MIRI Detector Performance

JWST MIRI's detector performance characteristics, such as read noise, dark current, saturation level, and persistence, have all been measured in flight during JWST Commissioning and Cycle 1 calibrations. 

On this page

The MIRI detectors offer relatively low dark current, read noise, and persistence. These types of detectors were used very successfully on all 3 of the Spitzer instruments. However, compared with detectors used in the optical and near-infrared, the MIRI detectors have a number of additional calibration challenges (Ressler et al. 2015, Rieke et al. 2015).

Table 1 provides a summary of measurements obtained in instrument commissioning and Cycle 1 calibration studies, followed by more detailed descriptions of the various parameters and measurement methods.

The values in Table 1 can provide some guidance, but users should ultimately use the Exposure Time Calculator for all sensitivity calculations. 


Table 1. Average MIRI detector properties measured in flight data

Pixel gain (e/ADU)~4

Dark current (e/s)

Read noise (DN)~6
Latent images (%) < 1

Full well (e)


Pixel gain

Conversion gain (e/ADU), is a fundamental parameter in detector characterization that is used to measure many detector properties, including quantum efficiency (QE), dark current, and read noise. Measuring these parameters to a higher precision is becoming more important as the demand for low signal observations increases and the scientific requirements evolve in complexity. The MIRI detectors have a gain of ~4 e/ADU.  

Dark current

On the ground and in flight, MIRI darks were obtained with the contamination control cover closed to make the instrument interior as dark as possible (although, as always, one can measure only upper limits to the true dark current, given the possibility of photon leaks). In processing the data, the first and last frames of an integration ramp were rejected, to circumvent the effects of the reset anomaly and last-frame effect. Dark currents were then determined by the slopes of the integration ramps over a 100 × 100 pixel region selected to avoid bad pixels. The slope calculations included all exposures in a test run; the effects of settling of the detector output artificially elevate the apparent dark current, again making the results upper limits. 

Measurements of the dark rate show that it is variable with time, ~ ± 0.5 DN/s from the median. This can have the effect of imprinting structure on images taken in the shorter wavelength filters that have low levels of sky background. Efforts are on-going to model this effect. 

Read noise

In ground testing, the read noise was measured by setting the detector bias voltage to 0 V; in this way, any dark or photocurrent was eliminated and only the voltage noise of the readout output amplifiers were measured. Experiments leaving the detector bias at 2 V were also performed to look at noise in the dark frames. The results were identical until the temperature climbed to 7.5 K. At this point, shot noise from the dark current began to dominate the noise, and the total noise began to exceed the requirement. 

In flight, the read noise was measured using the mean variation in correlated double samples (CDS) of dark images. This removes any structure in the dark rates and leaves only the voltage noise in the readout amplifiers. 


Quantum efficiency (QE)

Spectral quantum efficiency in impurity band conduction (IBC) IR devices is often referred to as "quantum yield" since it's only possible to measure the quantum efficiency times the internal gain. The QE of the detectors cannot be measured trivially in detector array form, so it's necessary to rely on measurements of test structures included on each wafer from which the flight detectors are selected.

Figure 1. MIRI detector quantum efficiencies

NIRCam detector quantum efficiencies

Measured responsive quantum efficiency of bare detector material (solid line). The dashed line is a computed result assuming the array has an antireflection coating applied optimized for 6 μm, and the dotted line is for an AR coating optimized for 16 μm. © Rieke et al. 2015.


Bright sources leave latents on the MIRI arrays, typically at a level of about 1% immediately after the source has been removed. The decay of these images shows multiple time constants, suggesting that there are a number of mechanisms that contribute to the effect. Further characterization of this complex behavior is needed to determine ways to correct it in the MIRI pipeline. 

Imaging properties

The response of the arrays is uniform, with pixel-to-pixel variations of no more than 3% rms. The MIRI imager has a small proportion of inoperative pixels (either hot, warm, or dead), of order <2%. In the MIRI imager detector, measurements indicate a level ~3% for the cross talk to the two vertical adjacent pixels around one receiving signal, and ~4% to the horizontal pixels. A plausible cause of this behavior is interpixel capacitance, although there may be secondary contributions from electron diffusion and optical effects. In addition, at 5.6 and 7.7 μm there is an additional cross-like imaging artifact known as a "cruciform."

The Mid-Infrared Instrument for the James Webb Space Telescope, VIII: The MIRI Focal Plane System

Rieke, G. H.  et al. 2015, PASP, 127, 665
The Mid-Infrared Instrument for the James Webb Space Telescope, VII: The MIRI Detectors

Latest updates
    Updated based on in-flight measurements.
Originally published