JWST Time-Series Observations Noise Sources

Determining the precision of time-series observations requires a thorough understanding of noise sources affecting the observations.

On this page

Detecting faint signal modulations in a time-series observation requires high precision (spectro-) photometric measurements. The highest achievable precision is to reach the photon noise floor of the observation, i.e., where the noise budget is dominated by the Poisson noise of the source and background.

Key to reaching the highest precision measurements is a thorough understanding of the noise sources affecting the observation, and the correlations between them. Sources of noise in any observation can be astrophysical or instrumental in nature. A faint exoplanet transit signal can, for example, be masked by flux variations from starspots or other types of stellar variability.

This article provides an overview of the noise sources arising from the spacecraft and instrumentation for time-series observations, and their observed temporal behavior. 

Spacecraft noise sources

See also: JWST Pointing Performance, JWST Communications Subsystems, JWST Background Model, JWST Fine Guidance Sensor

Spacecraft pointing drifts and jitter

Pointing control and slewing of JWST is performed by the attitude control subsystem (ACS). Fine guiding additionally involves the JWST Fine Guidance Sensor (FGS). Changes in the spacecraft pointing can cause the target to shift across the detector, causing correlated noise or discontinuities in the measured time series. Such motions can arise from a number of causes and over varying timescales. Pointing issues and their impact on time-series observations for other missions, such as Kepler and Spitzer, are provided in Beichman et al. (2014).

Once fine guiding has been established, the absolute pointing accuracy of JWST with respect to the celestial coordinate system will be determined by the astrometric accuracy of the Guide Star Catalog and the calibration of the JWST focal plane model. The spacecraft pointing accuracy is unlikely to be sufficient for high precision science observations; therefore, target acquisition is recommended (and in some cases, mandatory) for such observations. Where target acquisition is performed, the pointing accuracy is determined by the required slew distance, with smaller slews giving the highest accuracy. Latest information on pointing accuracy and stability are described in a dedicated page

The effects of pointing drifts can be further compounded by undersampling of the PSF, uncorrected intra-pixel gain variations, and flat fielding residuals (see below).

High-gain antenna moves

The high gain antenna (HGA) is part of the JWST communications subsystem that manages the required two-way communication between the observatory and the ground. The HGA requires periodic pointing adjustments to maintain its attitude towards Earth. This repointing will take place every 10,000 s, which imposes a 10,000 s limit on regular (non-TSO) science exposures. This limitation is waived for TSO observations, to allow observers to continuously monitor temporal events such as an exoplanet transit, which may occur over a longer period of time. The repointing of the HGA will, in this case, be performed during a science exposure, resulting in a small but measurable pointing jitter. Initial analyses performed on TSOs during commissioning suggest the impact of this movement is very small — almost negligible in some cases in the actual flux time series. By monitoring the position of the guide star used during TSO observations, these analyses reveal that while the HGA move produces some sub-pixel centroid fluctuations, the on-board mechanisms quickly recover the placement of the star on time-scales of a few minutes back to the same pixels as before the HGA move. Users are still encouraged to monitor this movement and any potential impact it might have on their observations using engineering telemetry data available via the MAST. See below for an explanation of how to access and navigate this data.

"Tilt" events

During wavefront sensing commissioning, a number of un-commanded changes of the primary mirror segment tilts and piston positions were observed. These “tilt events”  can create very small changes to the wavefront of the telescope adding just a few to 10s of nm of total wavefront error (depending on the magnitude of the event). These events can be easily corrected during the telescope routine alignment program. However, some of these events can occur in timescales of a few hours and, therefore, during a TSO exposure. Because of this,  it is recommended to inspect observations for the presence of these events.

“Tilt events” have sometimes been observed to give rise to small (few tens or hundreds of parts-per-million) flux "jumps" on the flux time-series due to the slight PSF change they imprint. The exact source of those flux changes is still under investigation; however, it was observed that these events can be easily found in at least three ways during a TSO: (1) monitoring sudden full-width at half maximum (FWHM) changes, (2) monitoring sudden centroid changes, and (3) measuring the counts observed by the FGS guide star (see below for details on how to extract this latter information from MAST). It is important to note that not all of those "tilt" events give rise to measurable changes on the target fluxes, and not all TSO exposures show these events. Still, monitoring these variables is recommended for every TSO in order to account for this possibility.


