Comparing low light performance of different cameras

[jiberlin]

...

To compare low-light performance the real aperture number, not the equivalent aperture number must used. The equivalent aperture is used to compare the DOF.

I have created a comparison of the different cameras with the ISO values adjust to the maximum (real) aperture:



I have chosen the Canon G15 as nearest relative to the Canon G10 because the G10 is not available for comparison, and i have rounded the ISO values to the available values. Now, the Canon G7X iii wins, an it has a size more comparable to your baseline Camera.

I'll suggest an alternative interpretation. Setting the A6500's ISO to 6400, and the other cameras to the highest ISOs, respectively, that come close to matching the A6500's visible noise, I get the following:

• A6500: ISO 6400
• G7XIII: ISO 3200
• RX100VII: ISO 1600
• G15: ISO 800

The APS-C A6500 can work with an exposure 1 stop weaker than the G7XIII, 2 stops weaker than there R100VII, and 3 stops weaker than the G15 and still make a usable image. That makes the A6500 better in low-light in my estimation.

There is no advantage of similar noise level with one stop higher ISO if the lens 'eats up' this one stop by a maximum aperture that is one stop smaller.

As far as i have understand the OP, he is interested in the low-light capability of the 'system' (camera+lens). To realize the advantage of the A6500 a 2.8 lens is needed and then the system is no longer compact/small.

 

[jiberlin]

To compare low-light performance the real aperture number, not the equivalent aperture number must used. The equivalent aperture is used to compare the DOF.

We had better clear this up before proceeding. Yes, the equivalent aperture can be used to compare DOF. But it is also the correct choice for comparing low light performance.

Depth of field is related to the aperture diameter. Low light performance is also related to aperture diameter or, more precisely aperture area, which is what determines how much light the lens captures for a given FOV. So what is good for one is good for the other.

That is wrong, because with a smaller sensor, the light thru the lens has to illuminate a smaller area.

 

[Physics1]

To compare low-light performance the real aperture number, not the equivalent aperture number must used. The equivalent aperture is used to compare the DOF.

We had better clear this up before proceeding. Yes, the equivalent aperture can be used to compare DOF. But it is also the correct choice for comparing low light performance.

Depth of field is related to the aperture diameter. Low light performance is also related to aperture diameter or, more precisely aperture area, which is what determines how much light the lens captures for a given FOV. So what is good for one is good for the other.

That is wrong, because with a smaller sensor, the light thru the lens has to illuminate a smaller area.

Relax. There is no contradiction.

I can assure that for a given FOV, the light captured by the lens scales as the aperture diameter or, more precisely, aperture area. How could it not? Every photographer knows that as you open the aperture (lower f-number) then you get more light through the lens and onto the sensor.

But if you shrink the sensor size, it is not the case that the same amount of light thru the lens now illuminates a smaller area. You have cropped the image, so now you are throwing away a lot of the light that was previously falling on the larger sensor.

Does that clear up your objection?

 

[Physics1]

You can see the difference here. Those are all APS-C cameras, and the display is compensated for differences in pixel size (i.e., pixel pitch). Also, in practice you can sometimes see a big difference if you have a very large range of brightness in the scene, so you need to raise the shadows a bit to compress the range to your display device. The read noise is one of two factors that limits the dynamic range of brightnesses that the camera can record -- the other factor being how many photons per pixel can be recorded without clipping.

Thanks for the "here" link. Just as you say, these are all APS-C camera with the same sensor size, and all at the same ISO, f-number and shutter speed so all sensors should have seen the same number of photons. If only shot noise then all images should be equally noisy, but I take your point that they are not, and the difference is due to other noise introduced by the sensor and readout electronics, primarily readout noise. Yes, very interesting. The difference is certainly real and significant. Other players could in principle be QE, and active area ratio but differences here are probably small. I note that a previous similar comparison of four full frame camera showed no noticeable difference in noise, presumably because the readout noise for those cameras was similar.

 

[Dem Bell]

...

As I've mentioned in other posts, the notion that an f/6.3 lens can equal or potentially better one of the exotic primes - the venerable 500mm f/4 - runs contrary to what most of us are taught.

