Astrophotography Talk Forum Forum
proamateur wrote:
Bill Ferris wrote:
proamateur wrote:
Bill Ferris wrote:
reinhard becker wrote:
But take another comparison: If you take a picture of the moon, you will always need the the same exposure if the aperture is the same, independent from the diameter of the lens or scope.
You've selected an example that perfectly illustrates the value of a larger lens aperture diameter. If you use the same camera and two different lenses - a 20mm f/6.3 and a 600mm f/6.3 - to make photos of the Moon at the same shutter speed, which lens collects more total light from the Moon?
That's right, the 600mm f/6.3. The longer focal length magnifies the subject so that it fills a much larger area of the frame. As a result, more of the light projected upon the sensor is moonlight. The exposures might both be f/6.3, 1/500 in both photos but the photo made with the 600mm lens - and its 95mm diameter entrance pupil - captures more light from the subject.
That is physical law for objects. Otherwise an exposure meter would make no sense in photography, because it only works with aperture, esposure time and ISO. I have never a an exposure meter where you can enter the diameter of your lens!
That's one way of looking at photography. Yes, any two cameras working at the same f-stop and shutter speed in the same light will be working with the same exposures.
But let's ask a question: what's more important, the exposure or the total light captured from the subject?
If exposure was the key, we could all use smartphone cameras to make single exposures of a scene and make amazing photos. However, the reality is that many photography enthusiasts eventually migrate to using a dedicated interchangeable lens camera because the photos made with that camera are higher in quality.
Why is that? How do two cameras working with the same exposure (and the same exposure settings) make such different photos?
The answer, at least in part, comes down to image quality being directly tied to the total light used to make a photo. At the same exposure settings of f-stop and shutter speed, the same light from the scene per unit area - the same exposure - is projected upon the sensor. However, the much larger surface area of a dedicated camera sensor allows it to capture more total light than the tiny sensor in the smartphone.
As a result, the image made with the dedicated camera has more dynamic range, less visible noise, and better color rendering. The dedicated camera photo also has more detail. That is a direct result of working with as larger entrance pupil, which allows the longer focal length lens to resolve finer details.
It's not exposure that determines image quality. Total light determines image quality. And this opens the door to an even deeper dive into the subject. Suppose we're not interested in photographing everything visible in the frame. Suppose our subject is something filling only part of the frame. How do we maximize image quality of that subject?
Let's return to lunar photography. Suppose you use the same camera to make photos of the Moon using two different lenses. For the first photo, you mount a $17,000 800mm, f/5.6 telephoto lens to the camera. It has an impressive 143mm diameter entrance pupil. For the second photo, you mount the camera to a $2,000 8-inch, f/10 Schmidt-Cassegrain telescope.
The 8-inch SCT is a 2,000mm f/10 optic with a 200mm physical aperture. The Moon is going to perfectly fill the field of view. At the same exposure, the photo made with the telescope as the lens will be more detailed. The larger telescope aperture will resolve finer detail than the exotic prime lens. That combined with the Moon filling more of the frame allows the camera to capture more total light from the Moon at the same exposure.
Exposure may be the photographic quality that gets talked about. But it's not the key to making a quality photo. That's determined by the total light used to make the photo. And when we're talking about the total light from a specific subject - as is often the case in astrophotography - it's the larger lens entrance pupil diameter that wins the day.
That or stacking multiple exposures of the same composition to capture more total light, greater dynamic range, better color fidelity...and less visible noise.
If we talk about pinpoint stars it is a little bit different. For pinpoint objects the rule is different. But for pinpoint objects except for color of stars there is just one rule: get as much light as possible….
You know what the goal is. If you open yourself to the multitude of paths available to accomplish that goal, you'll see that there are scenarios when lens aperture - not f-stop - is the key to achieving that goal.
The topic of this thread is about (improving) noise and it's the total amount of light captured per pixel that affects the noise in the image rather than the total amount of light from the subject.
Incorrect. Total light - not light per pixel - determines signal-to-noise ratio. To illustrate, here's a link to the studio scene comparison tool featuring pairs of cameras from the same generation but with significantly different megapixel counts: https://www.dpreview.com/reviews/image-comparison/fullscreen?attr18=daylight&attr13_0=nikon_z7ii&attr13_1=nikon_z6ii&attr13_2=canon_eosr5&attr13_3=canon_eosr6&attr15_0=raw&attr15_1=raw&attr15_2=raw&attr15_3=raw&attr16_0=6400&attr16_1=6400&attr16_2=6400&attr16_3=6400&normalization=compare&widget=1&x=0.5866147194943744&y=-0.14049307502905825
Notice that the noise visibility of the cameras with fewer megapixels and, therefore, receiving more light per pixel, do not show less noise than the photos made with their respective counterpart higher megapixel cameras which received less light per megapixel.
The noise visibility in the photo pairs is the same because the total light used to make the photos is the same.
No, it's precisely correct - light is the key, which I think we agree on, but you are forgetting about pixel scale.
