Image Sampling, F-ratios, and SNR in...
Transcript of Image Sampling, F-ratios, and SNR in...
Craig Stark, Stark Labs
Image Sampling, F-ratios, and SNR in Astrophotography
Borg 101 f/4QSI 5403.8”/pixel
Thursday, June 2, 2011
http://www.stark-labs.com
Sample according to your seeing or suffer the consequences
+ =
Long focallength
Small (color)pixels #@#$%&#@!!!!!
Thursday, June 2, 2011
http://www.stark-labs.com
Atmosphere, focus, tracking, and optics all blur before the CCD.
http://www.weatheroffice.gc.ca/astro/seeing_e.html
Thursday, June 2, 2011
http://www.stark-labs.com
Our CCD samples sections of the sky to make pixels.
Perfect seeingPerfect optics3” separation
High samplingrate
Low samplingrate
Thursday, June 2, 2011
http://www.stark-labs.com
What is my sampling rate and what should it be?arcsec/pixel = 206 x PixelSize ÷ FocalLength
4.3 micron pixels
Focal length arcsec/pixel
400 2.2”
1000 0.88”
2000 0.44”
3000 0.3”
For an 8” scope, f/5 and below!
Thursday, June 2, 2011
http://www.stark-labs.com
Why not run at high sample rates? What’s wrong with 0.3”/pixel?
For a given camera (a given pixel size), increasing the sampling rate cuts the FOV. You’ll fit less on the chip.
Unless you’re on Mt. Wilson or somewhere else with sub-arcsec seeing, you lost the details before they hit your chip. You’ll never record those details anyway.
High sample rates mean fewer photons per pixel which means worse SNR. You’ll have noisier images.
Thursday, June 2, 2011
http://www.stark-labs.com
Sampling beyond the seeing gives no extra detail in the image
Thursday, June 2, 2011
http://www.stark-labs.com
Sampling beyond the seeing gives no extra detail in the image
Thursday, June 2, 2011
http://www.stark-labs.com
Hubble M51Blur by 2 pixels
Hubble M51Blur by 2 pixels
Shrink 75%Rescale back
“Seeing” (2 pixel blur) limited the resolution. This meant I could sample it less (shrink) and not lose anything.
Thursday, June 2, 2011
http://www.stark-labs.com
C8 at 0.75”/pix C8 binned 1.5”/pix C8 binned 1.5”/pixresampled 2x
Does the 0.75” have any more spatial detail?Has it even lost something?
Thursday, June 2, 2011
http://www.stark-labs.com
Sampling rate affects your SNR100mm scope with 560 mm f/l (f/5.6)
100mm scope with 400 mm f/l (f/4)
Dropping one f-stop boosts SNR by 41%
Thursday, June 2, 2011
http://www.stark-labs.com
What is the f-ratio and whyshould I care?
F-ratio = Focal Length ÷ Aperture
Reducing the focal length of your scope drops the f-ratio
Increasing the aperture at the same f/l drops the f-ratio
F-ratio determines the number of photons per area of sky that make it to your CCD well.
Larger scopes at the same f-ratio give more magnification
Foca
l len
gth
Aperture
≈F-ratio
Thursday, June 2, 2011
http://www.stark-labs.com
But I heard f-ratio doesn’t matter. It’s a myth, right? Nope...
The “f-ratio myth” page by Stan Moore that started this said, “There is an actual relationship between S/N and f-ratio, but it is not the simple characterization of the ‘f-ratio myth’.”
His page uses a different definition of SNR. Instead of just “how noisy is the image” it measures “how much information is in the image” (including how much spatial detail).
When you look at “pixel SNR” (how noisy is the image), the f-ratio dominates. It’s not just in extreme or marginal cases.
Thursday, June 2, 2011
http://www.stark-labs.com
Images courtesy of Mark Keitel
FRC 300 @ f/7.8 FRC 300 @ f/5.9
Does the f/7.8 really not give up any SNR?
Thursday, June 2, 2011
http://www.stark-labs.com
Images courtesy of Mark Keitel
FRC 300 @ f/7.8 FRC 300 @ f/5.9
Does the f/7.8 really not give up any SNR?
Thursday, June 2, 2011
http://www.stark-labs.com
8” f/
5.75
4” f/
4
25% as many photons making it into the tube yet less noisy...In the end, almost 3x as many photons per pixel on the 4”.
Thursday, June 2, 2011
http://www.stark-labs.com
For extended targets, the photon count per pixel dictates the SNR
Also note effect of read noise for line filter work
Thursday, June 2, 2011
http://www.stark-labs.com
Take home messages on F-ratio
Aperture alone doesn’t “rule”. I’d rather image DSOs on an 8” f/5 than a 30” f/15.
F-ratio precisely encapsulates the photons per pixel (from extended objects) per unit time that drives pixel SNR (how noisy the image looks). It (along with exposure time) says if you’re photon poor.
Want higher SNR with the same image time? Drop the f-ratio.
Want more spatial detail with the same time and SNR? Keep the f-ratio and increase the aperture (and hope your skies give it to you).
Thursday, June 2, 2011
http://www.stark-labs.com
Sample according to your seeing or suffer the consequences
+ =
Long focallength
Small (color)pixels #@#$%&#@!!!!!
Aaaahh!I’m photon
poor!
Thursday, June 2, 2011
http://www.stark-labs.com
Suggestions1. Measure your sky’s seeing
On unstretched data, measure stars’ FWHM
from good nights
2. Figure the highest-resolution sampling
That FWHM divided by 2, 2.5, or 3
2.5” ~1”
3. Figure that focal length (or pixel size) and don’t exceed this
206 x PixSize ÷ FocLen
~1500 mm
4. Feel free to run even significantly less than this.
Don’t sacrifice coverage for wasted resolution
Thursday, June 2, 2011
http://www.stark-labs.com
“Undersampled” but smooth stars
~4 arcsec / pixel
Thursday, June 2, 2011
http://www.stark-labs.com
Suggestions1. Measure your sky’s seeing
On unstretched data, measure stars’ FWHM
from good nights
2. Figure the highest-resolution sampling
That FWHM divided by 2, 2.5, or 3
2.5” ~1”
3. Figure that focal length (or pixel size) and don’t exceed this
206 x PixSize ÷ FocLen
~1500 mm
4. Feel free to run even significantly less than this.
Don’t sacrifice coverage for wasted resolution
Thursday, June 2, 2011
Questions?
Borg 101 f/4 QSI 540 3.8”/pixelThursday, June 2, 2011