Comparing low light performance of different cameras

[ThrillaMozilla]

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.

We can add the noise measurements to this, but we have to make a correction. There's a theorem that states that random noise is a function of the size of the aperture. For a digital sensor it varies with the linear size of the pixel. Therefore small pixels inherently have lower noise levels than large pixels (but also lower S/N). The R5 and R7 have 4.39 and 3.19 micrometer pixels, respectively. So, to normalize the R7 data to the larger pixel size of the R5, multiply by 4.39/3.19. Claff's graph is presented in log2 units, so the correction is log2(4.39/3.19) = 0.46. Add 0.46 to the R7 data.

But noise can be structured, and is not always random, so it's good to have the visual confirmation of the DPR tests.

 

[Physics1]

Help. I have a mystery.

As I am seriously thinking of getting the Sony A6700 with 18-135 f/3.5-5.6 lens to replace my circa 2008 Canon G10, so I visited my camera store and took images in the shop to compare the two cameras. So which camera performed better in respect of low light performance? I hope to post the full results, and they are dramatic, but I have met an unexpected contradiction in the results that I would like to resolve first.

I took a number of shots with each camera. In all cases both cameras were doing the best they could with apertures wide open at f/4.0 Both cameras shot the same scene inside the shop at the same distance, and zoom was adjusted so that both saw the same area of scene, ie same FOV. ISO was set to many values over a wide range, and at each ISO, shutter speed was adjusted to get correct exposure as shown by the camera. I was able to adjust the shutter in 1/3 stops so was able to set it quite accurately. The Sony was actually fitted with an 18-105 lens with constant F/4 over the full zoom range, because that was the only zoom lens that the shop had available.

OK. Here are the ISO and shutter speeds that were used for the Canon G10.

• ISO Shutter
• 100 1/15
• 200 1/25
• 400 1/50
• 800 1/100
• 1600 1/200

Nothing fishy about that. For each doubling of ISO, the shutter speed was doubled in order to maintain the same correct exposure as shown by the camera. Perfect textbook result.

Then did the same thing with the Sony A6700.

• ISO Shutter
• 200 1/50
• 400 1/100
• 800 1/200
• 1600 1/400
• and so on for higher ISO

Again, this looks exactly as you would expect, except that it does not match the G10 results. At any given ISO, the A6700 required a shutter twice as fast for correct exposure. I could accept a discrepancy of up to say 30%, but a factor of two? What is going on? Assuming that I have not made an obvious mistake as a result of my brain going mushy late at night, this requires explanation.

 

[Dem Bell]

Help. I have a mystery.

As I am seriously thinking of getting the Sony A6700 with 18-135 f/3.5-5.6 lens to replace my circa 2008 Canon G10, so I visited my camera store and took images in the shop to compare the two cameras. So which camera performed better in respect of low light performance? I hope to post the full results, and they are dramatic, but I have met an unexpected contradiction in the results that I would like to resolve first.

I took a number of shots with each camera. In all cases both cameras were doing the best they could with apertures wide open at f/4.0 Both cameras shot the same scene inside the shop at the same distance, and zoom was adjusted so that both saw the same area of scene, ie same FOV. ISO was set to many values over a wide range, and at each ISO, shutter speed was adjusted to get correct exposure as shown by the camera. I was able to adjust the shutter in 1/3 stops so was able to set it quite accurately. The Sony was actually fitted with an 18-105 lens with constant F/4 over the full zoom range, because that was the only zoom lens that the shop had available.

OK. Here are the ISO and shutter speeds that were used for the Canon G10.

• ISO Shutter
• 100 1/15
• 200 1/25
• 400 1/50
• 800 1/100
• 1600 1/200

Nothing fishy about that. For each doubling of ISO, the shutter speed was doubled in order to maintain the same correct exposure as shown by the camera. Perfect textbook result.

Then did the same thing with the Sony A6700.

• ISO Shutter
• 200 1/50
• 400 1/100
• 800 1/200
• 1600 1/400
• and so on for higher ISO

Again, this looks exactly as you would expect, except that it does not match the G10 results. At any given ISO, the A6700 required a shutter twice as fast for correct exposure. I could accept a discrepancy of up to say 30%, but a factor of two? What is going on? Assuming that I have not made an obvious mistake as a result of my brain going mushy late at night, this requires explanation.

