How many megapixels are just right?

[HughR]

Originally Posted by SierraGary

What the human eye can see is important to me, because the images may be printed and enlarged or magnified. I don't have a quantitative measure of the acuity of vision for comparison to the resolution of a print, but some eyes, not mine so much, are very sharp. The difference in price (1000D vrs 500D) for one of the three key components (OTA, mount, camera body) of about $140 represents about five percent of basic system cost. The modification, which seems essential, appears to add about $300. The finances are beginning remind me of the time I purchased an used Peugeot, but the astroimaging is a lot more fun.

I do research on human vision, and the highest acuity I have ever seen reported is about 20/12, which is almost twice as good as 20/20. That works out to about 50 line pairs (alternate black and white lines) per degree of visual angle. Viewing a 13"x19" photo from 24" (approximately the diagonal), that requires 240 pixels per inch to represent all information a human can see. This ppi figure would obviously be larger for very close viewing or smaller for distant viewing. Hope this helps.

[SierraGary]

I'm still trying to determine whether there are important differences between the performance of the 1000D and 500D. I took another look at the Gary Honis page where he tests noise and sensitivity of the 1000D, 450D and 500D. (http://ghonis2.ho8.com/rebelmod500d/...0dcompare.html) Look at the liberty stamp photo.

http://ghonis2.ho8.com/rebelmod500d/...sLEVcurve2.jpg

As I understand it, the images of the liberty stamp were taken under very dark conditions. The 1000D is the first image; the 500D is the third image. Here is what GH concluded after applying photoshop levels and curves to 5 min exposures at ISO 1600:


"Conclusion:
"Based on the above ISO 1600 dark frame tests, my opinion is that the 500D camera tested displays less noise than both the 450D and 1000D cameras tested. The DPP "thumbnail with information" view comparisons for sensitivity testing shows a very slight edge in brightness for the 450D. The five-minute exposure sensitivity test images show a slightly better overall image for the 500D, both with and without a dark frame subtraction.

"My personal opinion is that other than the obvious difference of lower noise in the 500D image, all three cameras performed well in the comparison testing , provide surprisingly similar results and marginal differences in the sensitivity images are barely noticeable."



GH finds "marginal differences in the sensitivity", but to my eye, the 1000D image is significantly brighter (as spit suggested it would be). This also shows up in the images that haven't been photoshopped.

http://ghonis2.ho8.com/rebelmod500d/...calibrated.jpg

Less noise (500D), or more brightness and less resolution (1000D). How do you choose? Assume the object of interest is faint. You can get more brightness on the 500D with longer exposures or higher ISOs, but then you get more noise. You can get less noise on the 500D with shorter exposures, but then you get less brightness. The tests suggest the 500D is superior in noise reduction, so maybe we can push to higher ISOs or to longer exposures, but longer exposures aggravate errors introduced by tracking and temperature changes. Noise can be removed with software. Does the 1000D have the edge for imaging faint deep sky objects? How would this be affected by filters?

As I was finishing this posting, I saw HughR's post on human vision. I certainly helps, but it will take me some time to assimilate.

[spit]

add the 1100d to the batch, improvements over the 1000d are movie mode, larger LCD, higher ISO, less noise, 14 bit raw, probably other stuff, and its LIGHTER!!!

[SierraGary]

Originally Posted by HughR

I do research on human vision, and the highest acuity I have ever seen reported is about 20/12, which is almost twice as good as 20/20. That works out to about 50 line pairs (alternate black and white lines) per degree of visual angle. Viewing a 13"x19" photo from 24" (approximately the diagonal), that requires 240 pixels per inch to represent all information a human can see. This ppi figure would obviously be larger for very close viewing or smaller for distant viewing. Hope this helps.

Is there a significant difference in the ability to resolve vertical lines vrs horizontal lines?

[SierraGary]

HughR mentioned that the best human eyes can make use of only 240 pixels per inch in a 13" x 19" photo viewed from 24". If we use the 19" figure and multiply by 240 pixels, that works out to 4560 pixels. The canon 500D sensor supplies 4,752 pixels on its long axis. That suggests that the human eye viewing a 13" x 19" display from 24" can use almost all of the information provided by the sensor of the 500D. The same presentation of a raw image from a 1000D, with long axis resolution of 3,888 pixels, would lose some detail. A smaller format might restore detail, but would lose some field of view.

