An analyse of capture of an object with different time of exposition and ISO. They should contain the same data (amount of signal) although.
a) 120 seconds and ISO 800
b) 240 seconds and ISO 400
c) 480 seconds and ISO 200
All photos were conveted from RAW CR2 to JPG in Photoshop without any processing.
Diference of Histogram.
In animated GIF we can perceive:
1) less time exposition, more sharp the star.
2) less ISO, less noise
3) There isn't sifnificant color difference
4) There is loss of visual resolution when long time exposition.
Question:
The apparent loss of focus and shaprness with long time is due to:
a) RMS of guiding
b) capture of background glow (perhaps for humiditiy) around of star
c) capture of true data (signal) emited by star
d) contamination from excessive luminosity of adjascent pixels of sensor
e) others issues, inclusive those above ou some one of them.
The other animated GIF shows the difference having the 120 seconds 800 ISO as reference.
note: I will stack in DSS and compare the difference. I did a number of frames that will be the same total time of exposition for each one.
Hi Jorge. I'm always interested in photo exposure problems and I have just this morning been reading about this very topic. The noise in an image is located within the black (left-hand) end of the histogram so it is best to have your image peak data away from that end. The recommended value is about 25% across from the left hand side. On this basis, all your images are in fact under-exposed. However the best of your set would be the 480sec at 200 ISO.
Hope this helps.
Robert
480 seconds should show deeper stars than 120 seconds. The only explanatio0n that comes to mind why it doesn't is that your camera works much better at ISO 800. I'm pretty sure my camera picks up fainter stars at longer times regardless of ISO.
Mine doesn't seem to Kevin. Logically, why should it? you double the
exposure but cut the amplification in half, ending up with approx. the same result.[other than noise difference]. I know little about the digital world, but I know that was how it was for me in the days of film.
raymo
Historically, and through a number of tests by people like Craig Stark, Canon DSLRs perform best at ISO 800, so i'd forget shooting below that uness its the Moon. I shoot everything at ISO 800 and have no complaints.
Look carefully at your darks and include bias/offsets in the stack. Accurate guiding is essential I believe, regardless of sub time I always guide.
Now that its winter and the camera will be cooler the darks shoukd improve, make sure you allow a delay between subs for the sensor to cool. Craig Stark has an excellent write up on Canon dark noise comparisons - worth a read. Recent tests by RCheshire and myself using our cold finger cooled Canons show virtually no noise when the sensor temp reaches 0C, basically the darks look just like bias/offset frames so no more dark stacking needed. Cooling is extreme for most folks but mentioned to show Canons can be very low noise, especially on these cold winter nights. Let the camera cool down to ambient if you can-
Mine doesn't seem to Kevin. Logically, why should it? you double the
exposure but cut the amplification in half, ending up with approx. the same result.[other than noise difference].
That amplification creates noise which limits faint details. Get that lower and the limiting factor is sky fog.
I did a test with my Pentax a couple of years ago on M42. The low ISO longer time won over higher ISO shorter time. Then I did a test with different ISO's but same time. The higher ISO had more gain but about the same proportional noise increase. As long as the exposure wasn't black clipped, the ISO didn't matter much. Time was the greatest factor in limiting magnitude.
Of course this is Pentax but I also found the same thing with my old Nikon. Canon may be different. I don't have one of those to try.
That amplification creates noise which limits faint details. Get that lower and the limiting factor is sky fog.
This is why you put your exposure peak on the histogram at 25%. You get above the noise, but maximize detection of faint detail. Stacking multiple exposures then enhances the detail.
The true ISO (unity gain) of the T3/1100D is between 300 and 400, so I wouldn't go lower than 400 regardless, and it's then just a case of exposing sufficiently...that means, with the histogram away from the left edge and with no accumulation of pixels at the right hand edge, therefore clipping neither blacks or whites respectively.
The gain behaviour of this camera is non-linear with ISO however. If you double the ISO to 800, the noise only increases by about 30%. Double it again to 1600 and again doesn't increase by as much as it should. If you're imaging in cool conditions, with ambient temperatures around or below 5C, try taking some darks at ISO1600 and see how they look
Thank you for comments and attention to this issue.
Some new words about:
Quote:
about 25% across from the left hand side
.
False. This is a theoretical, laboratory-mathematics, knowledge.
When you have high skyglow this percentual and theoretical recomendations don't work.
Quote:
480 seconds should show deeper stars than 120 seconds.
It makes sense, only if the atmosphere is clean and with low refraction (mainly by humidity).
Quote:
is that your camera works much better at ISO 800
Not full time. If I need to shot nebula, ISO 800 catch more information from clouds. If I shot stars (cluster) ISO 800 saturate the stars. Look the shape and the borders of stars with 120 seconds ISO 800. After the DSS stacking ... it will be not natural. I will show later the stacking comparison.