You might want to play with https://www.photonstophotos.net/Charts/Measured_ISO.htm

and look up the difference between SOS and REI, e.g.

https://www.dpreview.com/forums/post/65221904

 

[Physics1]

Help. I have a mystery.

As I am seriously thinking of getting the Sony A6700 with 18-135 f/3.5-5.6 lens to replace my circa 2008 Canon G10, so I visited my camera store and took images in the shop to compare the two cameras. So which camera performed better in respect of low light performance? I hope to post the full results, and they are dramatic, but I have met an unexpected contradiction in the results that I would like to resolve first.

I took a number of shots with each camera. In all cases both cameras were doing the best they could with apertures wide open at f/4.0 Both cameras shot the same scene inside the shop at the same distance, and zoom was adjusted so that both saw the same area of scene, ie same FOV. ISO was set to many values over a wide range, and at each ISO, shutter speed was adjusted to get correct exposure as shown by the camera. I was able to adjust the shutter in 1/3 stops so was able to set it quite accurately. The Sony was actually fitted with an 18-105 lens with constant F/4 over the full zoom range, because that was the only zoom lens that the shop had available.

OK. Here are the ISO and shutter speeds that were used for the Canon G10.

• ISO Shutter
• 100 1/15
• 200 1/25
• 400 1/50
• 800 1/100
• 1600 1/200

Nothing fishy about that. For each doubling of ISO, the shutter speed was doubled in order to maintain the same correct exposure as shown by the camera. Perfect textbook result.

Then did the same thing with the Sony A6700.

• ISO Shutter
• 200 1/50
• 400 1/100
• 800 1/200
• 1600 1/400
• and so on for higher ISO

Again, this looks exactly as you would expect, except that it does not match the G10 results. At any given ISO, the A6700 required a shutter twice as fast for correct exposure. I could accept a discrepancy of up to say 30%, but a factor of two? What is going on? Assuming that I have not made an obvious mistake as a result of my brain going mushy late at night, this requires explanation.

You might want to play with https://www.photonstophotos.net/Charts/Measured_ISO.htm

and look up the difference between SOS and REI, e.g.

https://www.dpreview.com/forums/post/65221904

Thanks for that. Yeah, I get the general idea. Manufacturers can define ISO in different ways, and can potentially even be deliberately deceptive by defining in such a way as to make their camera give low noise images at "high ISO".

But if you test cameras in the real world as I have just done then it does not matter how the ISO is defined. The bottom line in the real world is that you point different cameras at the same dimly lit scene, all at fully open aperture, all at the same shutter speed and FOV, with ISO adjusted for the same observed image brightness, and whichever gives the best quality, least noisy image is the winner for low-light performance. What ISO is indicated is irrelevant. What is the source of the image noise, be it shot noise, readout noise etc is also irrelevant.

That is exactly what I have done. Assuming no one can find a major fault with the methodology, I will then present the results. And like I said previously, the results when comparing my ancient G10 with the Sony A6700 are dramatic.

 

[Bill Ferris]

Help. I have a mystery.

As I am seriously thinking of getting the Sony A6700 with 18-135 f/3.5-5.6 lens to replace my circa 2008 Canon G10, so I visited my camera store and took images in the shop to compare the two cameras. So which camera performed better in respect of low light performance? I hope to post the full results, and they are dramatic, but I have met an unexpected contradiction in the results that I would like to resolve first.

I took a number of shots with each camera. In all cases both cameras were doing the best they could with apertures wide open at f/4.0 Both cameras shot the same scene inside the shop at the same distance, and zoom was adjusted so that both saw the same area of scene, ie same FOV. ISO was set to many values over a wide range, and at each ISO, shutter speed was adjusted to get correct exposure as shown by the camera. I was able to adjust the shutter in 1/3 stops so was able to set it quite accurately. The Sony was actually fitted with an 18-105 lens with constant F/4 over the full zoom range, because that was the only zoom lens that the shop had available.

OK. Here are the ISO and shutter speeds that were used for the Canon G10.

• ISO Shutter
• 100 1/15
• 200 1/25
• 400 1/50
• 800 1/100
• 1600 1/200

Nothing fishy about that. For each doubling of ISO, the shutter speed was doubled in order to maintain the same correct exposure as shown by the camera. Perfect textbook result.

Then did the same thing with the Sony A6700.

