Why lower resolution sensors ARE NOT better in low light

[@Forthejoyofphotography]

You should do more of this 'myth' busting. I believe this together with your pragmatic reviews helps people to understand what best suits their photography and videography needs as well as just being interesting and entertaining.

[@dtibor5903]

They should have made a bit more extreme comparison like over ISO25600. That's where you start to have some benefits with the lower megapixels

[@SamueltloganVideos]

I agree. I would love to see the GRIII VS GRII and Leica m262 vs M10 vs M10r

[@the_original_skytiger]

Then someone would have to bust his myth.

[@colintoews7458]

Nice to see some images actually being printed 👍

[@jacksonrelaxin3425]

I’m inspired. I’m gonna go get some Walmart prints today.

[@mavfan1]

Same meaning without the word actually. It’s almost as if the word actually is a useless crutch.

[@anoopchaudhary8732]

@@ElementaryWatson-123 true, also one more thing that they agreed that we can see more noise one high mp camera when we zoom in , so if we shot a photo on high iso we can't crop in otherwise we can see extra noise.

[@MrAmchuk]

Wait you can do that?!

[@tkermi]

@@ElementaryWatson-123 Well, that's true, but I would say it's mainly due to noise (combination of luminance and chrominance noise). At least that's how it looks to me.

[@falsemcnuggethope]

This begs the question: how much else does the photography gear media get completely wrong? People working in creative fields are rarely very skilled in technology, so widespread misconceptions are common.

[@RainmakerAnton]

The bokeh shape and "quality" probably get too much unjustified attention.

[@dominiclester3232]

Yep, another one is: higher ISO automatically adds noise...it's the lack of light (usually a faster shutter speed) that adds noise! Test this with the same aperture/shutter speed and alter the ISO by one stop. Then in software alter the exposure to match and all 3 will be equally noisy. If you do a similar test with either ss or aperture then the noise will vary...

[@alphadog5676]

Another is that wide angle lenses are no good for portraits because they distort the face, whereas the real culprit is simply having the camera too close to the subject (I admit there is an indirect link here in that you are more likely to go in close with a wide angle lens in order to make the subject fill the frame)

[@Soundwave857]

@@dominiclester3232 Yea less light hitting the sensor = higher iso is needed, adding noise. the iso is the only way to get noise in the picture

[@dominiclester3232]

@@Soundwave857 Yes, I have tested this with a few cameras! Back to my first suggestion: keeping the ss/aperture the same, shoot 3 exposures with the ISO changing by one stop. In software, level up the exposures (+1 and -1) not altering the middle ISO shot. Then you will see at 100% the noise is the same. Then, since you are interested, keep the ISO/aperture the same for 3 shots, varying the ss. In software level up the (+1 and -1) shutter speed shots and you will see the faster shots are noisier. In other words the ss (or lack of light) is adding the noise, not the ISO.

[@alantuttphotography]

The myth did seem logical, but when you compare the final results (full image viewed normally), many differences disappear. I recall a few videos on the Northrup channel that showed this, and as you say, the main reason to get a lower pixel count is for speed, media storage, and keeping the camera cooler, which as I understand it, will help minimize noise as well.

[@perilthecat]

I love that Chris found an inventive way to get huge prints of his kiddo 😂 Joking aside, I know your view count is probably taking a hit but I’m actually enjoying this dearth of gear reviews due to a slow production cycle. You guys are coming up with some creative content and it’s fun to watch!

[@Zakna]

Chris just wanted free big pics of his kid for his home haha :p

[@JoATTech]

Isn't this Jordan daughter?

[@nobitaadidas]

@@JoATTech Chris's

[@sh8736]

Xmas presents for family too 😀

[@chirsd666]

Got the job done and some brownie points with the wife, eh?

[@topg2820]

The video actually convinced me that big prints from 12 MP cameras are still good

[@Maxime-ho9iv]

At this size, yea definitely.

[@topg2820]

@@Maxime-ho9iv billboard prints are not high resolution too bcoz they are viewed from afar

[@Maxime-ho9iv]

@@topg2820 billboard is very specific tho, because it is indeed viewed from very far. But go to any good art gallery and photography prints will be 40" by 60" or more. You can still print that with 12MP but you will definitely get better results with more MP, that’s just the way it is.

[@caleidoo]

@@topg2820 No professional billboard photographer is going to use a 12MP camera. They use medium format for pixel quality, color and tonality. Only those expensive bodies can offer that whole package.

[@nafnaf0]

Yeah it sort of depends, the rule of thumb is 300ppi. That would limit you to a 9.47in wide printed image without any crops. Yeah you can do things in post, but you start to see problems. I have printed 12MP iPhone pictures at 8.5" wide with minimal crops. They look 100% great, but I wouldn't go much bigger than that. So w/ 12MP you are basically limited to letter size (8"x10" standard) prints.

[@nyobunknown6983]

The myth comes from 100% viewing. To really compare a high res sensor to a low res sensor on your computer you should downsize the high res to the same resolution as the low res sensor. The downsized image will actually look better. Not to brag but I've been arguing this point since 2006.All the downvotes come from people who have low res cameras. :-)

[@MegaNardman]

Yup, the larger MP sensor image, when down-res'd, can average out the noise across several individual pixels. All things equal, it's an advantage when it comes to noise reduction.

[@Maxime-ho9iv]

Yep, me too. It’s a bit embarrassing that people in the field fought this for so long.

[@brianthompson9485]

This isn’t true… I have an A7SIII and an A7rv. The RV can’t hold a candle to the SIII while taking videos of the northern lights. The difference is massive.

[@shueibdahir]

@@brianthompson9485low read noise is the cause of that aswell as incredibly good noise reduction. But the sigma fp gets close to the a7siii

[@Astro-uv1xq]

Yes!! I agree that downsizing the image and comparing them in the same, lower resolution, will show that higher res is the same or better

[@oliverdecker1930]

Thanks, Chris and Jordan for another great video. You guys have found a perfect balance of being informative and entertaining at the same time.

[@davidpearson3304]

Just stay away from the DPreview message board. It’s so toxic with a bunch of knuckleheads arguing at who is the “smartest”

[@oliverdecker1930]

Yeah, the “Thoughts…” are strong on these ones… 🤢

[@mirrorlessny]

didn't see Jordan this time, was expecting him to talk about video benefits of lower mp sensors

[@blekenbleu]

Long ago, fill factor issues (relatively large gaps between photosensitive sensor pixels) did aggravate noise for high resolution sensors.

