DPReview TV: Why lower resolution sensors are not better in low light

That really changed my mind. I too thought that lower MP and thus bigger pixels would result in less noise, even if compared to a downsampled image of a higher resolution sensor. Too bad you don't have the D700 in the new studio scene comparison tool, but I take your word on it (vs the 5D Mark II with the "comp" option).

Previously I would have declared myself to be in the low-MP-low-noise camp. Now that this quality aspect "low-noise" doesn't seem to be real, it's just low-MP and thus probably equates as low-budget camp. Guess how I feel!? You Canadians are just mean people! Wolves disguised as sheeps.

Isn’t this just a RIP test?

no video comparison? 60MP downsampled to 4K as good as A7Sm3 4K? That thing does ok video at its highest ISO settings.

Why nikon Z6 has better ISO performance then Z7?

Quote from a good article: Similarly, large pixels collect more electrons than small pixels, so the point when the pixel flips from a 0 to a 1 can be much more precise. That in turn means far less data processing. With small pixels, the camera’s firmware is forced to extrapolate data.

If anything, the vid shows how usable a 12mp camera is. Imagine how many 30+mp cameras out there have never even been used to print a 12x18.

sad to hear you talking about being apprehensive about insults..so many trolls here :(

i watched the entire video, and i'm not sure i agree 100%. i think this video shows the advantage of lower megapixels in terms of _today's technology_ is getting close to negligible...but i think a lot factors have been left out to push the narrative...like in this, case: sensor size.

if the idea presented here actually holds up, then a 20MP 4/3 sensor is on par with a 20MP APS-C/FF in terms of noise, yeah? because we're still talking pixel density here, we just added the sensor size factor. *shrug*

but at the end of the day, i don't really care. i'm old enough that ISO 800 was totally unusable when I started getting serious with digital photography, and the technology today is mature enough that digital noise isn't really a problem anymore.

As long as the total light gathering area is the same, two sensors of same generational technology should perform the same w/r to noise and low light performance.
Now, the big question is: does a 60 MP sensor have the same light gathering area as a 12 MP sensor, given there is some, albeit minimal distance between each pixel?

I know there may be situations where it’s imperative to have as little noise as possible but I’ve always struggled with the obsession over it. I’ve taken shots with plenty noise that I love. I’ve taken shots that aren’t ‘technically’ perfect, hell, I’ve even taken shots that aren’t perfectly in focus that I’ve loved.
Just as well I’m easily pleased, just realised I take a lot of bad shots.
In all seriousness though, I’ve got a D800, a D3 and a D2x and I’ve loved using them all.
A local company recently used 2 of the hundreds of images I’ve taken at their premises over the last couple of years, these are now on their premises wall and are ridiculously large and neither of them were taken with the D800, because the pictures they chose were more pleasing to their eye….photography is a subjective thing and I think we do our profession/hobby/obsession, whatever it means to you a great disservice by picking at the details of it such as megapixles/camera make/noise.

Just my tuppence worth

There are many ways to compare the cameras with different resolutions. This video address the most common one: comparing two (printed) images of the same size. However, many other comments and suggestions in this thread are also legit and worth thinking. It's just that they are different experiments (unrelated to this video) and are addressing different conditions. To me, comments like "you don't get the point 1000%" are the most useless. /ending my useless comment

Where is Roger Cicala when you need him?

The entire comparison and conclusion was flawed from the beginning and in it's entirety is without any purpose. In the absence of examining and solving difference equations using the Classical methodology or Quantized linear sensor approximation, true light absorption into layered digital processors are without value.

Differentiation Properties and Symmetry Properties are necessary to examine the sampling of Non casual inputs. The conclusions were made without any disciplined step sequences which are essential for a proper comparative analysis.

I understand that the myth is debunked when photos are printed at the same size but it would be interesting to see how they compare at the same PPI. Is the noise for a high MP large print the same as a small MP small print for the same size sensor?

I'm interested in this experiment specifically for video. Using the same print size is akin to using the same TV size, but if one TV is outputting at a higher resolution (higher PPI like the prints), it would be tough to compare noise directly, no? Or am I crazy?

