Help
Options
Didn't know it?
click below
click below
Knew it?
click below
click below
Don't Know
Remaining cards (0)
Know
retry
shuffle
restart
0:00
Embed Code - If you would like this activity on your web page, copy the script below and paste it into your web page.
Normal Size Small Size show me how
Normal Size Small Size show me how
Lecture 17-18
| Question | Answer |
|---|---|
| bias correction | an exposure taken with 0 seconds of integration to represent the zero level or pedestal voltage applied to the detector |
| Dark current correction | an exposure taken with the same exposure time as the science frames but with the shutter closed to correct for internal thermal effects |
| A single bias frame is a single | readout of the detector with an exposure time of zero seconds thus only recording the pedestal voltage |
| The bias can also be calculated from an | overscan region, created by allowing the serial register to readout a few extra empty pixels after the physical pixels have been read out. |
| Every frame, including calibration frames downstream (darks, flats etc) must have the | bias level subtracted individually |
| Dark current correction | All OIR detectors will have some signal arise from thermal effects, even in the absence of illumination |
| To create master dark frames | first subtract master bias from each frame and then combine |
| The flat field correction is meant to correct for | sensitivity variations so that each pixel contributes equally to the final image. |
| How to make flat frames | first subtract master bias, then subtract master dark. Then combine into a master flat frame and normalize it so that the median pixel has 1.0ADU |
| Flat felids, we need to integrate on such a source for long enough to accrue | greater than or around 10^4 counts, and therefore reduce Poisson noise to the less than or around 1% level. |
| bias correction | an exposure taken with 0 seconds of integration to represent the zero level or pedestal voltage applied to the detector |
| Dark current correction | an exposure taken with the same exposure time as the science frames but with the shutter closed to correct for internal thermal effects |
| A single bias frame is a single | readout of the detector with an exposure time of zero seconds thus only recording the pedestal voltage |
| The bias can also be calculated from an | overscan region, created by allowing the serial register to readout a few extra empty pixels after the physical pixels have been read out. |
| Every frame, including calibration frames downstream (darks, flats etc) must have the | bias level subtracted individually |
| Dark current correction | All OIR detectors will have some signal arise from thermal effects, even in the absence of illumination |
| To create master dark frames | first subtract master bias from each frame and then combine |
| The flat field correction is meant to correct for | sensitivity variations so that each pixel contributes equally to the final image. |
| How to make flat frames | first subtract master bias, then subtract master dark. Then combine into a master flat frame and normalize it so that the median pixel has 1.0ADU |
| Flat felids, we need to integrate on such a source for long enough to accrue | greater than or around 10^4 counts, and therefore reduce Poisson noise to the less than or around 1% level. |
| Combining everything, pixel-by-pixel mean | computationally quick, but can mitigate the effects of cosmic rays |
| Combining everything, median | This is simple and more common than mean. It rejects cosmic rays effectively but is not as good at accurately representing the central values. |
| Combining everything, indescriminate rejection | simply rejects the highest value of a given pixel of the set of frames before taking the mean or median. Will reject cosmic rays, but skews toward underestimating the central value |
| Combining everything, selective rejection | or "k-sigma" clipping. Reject all values above/below a certain factor of the mean based on the standard deviation. Can be computationally expensive |
| Chopping technique for combining NIR (bc NIR sky is very bright, bright enough to outshine many targets) | chopping refers to a chopping secondary mirror that can rapidly tilt back and forth from the target to a nearby area of sky |
| Nodding technique for combining NIR | Nodding refers to physically moving the telescope slightly (so the secondary mirror remains fixed) to an offset sky area. |
| NIR observing general | For both nodding and chopping, the telescope will observe the target for less than 10 seconds then move to an adjacent area of sky for the same exposure time and back again in an object-sky-sky-object (OSSO or ABBA) pattern |
| Rather than counting how much charge builds up, an IR detector performs | double-correlated sampling: it reads out the voltage at the beginning and end of an exposure and reports the difference. THUS THE BIAD VOLTAGE CANCELS AND DOES NOT NEED TO BE SUBTRACTED |
| There is a practical limit on how long we can expose for to take a single image | The saturation point of our detector sets one hard limit, but varies from detector to detector, target to target. Cosmic rays set another limit that generally doesn't vary. Beyond around 1800 s, the build up makes longer exposures give diminishing returns |
| To identify source pixels we | run a filter (or convolutional kernel) over the image to find where isolated sources are. But we want to AVOID sources that are not astrophysical or transient |
| Our strategy to locate the positions of sources on each image | A given pixel must be N*(sigma) above the background (sigma is standard deviation of the background). There must be M contiguous pixels that meet this criteria. M is usually above 5-10 to avoid spurious sources |
| For identifying the positions of sources on each image, the choice of N in N*(sigma) and M is a dark art | High N and M will identify bright, reliable sources but risk having too few for reliable image combination. Low N and M will provide many sources, but risk letting spurious noise through |
| If necessary, calculate geometric transformations between the frames | pixel scale (differences in combining images from different instruments), rotations, distortions(coming from particularly wide field cameras), projection (effects due to the curvature of celestial sphere |
| Since centroids and distortions are continuous | mapping them back to a discrete integer framework requires interpolation and or resampling |
| Nearest pixel interpolation | simply rounds off the fractional pixel to shift it either up or down to the nearest integer in the new frame. |
| Nearest pixel interpolation is generally | easy and generally retains image detail, but the coarseness of the approach tends toward a loss in astrometric or positional accuracy |
| Resampling | is a collection of techniques that accompanies interpolation and in certain cases can improve the resolution of images by making the pixel of the combined image smaller than the original image pixels |
| The shift and add technique combines nearest pixel interpolation but adds an additional step to | preserve astrometric precision |
| Interlacing | is another resampling approach that only considers integer position of pixels. The original image is mapped to a rotated, finer space resampled image, by simply considering which output pixel maps to the center of which input pixel |
| Is Interlacing good for resampling a single image | No. Its terrible. Only a fraction of the new pixels receive "hits" so there will be gaps in the new image. Positional uncertainty is introduced |
| When is interlacing good | As more and more images are resampled to the same grid, this approach can actually improve positional uncertainty and improve detail for detector limited resolution |
| What is one limitation to interlacing | the pixels need to be precisely lined up on the sub-pixel level which is rarely the case when dithering images |
| Dithering | it simply refers to slight offsets between images to cover CCD/pixel gaps and facilitate cosmic ray/hot pixel removal |
| Variable-pixel linear reconstruction (AKA drizzling) | is another popular method of image resampling/combination. Rather than assuming the flux received by a pixel is spread across the entire pixel, drizzling assumes that the flux is uniformly concentrated within a smaller concentric square called the drop |
| Accuracy | How close to the truth is our measurement? We need to know the answer to gauge the accuracy of a measurement |
| Precision | to how many decimal places can we confidently report a measurement? The measurement may not be accurate but the value is very well determined |
Created by:
user-1996284