[@thomastuorto9929]

How long ago?

[@sabatiniontech7256]

This is why advanced sensors are so much more imporrant tban just raw pixel size. On earlier generation sensors the wires that connected up the pixels had to be routed around the edges of the pixel, significantly reducing true pixel size. With Back Side Illuminated sensors the light comes through the transparent back of the sensor to illuminate the pixels leaving the "front" free for use in wiring, each photo site now full size. A similar gain in dynamic range and high ISO perfotmance comes from stacking the sensor (putting the signal processing components on the sensor itself). Stacking reduces wiring length and thus lowers electrical noise on the sensor. Also smaller pixels reduce effects of radiation, cosmic and radioactive decay, since it then affects a smaller area of the image.

[@scb2scb2]

correct in the end it is/was about loss of data and how 'correct' the data is. On newer sensor less data is lost in the gaps so both FF sensors count the same amount of incoming light if the loss in the gaps is the same. But a low res. sensor will have less of a idea of where the light hit since its firing up a bigger area. In the end more data is always good, more data with less error is even better.

[@randomgeocacher]

Yes, and backside illuminated (electronics on the behind of the sensor) largely removes loss of light to gaps for electronics.

[@randomgeocacher]

@@ElementaryWatson-123 no, micro lenses does not fully resolve light collection problems due to metal wiring (unless Wikipedia description of backside illuminated sensors is all wrong.) Micro lenses are in front of the metal wiring. BSI is essentially exactly the same as previous sensors but the wiring moved away in order to not obscure light collected by the micro lenses. So, too the extent Wikipedia description is correct, BSI is an important improvement to increasing light captured ( reducing noise ).

[@sam80pr]

When you take pictures in normal to slightly low light situations your argument holds. But when testing in extreme conditions in real life you see a massive difference. I own both the a7s3 and a7r4. My first few days of experience with a7s3 was shocking. Pushing the sensor to environments with almost complete darkness. My friend could not believe what he was seeing. He said out of all cameras I own, and I own quite a few of them, this one was magic.

[@Jortio]

That is true but Chris didn't mention that there is a difference when read noise (not to be confused with shot noise) becomes dominant. That's where lower resolution sensors like the one mentioned here are still king. You'll see that with long exposures, files where you push high ISO shadows and ultra high ISO shadows. The color fidelity holds up longer (less blue/purple shadow cast) and less shadow noise. But yes, for most other low noise comparisons, the lower resolution sensor doesn't offer a noise benefit but rather less flexibility in terms of quality noise reduction after the fact.

[@richardbutler8532]

@@Jortio This is absolutely correct. I did discuss with Jordan and Chris whether to add a caveat about extreme conditions / very high ISOs, where the lower number of read noise events starts to come out in favour of the lower pixel-count camera, but it's difficult to include without getting into discussions of shot vs read noise. @sam80pr - You're quite right and I really respect that you acknowledge that Chris is addressing the general case, but not covering every detail. There are many people who'd be confrontational rather than nuanced.

[@Jortio]

@@richardbutler8532 Absolutely. For most people and most purposes, this is really the relevant takeaway. And as Chris said, there's only so much you can condense into a informative video without losing a good portion of your audience. At least we outlasted the TikTok target group.

[@Jortio]

@@Quoutub Based on my experience, even with smartphone cameras the pixel count for a given sensor size makes relatively little difference for noise, compared to other factors. But those other factors are a whole different can of worms to begin with. Massive differences in processing (compared to large sensor cameras from different brands) and especially the use of continuous stacking (or not). And then there's different (usually) fixed apertures etc etc. Sure, the 108 MP phone sensors for example aren't low light wonders, but neither are the 12 MP equivalents and some of the former still do better in low light than some of the latter, depending on implementation. The percentage of people shooting RAW with phones is still tiny anyway.

[@smaakjeks]

I'm looking at DPRev's picture comparison. a7s3 vs a7r4 @ ISO 100k viewed in "print". They look almost the same to me. I downloaded the RAW files and applied significant noise reduction, corrected white balance, and downscaled to similar size. There is a difference, but it's not massive. But yeah, if you find yourself regularly shooting at ISO +100k, then consider using a low pixel camera.

[@KiinaSu]

I had to explain this concept everytime I mention I would want a higher MP MFT camera because people always reply with the same thing: oh it will be unusable in low light. Finally I can just link a DPReview video

[@mcwaffles121593]

Well this doesn't help you. They're comparing 35mm sensors here with different MP, not MFT sensors which will still have more noise

[@gamingwithstand6886]

Lower megapixels are better for high ISO but only at super extreme high ISO you have to go past ISO 25,600 now to even notice a difference and the everyday photographer does not use those high ISO. Also DXO PhotoLab 4 Deep Prime and DXO PureRaw are the best noise reduction programs. I export to DNG and edit in other editors.

[@CallMeRabbitzUSVI]

You can take a look at high-res Smartphones he's like the Mi 10 Pro, and shooting in RAW with that admittedly much smaller sensor still looks sharp and isn't a noisefest

[@frstesiste7670]

Agree, don't know how many times I've had this discussion and some people just wont accept the facts no matter how much documentation I provide. Maybe this video will help.

[@coolerdaniel9899]

@@mcwaffles121593 The principle is exactly the same. When printed such that display size/print size are proportional to sensor size between cameras, noise is equal because the only factor that matters is total light gathering area, not pixel area. MFT sensors work no different to full frame sensors, they're just smaller.

[@thesharpercoder]

Nice video. I agree with everything in it. Except, your tests were a little misleading because you stayed within the ISO range of the high resolution camera, ISO 6400. I would like to see comparisons between images taken at progressively higher ISO settings. To me, this would be a better low light performance test.

[@CallMeChato]

Yes, because EVERYONE is still complaining about my 3' prints that I made back in the 35mm film days which have an equivalent resolution of 8 megapixels.

[@jacksonrelaxin3425]

7-8 mp cameras are criminally underrated. I have an old 7mp powershot that still takes amazing photos and even has a 60fps setting.. in 360p but still

[@danielweber3433]

Up to this day the best photo I ever took was with a Nikon D80. 10MP. It is sharp even printed in A4 size. There is no noise and the colours are great without any work on the file. And there wasn't too much light around either. You don't always have to have the latest gear, but I have to admit that my z50 makes every aspect of taking a Foto so much easier.

[@karlrichards]

Shit, everyone might jump back on the micro 4 thirds cameras again now eek lol 🤣🤣🤣😆

[@danielweber3433]

@@karlrichards Phone sensors are the thing to use!