The most curious is that many people talk about the response of sensor used by Canon. And they are talking not only about sensor, they are talking about the sensor with the electronic Canon inside the camera.
However, none of them work inside the Canon Inc. None of them work inside the laboratory that produce the cameras to Canon Inc. And Canon Inc never sign below of any document about theirs cameras. All technical documents from Canon Inc are laboratory measurements of isolated parts of theirs cameras.
It seems that Canon Inc doesn't know really theirs equipments. They make a camera. Measure its responses ... and sell it !
Question:
Who use cameras by cameras only ? Or better, you will use cameras always in enviroment that cause distortions of camera responses ! They are not out in the space, as the cameras of Hubble and others. There, the most problem is cosmic rays.
So what works for me, here aside the beach in Brasil, can not work well there, Australia near the beach, also.
Two groups of capture:
1) 240 seconds ISO 400
2) 120 seconds ISO 400
Two stacking:
1) only 11 x 240s 400iso
2) 11 x 240s + 21 x 120s
Comparasion between Autosave.tif 32 bits DSS - linear - with magnify of 200x - screen copy by Photoscape. None graphic processing.
photo 1
Comparasion between Autosave.tif 32 bits DSS - autolevel photoshop - with magnify of 200x - screen copy by Photoscape. No other graphic processing.
photo 2
Comparison of histogram
photo 3
note: stacking all photos give appearance of better color of stars, noise reduction, and reduction of "false Airy disk".
Useful if we only want reduce the noise when capturing star cluster with strong stars near to saturation. We don't need 32 photos of 4 min. Only 11 x 4 min e others 21 x 2 min !
Comparison of my normal convertion to 16 bits from Autosave 32 bits - linear - by HDR in photoshop - CS3.
photo 4
The most curious is that many people talk about the response of sensor used by Canon. And they are talking not only about sensor, they are talking about the sensor with the electronic Canon inside the camera.
Regarding unity gain...I've read and understood the many articles on the subject including the one you posted and I don't agree 100%. Just my assessment. In an ideal world, I'd rather have the raw signal with no digital gain applied or any other monkey business as we don't know what else is happening to the signal, but since none of the manufacturers of DSLRs wish to give us that we have to take what we can get.
However...unity gain or thereabouts is only good under the right circumstances. As I said, you have to try out the options under your own conditions and see what works best for you - which is what I believe you are doing
Personally, I find that ISO1600 works best under most circumstances, since it seems to be a sweet spot between signal and noise. In cool conditions (5C or less ambient) then the darks are practically clean from my camera. Under warm ambient conditions it may not be ideal because of the red snow storm that appears on images. But with the shorter exposure times it is possible to take a larger number of subs and reduce this noise with stacking. Lower ISO is certainly less noisy, but also requires longer exposures. One of my experiments last summer was taking 50x 2 minute subs of M42 to capture the faint nebulosity from a dark site and it was quite effective. It obviously isn't ideal since it was summer time and therefore warm evenings and noisy individual subs, but I take what I can from the equipment I've got
At this moment I am thinking that this issue hasn't a global answer. Many others things can affect the ideal time and ISO.
Some people consider as same thing ISO and gain of amplifier. Since digital camera is the main camera sold, why we haven't a potentiometer to adjust linear or logaritimic this gain ? Similar to audio amplifyer. Why it is in step: 100, 200 , 400 , ... , yet ? Would be because ISO isn't only gain ? There is some more work with the signal to correspond with the old ISO from film results ?
No matter. I will see my Canon as a black box. I have few steps of ISO to use: 200, 400, 800. ISO 1600 with my normal camera temperature about 35 degrees celsius is not useful. Therefore my main option will be 400 and 800.
I have a limit of signal on the histogram. The ideal would be around 25 % to 50 %. But my normal sky glow can kill it. So the maximum time of exposition with those two ISOs can be found fastly. Perhaps with 2 or 4 captures, since I work with steps of time: 30s, 60s, 120s, 180s, 240s ... Less dark files to store in my hard disk.
Therefore, do what I did: take 2 or 4 photos. Analyse the shape of stars or details from a cloud. Analyse the colors of stars or clouds. And ... the end ! It's done.
note: amount of noise can be reduced by dithering and dark files subtraction.
Undoubtly with CCD this is different. With them we have linear control of the gain ! And, perhaps, here is an other misunderstanding. Some people think the DSLRs as CCD. "you need more dark, you need more time of expositon, you need more ISO (gain), ..."
CCD and DSLRS are two different worlds ! Mainly if you are talking of stock color DSLR.
That extra 21x2 minutes hardly seems worth it Jorge. Just a little less noise. Doesn't seem to go any deeper in magnitude. Probably better with 22 x 4 minutes. Same total time.
Since you have clear sky. Want to do an experiment for me? Compare 1 x 4 minute to 4 x 1 minute stacked. Stretch the stack to equal the same background noise as the single frame. The theory is that 1 x 4 minute is better than 4 x 1 minute.