• ISO Shutter
• 200 1/50
• 400 1/100
• 800 1/200
• 1600 1/400
• and so on for higher ISO

Again, this looks exactly as you would expect, except that it does not match the G10 results. At any given ISO, the A6700 required a shutter twice as fast for correct exposure. I could accept a discrepancy of up to say 30%, but a factor of two? What is going on? Assuming that I have not made an obvious mistake as a result of my brain going mushy late at night, this requires explanation.

You might want to play with https://www.photonstophotos.net/Charts/Measured_ISO.htm

and look up the difference between SOS and REI, e.g.

https://www.dpreview.com/forums/post/65221904

Thanks for that. Yeah, I get the general idea. Manufacturers can define ISO in different ways, and can potentially even be deliberately deceptive by defining in such as way as to make their camera give low noise images at "high ISO".

But if you test cameras in the real world as I have just done then it does not matter how the ISO is defined. The bottom line in the real world is that you point different cameras at the same dimly lit scene, all at fully open aperture, all at the same shutter speed and FOV, with ISO adjusted for the same observed image brightness, and whichever gives the best quality, least noisy image is the winner for low-light performance. What ISO is indicated is irrelevant. What is the source of the image noise, be it shot noise, readout noise etc is also irrelevant.

That is exactly what I have done. Assuming no one has a better idea for the best methodology, I will then present the results. And like I said previously, the results when comparing my ancient G10 with the Sony A6700 are dramatic.

You may find this sample page from ISO 12232:2019 of interest: https://www.iso.org/obp/ui/en/#iso:std:iso:12232:ed-3:v1:en

It's the ISO standard that defines digital ISO. The previous version was published in 2006, about two years before you got your G10. The G10 was built around a CCD sensor. I'd be interested in knowing more about Canon's method in the mid-2000s for determining ISO. Was it REI, SOS, or something else?

The current 2019 standard does not require manufacturers to implement ISO in a defined manner. It does identify two methods for reporting ISO: REI and SOS. Canon, Nikon and Sony use the Recommended Exposure Index (REI). Fujifilm is among the companies using Standard Output Sensitivity (SOS).

Both methods define ISO as equal to 10/Hm, where Hm is the average focal plane exposure. SOS uses a fixed lightness reference (e.g. 18% gray) for tonality (lightness) matching a reading of 0 on the in-camera meter. REI allows the manufacturer full flexibility to determine which lightness standard to apply. Commonly, a brighter standard (e.g. 12.5% gray) is used.

I've got a ca. 2017 Fujifilm X-T20. As a standard practice, I choose exposure settings that, combined with my chosen ISO, produce an in-camera meter reading of +2/3 to +1 stop from 0. That reliably results in an image having a pleasing lightness to my eye. My Nikon cameras (REI) don't need that kind of mental exposure compensation applied.

It's not that manufacturers are being "deliberately deceptive". They observe the industry standard for ISO. Results can vary depending on the methodology a manufacturer uses and how they implement tools such as in-camera noise reduction.

The one-stop difference in exposure settings used by the G10 and A6700 is intriguing. Do the photos from both cameras have the same overall lightness? If not and if that difference matches a one-stop difference in settings, I would ask if either camera had exposure compensation other than 0 dialed in.

In a previous post you mention, "at each ISO, shutter speed was adjusted to get correct exposure as shown by the camera." How did you determine what the correct exposure was? Were you referencing the in-camera meter and going for a reading of, 0? Were you using another reference?

 

[Physics1]

The one-stop difference in exposure settings used by the G10 and A6700 is intriguing. Do the photos from both cameras have the same overall lightness? If not and if that difference matches a one-stop difference in settings, I would ask if either camera had exposure compensation other than 0 dialed in.

In a previous post you mention, "at each ISO, shutter speed was adjusted to get correct exposure as shown by the camera." How did you determine what the correct exposure was? Were you referencing the in-camera meter and going for a reading of, 0? Were you using another reference?

Yes, the one stop difference is more than I would expect. When you look at images from different cameras on the DP Test Scene then you observe that generally the exposure settings are the same or at least similar - I was not able to find a full stop of difference in DP images.

Neither camera had exposure compensation dialed in and, even if they did, it does nothing when in fully manual mode, at least for the G10. But if one camera had a full stop of exposure compensation it would stand out like a sore thumb in different lightness of the images and this is not observed.

Exposure was set using the in-camera meter. The observed lightness in the images is similar, as you would expect, with any difference being nothing like 1 stop. Very soon I will post a pair of images that dramatically shows the superiority of the A6700 with the 18-105 lens. Of course, with this methodology it is meaningless to say the the A6700 is superior, for the result depends (almost entirely actually) on the lens.