[@thomastuorto9929]

@@jacksonrelaxin3425 I still have but don't use a 5.something mp point & shoot purchased around the year 2000. You can do some serious cropping with that camera for what it is.

[@TheFraziers]

Great comparison. This is the perfect example as to why (in my opinion) 24 mp sensors are basically the perfect compromise between resolution detail, scanning speed, low light performance, and file size management. My last 4 cameras have been full frame 24 mp and frankly, I would never want anything different.

[@Bayonet1809]

Actually this proved the opposite of what you claim, that instead higher resolution is always better, as long as your workflow can cope with the added processing power and storage requirements. The point being that as technology improves 24mp will not be a sweet spot any more, and arguably has already been replaced with 40-50mp, as evidenced by the Sony A1 and Nikon Z9.

[@jaekimchi]

@@Bayonet1809 his point includes file size management and sensor readout speed. Two points that Chris admits are benefits of low resolution sensors.

[@Bayonet1809]

@@jaekimchi And Chris also said that those things are changing, so are less applicable.

[@jaekimchi]

@@Bayonet1809 still applicable.

[@burritobrosvideos8060]

45mpx is the sweet spot

[@mothaybabonnam5632]

The REAL takeaway from this video is 120ppi (capture density) is a ton of pixels, even printed on that gigantic files. As contrast to internet warriors claiming they need 36-48MP to make the occasional 8x11” prints of their cats. Some of my best A3+ sized prints were made from 12MP, some even cropped.

[@codeofcodedotorg]

I think people believe this myth because of low light video tests where most high res sensors line skip or crop, therefore a low res sensor with full sensor readout obviously wins. Processor speed at this point is a non factor for photos so but I agree for video processor technology still has ways to go. That’s why I got the FX3 to handle all my video needs personally.

[@Leptospirosi]

It is not a "myth" it's phisics: when you print you are turning an digital image into an analogue one and the "grain" (which is "noise" in colour image) will be more important than light sensitivity of pixels Large photodiodes will always have an edge over small one because they get 4 times the amount of light for every nm reduction, but at these resolutions it does not matters and pixel density is what gives you details and colour blending.

[@semahj]

@@Leptospirosi So, if it doesn't produce the myth results regardless of physics of it its still a myth

[@Leptospirosi]

@@semahj try lifting up a severely underexposed image: think of an almost black screen and push it up like 6 stops. Think of a deep sky image. Any low res camera would destroy a high MP one every day of the week. Thats why people uses image staking: you take dozens of the same pictures and stack them together to "average" the colour errors (which si what gives the colour pattern to the noise) registered by the sensor. The larger the pixel, the more photon it gets, the less probability it is it records a faulty colour introduced by heat or electronic noise in the chip circuitry. The fact is that modern softwares are getting VERY good at correcting these errors and what you you see on the back of your camera is a software filtration of the noise to signal ratio of the pixels on your sensors. Unless you need EXTREME performances, go high res.

[@Your_Paramour]

@@Leptospirosi While it's true that larger photosites receive more light, the total amount of light received by the sensor in either case is equal which is the most important metric for most modern sensors. With smaller photosites, there is more overall read noise from the sensor, but when you downsample the higher resolution into lower resolution for modern cameras, the averaging of photosite values cancels out that higher read noise vs the lower resolution sensor. What has changed over the years is sensors have had their read noises continuously lowered so higher resolution cameras get better comparative noise performance. Obviously there will come a point where you push the iso enough you are amplifying the read noise so much it will be worse than the lower resolution sensor. For example, iso 12800 is 128x amplification of the base iso (assumed 100), while 102400 iso is 1024x amplification, 10x higher.

[@codeofcodedotorg]

Ugh, this is why they were wary of making this video, some people just don’t understand. If you got a 4K monitor (8.1mp) and look at a 12mp image and a 48mp image, but you crop in to view only 8mp, of course the 12mp image will look better. But if you fit to screen you will not see any difference. On the contrary, if you cropped in 800%, the 48mp image will destroy the 12mp one in every way.

[@SlaveNCMC]

There is another option: AI noise reduction. It works very well for a high-MP sensors, and after the downsize the picture looks nice.

[@arda_]

Watching this on my TV at 4K. What a beautiful video Jordan you are killing it. Also did the intern do the thumbnail what is with that? I almost didn’t click until I see the dpreview below.

[@niccollsvideo]

I'm not an intern!?

[@arda_]

@@niccollsvideo Made me chuckle :) Great informative video as always. One thing I would like to learn, does this still apply in much darker situations?

[@niccollsvideo]

@@arda_ To the extent that the noise still looks similar, yes absolutely. Now at extremely high ISOs there will be more variation but at ranges we would consider more commonly used, we feels this still holds true.

[@sfayers3361]

At Last! About time a reviewer engaged their brain regarding low MP/high MP sensor noise. Reviewing noise at 100% without taking into account the sensor MP has been a crazy trend for so long. Great video chaps, thank you.

[@TL-vt8uk]

With today's technology it is difficult to see an advantage to larger photo sites, but in the past, before technology evolved, it was easier to see an advantage. Good video, thx.

[@Maxime-ho9iv]

No, it has always been exactly the same. Even when the first a7s got released, what’s explained in the video was already true. People simply did not compare things equal. If you compare a 12MP file to a 36MP file, then the 12MP file will of course have less noise. But if you downscale the 36MP file to the smaller 12MP resolution, then the difference is non-existent, or very very small. It is true now and it was true back in the days. People completely misunderstood this topic and spread the wrong information over and over.

[@-OzSteve]

The per-pixel noise is higher for high res cameras because the pixel surface area and thus the light absorbed per pixel is less. However, the total sensor surface area is the same for both sensors, so both sensors will receive the same amount of light overall. As a result, if you were to downsample the hi res image to the same resolution as the low res image, you would see about the same level of noise. In theory, the high res image should still be a little noisier due to the space between pixels reducing the total surface area of the photosites more for the high res sensor than the low res, but in practise this doesn't make much difference with modern sensors.

[@gregoryfricker9971]

And you can always trade detail for better noise so that always helps

[@Quoutub]

No, light isn't lost, when the pixels are small, otherwise high resolution sensors would have pixels with a smaller fill factor and that's just speculation. I have never heard that small pixels have a noticeably smaller fill factor and somebody tested that many years ago. Maybe it was different 20 years ago. And by the way, this video is wrong due to read noise.