I think we all agree on "potentially better"

On paper, the difference is minor. In reality, that concept challenges the status quo.

In reality? In reality the Nikon 500 mm f/4 E FL ED VR is actually 490 mm f/4.12 and has the entrance pupil of 118.9 mm at infinity focus, dropping down to 464/4.12 = 112.6 mm at 15 m focus, dropping down to 385.5/4.31 = 89.4 mm at 3.2 m focus (she is a heavy breather. Bill's Optical Bench does not have the 800 f/6.3 yet) and the whole 125 mm vs 127 mm argument looks ridiculous.

 

[John Sheehy]

:-

To get an overview of low-light performance use the test-scene compare in the DPreview tests.

Despite being a useful and wonderful resource in general, do we all agree that the images are not in general useful for ranking cameras in terms of low light performance?

We do not. It is as perfect a comparison of sensor low light performance at all ISOs as you're going to find. Between the Big Three, there is very little difference at high or low ISO for any of their APSC cameras. They are all good and will make a much better picture than your old camera. All three have similar zoom lenses available.

The thing is, the OP was asking about something very relevant that the tool is not directly set up to do; the OP wants to know what noise is like with the largest aperture that you can use on the camera, and I assume with the same AOV and shutter speed. That would require different ISOs for comparison. The tool is ISO/exposure-centric, which tells us nothing about optics, and in fact, smaller-sensor cameras are pushed more into torturous f-ratios in order to normalize exposure. Look at the ratios of f-ratio to sensor diagonal in the tool; they are much higher with smaller sensors. So, smaller-sensor cameras are generally using smaller apertures than are available with the camera or available ILC lenses, with an ISO that is probably higher than it would be in a real world usage where the lenses are used at lower f-ratios than in the studio comparison tool. Higher-than-practical f-ratios not only cause gratuitous noise; they cause softer images where the details compete more poorly with the noise. There is also a wild card in AOV; the tool uses different AOVs and changes distance to compensate, which goes against the OP's demand.

The tool was really only set up to compare noise of sensors by controlled exposure; it becomes inconsistent and full of wild-cards when you try to do anything else with it. It is great for comparing different sensors of the same size, assuming similar optics are available, but as soon as you start comparing sensors of different sizes, especially when you are interested in full system (sensor and lens) performance; then the relevance of the comparisons become unreliable and arbitrary. I can guarantee you that people shooting smaller-sensor cameras are generally not using apertures as small as the ones used in the tool, unless they specifically need DOF.

--
Beware of correct answers to wrong questions.
John
http://www.pbase.com/image/55384958.jpg

 

[John Sheehy]

So what does this methodology measure then? Basically, all it does is rank the cameras on the basis of their sensor size, which may or my not correlate with low light performance.

If this was true, all full frame shots at any given ISO would look exactly the same. Do they?

Just for fun of it, let's look at four APS-C cameras at ISO 3200. One can conclude that the discontinued Fuji camera trading for about £200 on ebay is better than any recent offerings from Canon, Nikon and Sony all sold for about 3 times as much. Interesting.

Yes, in terms of shot noise in the RAW Test Scene images, then I would expect all shots taken with same sensor size, at same ISO, to be similar, as all sensors should receive same number of photons, and thus shot noise should be the same.

I fully approve of, and thank you for, your link to four APS-C cameras at ISO3200. And indeed, when all are viewed RAW, then at least to my eye, the image noise does seem similar. Phew! I would have been very worried if the theoretical prediction was seriously wrong!

JPEG is as much or more about the firmware's artistic license than it is about the sensor. JPEG is only worth looking at if you are going to be dependent upon OOC JPEGs with default styles. Look at the 90D style as JPEG; soft as hell. Yet, when you go to raw, the 90D is as full of detail as you can expect with an AA filter, which the other cameras don't have, and competes very well with them. Even the raw versions have some interpretation, but much less so than the JPEGs. For example, the tool shows no color moiré in raw mode with some cameras, but that is fiction, because all cameras will get moiré in the sketch, because of relatively low diffraction and aberration at the chosen f-numbers.