 

[Physics1]

Low light comparison of Canon G10 with Sony A6700 with Sony 18-105 f/4 lens

• Same scene and same distance from scene
• Same FOV
• Same shutter speed (as required for comparing low light performance)
• Apertures both wide open, which by chance meant both were at f/4.0
• ISO set for correct exposure as judged by camera
• JPEG images, because that's what I shoot
• Images shown here are a small crop from a much larger image

The result is chalk and cheese, as you would expect given the much larger lens on the Sony. To properly see the difference you will need to view full size. Given just how clear and beautiful is the A6700 image compared to my G10, I fear that I just have to have one. Yeah, it's bigger and heavier to carry when I'm out adventuring, but the engineer in me demands that I have it anyway! And as a bonus, I can zoom out to 200 mm equiv versus 140 mm for the G10. That's having your cake and eating it too, but the price, as always, is the much bigger lens ....

However, comparison of these 2 images gives only a qualitative measure of how much better is the A6700. I will later post a different pair of comparison images that gives a quantitative measure of the superiority, in f-stops. To do that, two images are chosen that are judged to be of equal quality, especially in respect of noise. Then the ratio of shutter speeds is the factor by which the A6700 is superior in conditions limited by low light. The result is remarkable.

 

[ThrillaMozilla]

Oh. Now I know why the difference in exposure. You have completely different color balances. And oh, yes, that can make a big difference.

But more than that, the image lightnesses are very different. Here is a comparison of the RGB composite histograms. (The situation is similar if you consider R, G, and B separately.)

Sony on left; Canon on right. These images are not similar.

 

[jiberlin]

Help. I have a mystery.

As I am seriously thinking of getting the Sony A6700 with 18-135 f/3.5-5.6 lens to replace my circa 2008 Canon G10, so I visited my camera store and took images in the shop to compare the two cameras. So which camera performed better in respect of low light performance? I hope to post the full results, and they are dramatic, but I have met an unexpected contradiction in the results that I would like to resolve first.

...

Again, this looks exactly as you would expect, except that it does not match the G10 results. At any given ISO, the A6700 required a shutter twice as fast for correct exposure. I could accept a discrepancy of up to say 30%, but a factor of two? What is going on? Assuming that I have not made an obvious mistake as a result of my brain going mushy late at night, this requires explanation.

As others have written, the exact usage/definition of the ISO setting can be different between cameras. And the metering will/can also be different, best is to use spot metering on a uniform area.

 

[Physics1]

Oh. Now I know why the difference in exposure. You have completely different color balances. And oh, yes, that can make a big difference.

Good point. Yet another real-world variable, though it won't change the conclusion that the G10 image is far noisier. I suppose that ideally I should have tweaked the white balance on one or both cameras to make the images look the same. Maybe next time, but it takes time, and I was already stretching the friendship with the stop staff.

 

[ThrillaMozilla]

Oh. Now I know why the difference in exposure. You have completely different color balances. And oh, yes, that can make a big difference.

Good point. Yet another real-world variable, though it won't change the conclusion that the G10 image is far noisier. I suppose that ideally I should have tweaked the white balance on one or both cameras to make the images look the same. Maybe next time, but it takes time, and I was already stretching the friendship with the stop staff.

Please reread my post. You quoted it just before I edited it and added a lot of material. Your measurements are all out of whack.

I don't doubt that the Sony is hugely superior, though. The Canon has a tiny sensor the size of a large matchhead. Better make a purchase from that camera store.

On an unrelated subject, I do wish someone would figure out the ISO standard mess, though, and explain it in a way that I could understand it.

Standards? We've got lots of standards. We've got so many standards that there is no standard.

 

[Physics1]

Oh. Now I know why the difference in exposure. You have completely different color balances. And oh, yes, that can make a big difference.

Good point. Yet another real-world variable, though it won't change the conclusion that the G10 image is far noisier. I suppose that ideally I should have tweaked the white balance on one or both cameras to make the images look the same. Maybe next time, but it takes time, and I was already stretching the friendship with the stop staff.

Please reread my post. You quoted it just before I edited it and added a lot of material. Your measurements are all out of whack.

I don't doubt that the Sony is hugely superior, though. The Canon has a tiny sensor the size of a large matchhead. Better make a purchase from that camera store.