[@burritobrosvideos8060]

@@Quoutub the video isn't wrong, he literally just shows the results lol

[@Quoutub]

@@burritobrosvideos8060 Amazon's video is wrong indeed as it creates the myth that a lower resolution sensor never leads to less noisy results and that's wrong due to read noise.

[@burritobrosvideos8060]

@@Quoutub give it up man, literally no one believes you.

[@JamesonsTravels]

great video. that really made sense. thanks for demystifying a confusing topic.

[@Quoutub]

Amazon is wrong. The reason is read noise and it makes a huge difference. Take a photo of a distant object with a 500mm f/20 lens and with a 50mm f/2 lens in low light conditions. The colors of the object (for example a distant tree) will look much better with the f/2 lens. A 50mm f/2 lens leads to more light per time per mm² of a camera sensor , but it captures the same amount of light per time from an object as a 500mm f/20 lens because the amount of light from an object depends on focal length/f-number, that's why telescopes are so large (the Hubble Space Telescope has a 57600mm f/24 objective). But with the 500mm lens the tree consists of 100 times as many pixels. Each pixel is not able to measure the amount of light 100% correctly, that's called read noise. The more pixels, the higher the total read noise. That's why the tree will have worse colors with the 500mm f/20 lens even though the 500mm f/20 lens captures the same amount of light from the tree. That's also the reason why smartphone high resolution sensors use analog binning in order to read out multiple pixels together, otherwise the results would be much worse. It's also the reason why smartphone manufacturers often crop the image of the wide angle camera instead of using the tele camera in low-light conditions, even though the tele camera doesn't have a worse focal length/f-number ratio. It's also a reason why astrophotographers don't want too many pixels per galaxy as this leads to more read noise per object and this can be compensated only when the telescope is even larger or when the exposure time is even longer.

[@Seanonyoutube]

@@Quoutub so how do you explain the print experiment? All of this science seems to not make a difference when it comes to actually printing your photos…

[@dtibor5903]

@@Quoutub yeh, but still, high resolution sensors perform pretty well against the lower resolution ones, there's not a lot difference up to about 25600. At the extreme upper ISOs low res sensors clearly have an advantage.

[@trym2121]

@@Quoutub please make a video explaining this and the real world result.

[@scipioafr]

demystifying for noobs

[@MattisProbably]

It comes down to the actual quality of the sensor. My previous D7100 I used up to ISO 3.200. That was the limit I was comfortable with. My new Z50 does have a lower resolution but the images aren't cleaner because of that. I can relatively comfortably use it up to ISO 25.600 because the sensor and processor are newer and better. It's similar to the myth that with a larger sensor your images will be brighter compared to the same scene captured with a smaller sensor at the same settings because you capture more light. Well, yeah. You might capture more photons on the larger sensor in total but those are used for that part of the frame the smaller sensor doesn't cover. The amount of photons per area on the sensor are identical.

[@ktrev34]

Old video and old comments, but this is the truth that I've noticed, my cousin has the a7siii and I just got the a7iv and at 12800 ISO his image is cleaner because that's his camera's second base ISO (I think that's what it's called) where as my camera's second base ISO is 3200 so there for if his camera is in a situation where 12800 ISO is needed he will have the low light advantage, I could be wrong but that's what I've noticed and it makes sense to me.

[@BillFerris]

Well done, guys. The element of doing the comparison using images of the same size and viewing them from the same distance is too often overlooked when discussing this and other related topics. In still photography, it's the total light gathered - regardless of pixel size - and used to make a photo that determines shot noise. Fair real world comparisons such as yours confirm this.

[@zaneclone]

I've always found that all else being equal- when a hi-res image is reduced in size to match the lower res image, noise in the images are very similar. Also- the high-res sensor does allow more tweaking prior the re-size to often result in a superior image to the lower-res camera.

[@caleidoo]

I take photos in very difficult and challenging light conditions and I occasionally have to light up shadows and blacks. There will be banding, hot pixels and more noise in high-res sensors in those areos. And it won''t go away when downsizing, not completely. I think 24MP is still the sweet spot.

[@zaneclone]

@@caleidoo Personally I found older (lower res) cameras suffered badly with banding when pulling up shadows. Now- is that because the lower resolution cameras were also utilizing the older processors ? What could be done today were they to make a low res sensor with a modern processor ?

[@Mark_James_Hill]

@@caleidoo ​ @Simon Murray maybe I am ignorant but banding can come from flicker of the artificial lighting?

[@RicochetForce]

Thank you guys for helping kill this ridiculous myth. I'm glad you guys were honest and admitted making the mistake in the past.

[@sexysilversurfer]

Personally I think they were right, the Canon 5Dmkii had horrible banding and red splotches at high iso which noise reduction could not get rid off. The cleaner lower MP D700 would have still been more preferable. The sensor tech was very different between the cameras where as today they are becoming more similar plus the noise reduction software will work better with more megapixels.

[@TechnoBabble]

@@sexysilversurfer yes, different sensor technology, which is why they were incorrect. The better low light performance on the D700 was due to better sensor tech, not because of the lower resolution.

[@Pixelpeeps-69]

I’ve bought a lower megapixel camera due to this phenomenon ☹️ and now I’m selling it , thanks for highlighting it

[@Maxime-ho9iv]

We’ve been telling you for years.

[@Quoutub]

Amazon is wrong. Yes, the number of pixels would basically make no difference regarding noise, if there was no read noise. That's obvious because the remaining noise is nothing else than the varying number of photons that are emitted/reflected every millisecond from an object and enter the lens (for example the sun doesn't emit the same number of photons every millisecond). This variation is perceived as noise. Obviously, the number of pixels does not influence this process. But the reality is that every pixel doesn't measure the number of photons 100% correctly. There is an absolute deviation per pixel (read noise) (and the deviation doesn't necessarily depend on the pixel size). This means that the total deviation is larger, if the object consists of many pixels. That's the reason why a 200mm f/20 lens will deliver worse colors from an object than a 20mm f/2 lens in low-light conditions. Many people think that this is due to capturing less light, but that's wrong. A 200mm f/20 lens captures the same amount of light per time from a distant object as a 20mm f/2 lens (no matter which sensor size). The colors are worse because the object consists of (200/20)² = 100 times as many pixels, so the same amount of light is distributed over 100 times as many pixels. This leads to much more read noise (not more read noise per pixel, but the total deviation is larger). Therefore, Dpreview's video is quite misleading and rather belongs to the future, when there are cameras without read noise. Read noise is also the reason why many high resolution smartphone cameras have a low resolution mode where they can read out multiple pixels together, this is called analog binning and this makes a high resolution sensor behave more like a low resolution sensor, so the read noise (the total deviation) is lower. The low light image quality is much worse indeed, when high resolution smartphone camera sensors don't utilize this mode. Astrophotographers also use focal reducers (also called speed booster) (that's basically a wide angle converter) in order to reduce the number of pixels of an object. Then the object has less read noise, so a significantly smaller total exposure time is necessary in order to capture galaxies, etc in order to achieve the same signal to noise ratio of the object. By the way, even without read noise a high resolution sensor would look slightly noisier because one large pixel leads to a spatially averaged result, so it hides the noise that occurred at the center and at the edge of the pixel, so you would need to downsample (this averages the smaller pixels) and upsample the high resolution image in order to hide the noise as well.