I'm not forgetting about pixel scale. I'm internationally ignoring it because it doesn't determine SNR and noise visibility. Total light does.
An image sensor comprises millions of discrete photosites and each one contributes to the overall noise.
The sensor acts as a complete system; not as millions of discrete systems. Also, noise has been known to be proportional to the square root of signal since long before CCDs and CMOS sensors. The same physics of optics and there nature of light that determined noise in the film era didn't change just because cameras went digital.
To achieve comparable noise per pixel you had to down scale the larger image, which I already mentioned in my previous comment, thereby normalizing the pixel scale.
Any comparison of photos for noise, resolution, etc. needs to be done with photos that are normalized to the same size and viewed at the same distance. Your method of comparing photos puts a thumb on the scale to unfairly favor one image over another.
Once you normalize the pixel scale, as you did in your example, then the noise levels will be comparable because the amount of light per-pixel will also be comparable (the actual result will depend on the algorithm used to down sample the larger image) but you are no longer comparing the pixel level noise of the original image.
The reason the noise levels are the same in the fair image comparison I shared, is that total light used to make the images is the same. Therefore both images have the same SNR and noise visibility.
Furthermore, the process of down sampling comes with an inevitable trade-off in resolution, which might not be an option depending on the application.
Again, a fair comparison of photos must be done with photos of the same size and viewed at the same distance.
By applying the extra step of scaling the image, your comparison does not illustrate the true per-pixel noise in the original image as captured by the sensors.
Per pixel noise doesn't determine noise. Total light does. Here's a link to an article I think you'll find instructive on the topic: https://www.dpreview.com/articles/8189925268/what-s-that-noise-shedding-some-light-on-the-sources-of-noise
To do this, images should be viewed at their original size (i.e 100% scale). For example, let's compare the studio scenes of the 24.5MP Z6II and the 45.7MP Z7II full-frame cameras.
Incorrect. A fair comparison is of photos of the same size and at the same distance.
The studio scene is shot with the same Nikkor S 85mm f/1.8 lens @ f/5.6, ISO 6400 and 1/20 sec - aperture, f/ratio and field-of-view are all the same; therefore, both sensors capture the same overall total amount of light from the scene when all the photosites are combined. However, the Z7II captures more detail because of it's higher resolution sensor while the Z6II captures almost twice as much light per pixel (~1.9x). To illustrate how this affects noise on the pixel level, consider the studio scene comparison below at original size (i.e 100% scale). To make the comparison more obvious and relevant for the OP, I selected the low-light option and highlighted a darker part of the scene.
Z6II vs Z7II pixel noise comparison - low-light original size
If you select the "Comp" view for a fair & meaningful comparison, both cameras show there same noise visibility, which is what we would expect considering that both photos are made with the same total light.
Bearing in mind that the total amount of light captured by the sensor has not changed, which image looks noisier now?
When you zoom in further on the image made with the higher megapixel camera, you make noise more pertinent. The area observed in that deeper zoom is smaller. It was made with less total light. Noise is more prominent. Scaling the images to the same size takes the thumb off the scale and shows for a fair and meaningful comparison.
As expected, the Z7II captures more detail from the scene but is clearly noisier because the sensor captures less light per pixel.
It's clearly noisier in the "Full" view because we're zoomed in further and seeing a smaller area of the image; an area made with less total light and, therefore, in which noise is more prominent.
Now, consider what happens if the image sizes are normalized by (down) scaling the larger image.
A fair and meaningful comparison is only possible when viewing photos of the same scene at the same size and distance.
Z6II vs Z7II pixel noise comparison - low light down-scaled
Again the total amount of light from the scene has not changed but the noise per-pixel is now comparable because the amount of light per pixel is similar;
Total light used to make the photos is the same. In this fair comparison, noise visibility is the same, just as we'd expect.
however, image detail has been lost in the process.
Incorrect. A higher megapixel image will show correspondingly more detail in a fair comparison of same-sized images viewed at the same distance.
In order to achieve comparable noise levels per pixel while maintaining the benefit of the higher resolution sensor, one would need to expose with the Z7II almost twice as long.
Incorrect. The noise equivalence and subtle resolution advantage of the higher megapixel sensor in the "Comp" view are as one would expect.
This can also be tested with the studio scene. Consider, the Z7II shot at 1/10 sec compared to the Z6II shot at 1/20 sec.
Z6II vs Z7II pixel noise comparison - 1/20 vs 1/10 sec low-light original size
As expected, the noise levels per pixel are now comparable and the detail provided by the Z7II sensor is preserved.
Zooming in on the higher megapixel image unfairly views a smaller area of the photo in comparison with the larger area viewed of the lower megapixel photo.
To overcome the disproportionate viewing scenario, a higher exposure and more total light are needed. This comparison is essentially meaningless, other than confirming the inequitable nature of the comparisons you've shared.
The noise levels are comparable not because ISO or aperture changed
ISO isn't a noise source.
but because each of the Z7II's pixels captured twice as much light per unit area.
