Automatic aperture

Is automatic aperture 1.5 times FWHM applied for each star in the sequence? Why different stars on small previews (Color Photometry page) have equal apertures and annulus?

All stars in a single image require the same aperture. VPhot calculates the average FWHM of the stars in the image and applies this average aperture (1.5xFWHM) to all measured stars.

Ken

It’s not the best news. I beleived, each star automatically get optimal aperture :frowning:
Well, what to do now? Use maximal circles or short sequences?

Mikhail:

For clarification, I assume you are referring to ‘Single Image Photometry’ and NOT ‘Time Series Photometry’?

Ken

Yes, yes. Two color transform.
For untransformed images I choose recommended aperture, but in this case I blindy hoped on automation…
It will be better to use manual set.

I wonder if the confusion here is wanting to measure faint stars with a smaller aperture than a bright star. If image-quality is reasonably uniform across an image, however, it turns out the full-widith-at-half-maximum is the same for all the stars. I agree it is very difficult to gets one’s eye/mind around that, but that’s how it works out. So the VPHOT scheme of using a uniform aperture is correct.

\Brian

2 Likes

Of course, it should work correctly, but why automatic aperture is so small???

Mikhail:

Actually, what you have shown is normal and correct. The aperture is not too small! What you see visually is a result of the screen stretch of the thumbnails. I suspect you think the aperture of a ‘brighter’ star is not big enough. That is incorrect.

Watch the attached video from my VPhot course. Look at the ‘Measurement Details - Star Profiles’ for some faint to bright stars to prove this fact to yourself. You will find that the aperture where the star profile returns to background is the ‘same’ for both fainter and brighter stars!

Ken

PS: I shared a google drive link with you by email. I also shared to others below.

3 Likes

Ken is correct. It does look too small for the bright stars, but the aperture is taking in the same fraction of the light profile for all the stars. There is very little information beyond about 2.5-sigma in an assumed Gaussian profile. One could quibble a bit (from your image) and make the gap between the ‘star’ aperture and ‘sky’ annulus a bit larger, just to get the sky measurement out farther from the wings of the profile, but that should make very little difference.

\Brian

Thank you, Ken and Brian, for the explanations! I thought, thin gap between star and circle is needed for avoiding of exposed pixels cutting. Of course I believe to VPhot, but these previews…

Mikhail:

Yes, you do want a reasonable ‘gap’ to make sure your sky annulus avoids most of the target wings as Brian mentioned. The gap you set in “Tools” is a radius not a width so if you make this setting too small it is possible that poor seeing (large FWHM) may cause the auto-adjusted measurement aperture to get quite large and approach the gap outer radius. Obviously not wanted but usually not a problem.

With a little judgement and typical seeing, this should not present a problem but you do need to readjust the gap setting if it causes such a problem. Just check a few times and adjust the gap radius and you should avoid any problem. I rarely need to adjust my settings.

Ken

Hi!

I have a similar question, but in regards targets. I do have 3 images of V0605_Cas: BIV
I image shows a larger target than B and V. What ever aperture settings that fit B and V won’t fit I.

Do I choose an aperture that fits B and V and just stack them for photometry?

I tried to modified the centroid for I but it changes B and V also in the Pixel/AUD mode, ut not sure how to change X/Y.

Should I submit 2 reports? BV and I?

Thanks!

Just checked star profile for image I and B, and they are very different.

- B

I image

Looking at the image is NOT the same as looking at the star profile plot on the measurement details page.

What is the radius at which the profile returns to background for the B vs I image?

Provide a screenshot of the star profiles not the images and report whether you still think they are very different?