[@westernexco1]

I knew it! In all my experience using all levels of pixel count, full frame, pro cameras... i never ever witnessed the so called" low light advantage" of lower pixel count! It never made any sense that less pixels (though larger) gathered more light, through the same aperture, same lens, to the same panel size.

[@GiesbertNijhuis]

Eh, bigger pixels are better pixels (up to a point). See every pixel as a bucket that collects photons. The smaller pixel gets less photons in = less information. Also will errors from the lens show more on a smaller pixel. Also-also between each pixel is a wall, and the more pixels the more of the image sensor is wall. The total amount of light on a same size sensor stays the same of course and I do totally agree with this video.

[@noahtr7244]

Makes sense Chris, some interesting points. Still wondering though, don’t lower resolution sensors handle high ISO better than high resolution sensors? For example ISO 10,000 on the Canon R6 appears better than on the R5?

[@Noojtxeeg]

ISO has no standard. A lot of the times it's just some random number w/ noise reduction.

[@youuuuuuuuuuutube]

@@Noojtxeeg Exactly. And the Canon R5/R6 apply noise reduction even at lower ISOs, baked in the raw files ...

[@CrazyWeeMonkey]

@@youuuuuuuuuuutube I think denoising raw files could be an extremely useful feature if it was (optional) and temporal. The DJI Air 2S apparently does this with its raw files, and looks pretty clean at higher isos because of it. Maybe a mode that automatically takes 2-3 burst photos in camera and processes them into a clean RAW (DNG maybe) could do this. My pixel 4xl does this with its raw files and as a result they're very clean looking for having such a small sensor meaning I don't need to apply as much noise reduction in post.

[@DustinBKerensky97]

No, it's about the same. www.dpreview.com/reviews/image-comparison/fullscreen?attr18=daylight&attr13_0=canon_eosr5&attr13_1=canon_eosr6&attr13_2=canon_eosr5&attr13_3=canon_eosr6&attr15_0=raw&attr15_1=raw&attr15_2=raw&attr15_3=raw&attr16_0=100&attr16_1=100&attr16_2=102400&attr16_3=102400&normalization=compare&widget=1&x=0.13768067226890757&y=1.001341440481414

[@TangerineTux]

@@youuuuuuuuuuutube You say “even at lower ISOs” but in fact it’s *only* at lower ISOs.

[@mika2666]

The gap has also been filled due to BSI sensors losing less light to the edges of the pixel, thus penalizing adding extra pixels less with FSI sensors.

[@planetfun85]

R6 and R5 dont have BSI sensors.

[@youuuuuuuuuuutube]

@@planetfun85 The R6 and R5 were not compared in the video, the A7siii and A7R4 were, and they both have BSI sensors. BSI helps a lot when the MP goes high.

[@mizinoinovermyhead.7523]

The issue is you aren't actually comparing low light images. Go try the same with astro shots and see if you still think the same thing.

[@djole02]

ISO 6400 is low light.

[@mizinoinovermyhead.7523]

@@djole02 lol no it’s not. That’s not even half the native iso of most modern sensors. You’ll hit 6400 doing wildlife with an f4 lens quite a lot. Low light is twice that iso normally.

[@djole02]

@@mizinoinovermyhead.7523 ISO 6400 is low light and it proves the point of the video. Higher ISOs would still prove the point and you can verify yourself on dpreview. There will be some variations at the extreme low end of the cameras but you would never use that for serious work anyway.

[@mizinoinovermyhead.7523]

@@djole02 Lol astro starts at 6400 iso. People use it for professional work all the time.

[@coreylowey]

Funny, I took 200 lights @ 2.5 secs, 800 iso, f2.8 on a crappy FZ150 and managed okay shots. Any more and the images are far too noisy so why would I increase that? Typically, there's a combination of elements that determines low light iso performance that seems completely unrelated to megapixels. I'll admit I can't point out exactly why other than say an old Nikon D800 can still whoop the majority of modern cameras asses, bar the new top dogs, Sony A7, etc. Probably due to a combination of elements such as sensor size, pixel area, mp, processing and dynamic range. But that's just my take and I'm clearly not an engineer yet

[@JavisandLea]

Man weaving these informative/myth busting type videos into you guys’ already amazing library of comparisons and reviews seriously puts this channel above the rest!!!!

[@jpdj2715]