On an unrelated subject, I do wish someone would figure out the ISO standard mess, though, and explain it in a way that I could understand it.

Standards? We've got lots of standards. We've got so many standards that there is no standard.

Yes, apart from the color balance, the image lightness is not the same. I had no way of knowing or improving such matters while hurriedly taking measurements in the shop. But, ya know, the key point would not be changed. It is obvious to blind Freddie that the G10 image is vastly noisier and inferior when shooting the same scene at the same FOV at the same shutter speed with both apertures wide open. That is the real world bottom line when shooting in low light, where the rubber meets the road, as Bill Ferris once put it. When I look at those two images, I know which camera I want!

You mention the small G10 sensor size, but I would be surprised if the major factor is not just that the A6700 lens is vastly bigger and collects more light.

I'm an amateur industrial archaeologist and often find myself taking photos inside of old abandoned building, factories, mines etc where the light is very poor, but flash completely ruins the shot. If I was serious I would bring a tripod to allow long exposure times, but I don't, so low light performance of the camera is important.

 

[ThrillaMozilla]

You mention the small G10 sensor size, but I would be surprised if the major factor is not just that the A6700 lens is vastly bigger and collects more light.

Yes, absolutely. They go together.

I'm an amateur industrial archaeologist and often find myself taking photos inside of old abandoned building, factories, mines etc where the light is very poor, but flash completely ruins the shot. If I was serious I would bring a tripod to allow long exposure times, but I don't, so low light performance of the camera is important.

Then maybe a fast lens too. I have a Canon EF-M 32 mm/1.4 lens that I like very much. It's only one focal length, but I amazed at how well it does in low light.

 

[Horacecoker]

Low light comparison of Canon G10 with Sony A6700 with Sony 18-105 f/4 lens

• Same scene and same distance from scene
• Same FOV
• Same shutter speed (as required for comparing low light performance)
• Apertures both wide open, which by chance meant both were at f/4.0
• ISO set for correct exposure as judged by camera
• JPEG images, because that's what I shoot
• Images shown here are a small crop from a much larger image

The result is chalk and cheese, as you would expect given the much larger lens on the Sony. To properly see the difference you will need to view full size. Given just how clear and beautiful is the A6700 image compared to my G10, I fear that I just have to have one. Yeah, it's bigger and heavier to carry when I'm out adventuring, but the engineer in me demands that I have it anyway! And as a bonus, I can zoom out to 200 mm equiv versus 140 mm for the G10. That's having your cake and eating it too, but the price, as always, is the much bigger lens ....

However, comparison of these 2 images gives only a qualitative measure of how much better is the A6700. I will later post a different pair of comparison images that gives a quantitative measure of the superiority, in f-stops. To do that, two images are chosen that are judged to be of equal quality, especially in respect of noise. Then the ratio of shutter speeds is the factor by which the A6700 is superior in conditions limited by low light. The result is remarkable.

Considering that your whole premise was to see the difference in image quality/noise when camera/lens combos are set to the best results they can achieve in low light, I'm baffled as to why you didn't have the G10's aperture set to f/2.8?

 

[Physics1]

Low light comparison of Canon G10 with Sony A6700 with 18-105 f/4 lens

• Same scene and same distance from scene
• Same FOV
• Different shutter speed, judged so as to get same amount of noise in the image
• Apertures both wide open, which by chance meant both were at f/4.0
• ISO set for correct exposure as judged by camera
• JPEG images, because that's what I shoot
• Images shown here are a small crop from a much larger image

I took a number of shots with the Sony A6700 at different shutter speeds, with the ISO set by the camera. Then I chose the Sony photo which visually gives the same image quality as the G10 at ISO 1600. Sure, that process of choosing an equally noisy image is a bit rough and subjective, but looking at the two photos above, I reckon they are approximately equally noisy, the image quality is (very roughly) similar.

So what this means, is that in low light, the A6700 can produce a similar image quality at 1/4000 sec as the G10 could at 1/200 sec. In other words, it is better in low light conditions by approximately a factor of 4000/200 = x20!

That is extraordinary. That mighty handy x20 benefit can be used to get same image quality at 20 times the shutter speed, as above, or put 20 times more light on the sensor for the same shutter speed, which explains why in the previous pair of photos, the A6700 image was so dramatically better. Or get the same image quality, with same shutter speed, with a light level x15 less.

I suspect the the true factor will be a bit less, because I would judge that the G10 image is lighter and has a slightly higher exposure. OK, so call it a x15 low light benefit.