What’s amazing is that even after a relatively simple demonstration and clear explanation, a number of DPR commenters still refuse to understand it.

I think the way to go is a "transformer" sensors. As SUPERCCD was back in long-ago days. Now we have new technology called QuadCMOS. Imagine having a 80Mp camera that can easily become a 20MP one with much better low light performance a OR with much better dynamic range (your choice). For me this would be a very tempting technology. It already began making it first steps on mobile market so I think it can appear in future cameras also... Hey, Fuji! You're THE best at non-standard wondrous sensor AND you said you're creating something new for us in 2022 in that field, maybe this is what it's all about?

Any difference in dynamic range?

How about autofocus performance in extreme low light though, does the lower MP count help here? From what I read in the specifications, the A7sIII focuses down to -6EV while the A7rIV will AF down to -3EV.

I honestly wouldn’t mind an A7sIII for photography for the smaller files, since Sony unfortunately doesn’t offer small-RAW files. 12mp is more than enough for my purposes.

Something everyone should keep in mind is that resolution is not always set by the sensor, but by diffraction in the lens. Pixel dimensions are more or less in the range of 3 to 6 microns on a side. That must be considered in comparison to lens resolution sizes that are depedent on the f-stop in use. Using the Rayleigh condition, a "perfect point" is diffracted to 1.22 W/D where W is the wavelength of light considered and D is the diameter of the aperture opening. That's an angular "size" in radians. To convert that to a distance on the sensor, multiply by the lens to sensor distance, (m + 1)F, where m is the magnification from real life to the sensor, and F is the focal length. Noting that F/D is f-number (N), the best lens resolution possible is 1.22WN(m+1). m in most cases (macro shooting is the big exception) is well below one, so we can simplify to 1.22WN in those cases. Macro shooters usually top out with an on-sensor m of 1, so they may wish to use 2.44WN.

Thanks TV-team, this made a lot of sense to me!

Costas did mention that he could work with the 12 mp photo, and get it very close to that from the a7r IV at a 22 X 33 print size. And Chris said the larger format image could be equally improved as well. Thing is, the 17R IV is already overkill for most prints up to 30X40, so no one would normally up res them anyway, where as the 12 mp camera is still good enough for fairly large prints, and with Gigapixel, an amazingly efficient software, for instance, will create a close match. A 20-24 mp sensor is really a sweet spot. High resolution cameras are very expensive, many photographers are fine with the details they're getting even from 12 mp cameras, its only a photograph after all, and getting into the halls of MOMA has nothing to do with resolution.

Modern Low Resolutions Sensor Cameras are mostly if not solely focused(Engineered) on Video aka the 10.28-megapixel high-sensitivity Live MOS sensor LUMIX GH5S or the Alpha 7S III 12.1MP backside illuminated full-frame sensor. So IMO, we shouldn't get all caught up in their comparisons to the far higher Modern Resolutions Sensor Cameras in which many are far more focused (Engineered) on shooting stills. IMO.

A good anaolgy I use to explain this to people is cropping. Using a low resolution sesor with bigger pixels is like using a digital crop. The sensor averages out noise at the point of taking the picture but the resolution is lost forever. A high resolution small pixel sensor retains the resolution information but can be digitally averged in post by applying noise reduction. In both cases the result looks the same but in the latter case the photographer has more ccontrol and retains the resolution information.

Strange article heading: "Why lower resolution sensors are not better in low light". It's got nothing to do with resolution and everything to do with "pixel size" (the more accurate term being photosite size - sensors don't have "pixels").

This would be a very good time to thank renowned physicist Emil Martinec, whose 2008 paper we used to convince DPR editors and readers about the relationship between noise and pixel size.

http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/

Good, straightforward, clear video - thanks. Judging from some of the comments though, people don't change deep-seated beliefs so easily, despite clear evidence to the contrary. Bit like climate change...