Chris self-incriminates about having been wrong or naive in the past. The problem today is that the comparison still is naive. Printed at 550 ppi (0.3 M square dpi) for the hi-res camera and 250 ppi (0.06 M square dpi) for the low one, the Canon printer still operates its nozzles at 1,200 DPI (1.44 M square dpi). Per square inch, the software layers in the computer (driver) and printer ("print engine") have to "imagine" a lot. We just don't know what happens there and we all are "naive" in that sense. Between the 60MP and 12MP camera, the 60MP will not have a fuzzy filter (*) and the 12 MP will have one. Both the raw files are raw-processed and this is where noise is created, or not created because it was done smarter. The fuzzy filter in the 12 MP camera reduces contour sharpness, but makes raw processing easier. The question for the 60MP raw shots is if a different raw processing algorithm was applied that acknowledges the absence of the fuzzy filter. We don't know and all are "naive" in that sense. A dead giveaway is DxO PureRAW that pre-processes our raw files so they come better out of Lightroom's (= Camera Raw's) raw processing. We probably subscribe to anachronistic sub-optimal software - all being "naive". As the 60MP camera does not have the fuzzy filter, it may gain up to one EV dynamic range and make its lenses suffer less from the vignetting all the filter tunnels cause (because there also is the Bayer filter tunnel over each photosite). We don't exactly know and all are "naive". And when DxO Mark (separate entity from DxO software) reports that they test "sensors" we are all fooled. They measure a raw file (produced by a camera, not a "sensor"), they state, "before demosaicking" and this implies "after deBayerization". As "deBayerization" is where the (RGB) pixels are faked from the raw monochrome photosite data and this is where we lose 5 or more of our bit-depth, the whole "sensor testing" process depends on the algorithms (*****) used in there. What is called "noise" is 99.9% "deBayerization" and neither sensor nor camera. And we are all "naive" here. "Color noise" in Lightroom, is a deBayerization artifact. "Luminance noise" in Lightroom, is a deBayerization artifact. When these artifacts become a pattern, we call it Moiré and some less naive understand where that comes from. When we do not recognize patterns, we call it "noise" and blame the camera. What The Fotography (WTF)?!! "We" need to discuss the Bayer paradigm (or variants like Fujifilm's T thing) more and how it impacts everything we "see" from our photos, and the algorithms that our apps apply. My thoughts go out to the Foveon history of a sensor that tried to mimic color film. It would have been so nice to have a camera creating a 16 bits per channel TIFF file that does not need the wild-assed color guessing that is raw processing. Imagine a JPEG as a simple data compression from that. And "we" need to discuss time parallax that started with focal plane shutters (that started as curtain or slit shutters in front of lenses), continued through television development and still bites us in the photographic [donkey] today. If "we" parrot marketers and naive influencers, we slow down innovation. Of electronics components developments and of improved raw processing. (*) It's a low-pass or anti-aliasing filter. The idea to make the wild-assed guessing of curves and details easier by having a bit of fuzziness (dispersion) in the hardware (in the digital camera a "glass" filter) was already applied by developers of scanning tunneling electron microscopes in the 1970s. In digital cameras a complexity is added over the STEM image processing in that we have colors that need to be guessed. Note, if the dispersion filer causes "noise" then it's not the sensor doing that. If the raw processing algorithm does not remove that "noise" then it is a naive product of lazy programming. (**) The sensor does not have "pixels" but rather analog and colorblind photosites. Taking a photo (or a frame in a movie) means the photosites are scanned - metered - and have their analog EV converted into a digital number of the bit-depth you set on your camera. As each photosite only "sees" a single color band of the visible spectrum because of a color filter over it, that, say, 14 bits "word" is monochrome (mono-one, chrome-color) and either red, green or blue. These 14 bits are generally surrendered into the raw file in a 16 bits package - we can imagine that the two added bits are either void or carry the color code of the filter over the photosite. Raw processing must convert these monochrome "words" into RGB pixel words. As we have 14 bits R words, 14 bits G words and 14 bits B words, each needs to be converted into RGB. As DxO Mark report that the best cameras have 27 bits color space, this means 9+9+9 bits and the invention (faking) of the two missing colors (18 out of 27) have cost us to loose 5 of the 14 that we started with. We cannot see that "color space" separate from dynamic range. Where color space and bit depth relate to gradation or nuance resolution, the question in digital photography is what is better: 14 bits depth into 16 EV dynamic range or 12 bits into 10 EV dynamic range. As we end up with a fraction of that after raw processing, we are taken farther down by a display that may be just able to do 6 bits gradation per color channel, or photo printing paper that maxes out at 5 EV dynamic range. And yet, taking a raw processed image into Photoshop for processing at 32 bits gives a clearly visible improvement - the print driver and printer engine operate in that same space. (***) When we display an image at larger than 100% and we can argue that printing at 1200 DPI does precisely that, then pixels (on a display) or dots (in a printer) must be invented. And this is where we are fooled by simple algorithms or smart AI. When I scaled a set of 45MP shots up so as to match my printer's 1,440/2,880 DPI resolution, LrC showed hikers in the far distance as woolly fluffy colored balls. Gigapixel AI revealed their arms and legs. We, "naive" people, are fooled all the time. And that includes fooling by influencers and sensor bakers that call photosites "pixels" which they by definition are not. And it includes the "16 bits" [male bovine's excrement] of sensor bakers that bundle an Analog-to-Digital circuit with the sensor and claim the sensor is 16 bits - look for the tiny print that says that 14 out of the 16 are photographically meaningful: present gradation resolution. When the camera generates JPEG (or MPEG) then it does the raw processing. 27 bits RGB compared to the 24 bits in JPEG is not a big deal. The reason we shoot raw is because we can influence how the raw processing is done, not because 27 bits are shockingly better than 24, which they are not. (****) Humans see detail resolution - sharpness - along linear magnification lines, not simply related to area numbers. MP are area numbers. If we want to perceive 2x more resolution, we need 2 times more linear detail. Starting with a 6,000 x 4,000 = 24MP sensor, 2x linear means 12,000 x 8,000 = 96MP are required for that. Because 2 times linear applies to both the x and y number and 2x2=4. This is to say that the 60MP camera in theory only has a bit more than 2 times the detail resolution of the 12MP one (2 times linear applied to the 12MP gives 48MP). (*****) Stating "before demosaicking" implies that demosaicking is done after deBayerization as a way to address the artifacts caused by deBayerization. In terms of software engineering, this means an architecture where first faults are created that next are removed. This may have been the only feasible way with computational powers of 40 years ago but under Moore's Law we gained a doubling of transistors in ICs (like CPU) per area that gave more compute power. If "we" got only 1% of these Moore's Law cycles working for raw processing, then instead of 106.5 million times more power in 40 years, we got 1 million times more computing power. In our cameras, our notebooks, smartphones and desktop workstations. Isn't it time to use that properly? Between a million and 100 million times more compute power.

[@unfrostedpoptart]

Tl;Dr version (but great thanks to your post!!!) - there are so many other places in the chain from the photosensor (and its color layout) to RAW to the printer to paper that are up and down-sampling pixel depth and resolution - some of which is based on old cameras that had far less computing power - that it isn't as simple to isolate the issue as the video does. I hope I got that right.

[@doghouseriley4732]

That's actually a realy good read. Not sure how many times people can make a fool of themselves before viewers wake -up. I get bored with the channels making videos for the sake of making videos

[@jpdj2715]

@@doghouseriley4732 - Thank you. Another thing along these same lines to be aware of: your raw file only contains "data" about an EV, a monochrome color and a coordinate. That data thing with a photosite in the sensor has no physical dimensions and is in a sense a mathematical abstract thing like a by axiom dimensionless "point". What happens with these "points" is a software story. That's a bit nerdy or academic, maybe, but in film we had physical grain and pigment and in the raw file, only data.

[@joefaracevideos]

Thanks for showing us, once again, why you two are the best photo channel on KZbin.