Ken

3 Likes

This is obviously a widespread misconception, and I will admit it is difficult to wrap one’s mind around the idea that faint stars have the same (or nearly identical) profile sizes as bright stars. If you have software that puts out magnitudes or total counts above sky and the resulting signal-to-noise, try the following. Using a single good image, measure a pretty bright but unsaturated star in several measuring apertures starting very small, say two or three pixels, and increase that until you go well past what you think is the outer edge of the profile. Be sure in all this to make the sky annulus plenty big and well outside whatever you end up measuring. As you go, make a list of the magnitudes and the S/N as function of aperture. Repeat this process on the same image with a medium-bright star, and then a rather faint star. Now make a plot of the results (magnitudes versus S/N). What you’ll see is that the S/N rises very fast from the tiny aperture up so some maximum, fairly broad, then declines gradually outward — and the peak of the plots will be nearly the same for the very bright star as for the faint star. That is the demonstration that the optimal measuring aperture is the same for all the stars, no matter how they look with the particular brightness scaling shown on the computer screen (which is where the confusion starts). The same fraction of the total light is being included in the measuring aperture for the bright/faint stars.
You can do the same exercise for an ensemble of comp stars versus some target. You’ll find that if you make the aperture too tight, the scatter on the comp stars will get worse even though you’d think you’re including just the easily visible part of the fainter images. This is because an aperture small enough to minimize the random errors will in general have unacceptably large systematic errors due to the excluded wings of the star image (a lot of pixels in the wings). If you make the apertures too big, the data get noisy because there is so much sky background underneath the star image. The trick is to find the optimal aperture between those extremes, which you’ve done with the plot-exercise just described.
The ‘optimal’ spot is somewhere near that peak. Due to the statistical nature of the data, the ‘optimal’ spot is not sharply defined and thus forgiving of not being in exactly right (the plot of the S/N has a broad maximum). The conventional wisdom of around 2.5x the full-width-half-maximum is a good round value to adopt. For an ideal Gaussian profile, the theoretical peak S/N is at 1.6 sigma, which turns out to be 2.35x fwhm. Since one’s images are usually not strictly Gaussian (tracking/focus errors, crummy optics, bad seeing, etc), the lore is to err a bit on the larger side of that perfect size.
The suggestion here is to experiment with your data to see what works well and how things fail or get worse. You’ll likely find that the conventional wisdom is about right, and in the process you’ll learn a lot about your images that might lead to better data-taking and how to analyze data of all kinds.

\Brian

2 Likes

Ok,

For B image:
image

For I image:
image

Radius still 9 :upside_down_face:

1 Like

Celsa:

So yes, for most stars on the image, both bright and faint, the star profile is the same (return to background about 9 in this case}. As Brian mentioned, this is ‘hard to believe’ based on what you ‘see’ on the image. Screen stretch plays games with your eyes so you need to look at the data plot! The main exception is a saturated star where the star profile will be flat at the centroid (0 radius) and the return to background will be much later and the star ‘looks’ bloated. Don’t trust it! You should always ignore such saturated stars.

It is also easy to check the FWHM reported for all the stars, both bright and faint. They will all be about the same (your seeing is the same across at least a small portion of the sky), again except for saturated stars.

Ken

3 Likes

Thanks!

So I’ll stack them and then create a sequence choosing appropriate aperture values and submit magnitude. :+1:

Celsa:

I noted that you put this post in a topic that discusses ‘automatic aperture’.

IF you run a long time series, seeing (FWHM) may change somewhat during the night. Perhaps a fixed aperture may not be ideal during the entire run?

The automatic aperture tool is designed to address this issue. IF FWHM increases during the night, would you want to use a slightly larger aperture to collect the same ‘fraction’ of light as collected when seeing was better and FWHM was smaller? I think the answer is yes, so I often use 1.7xFWHM as my default aperture.

In the VPhot course, the students run an experiment where different apertures (n x FWHM) are tested and the precision and accuracy (O-K of check) are measured. Based on the results year after year, we find the 1.5-2 x FWHM (radius) seems to produce optimum results. As Brian mentioned, the choice is often quite forgiving!

So, in your case, you could alternatively run a time series of all your images (not stacked) and subsequently combine/average the magnitudes.

Either process works. Use what ever tool you find most efficient.

Ken

Ken

3 Likes

I stacked all BIV (3 images total) and got a black image.

I selected the 3 images and then “stack”.
Click on the stacked image and got this blank.
What went wrong?

Thanks!