But this does bring up an interesting point. When viewed as JPEGs, then (at least in this example) we do see significant difference in apparent image quality and noise on account of the different in-camera processing algorithms. So for those like me that shoot in JPEG, comparing Studio Scene images of cameras with same sensor size at same ISO is arguably meaningful for evaluating the effectiveness of the in-camera JPEG image processing.

That's when the tool is most meaningful. When it is comparing different sensor sizes, it goes off into la-la land as far as practicality is concerned.

I am not considering read noise, as in most situation the shot noise is dominant.

in the higher tones of lower ISOs, but read noise is still the biggest variant in image noise in deep shadows and high ISOs.

 

[lumigraphics]

:-

To get an overview of low-light performance use the test-scene compare in the DPreview tests.

Despite being a useful and wonderful resource in general, do we all agree that the images are not in general useful for ranking cameras in terms of low light performance?

We do not. It is as perfect a comparison of sensor low light performance at all ISOs as you're going to find. Between the Big Three, there is very little difference at high or low ISO for any of their APSC cameras. They are all good and will make a much better picture than your old camera. All three have similar zoom lenses available.

The thing is, the OP was asking about something very relevant that the tool is not directly set up to do; the OP wants to know what noise is like with the largest aperture that you can use on the camera, and I assume with the same AOV and shutter speed. That would require different ISOs for comparison. The tool is ISO/exposure-centric, which tells us nothing about optics, and in fact, smaller-sensor cameras are pushed more into torturous f-ratios in order to normalize exposure. Look at the ratios of f-ratio to sensor diagonal in the tool; they are much higher with smaller sensors. So, smaller-sensor cameras are generally using smaller apertures than are available with the camera or available ILC lenses, with an ISO that is probably higher than it would be in a real world usage where the lenses are used at lower f-ratios than in the studio comparison tool. Higher-than-practical f-ratios not only cause gratuitous noise; they cause softer images where the details compete more poorly with the noise. There is also a wild card in AOV; the tool uses different AOVs and changes distance to compensate, which goes against the OP's demand.

The tool was really only set up to compare noise of sensors by controlled exposure; it becomes inconsistent and full of wild-cards when you try to do anything else with it. It is great for comparing different sensors of the same size, assuming similar optics are available, but as soon as you start comparing sensors of different sizes, especially when you are interested in full system (sensor and lens) performance; then the relevance of the comparisons become unreliable and arbitrary. I can guarantee you that people shooting smaller-sensor cameras are generally not using apertures as small as the ones used in the tool, unless they specifically need DOF.

The OP can only use the information that is available. The choices are to just guess or to get a rough comparison with test images.

I suspect that its going to be a fairly moot question in real-world use, that other factors would be more important when making a buying decision.

 

[Physics1]

The thing is, the OP was asking about something very relevant that the (DP Test Scene) tool is not directly set up to do; the OP wants to know what noise is like with the largest aperture that you can use on the camera, and I assume with the same AOV and shutter speed. That would require different ISOs for comparison.

Correct.

The tool was really only set up to compare noise of sensors by controlled exposure; it becomes inconsistent and full of wild-cards when you try to do anything else with it. It is great for comparing different sensors of the same size, assuming similar optics are available, but as soon as you start comparing sensors of different sizes, especially when you are interested in full system (sensor and lens) performance; then the relevance of the comparisons become unreliable and arbitrary. I can guarantee you that people shooting smaller-sensor cameras are generally not using apertures as small as the ones used in the tool, unless they specifically need DOF.

You put this more eloquently and succinctly than I could have.

There are many here like yourself that do understand what the DP Test Scene is intended for. But I'm pretty sure that there are also many that don't, and go ahead and use it for purposes and in ways that lead to incorrect conclusions. That's a pity, so I think the discussion has been valuable.

 

[Physics1]

The OP can only use the information that is available. The choices are to just guess or to get a rough comparison with test images.

If you are interested in the low noise performance, with aperture fully open, of complete cameras (camera+lens), then the DP test Scenes are not directly suitable. And if used naively you will not get a "rough comparison" but a completely wrong conclusion.