[@JamesBoyer-plus]

Thanks for a wonderfully provocative (especially for Canadians) experiment! The root of this myth seems to be the perception that larger pixels/photosites outperform smaller ones. That's also a part of the thinking when people argue the advantages of full-frame vs. crop sensors. Please consider a similar experiment judging the actual printed output from the most recent generations of FF, APS-C, and MFT sensor cameras with similar resolutions (and thus differing pixel sizes). 🙂

[@Razor2048]

One thing that is often ignored, is the flexibility of the files. If you supersample a higher res and higher noise image down to a lower res, with no other adjustments, both will look similar, but if you push the files with some curve adjustments as well as shadow recovery, and the sensor with lower SNR per pixel will fall apart. For example on the high res cam, at around 12800, shadow recovery looks far worse. Ultimately it depends on your work flow, if you are doing many adjustments, then you will get more artifacts that noise reduction can't remove.

[@AspectedVisages]

I was surprised this was not covered in the video. Chris / DP Review, could you do another video comparing pushing shadows and clamping highlights on a lower res sensor vs hi res sensor?

[@PatrickAdairDesigns]

Super interesting video, and all of the points you made were valid and interesting!I’d like to know, is it always a fairly equal outcome? I think there would probably be some scenarios where you could see more pros/cons? Perhaps cranking iso a lot higher so the differences start to become more apparent and actually relevant? This video almost makes me think there’s absolutely no benefit to less megapixels, but there’s gotta be a scenario where the larger pixels make an appreciable difference??

[@LimitedWard]

I mean there are still plenty of other reasons why a lower megapixel sensor makes sense. Apart from cost savings, you're also getting smaller image files, which means you can use up less storage, and they are easier and faster to process.

[@__dm__]

He explains why there are benefits at the very end…

[@zaneclone]

I guess the argument was that more pixels require more processing as there's simply more data- which in theory equates to more "noise" during the processing. Also one might argue why a D700 has better noise performance than a D300...

[@Gjarllarhorn1]

Would love to see more printing related videos!

[@JamesClark1991]

+1000

[@DanielBristotdeOliveira]

your guys are the best on reviews, no doubt about it.

[@KikujiroChan]

2:24 The guy who bought a Nikon D700 on the recommendation of chris and jordan 13 years ago: "I want my money back!"

[@CrazyWeeMonkey]

D700 was and still is an excellent camera at least

[@topg2820]

D700 is a legendary camera, photography is not just pixel peeping

[@zenden6564]

And an excellent personal defence tool as well :)

[@JoseTheRover]

I’m thinking Sony kept the a7siii at 12mp is to keep the heat low, reduce rolling shutter and be able to do 4k120

[@TechnoBabble]

That is exactly why it's still 12mp. The lower resolution sensor means no extra math for 4k (even though it does actually scale down a tiny bit from a slightly higher resolution) and faster readout.

[@CrazyWeeMonkey]

@Tom Hardy Yep. This video likely would've had different results if they compared higher iso like 12,800iso+. Is that usable? Not really, but the noise *will* be less apparent on the camera with bigger pixels (a7Siii), meaning you have a better chance of high iso being usable for photos/videos. Also, the file size difference is huge with higher resolution sensors (especially Sony ones); I have an a7iii and the 45-50MB RAW files are already large enough to make me think about their size every time I shoot.

[@TechnoBabble]

@Tom Hardy What does file size have to do with the subject of this video? Wow, what a shocker, having more detail means larger files, who would have thought?

[@CrazyWeeMonkey]

@@TechnoBabble Because most of the time the quality difference is effectively indistinguishable, but the workflow working with large files can be a hassle. Managing storage for 50MB/photo is already a pain, 150MB/photo is even worse (especially if you do HDR bracketing like I do, meaning every photo is basically 150MB+ already from 3 exposures from an A7iii, or close to 0.5GB with something like the A7Riv)

[@CrazyWeeMonkey]

@@TechnoBabble You usually won't get enough use out of higher resolution sensors to make up for the processing+storage hassle, unless you are a photographer in one of the very few ways that will benefit from a higher resolution (astrophotography, product photography, landscape photography, digitizing, and/or very high end portrait photography).

[@JasonTheZoologist]

Great video, but one correction from the wildlife photographer community. When shooting animals like wild birds, it is common to be under the canopy where there's low light and not being able to fill the frame. Hence, for us, 100% crop is still very commonly applied. That is why our community still sometimes prefer a slightly lower resolution camera. In short, we want it to be balanced and also why we don't always go for the highest resolution cameras. With current technology 20-30mp is my sweet spot. 60 maybe a little too much, but this might change with new technology.

[@TechnoBabble]

That doesn't make any sense, cropping into 100% changes the amount being cropped based on the resolution of the camera. So if you're often cropping to 100% on a 24mp sensor you would probably greatly benefit from a higher resolution camera as it would give you more detail.

[@JasonTheZoologist]

@@TechnoBabble You misunderstood my message. I am not talking about the "we want lower resolution", I am talking about " we want the low light superiority of lower resolution". Watch the video again, he mentioned that there is low light superiority in lower resolution (which is commonly understood), but argued there is no need for it since such low light conditions are rare. I am countering by saying this is absolutely not rare for wildlife photographers, especially when you are in the tropics where only a small portion of light passes the tree canopy. In short, try photographing with two cameras with same sensor size and same technology. One with ultra high resolution (64MP), one with lower resolution (24MP). You will notice the one with 64MP is so noisy in low light, it does not matter even if you can crop in. The noise totally ruins the photo.

[@80-80.]

From what I’ve seen, you need to push the iso to extreme levels to see the very small benefit of lower resolution sensors nowadays. Pics from the high-res camera will probably fall apart 1 or 1/2 a stop sooner. For almost all normal photography, high-res sensors is the way to go. (Not talking about video).

[@JCANPetros]

Are you doing that or need to for your regular shooting?

[@MindMeddler]

For shooting in moonlight without a flash, yes you need an a7S3

[@neilholmes152]

Yes. Shooting light shows (that are very popular these days) with projections on buildings and the like, you often have to keep the shutter speed up and if using a slow tilt shift lens for instance, you can very quickly get into ISOs over 25600. Not really going to print that at large size but for things like newspapers and websites, an A7s is still a very nice camera (even the first version).