Jorge, consider cooling your camera. Even a single stage TEC will drop your sensor temps by -25C. Check out RCheshire's cold Finger thread here on IIS. If you can shoot near 0C any noise will disappear.
That extra 21x2 minutes hardly seems worth it Jorge. Just a little less noise.
yes, and I must be glad with this. Less noise only by some 2 minutes frames. Wow ! It is very cheap ! I must try with 1 minute, would have the same performance ?
With star clusters, the secret for a good shot is don't saturate the brightness stars. If they are saturated (all white color or exploded appearance) the photo will be lost, since that to reduce those informations in graphic process is very hard. Mainly if you will stack them in DSS.
Therefore, I see that sometimes 4 x 1 minute give better final results than 1 x 4 minutes.
Quote:
The theory is that 1 x 4 minute is better than 4 x 1 minute.
Yes, you are right. But for the theory ! You must adjust these theories to work with your true reallity. Think about high sky glow.
Another issue: with high camera temperature the read noise will nothing with reference to thermal noise.
Quote:
consider cooling your camera
I am thinking about. I have some difficulty to find a way to make it, because of accessories: filter wheel and OAG. I can't do a common modification.
The weight and fan vibration would be another issue.
Undoubtly this is the most important modification that I must to do. My thermal noise is absurd !
A question: With RAW mode, the sliders to adjust bright, contrast, saturation, sharpness, color tone is done on sensor ? Or it is done after the sensor, in analog signal, before the Analogic/Digital converter ? After that convertion, on digital information ?
Well, if they are not acting in/on the sensor ... the RAW signal of sensor isn't the Raw signal provided by the camera.
This is only for curiosity. We will work with raw signal inside the file provided by the camera. Real RAW, false raw ... no matter.
----------------
Someone can tell me, Jorge, these informations are in the file RAW as information for debayer task. They don't change the RAW data. Like white balance information.
Good, but I can't access to the signal RAW without debayering task. At least with Photoshop or DSS. I read that there are some debayer algorithm that are better than others.
Better than others ?!?! It will be right to say that if I debayer with one or other software I can have different photos ? Different data infomations ?
It seems that yes. Should be my worse photos problems of debayer algorithm ? It is the guilty !
A question: With RAW mode, the sliders to adjust bright, contrast, saturation, sharpness, color tone is done on sensor ? Or it is done after the sensor, in analog signal, before the Analogic/Digital converter ? After that convertion, on digital information ?
Hi Jorge, as far as I understand it, RAW mode is more-or-less the raw signal from your chip ... but not quite! Craig Stark's articles are a good mine of information about it, and some of the reasons it's not quite the pure raw chip readout we would like is a small amout of pre-RAW processing by Canon that he identifies in this article which also has loads about the Canon RAW format. I'm not that frustrated that it's non-ideal - getting such a decent CCD for a fraction of the cost of an astro CCD, and the ability to take good terrestrial snaps is worth it!
More generally, all your brightness, contrast, white balance, sharpness etc are post-processing, not done 'on the chip'. When you take a photo in RAW format, the camera stores your settings for this along with the raw image data. Your chosen RAW viewer (whether Canon's own Digital Photo Professional [DPP], DSS or Photoshop) first debayers the image with the RGGB filter, then either applies the settings you chose when you took the image, or allows you to change those settings to whatever you like. In DPP for example, you can alter any of the post-processing settings like white balance, saturation, sharpness etc, and the result is identical to that if you had changed them on the back of the camera, then snapped the picture. If you are lobbing your image straight into DeepSkyStacker, you can do some adjustments in DSS. But it's easier for DSS if you set up your camera with the settings you'd like (daylight white balance, neutral sharpness, contrast, colour settings etc), then the RAW file should have those settings when you open it in DSS.
As an aside, I never got my stars to look good from RAW files in DSS - the star edges in stacks were ridiculously hard, like a bad sharpening had been applied, despite the same files converted to TIFF and stacked producing softer edges. On reason I bought PixInsight, for my sins !
Thank you very much for your words and link, Andyc. It confirms what I had as idea. And I agree that Canons are not expensive and it has good performance for astrophotos. Not perfect as some CCDs, but it worth by its price.
I did more tests and did a conclusion:
1 ) for star clusters, not globular, 10 - 20 frames is ok for reduce noise and a total time of 20 - 40 minutes will work well
2 ) if you want more information about the background, like nebulas or galaxies near the cluster, you need much more time: hours
3 ) for nebula, you need much time to catch the faint areas: hours
I will post later some images of experience with 60 frames of NGC 3766 - opent cluster - 180 seconds ISO 800. I processed all them and only 10 frames and did a comparison.
I am finishing another test with NGC 3324 - Gabriela Mistral nebula. A faint one. I am doing 4 hours - 60 frames with 240 seconds ISO 800. I will do the same process above: all them and only 20. I will compare them, also.
!