The other choice is not to just guess, but to rank the cameras on the basis of equivalent f-number, as I did in a previous posting. This will be a fairly good guide to image noise for pictures taken under the same conditions, with each camera doing the best it can at full aperture. Not perfect on account of differences in sensor read noise, but a lot better than guessing.

 

[Bill Ferris]

...

As I've mentioned in other posts, the notion that an f/6.3 lens can equal or potentially better one of the exotic primes - the venerable 500mm f/4 - runs contrary to what most of us are taught.

I think we all agree on "potentially better"

On paper, the difference is minor. In reality, that concept challenges the status quo.

In reality? In reality the Nikon 500 mm f/4 E FL ED VR is actually 490 mm f/4.12 and has the entrance pupil of 118.9 mm at infinity focus, dropping down to 464/4.12 = 112.6 mm at 15 m focus, dropping down to 385.5/4.31 = 89.4 mm at 3.2 m focus (she is a heavy breather. Bill's Optical Bench does not have the 800 f/6.3 yet) and the whole 125 mm vs 127 mm argument looks ridiculous.

Thank you for including the link to the "Bill's Optical Bench" page for the Nikkor 500mm f/4. I'd not yet seen that page or the listing of lenses Claff has tested. I just spent a few minutes looking at the pages for several of the lenses. That's an interesting resource. Do you happen to know if there are any lenses for which Claff has measured a large enough sample size to draw evidence-based conclusions about the degree of variability in the production process?

It is ridiculous that anybody would argue that bench-measured focal lengths and maximum f-stops for prime or zoom lenses of this caliber will be so far removed from published specs as to invalidate a photographer's analysis based on the published numbers. The actual numbers won't be a match for what's stamped on the lens but they'll be close enough.

 

[chrisfisheye]

The thing is, the OP was asking about something very relevant that the (DP Test Scene) tool is not directly set up to do; the OP wants to know what noise is like with the largest aperture that you can use on the camera, and I assume with the same AOV and shutter speed. That would require different ISOs for comparison.

Correct.

The tool was really only set up to compare noise of sensors by controlled exposure; it becomes inconsistent and full of wild-cards when you try to do anything else with it. It is great for comparing different sensors of the same size, assuming similar optics are available, but as soon as you start comparing sensors of different sizes, especially when you are interested in full system (sensor and lens) performance; then the relevance of the comparisons become unreliable and arbitrary. I can guarantee you that people shooting smaller-sensor cameras are generally not using apertures as small as the ones used in the tool, unless they specifically need DOF.

You put this more eloquently and succinctly than I could have.

There are many here like yourself that do understand what the DP Test Scene is intended for. But I'm pretty sure that there are also many that don't, and go ahead and use it for purposes and in ways that lead to incorrect conclusions. That's a pity, so I think the discussion has been valuable.

Many people want to know if the sensor is good (efficient if you prefer) and you can have a good idea by looking at the studio scene. And even compare sensors of different size if you know how equivalence works.

In every case (smartphone also..), I prefer to have an efficient sensor.

 

[Dem Bell]

...

As I've mentioned in other posts, the notion that an f/6.3 lens can equal or potentially better one of the exotic primes - the venerable 500mm f/4 - runs contrary to what most of us are taught.

I think we all agree on "potentially better"

On paper, the difference is minor. In reality, that concept challenges the status quo.

In reality? In reality the Nikon 500 mm f/4 E FL ED VR is actually 490 mm f/4.12 and has the entrance pupil of 118.9 mm at infinity focus, dropping down to 464/4.12 = 112.6 mm at 15 m focus, dropping down to 385.5/4.31 = 89.4 mm at 3.2 m focus (she is a heavy breather. Bill's Optical Bench does not have the 800 f/6.3 yet) and the whole 125 mm vs 127 mm argument looks ridiculous.

Thank you for including the link to the "Bill's Optical Bench" page for the Nikkor 500mm f/4. I'd not yet seen that page or the listing of lenses Claff has tested. I just spent a few minutes looking at the pages for several of the lenses. That's an interesting resource. Do you happen to know if there are any lenses for which Claff has measured a large enough sample size to draw evidence-based conclusions about the degree of variability in the production process?