Many observations with JWST will be background limited, meaning that the noise will be dominated by the background emission, and not by photon noise from the target or detector read noise. The overall background seen by JWST instrumentation has contributions from several astrophysical and observatory sources; articles on the observatory background describe these sources in more detail. At wavelengths greater than 15 μm, the background seen by JWST is expected to be dominated by thermal emission from JWST itself. 

Of particular importance for TSO is the variability of the thermal background over the duration of a typical observation. An exoplanet transit observation may last several hours. A full phase curve reconstruction could perhaps span a full day. A detailed understanding of the time variability of background levels during TSOs is under investigation.  Modeling suggests that the thermal constant of the mirrors is long (over days), and the thermal constant of the sunshield is short (over minutes). Note that the JWST Exposure Time Calculator assumes a static background model that varies over several parameters such as wavelength and target position, but it does not capture potential short term variability that the real observatory may experience.

See the JWST Background Model article for more detailed information.

Monitoring spacecraft telemetry during a TSO 

As mentioned above, jitter and even small mirror segment shifts could give rise to changes in the observed relative fluxes during a TSO.  One way of monitoring the state of the observatory is through the recorded and downloaded on-board telemetry of the spacecraft, which includes information about, e.g., times at which the HGA movements happen, guide star centroids and flux counts, among other data. STScI has developed software to retrieve this information from MAST which, once installed, allows a straightforward way to access this data. To obtain the FGS centroid and flux counts during a TSO, as well as the flags for HGA moves, for example, one would do the following query in MAST:

from jwstuser import engdb

# Define a MAST API token:
api_token = 'Paste your personal token here'

EDB = engdb.EngineeringDatabase(mast_api_token = api_token)

# First, get Guide Star X and Y positions over a pre-determined range of time:
start = '2022-05-30T01:00:00'
end = '2022-05-30T07:30:00'

# Get x-centroid data:
mnemonic = 'SA_ZFGGSPOSX'
cen_x = EDB.timeseries(mnemonic, start, end)

# Get y-centroid data:
mnemonic = 'SA_ZFGGSPOSY'
cen_y = EDB.timeseries(mnemonic, start, end)

# Get flux counts for FGS guide star:
mnemonic = 'SA_ZFGINSTCT'
gs_flux = EDB.timeseries(mnemonic, start, end)

# Get HGA move flags:
mnemonic = 'SA_ZHGAUPST'
hga_move = EDB.timeseries(mnemonic, start, end)

This example will retrieve the time-series for the FGS X and Y centroid of the guide star used by FGS, along with its flux counts and flag indicating HGA moves from May 30, 2022 from 01:00 UT to 07:30 UT. Note a MAST API token is needed to connect to the database. For more details on how to use this software, please read its documentation. A full list of the most useful mnemonics for TSOs will be eventually provided for the community to use.

Monitoring FGS guide star data & photometry 

While the on-board telemetry is a fast and reliable way of tracking the spacecraft state during a TSO, the data that are monitored is limited to that portion relevant for on-board operations. This implies that variables of potential interest for TSO analyses such as, e.g., the FWHM change as a function of time of the FGS guide star, are not recorded with enough precision in the telemetry to be of practical use. However, the FGS guide star images are also available for download and analysis in MAST. The unprocessed FGS data has a  *_stream  extension that identifies them, and can be found in MAST by selecting "Guide Star" from the "JWST Instrument Keywords" option on MAST, and filtering by the visit filename prefix (e.g., "jw01118005001*") using the "FileName" filter option in MAST.

See also discussion on the JWST Pipeline Tips & Tricks page.

Detector gain variations

The near-infrared and mid-infrared detectors onboard JWST are all subject to inter-pixel gain variations, i.e., small pixel-to-pixel differences in the digital numbers generated per electron by the detector electronics. Gain variations are typically a few percent in magnitude for JWST's IR detectors. Such gain variations are traditionally measured and calibrated using a detector flat field. They are multiplicative in nature, and the effect is therefore corrected by division by the flat field. 

Uncertainty in the flat field measurement, or changes in pixel gain over time, can leave residual flat field noise in the reduced data.