Anyway, I think these educational pieces are good and make a nice change from 'which gear is better' type content. I feel like another piece on equivalence is due! Yes it's a minefield I know, but actually much needed because it's a complex and much-misunderstood area. If you use a range of formats, then you end up thinking about it a lot - or at least, I do.

It should probably be covered in stages rather than all at once, because there are many aspects, and ideally to include some content in the written format.

@Commentators
It's not about the noise. It's about what's left after we filter it out. Counting noise specks at 100% doesn't tell us which picture is higher quality. Higher resolution will always produce more noise. AND at the same time it will produce more data as well. More data = higher quality.

You don't make it clear that this largely because of BSI and microlens improvements and the fact that full frame pixels are still enormous in comparison to smaller formats. At smaller pixels sizes with FSI quality really started to drop (just compare the Nikon J1 to J4). BSI fixed this by moving wiring to the back and improving the angle response (see the Nikon J5). Then there was higher cross talk at very high resolutions which was again solved in phone sensor with well isolation and will most likely be moved to cameras when they reach that density. So it was something that was very real but is less of an issue now.

A very welcome exercise, and fully confirmed by my own personal experience (currently Z7 and Df). At a given sensor size and generation, more pixels has significant advantages when it comes to cropability, size potential of final prints, level of fine detail in very large prints observed at close distance. At same sensor size and generation, less pixels has significant advantages in terms of device and storage costs, readout speed, video implementations, editing speed. Noise is the same when you look at same size output up until the final print breaks down (and nobody routinely prints at sizes where 12 MP break down). Pixel peeping shows less noise per pixel for lower res bodies but that has very little to do with the end product (a displayed or printed whole image).

I would think the difference would depend on the pixel area efficiency, not the number of pixels, pixel resolution that is. By that I mean how many % of the sensor area actually is light sensitive.

You did say you were going to keep the test simple( shooting a portrait). now go and shoot a tree with the sun behind it with a 1/4 ff sensor and full m43 sensor and then post the colour of the leaves, the m43 leaves will not be green but grey, and the 1/4 FF sensor are still green. no one is interested in noise but colour reproduction.

In a nutshell, noise tends to be random. A higher resolution camera has smaller pixels than one with lower resolution, at the same overall magnification (i.e., print size), so the noise is less visible. Grain/noise is less visible in video because it is visually averaged over multiple frames, as long as the video (or movie) is running.

In a practical sense, the best comparison is in terms of results, the same print size, as used in this video. In terms of signal to noise ratio, the best comparision is at 100%, or pixel=pixel.

I have both cameras, an A7Riv and and A7Siii, and noise has not been an issue in either. Understand that in a situation which requires ISO 25,600 (a long way from the top in either camera), you probably can't read the dials on the camera without assistance.

Few weeks before Sony a7sIV release : why low resolution sensors are better in lowlight

I thought the camera comparison was apples and oranges. The comparison should be of same megapixel count sensors of different aspect sizes, so same MPX in 35 full, APS-C, 4/3rds. If you review the "studio" camera tests looking at raw files, it looks to me that full frame 35mm has a near two stop advantage noise-wise over APS-C, that is, at the point where you start seeing noise, "ASA" 800 on APS-C looks like ASA 3200 (two stops) on full frame.

Bigger pixels especially in full frame lower megapixel sensors are more sensitive to light and can gather more, this video doesn't dispel anything. The high megapixel version simply has a modern processor that deals with noise better than older cameras.

The point still remains, larger pixels gather more light which you can't argue with physics.

I have a Nikon 1 V2 camera with its dinky little 1 inch sensor. I agree with the point of this article, number of pixels has little to do with noise. I found very little difference in noise levels with all the Nikon 1 cameras I own and have tried. I did, however, find a difference with larger vs smaller number of megapixels. The 10 megapixel cameras (Nikon 1 V1, and Nikon 1 J1), definitely had a lower resolution, and less detail, compared to V2, V 3, J4, J5 models, all with 14 or more megapixel sensors. This obviously flies in the face of the "megapixel myth". For the responses, I have plenty of popcorn.