These are the design specs from the patents. They describe what a perfectly manufactured lens should be like.

It is ridiculous that anybody would argue that bench-measured focal lengths and maximum f-stops for prime or zoom lenses of this caliber will be so far removed from published specs as to invalidate a photographer's analysis based on the published numbers.

Welcome to the real world. The published focal length is typically within 5% of the actual focal length. There might be more leeway on the f-stop.

The actual numbers won't be a match for what's stamped on the lens but they'll be close enough.

Are you saying that the entrance pupil changing from 89.4 mm to 118.9 mm depending on focus distance is close enough to 500/4 = 125 mm? It is what it is.

 

[Bill Ferris]

...

As I've mentioned in other posts, the notion that an f/6.3 lens can equal or potentially better one of the exotic primes - the venerable 500mm f/4 - runs contrary to what most of us are taught.

I think we all agree on "potentially better"

On paper, the difference is minor. In reality, that concept challenges the status quo.

In reality? In reality the Nikon 500 mm f/4 E FL ED VR is actually 490 mm f/4.12 and has the entrance pupil of 118.9 mm at infinity focus, dropping down to 464/4.12 = 112.6 mm at 15 m focus, dropping down to 385.5/4.31 = 89.4 mm at 3.2 m focus (she is a heavy breather. Bill's Optical Bench does not have the 800 f/6.3 yet) and the whole 125 mm vs 127 mm argument looks ridiculous.

Thank you for including the link to the "Bill's Optical Bench" page for the Nikkor 500mm f/4. I'd not yet seen that page or the listing of lenses Claff has tested. I just spent a few minutes looking at the pages for several of the lenses. That's an interesting resource. Do you happen to know if there are any lenses for which Claff has measured a large enough sample size to draw evidence-based conclusions about the degree of variability in the production process?

These are the design specs from the patents. They describe what a perfectly manufactured lens should be like.

It is ridiculous that anybody would argue that bench-measured focal lengths and maximum f-stops for prime or zoom lenses of this caliber will be so far removed from published specs as to invalidate a photographer's analysis based on the published numbers.

Welcome to the real world. The published focal length is typically within 5% of the actual focal length. There might be more leeway on the f-stop.

The actual numbers won't be a match for what's stamped on the lens but they'll be close enough.

Are you saying that the entrance pupil changing from 89.4 mm to 118.9 mm depending on focus distance is close enough to 500/4 = 125 mm? It is what it is.

Focus breathing is a common aberration. It's not a disqualifier to see it in a lens. If the equivalent Canon prime had a focus distance slider, it would be interesting to compare the two lenses. The Canon 500mm f/4L IS USM measured within a millimeter of the Nikkor in focal length when focused at infinity. Both were measured as f/4.1 at infinity. However, there's no focus distance adjustment option for the Canon that I could find.

Out of curiosity, I checked the Nikon and Canon 600mm f/4 primes. Again, the Nikon has a focus distance slider but the Canon does not. Both measured within 3mm of each other in focal length when focused at infinity. The Nikkor measured at f/4.08 and the Canon at f/4.23 at infinity.

While it's unlikely an owner would be pleased to learn their $12-14K lens operates at a much shorter focal length when focused at the minimum distance, it's pretty common. It's not necessarily uniform across brands.

My recollection from 6 or 7 years ago is that the Nikkor 70-200mm f/2.8G ED VR was noticeably shorter at minimum focus distance than the equivalent Canon 70-200mm f/2.8 zoom. So there are scenarios when competing products can have noticeable performance differences in this regard.

Again, that's an interesting page. Thanks for posting the link.

 

[ThrillaMozilla]

:-

To get an overview of low-light performance use the test-scene compare in the DPreview tests.

Despite being a useful and wonderful resource in general, do we all agree that the images are not in general useful for ranking cameras in terms of low light performance?

We do not. It is as perfect a comparison of sensor low light performance at all ISOs as you're going to find. Between the Big Three, there is very little difference at high or low ISO for any of their APSC cameras. They are all good and will make a much better picture than your old camera. All three have similar zoom lenses available.