Some of that benefit may come from superior JPEG processing in the modern A6700, and some may come from lower readout noise, some from better QE and a better active area ratio, and we already know that a lot of it (almost certainly most) comes from the larger aperture diameter of the Sony lens, consistent with the fact that Sony lens is massively bigger than that on the G10.

But for all inevitable "roughness" in this comparison, it is undeniable that the A6700 with 18-105 lens performs vastly better than the G10 in conditions of low light.

I will spare you the algebra, but it is fairly easy to calculate that for this example with both lenses flat out at f/4.0, and the Sony sensor being 3.07 times bigger than the G10 sensor, then the ratio of lens aperture diameters is x3.07, and therefore we would expect the Sony camera camera to perform 3.07^2 = 9.4 times better than the G10.

But to my surprise it seems to do significantly better, perhaps x15 better as estimated above. As suggested previously, possible reasons could be better JPEG processing, lower read noise, higher sensor QE and higher sensor active area ratio. The extensive scratching on the front of my G10 lens would not helping either.

I am convinced that the low light performance of the A6700 with 18-105 mm lens is dramatically better than my faithful G10, so much so that I need one!

I would actually get the 18-135 mm, f/3.5-5.6 lens which is a tad slower over most of the zoom range compared to the 18-105 mm f/4 lens that I did this testing with, but there is not much in it.

 

[John Sheehy]

We can add the noise measurements to this, but we have to make a correction. There's a theorem that states that random noise is a function of the size of the aperture. For a digital sensor it varies with the linear size of the pixel. Therefore small pixels inherently have lower noise levels than large pixels (but also lower S/N). The R5 and R7 have 4.39 and 3.19 micrometer pixels, respectively. So, to normalize the R7 data to the larger pixel size of the R5, multiply by 4.39/3.19. Claff's graph is presented in log2 units, so the correction is log2(4.39/3.19) = 0.46. Add 0.46 to the R7 data.

But noise can be structured, and is not always random, so it's good to have the visual confirmation of the DPR tests.

Yes, I've always found that smaller sensors tend to have better spatial randomness to their noise, in general. Many times DXO DR trends (adjusted for saturation) would suggest a little bit less area-normalized standard deviation for larger sensors, but actual images actually flipped that.

BTW, Bill's "input" that read noise is being referred to is electron charges after capture, not pre-capture photons projected by the lens, so it is subject to QE differences, but most current sensors are very close in QE, AFAIK, except the new Sony A9-III, which seems to have a QE about half of other current sensors or even less, back to the QEs of the 25 years ago.

 

[Horacecoker]

Low light comparison of Canon G10 with Sony A6700 with 18-105 f/4 lens

• Same scene and same distance from scene
• Same FOV
• Different shutter speed, judged so as to get same amount of noise in the image
• Apertures both wide open, which by chance meant both were at f/4.0
• ISO set for correct exposure as judged by camera
• JPEG images, because that's what I shoot
• Images shown here are a small crop from a much larger image

I took a number of shots with the Sony A6700 at different shutter speeds, with the ISO set by the camera. Then I chose the Sony photo which visually gives the same image quality as the G10 at ISO 1600. Sure, that process of choosing an equally noisy image is a bit rough and subjective, but looking at the two photos above, I reckon they are approximately equally noisy, the image quality is (very roughly) similar.

So what this means, is that in low light, the A6700 can produce a similar image quality at 1/4000 sec as the G10 could at 1/200 sec. In other words, it is better in low light conditions by approximately a factor of 4000/200 = x20!

That is extraordinary. That mighty handy x20 benefit can be used to get same image quality at 20 times the shutter speed, as above, or put 20 times more light on the sensor for the same shutter speed, which explains why in the previous pair of photos, the A6700 image was so dramatically better. Or get the same image quality, with same shutter speed, with a light level x15 less.

I suspect the the true factor will be a bit less, because I would judge that the G10 image is lighter and has a slightly higher exposure. OK, so call it a x15 low light benefit.

Some of that benefit may come from superior JPEG processing in the modern A6700, and some may come from lower readout noise, some from better QE and a better active area ratio, and we already know that a lot of it (almost certainly most) comes from the larger aperture diameter of the Sony lens, consistent with the fact that Sony lens is massively bigger than that on the G10.

But for all inevitable "roughness" in this comparison, it is undeniable that the A6700 with 18-105 lens performs vastly better than the G10 in conditions of low light.