If one is trying to reach out to, say 23 million light years with an 80mm SV80ED, the Pinwheel Galaxy would only occupy 1/4 of the width of the 500D sensor. You would probably want to zoom in (blow up the image) until the image occupied the full print or screen, so to experience the same sharpness, the viewer, from 24", would be limited to a smaller format. While the smaller image will appear to be sharper, it will not contain any more detail.

I am trying to work out the implications of having more or fewer megapixels. I think the above is correct, but I trust readers will not take it for granted.

[HughR]

Originally Posted by SierraGary

Is there a significant difference in the ability to resolve vertical lines vrs horizontal lines?

Not unless you have what is called "meridional astigmatism", which means that you have just such a difference. Otherwise, there is no significant difference in horizontal and vertical resolution. There is, however, a difference in ability to resolve diagonal lines, which is worse by about a factor of 2. That's referred to as the "oblique effect".

[HughR]

Originally Posted by SierraGary

HughR mentioned that the best human eyes can make use of only 240 pixels per inch in a 13" x 19" photo viewed from 24". If we use the 19" figure and multiply by 240 pixels, that works out to 4560 pixels. The canon 500D sensor supplies 4,752 pixels on its long axis. That suggests that the human eye viewing a 13" x 19" display from 24" can use almost all of the information provided by the sensor of the 500D. The same presentation of a raw image from a 1000D, with long axis resolution of 3,888 pixels, would lose some detail. A smaller format might restore detail, but would lose some field of view.

If one is trying to reach out to, say 23 million light years with an 80mm SV80ED, the Pinwheel Galaxy would only occupy 1/4 of the width of the 500D sensor. You would probably want to zoom in (blow up the image) until the image occupied the full print or screen, so to experience the same sharpness, the viewer, from 24", would be limited to a smaller format. While the smaller image will appear to be sharper, it will not contain any more detail.

I am trying to work out the implications of having more or fewer megapixels. I think the above is correct, but I trust readers will not take it for granted.

This is basically correct.

[SierraGary]

This thread has taught me quite a lot and I will continue to monitor it. I appreciate the contributions from practical experience and specialized knowledge.

To summarize, number of megapixels of the camera is probably not a critical consideration in the range from 10 to 15 MP, but it will very likely have a small effect on the sharpness of images of objects in the galaxy and local group neighborhoods, which I would like to target with my 80mm scope. One needs also to consider bit width of the processor (12 vas 14 bit). Noise is probably not a big factor when shooting with the 1000d, 1100D, 450D, or 500D, except when low intensity images are severely stretched.

It is possible to get impressive images from the least expensive of the group, the 1000D. All other things being under control, I might be able to print sharp images from a 500D in formats up to 13" by 19" viewed from 24" of targets out to 20 some million light years away. Larger formats would require viewing from a greater distance.

Now I can stop worrying about pixels and think about filters, alignment, tracking, guiding, temperature, humidity, pollution and exposure schedules. The weather hasn't given many opportunities for actual practice lately, but I look for improvements in June and July. I will continue to monitor the thread if any should care to carry it on or to jump in. Many thanks. I may start a new thread on the use of filters for astroimaging with the same Canon series, assuming I don't find one already posted.

[SierraGary]

This thread was a great help to me when I was trying to get some orientation in astrophotography, so here is a follow up. We bought a Canon T1i, which so far has performed well. We are still kind of thrashing around trying to figure out what objects can be imaged at particular times of the year at our somewhat restricted window (big trees), how to avoid moon glow, what exposure schedules to use, how best to pre and post process them, and how to get smoother guiding. But we have made a few images and gotten a big kick out of them. For Andromeda, just seeing that first smudge on the RAW frames was occasion for high fives, a terrific thrill. There is something about finding an object hidden in the sky and seeing it emerge in your own images that inspires interest and close observation more than looking at professional images ever did.

We can't match the images we see on the internet, but we are making progress and I have high hopes for the next few clear and moonless nights that come along. Contributors to the thread may recall that we wondered if we could image something as faint and distant as M101. After some struggle, we managed to do it, but it's a little noisy and we need more sub-exposures. By the time the rain quit this Spring, M101 was in a bad position for us, close to the tip over point. We will be imaging faint objects with more and shorter subs in future attempts. What we have so far can be viewed at

faculty.unlv.edu/gbp/astro

Thanks to all for the boost and good advice.

Gary