The thing is, the OP was asking about something very relevant that the tool is not directly set up to do; the OP wants to know what noise is like with the largest aperture that you can use on the camera, and I assume with the same AOV and shutter speed. That would require different ISOs for comparison. The tool is ISO/exposure-centric, which tells us nothing about optics, and in fact, smaller-sensor cameras are pushed more into torturous f-ratios in order to normalize exposure. Look at the ratios of f-ratio to sensor diagonal in the tool; they are much higher with smaller sensors. So, smaller-sensor cameras are generally using smaller apertures than are available with the camera or available ILC lenses, with an ISO that is probably higher than it would be in a real world usage where the lenses are used at lower f-ratios than in the studio comparison tool. Higher-than-practical f-ratios not only cause gratuitous noise; they cause softer images where the details compete more poorly with the noise.

Yes.

There is also a wild card in AOV; the tool uses different AOVs and changes distance to compensate, which goes against the OP's demand.

Yes! That's definitely one to watch out for. That could easily lead someone astray. I don't recall it being mentioned before on DPR, but it underscores the requirement that you know what you're doing when you try to make such comparisons. Even so, sometimes the best you can get is an approximation or an educated guess.

The tool was really only set up to compare noise of sensors by controlled exposure; it becomes inconsistent and full of wild-cards when you try to do anything else with it. It is great for comparing different sensors of the same size, assuming similar optics are available, but as soon as you start comparing sensors of different sizes, especially when you are interested in full system (sensor and lens) performance; then the relevance of the comparisons become unreliable and arbitrary. I can guarantee you that people shooting smaller-sensor cameras are generally not using apertures as small as the ones used in the tool, unless they specifically need DOF.

Yes again.

 

[ThrillaMozilla]

To compare low-light performance the real aperture number, not the equivalent aperture number must used. The equivalent aperture is used to compare the DOF.

We had better clear this up before proceeding. Yes, the equivalent aperture can be used to compare DOF. But it is also the correct choice for comparing low light performance.

Depth of field is related to the aperture diameter. Low light performance is also related to aperture diameter or, more precisely aperture area, which is what determines how much light the lens captures for a given FOV. So what is good for one is good for the other.

Surely you must have noticed that cameras with small sensors such as the Canon G15 with 1/1.7" sensor also have amazing low f-numbers, given that the lenses are physically very small. For example, the G15 sports a 5:1 zoom lens with f/1.8-2.8!! And it's small to boot. Wow. If that sounds too good to be true then it's because it is. As far as low light capability is concerned, you cannot validly compare f-numbers across different sensor sizes, but must instead convert to "35 mm equivalent" f-numbers by multiplying by the sensor crop factor. That is what I did, and it is the correct procedure for comparing low light performance.

One can alternatively calculate the lens aperture diameter (D=L/fnumber) and use that to compare low light performance. The two methods are mathematically equivalent. Sometimes one method is easier and more natural, sometimes the other.

I'm in a photo blind with another photographer. It's morning twilight and we're both focused on the same animal. We have the same make & model full-frame camera bodies but are using different lenses. I'm using an 800mm f/6.3 and my friend is using a 500mm f/4.

Which system captures more light from the animal?

I'd say the the 800mm lens due to the animal being much bigger in the frame. 60% bigger in all linear directions.

Well done. The increased magnification and slightly larger entrance pupil give the longer slower prime a subtle edge.

This is one scenario in which the full-frame equivalent f-numbers do not predict the outcome.

I would word things differently. The 800 mm lens has a slightly larger aperture diameter (127 mm vs 125 mm) so it will collect more light for a given field of view (FOV). Period.

Again, that would be a foolish basis for a choice, because you don't have the numbers, and in any case that difference is insignificant. See Dem Bell's post and Bill Ferris' lastest post for some actual numbers, which demonstrate that lens specifications are approximate. They also omit transmission losses.

Sorry to say, the 127 vs. 125 mm comparison is indeed ridiculous.

 

[ThrillaMozilla]

The point you're missing, is that a superficial comparison of f-number would lead many to assume the 500mm f/4 would be significantly better in low light.