Actually, more total light was used to make the image. As a result, an inequitable comparison has the false appearance of showing comparable noise. In fact, provided one understands how to interpret such things, the comparison confirms that the Z7II photo has less prominent noise due to having been made with more total light.
This also means that Z7II captures twice as much light overall so the normalized image sie should have less noise compared to the Z6II - and it does.
Z6II vs Z7II pixel noise comparison -1/20 vs 1/10 sec low-light down-scaled
This is an equitable comparison of photos scaled to the same size and viewed at the same distance. It shows what one would expect: the Z7II photo has less prominent noise due to having been made with more total light.
All else being equal like quantum efficiency etc, capturing more light per pixel is the key to lowering noise and there are many ways to go about achieving this goal.
Incorrect, for all the reasons provided. Please, read the DPR article. It's a good primer on noise for the layperson.
For example, there is absolutely nothing wrong with down scaling an image to improve the signal-to-noise-ratio, but this changes the pixel scale which can have some disadvantages depending on the application.
Scaling photos to the same size is required to perform a fair comparison.
For a wide-field image of the Milky Way where resolution is not the biggest concern, down sampling, which it's required anyway to view the full image on a normal screen, can certainly improve noise. However, for an image where resolution is the priority, say a picture of the moon, then down sampling may not be an option. In this case, lowering the noise with the same lens or scope would require longer exposure time, stacking or a combination of both.
Or use a longer focal length and larger entrance pupil diameter to fill more of the frame with the Moon and capture more total light from the Moon. When both Moon photos are scaled to the same size and viewed at the same distance, the photo made with the larger entrance pupil diameter will display less noise...because that photo was made with more total light from the subject.
In your example above, the 600 mm f/6.3 will capture more light from the moon and more detail compared to the 20mm lens. However, the light from the 600mm is spread over a proportionally larger area of the sensor so the amount of moon light captured per pixel is the same compared to the 20mm f/6.3; hence, the noise per pixel will also be the same. Unless something else is done with that light, like binning, then the larger diameter of the the 600mm aperture does not help to lower noise despite capturing more total light from the moon.
Compare the Moon images at the same scale an distance. The Moon image made with the 20mm f/6.3 lens, when scaled tp match there size of the moon image made with the 600mm f/6.3 lens will show substantially more noise.
Of course, scaling up an image will also scale up the noise. Scaling up also does not add any extra detail to the image. I think this is obvious to most photographers so I don't understand the point you are making. I will reiterate though that the noise on the sensor pixel level (when viewed at 100% without scaling) will be comparable as I demonstrated above.
Similarly, the 8-inch SCT may be able to capture more detail of the moon than the 800mm f/6.3 but it's still an f/10. The faster f/6.3 lens will still deliver more moon light per pixel for a given exposure time despite the larger aperture diameter. In order to capture the same amount of moon light per pixel with the 8-inch scope, an exposure time almost four times longer is required.
Scale the Moon images to the same size and view them from the same distance. The Moon image made with the shorter lens and smaller aperture diameter, when scaled to match the size of the Moon image made with the longer focal length and larger aperture diameter, will display more visible noise.
Same as above. However, you originally stated that the image from the 8-inch SCT f/10 will be more detailed than the 800mm f/6.3, which is certainly true. But it's also true that the image will be noisier on a pixel level (at the original size) for the same exposure settings as I demonstrated above.
However, pixel level illumination doesn't determine noise in a photo. That's an unfortunate and pernicious myth of digital photography.
To achieve comparable noise per pixel, as you did with the studio scene, requires down scaling the original image by approximately 2.5, which wipes out the extra detail captured by the 2000mm focal length.
Scale the photo made with the 800mm lens to the same size as the photo made with the SCT. The differences will be obvious and the SCT - slower f-ratio but larger aperture - is the clear winner.
There is a very good reason why that 800mm f/6.3 exotic prime costs so much - one just has to evaluate whether the extra cost is worth it for their use case.
Here's another proof that "light per pixel" does not determine noise.
It's a comparison of the Nikon D850 and Nikon D500. Both cameras have essentially the same pixel size and are working with the same exposure. The light outer per pixel is the same. However, the D500 image is obviously noisier. Why? Because the D500 is an APS-C camera and the D850 is a full-frame camera. At the same exposure, the APS-C sensor captures less total light. Therefore, it's photo displays more prominent noise.
But according to you, the noise visibility should be the same. The pixels are the same size and illuminated by the same exposure. Light per pixel is the same. So, noise visibility should also be the same.
However, the photo made with less total light is obviously noisier. That is as we would expect given the fact that noise is proportional to the square root of the total signal.
I doubt there is anything inherently wrong with the OP's camera or lens. To lower noise in the image, the OP should simply capture more light per pixel. For the same camera and lens, this may be achieved by:
- Increasing the exposure time of a single image (i.e capture more light)
- Stacking multiple images (i.e capture more light)
- A combination of 1 & 2 (i.e capture more light)
As I recommended and explained, capture more light.
--
Bill Ferris Photography
Flagstaff, AZ
http://www.billferris.photoshelter.com