Intra-pixel gain variations

More challenging than pixel-to-pixel gain variations are changes in pixel responsivity or quantum efficiency across the pixel area, the so-called intra-pixel gain variations. This issue is compounded when the point spread function (PSF) of the instrument is undersampled by the detector pixels. Ingalls et al. (2012) describe the effect in detail for the IRAC instrument on board the Spitzer Space Telescope, where this was found to be the dominant source of correlated noise. Various methods for correcting this effect can be found in the literature, e.g., Ballard et al. (2010), Stevenson et al. (2011), Krick et al. (2016).  We do not expect this issue to affect the JWST instruments to the same extent as previous IR missions since a number of factors will aid to mitigate their effect:

  • Improved pointing stability of the telescope
  • Better sampling of the PSF
  • Higher precision astrometry and proper motion information on guide stars
  • Improved detector technology

More detailed measurements will be analyzed after commissioning, as time and resources permit.

1/f noise

The JWST near-infrared detectors are connected to a device called a SIDECAR ASIC that controls, among other things, how pixels are read. The SIDECAR ASIC, however, also injects some signatures on the frames that are read. One of the signatures having the largest amplitude in the near-infrared detectors (i.e., those used by the NIRISS, NIRCam and NIRSpec instruments) is the so-called 1/f noise. This type of noise is present on every pixel, but due to its spatial properties on the detector, the most striking property is that of an apparent "banding" in JWST detector frames. In an investigation of random noise sources for the NIRCam instrument, Schlawin et al. (2020) found 1/f noise to be the biggest contributor to the noise floor in TSOs with the NIRCam grism time-series mode. 

It is important to note that removing its impact on acquired data is quite complex given the stochastic nature of the effect (every frame will have its own, unique signature — however, they will all share similar statistical properties). While the reference pixel correction helps diminish its effect in some cases, in general extra steps should be taken to try to correct for it. One simple way of removing part of its effect is to use background pixels along the "banding" to remove the high-frequency components of this noise. However, care must be taken to handle real background flux structure when performing such corrections.

MIRI's mid-infrared detectors use different readout electronics from the near-infrared instruments, and are much less affected by 1/f noise. 

Figure 1. NIRISS/SOSS frame showing the banding effect along the "fast-read" direction, distinctive symptoms of 1/f noise

This NIRISS/SOSS frame shows one of the most obvious features of 1/f noise on JWST detector frames: that of an apparent banding along the "fast-read" direction (columns, in the figure above). The same banding is prominent on NIRCam and NIRSpec frames, and much less prominent on MIRI. Note the banding is just one of the symptoms of this noise — all pixels are impacted, not only the "columns" as shown above.

An interesting effect observed during commissioning observations and during Cycle 1 at least in the near-infrared detectors (i.e., in NIRCam, NIRISS and NIRSpec) that is most likely related to 1/f noise is the fact that the common procedure of binning spectral channels to increase the signal-to-noise ratio (SNR) on time-series observations typically leads to a much lower increase in SNR than what one would otherwise expect (see Espinoza et al. 2022). In other words, the noise level by binning spectral channels does not decrease with the square root of the number of channels. One strategy employed during commissioning to analyze TSOs was to work at the resolution level of the instrument, binning observables after performing analyses on that "high"-resolution level data. For example, if a transit spectrum is the science objective, the recommendation would be, if possible, to perform transit lightcurve fitting at the resolution level of the instrument and then bin the extracted transit depths in spectral channels instead of working with binned transit lightcurves. In our TSO experiments, this strategy yielded the best results.

Quantum yield

The quantum yield is the average number of electrons created by an incident photon in a detector. For the 5.2 μm cutoff HgCdTe detectors in NIRSpec and NIRISS, the quantum yield below wavelengths of about 2 μm can be larger than one: more than one electron might be generated by each incident photon in a pixel. The main consequence of this effect is that electrons no longer follow classic Poisson statistics (see, e.g, McCullough et al. 2008). The JWST pipeline currently does not consider this effect when estimating errorbars on the number of electrons, which in turn can lead to underestimated errorbars, which is particularly important for photon-noise dominated measurements.

