Pixel Behavior Families

The majority of pixels in the Roman Wide Field Instrument (WFI) detectors operate as expected. However, each detector contains a small subset of pixels that exhibit anomalous behavior and are excluded or flagged during pipeline processing. This article describes the different classes of these anomalous pixels.

The Roman Wide Field Instrument (WFI) contains a focal plane array of eighteen H4RG-10 detectors, which are HgCdTe 4096 pixel by 4096 pixel photo-diode arrays. Each have undergone vigorous testing before launch to ensure that they will operate at peak performance during the primary five year mission and will be able to meet the stringent science requirements (see WFI Performance for more information about the performance characterization activities of the WFI detectors). Over 98% of the pixels for each of the eighteen detectors maintains a stable, predictable response to illumination. As with any type of detector, there are a very small subset of pixels whose behavior deviates from the norm due to electronic and manufacturing imperfections. These result in three different behavior families: 

  1. Pixels with low signal accumulation
  2. Pixels with higher signal accumulation
  3. Other anomalous pixels

The Science Operations Center (SOC) at STScI has analyzed data from the WFI-level Thermal Vacuum Testing campaigns (TVAC) to create preliminary bad pixel masks that will be utilized during pipeline processing. These bad pixel masks are used in the romancal.dq_init.DQInitStep()  to mask certain classes of unusable bad pixels. More information about how romancal  applies these masks, as well as the algorithms for creating the bad pixel masks, can be found in the Exposure Level Pipeline article in the Data Handbook Home.

Note

This page reflects the current understanding of bad pixels for WFI. It will be updated in future publications of RDox as analyses continue.



Pixels with Low Signal Accumulation 

The SOC has developed algorithms to identify pixels that accumulate less signal than expected; a normalized slope image was created using averaged TVAC flat-field data. Two different classes of bad pixels were identified from this image.

Dead Pixels 

Pixels with essentially zero (or very low) signal response to illumination are classified as dead pixels. Dead pixels consistently do not accumulate counts throughout an exposure. The Figure of a Dead Pixel provides an example of a dead pixel in WFI09. Dead pixels are not used for science analysis. 

Figure of a Dead Pixel

This 10 by 10 pixel cutout shows an example of a dead pixel on detector WFI09. The data shown is from a normalized slope image. The dead pixel has a normalized count rate value close to zero indicating little to no response to illumination.

Low Quantum Efficiency (QE) Pixels 

Pixels that have a lower response to light are classified as low quantum efficiency pixels. These pixels are flagged when a pixel has < 50% of the normalized count rate of neighboring pixels. An example of a low QE pixel are show in the Figure of a Low QE Pixel. Low QE pixels are still sensitive enough to be used in pipeline processing once the low QE is taken into account.

Figure of a Low QE Pixel

This 10 by 10 pixel cutout shows an example of a low QE pixel on detector WFI09. The data shown is a normalized slope image. The low QE pixel has a normalized count rate of about 0.4 counts per second. This is roughly 50% of the count rate in the neighboring pixels.

Pixels with Higher Signal Accumulation 

Pixels that accumulate signal in the absence of light can be identified by using long dark data from the TVAC tests. There are multiple classes of these types of pixels and can be differentiated between by fitting the ramps to different models and calculating the dark current of each pixel. In long dark exposures, a "good" pixel should have a linear ramp with a slope (i.e., dark rate) of less than 0.005 e-/sec to be used for science analysis (see the WFI Characterization Activities article for more information about the characterization of the detectors).

Hot Pixels 

Hot pixels are identified as having a linear ramp and a dark current rate of more than 0.5 e-/sec. An example of a hot pixel ramp can be seen in the Figure of a Hot Pixel. Hot pixels are excluded for science analysis.

Figure of a Hot Pixel

The counts accumulated in a hot pixel on detector WFI09. Each black point is a single read, and the read time is 3.16 seconds. A linear model was fit to the ramp, and the slope of this is the dark rate of the pixel. Since the dark rate was determined to be 0.919 e-/s, it exceeds the threshold of 0.5 e-/s and is classified as a hot pixel. 

Warm Pixels 

Similarly to hot pixels, warm pixels also have a linear ramp, but have a lower calculated dark rate than hot pixels. Pixels with a dark rate that is greater than 0.1 e-/s but less than the hot pixel threshold of 0.5 e-/s are classified as warm pixels. An example of a warm pixel's ramp can be seen in the Figure of a Warm Pixel. 

Figure of a Warm Pixel

The counts accumulated in a warm pixel on detector WFI09. Each black point is a single read, and the read time is 3.16 seconds. A linear model was fit to the ramp, and the slope of this is the dark rate of the pixel. Since the dark rate was determined to be 0.253 e-/s, it exceeds the threshold of 0.1 e-/s and is classified as a warm pixel.

Resistor-Capacitor (RC)/Inverse RC Pixels

RC and Inverse RC pixels are primarily identified based on the shape of the signal accumulated over non-destructive reads when exposed to no illumination. These pixels accumulate signal in the absence of light, but the ramp is highly non-linear and resembles the charge curve of an RC circuit. An image of the ramp can be seen in the Figure of an RC Pixel. RC and IRC pixels are not used in science analysis.

Figure of an RC Pixel

The counts accumulated in an inverse RC pixel's ramp on detector WFI09. Each point on the plot is a single read, with a read time of 3.16 seconds. The ramp fits a double exponential model, shown by the red line. The fitted time constants are 15.5 seconds for the fast component, and 108.4 seconds for the slow component.

Telegraph Pixels 

Telegraph pixels do not exhibit linear or double exponential ramps. Instead, telegraph pixels experience jumps between different states throughout the exposure. An example can be seen in the Figure of a Telegraph Pixel. Telegraph pixels are not used in science processing due to their jumps.

Figure of a Telegraph Pixel

The counts accumulated in an telegraph pixel on detector WFI09. Each black point is a single read, and the read time is 3.16015625 seconds. This pixel flips between three different states throughout the exposure, with each state highlighted in blue, red, and yellow.

Other Anomalous Pixels

Open and Adjacent Pixels 

Open and adjacent pixels are clusters of pixels that exhibit differences in signal accumulation. The center pixel (called the "Open" pixel) is low in sensitivity (but not low enough to be considered a Dead pixel), and the pixels adjacent to it have a higher than average signal. All four of the adjacent pixels each have a normalized count rate value 10% greater than the mean of the image. These adjacent pixels are theorized to "take" the signal from the center pixel. An example of an open and adjacent pixel cluster can be seen in Figure of Open and Adjacent Pixels. These pixels are identified using the same slope image as described in the Pixels with Low Signal Accumulation section. Open and adjacent pixels are not used in science processing.

Figure of Open and Adjacent Pixels 

Count rate behavior for open and adjacent pixels. Marked with a blue circle, the open pixel has significantly less signal than the four adjacent pixels. The four adjacent pixels have greater than 10% the normalized count rate value, which is why this pixel is flagged as OPEN/ADJ rather than dead.



For additional questions not answered in this article, please contact the Roman Help Desk.



Latest Update

 

Minor updates due to Instrument Handbook reorganization.
Publication

 

Initial publication of the article.