Indeed, many would make that mistake, and I certainly did not miss it.

In fact, if a photographer needs 800mm of reach for the photography they do, the 800mm f/6.3 PF will be just as good in low light and, perhaps, even a touch better.

We're going in circles here (for the last time), but as I stated previously, the 800 mm will certainly get you more pixels per duck, and should give approximately the same low light performance.

 

[John Sheehy]

Many people want to know if the sensor is good (efficient if you prefer) and you can have a good idea by looking at the studio scene. And even compare sensors of different size if you know how equivalence works.

It takes a bit of math to compare different sensor sizes, and the exposures used do not have the granularity necessary for direct comparisons. Most exposure ranges have 1-stop granularity, with the occasion top ISO in a 1/3-stop granular exception.

So, I often wind up using 3 windows in "Comp" mode, one for one sensor, and two for another sensor, one below what the equivalent would be for the first one, and one above it, a stop apart. I then think about how far between the two the one would be, to get an idea on "visible noise per unit of sensor area", which is really the only important noise factor if one is already completely aware of the role of sensor size, itself, in the context of available optics.

So, if I choose the Canon R7 (1.6x-APS-C) as a single window, and the R5 for comparison, I will make one window R7 at ISO 25.6K, and one window R5 at 51.2K, and the other R5 at 102K, knowing that in equivalence of the crops in the windows in the tool, the R5 at 65.5K would be equivalent to the R7 at 25.6K. So, I ask myself, how noisy would 65.5K on the R5 be, interpolating from 51.2K and 102K. With these particular cameras, the answer is simple; the R5 is less efficient with noise to the point that the R7 at 25.6K is better than the R5 at 51.2K, with no need for visual interpolation, and I can then say that at high ISOs, in the shadows, the R7 is about 1/2 stop less visibly noisy, with equal total light.

4:3 sensors slightly crop the target on the left and right, to lose 1/9 of the target area. m43 sensors are not exactly 1/4 the area of FF sensors, etc. Doing more detailed math, I found that m43 sensors are actually getting about 1.75 less stops of total light with the same exposure, not 2 stops less per window in "Comp" and "Print" modes. So, some of my claims that very small sensors were better than the best FF ones did not take this 0.25-stop discrepancy into question, and may have been more a case of being very close, but it is still clear that smaller pixels on a smaller sensor do NOT necessitate lower visible noise efficiency.

A more advanced version of the tool could theoretically have the sensor and pixel parameters available for each sensor, and scale things for target comparison types based on those numbers, in order to effect something like "the same total light" for each of the four windows, all upsampled, to avoid loss of captured detail.

In every case (smartphone also..), I prefer to have an efficient sensor.

--
Beware of correct answers to wrong questions.
John
http://www.pbase.com/image/55384958.jpg

 

[Ellis Vener]

Pick the one that

a) is inside your budget range

b)you enjoy using the most

c) yields results that satisfy you and what you mostly do with a camera

d) fits your budget

Everything else is just argument.

--
Ellis Vener
To see my work, please visit http://www.ellisvener.com
I am on Instagram @EllisVenerStudio
“It's not about the f-stop." -Jay Maisel
Don't be "a photographer.” Be photographing. (Paraphrasing William Faulkner's advice to writers.)

 

[Jack Tingle]

There's nothing wrong with your logic. I suggest a screening step first.

Put on a nice lens you'll want to use. Take a series of photos in lighting you'll encounter. Step ISO up and find the highest ISO you can tolerate on each camera/lens. Now rank the cameras you compared and see which ones you like best. Then make your science project of the ones that make the cut.

With my older camera, I find no impact below ISO 800, so the camera stays on that. I can, of course increase it as needed, but for 90% of what I do, 800 is fine.

My newer camera can handle ISO 1600 with no degradation. Again, I can raise it as needed.

I'm not picky about a bit of noise in deep shadows. If I object, I can always wipe it out in post. Similarly, I don't mind bright whites blowing out a small amount. Someone saying, "The highlights of your white clouds are blown out, you cretin," would not make me sad.