Analyses using commissioning data are underway to estimate the exact impact of this effect on the estimated error bars calculated by the JWST pipeline. Using pre-flight estimates of the quantum yield (Rauscher et al. (2014); for NIRSpec detectors), photon-noise error bars at 1.5 μm can be found to be likely underestimated by 10%–30% — the same effect implies an underestimation of those error bars of 30%-50% at 0.6 μm (Espinoza & Volk, 2023; Jakobsen, 2022). Early analysis of commissioning data point to these percentages being at the right levels at least for NIRISS/SOSS. This implies that users might observe white-noise light curve scatters that are larger than those computed by the JWST pipeline by those amounts. 


Persistence is a memory effect in the instrument detectors, leaving a weak positive image that decays exponentially over time. The strength of the image is related to the brightness of the source that illuminated the detector. The physical mechanism causing persistence is thought to be the trapping of a small number of photoelectrons in the detector substrate, which are then released over time after the illumination has ceased (Regan et. al. 2012). This slow release manifests as a weak latent image that decays with time.

Because the effect depends on the previous illumination of the detector, it is hard to characterize and correct in a given observation. The impact on precision photometric measurements is most severe when observing a very faint target in the same detector region that previously held a very bright source. It will pose less of a problem for lengthy continuous observations of bright targets at a fixed detector location. Because TSO observations are predominantly targeting bright objects—to maximize SNR over time—TSOs requiring the highest precision may need to monitor and correct for persistence during their data processing.

Whilst the impact of persistence on JWST time-series observations will depend on the observing mode, the target brightness and the observing strategy, detector modeling and testing will help define mitigation strategies or pipeline correction steps. Leisenring et al. (2016) provide detailed results for the NIRCam detectors. Analysis of the effect for the MIRI detectors has shown multiple time decay constants, which makes characterization challenging; based on testing experience, the time between visits (timescales of minutes) should allow for any persistent images to decay to <1% of their initial levels. However the effect will require further characterization during Cycle 1.

Further background on the physical nature and predicted extent of this effect on JWST instruments are provided in, e.g., Regan et al. (2012), Leisenring et al. (2016).


Ballard, S., et al. 2010, PASP, 122, 897
Search for a Sub-Earth-Sized Companion to GJ 436 and a Novel Method to Calibrate Warm Spitzer IRAC Observations
ADS  Arxiv

Beichman, C., et al. 2014, PASP, 126, 946
Observations of Transiting Exoplanets with the James Webb Space Telescope (JWST)
ADS  Arxiv

Espinoza, N., et al. 2022, PASP, in press
Spectroscopic Time-series Performance of JWST/NIRSpec from Commissioning Observations 

Ingalls, J. G., et al. 2012, Proc. of SPIE, 8442, 84421Y
Intra-Pixel Gain Variations and High-Precision Photometry with the Infrared Array Camera (IRAC)

Jakobsen, P., 2022, ESA Instrument Report, ESA-JWST-SCI-NRS-TN-2022-003

Krick, J. E., et al. 2016, ApJ, 824, 1
Spitzer IRAC Sparsely Sampled Phase Curve of the Exoplanet Wasp-14B
ADS  Arxiv

Leisenring, J. M., et al. 2016, Proc. SPIE, 9915, 99152N
Characterizing persistence in JWST NIRCam flight detectors

McCullough, P. R., et al. 2008, PASP, vol. 120, issue 759
Quantum Efficiency and Quantum Yield of an HgCdTe Infrared Sensor Array

Rauscher, B. J., et al., 2014, PASP, vol. 126, issue 942
New and Better Detectors for the JWST Near-Infrared Spectrograph

Regan, M., et al. 2012, Proc. SPIE 8442, 84424W
Count rate nonlinearity in near infrared detectors: inverse persistence

Schlawin et al 2020, AJ 150, 5, id. 231
JWST Noise Floor. I. Random Error Sources in JWST NIRCam Time Series
ADS Arxiv

Stevenson, K. B., et al. 2012, ApJ 754, 2 
Transit and Eclipse Analyses of the Exoplanet HD 149026b Using BLISS Mapping
ADS  Arxiv

Zhou, Y., et al. 2017, AJ, 153, 6
A Physical Model-based Correction for Charge Traps in the Hubble Space Telescope’s Wide Field Camera 3 Near-IR Detector and Its Applications to Transiting Exoplanets and Brown Dwarfs,  
ADS  Arxiv

Notable updates
    Updated for post-commissioning release.
Originally published