I will spare you the algebra, but it is fairly easy to calculate that for this example with both lenses flat out at f/4.0,

If your term 'flat out' means both lenses were set to their maximum aperture, this is not true. They were both set to f/4. The max aperture of the Sony lens is f/3.5 and the G10's maximum aperture is f/2.8.

and the Sony sensor being 3.07 times bigger than the G10 sensor, then the ratio of lens aperture diameters is x3.07, and therefore we would expect the Sony camera camera to perform 3.07^2 = 9.4 times better than the G10.

But to my surprise it seems to do significantly better, perhaps x15 better as estimated above. As suggested previously, possible reasons could be better JPEG processing, lower read noise, higher sensor QE and higher sensor active area ratio. The extensive scratching on the front of my G10 lens would not helping either.

I am convinced that the low light performance of the A6700 with 18-105 mm lens is dramatically better than my faithful G10, so much so that I need one!

I would actually get the 18-135 mm, f/3.5-5.6 lens which is a tad slower over most of the zoom range compared to the 18-105 mm f/4 lens that I did this testing with, but there is not much in it.

 

[ggbutcher]

https://www.photonstophotos.net/Cha...erShot G12,Canon PowerShot G16,Sony ILCE-6700

Couldn't find the G10 in Bill's collection, so I just selected all the G1x series as a "class-indicator"

Got tired of all the "compare photos" pet tricks...

 

[ThrillaMozilla]

So what this means, is that in low light, the A6700 can produce a similar image quality at 1/4000 sec as the G10 could at 1/200 sec. In other words, it is better in low light conditions by approximately a factor of 4000/200 = x20!

That is extraordinary. That mighty handy x20 benefit can be used to get same image quality at 20 times the shutter speed, as above, or put 20 times more light on the sensor for the same shutter speed, which explains why in the previous pair of photos, the A6700 image was so dramatically better. Or get the same image quality, with same shutter speed, with a light level x15 less.

I suspect the the true factor will be a bit less, because I would judge that the G10 image is lighter and has a slightly higher exposure. OK, so call it a x15 low light benefit.

Some of that benefit may come from superior JPEG processing in the modern A6700, and some may come from lower readout noise, some from better QE and a better active area ratio, and we already know that a lot of it (almost certainly most) comes from the larger aperture diameter of the Sony lens, consistent with the fact that Sony lens is massively bigger than that on the G10.

But for all inevitable "roughness" in this comparison, it is undeniable that the A6700 with 18-105 lens performs vastly better than the G10 in conditions of low light.

I will spare you the algebra, but it is fairly easy to calculate that for this example with both lenses flat out at f/4.0, and the Sony sensor being 3.07 times bigger than the G10 sensor, then the ratio of lens aperture diameters is x3.07, and therefore we would expect the Sony camera camera to perform 3.07^2 = 9.4 times better than the G10.

But to my surprise it seems to do significantly better, perhaps x15 better as estimated above. As suggested previously, possible reasons could be better JPEG processing, lower read noise, higher sensor QE and higher sensor active area ratio. The extensive scratching on the front of my G10 lens would not helping either.

There is also considerable and unknown measurement error, namely, there are actually no measurements. Also, don't forget that jpeg files are not linear.

I am convinced that the low light performance of the A6700 with 18-105 mm lens is dramatically better than my faithful G10, so much so that I need one!

No surprise, really, is there?

I would actually get the 18-135 mm, f/3.5-5.6 lens which is a tad slower over most of the zoom range compared to the 18-105 mm f/4 lens that I did this testing with, but there is not much in it.

 

[Physics1]

But to my surprise it seems to do significantly better, perhaps x15 better as estimated above. As suggested previously, possible reasons could be better JPEG processing, lower read noise, higher sensor QE and higher sensor active area ratio. The extensive scratching on the front of my G10 lens would not helping either.

There is also considerable and unknown measurement error, namely, there are actually no measurements.

Agreed 100%. Nothing I did was a "measurement", and the subjective nature of the image comparison does indeed represent a considerable "error". But then again, the images for different cameras on the DP Review Test Scene are not measurements either, and these appear to be respected as a useful means of comparing camera sensor performance.

From my perspective, the procedure provided evidence that the A6700 with zoom lens does somewhat better in low light (compared to the G10) than the calculation based on lens aperture diameter suggests, and for me that is interesting and significant. I would not hang my hat on stating numerically just how much better though.