full well before A/D saturation, typically between 95,000 and 105,000 e. Due to the non-linearity effects, the raw full well value in raw ADU is generally in the range between 60,000 and 65,535 ADU. After subtraction of the pedestal, the saturation values range from 42,000 ADU to 57,000 ADU as there is significant structure in the pedestal.


Robberto, M., 2009 JWST-STScI-001853
Derivation of the correct noise equation for general MULTIACCUM readout

Latest updates
Originally published