[@ElGrecoDaGeek]

Extreme is relative to do yo want to capture the image/moment/memory. On my 24MP z6 I'm willing to go to max if the situation requires it. This distaste of noise over capturing the moment needs to stop for at least event photography. Of course we want clean images, but more importantly, why do we take photographs? To capture the moment... That obviously doesn't apply to studio or landscape the way it does say at a wedding or nighttime outing. While you want clean photos you also want mood lighting and to avoid blinding everyone with flash. z6 does that relatively cleanly when viewed at full screen up to 50k ISO. My uncle's D850 to 12800-25k. 1 to 2 stops difference. Probably 1 like you said. That said for low light event no flash 51k is very usable and that extra stop can make a big difference. The D850 can get a colorcast at 51k...

[@burritobrosvideos8060]

Um, nope, you are wrong haha. How many times do people have to prove that is a myth. You just watched a video on it.

[@Clint_the_Audio-Photo_Guy]

I got the new Fuji X-H2 and immediately thought it wasn't great in low light. I have to keep reminding myself to not pixel peep at 100%, because it's going to look worse but that doesn't equate to real-world differences in print. 40 mp is magnified a lot more than 26mp when at 100%. Then why are medium format sensors so good, when the pixels are bigger? I know you've shown the "medium format look" is just a combination of equivalent focal length, aperture, etc but if you view the files at the same size on the screen the medium format files are amazingly detailed and color rich. I've seen side by side comparisons between similar res Nikon full frame (Like a D850 or Z7) and just the Fuji 50mp GFX camera even, and the GFX is amazingly better at detail and color. I'd love to see a video dial in those differences more and include an X-H2 (or X-T5) for an APSC comparison too. Weather there are differences or not. Your video of the GFX 50S II made it sound like who in their right mind would buy one, but the images it produces that I've seen, look like they're on another level...or do they? If they didn't, we'd all be fine with using the highest res, smallest sensors we can find. Shoot them all with a 50mm equivalent at a F/8, so the differences are in the format. Just an idea.

[@0ecka]

Well done, guys! A picture says more than a thousand forums of endless debates :)

[@bobsykes]

This is a good one. Making real prints like that at a quality "lab" pretty much takes the confusion out of this. I didn't know about that comparison tool on your associated website. Glad I stayed until the end.

[@vikmanphotography7984]

I'd be really curious to see how those two cameras stack up at a higher ISO like 25600. Regardless, I'm usually satisfied with the detail from my 24mp sensor with clients basically never needing higher (and then I rent cameras)

[@BrentODell]

Using the tool they mention from the site, and even downloading raw files from there, I've found that higher mp almost always does as well or better, given the same generation of tech.

[@stjepanjina]

Who cares, nobody uses those numbers

[@toddysurcharge771]

@@stjepanjina no professionals anyway…video could be a diff story if you’re shooting a dark ride at a theme park

[@GiesbertNijhuis]

Wow; I totally agree, while up to this video I was in the old camp of "bigger pixels = better pixels = better picture quality", but the last part is not true. Bigger pixels do make better pixels, but when printing on the same size, a size big enough to show the difference in sharpness, the higher resolution wins the picture quality! Also: the higher resolution picture its noise is practically reduced because on a same size print the size of the noise is reduced. Thanks for this excellent test!

[@jamespeirce2582]

Nice to see someone publishing correct information about this. High resolution has another advantage in post processing. Because the noise is smaller relative to important resolved detail, it is easier to clean up noise on the higher resolution sensor without destroying important detail.

[@GalaxyArtMedia]

Hello. Higher megapixels camera win in image resolution but they lose in light gathering, always is a compromise, regarding the test you need more low light then that, push iso higher and have visible noise in the images to be able to benefit from the light gathering capacity of larger pixels, a video test is a must since on photography we can use longer exposures to gather the light we need.

[@ma76bball68]

“It’s probably in our anthem somewhere “🤣🤣🤣I’m a proud Canadian and that made me lol!

[@artemholstov9207]

You guys are probably the most unbiased reviewers on youtube and, usually, your videos are amazing. But here, you disregarded two very important considerations - storage requirement and PC processing/editing requirement that depend hugely on resolution. Surely as the technologies progress, the megapixel count can keep increasing, but as the resolution grows, the practical difference it makes in terms of IQ is diminishing and at some point (I think very soon) will be completely insignifficant/imperceivable for non-specialist applications. Yet, the effect the increased resolution has on the storage and processing requirements is not diminishing. For this reason, imho, further progress in this direction will quite soon be pointless and this megapixel race will just unnecessarily be contributing to our problems with waste etc.. In my opinion, very soon, the focus in this sphere should fully shift towards the development of sensor technologies with improved sensitivity, readout speed etc., not higher resolution.

[@RicochetForce]

It's not really much of a problem. We have very cheap, high capacity storage devices (that are getting cheaper by the year), our CPUs and GPUs can easily handle 100MP files. I do agree that companies will switch their R&D to other things that they can sell users on once megapixel advantages become like splitting hairs. A big one is going to be things like internal downsampling of phone-style HDR multiexposures leading to easy absurd IQ boosts to sell people on.

[@artemholstov9207]

@@RicochetForce yes, the computational thing where several exposures are used to create cleaner images and increase detail will be a welcome feature!

[@andreasseibt8238]

I appreciate your work. However, two things to consider: 1. You only tested for noise and did not take dynamic range into account. At iso 6400 the a7siii should have about a stop more dynamic range than the a7riv. I would argue thats important when talking about low light capabilities. Obviously, you did not reach that limit in a studio situation. 2. If you want to take a fair comparison why not choosing a higher iso value? You chose a scenario where the a7s had no chance to shine, since the small prints where not pushed far enough. Maybe the results would have been different at a higher iso and at the same ppi? Maybe not, but we don't know.

[@burritobrosvideos8060]

This myth has been busted, give it up

[@BensHacks]

But ... what about dynamic range? The higher pixel count might catch more variations, but averaging will put on an statistical pressure toward the middle. On the other hand this might be easy to compensate since the standard deviation is more predictable with more data ... ?

[@fredericbeudot822]

What is that weird thing you are doing with your photos. They seem to come out of a strange contraption on some sort of paper... I thought pictures could not be looked at if the loupe doesn't show at least 400% - I feel you guys are on some kind of metaphorical limb here, maybe that's why Chris is sitting on a limb over the water at the end. All joking aside, glad to see you guys be brave enough to reset the narrative; what you are saying is what anybody who prints pictures larger than 8x10 knows but for some reason is ignored by the vast majority.

[@michaelangelosiracusa]

I know this is maybe being pedantic, but (at least for me anyway) when you think of low light performance, you think about noise levels at higher ISOs - something that you often need to move into in low light conditons - and how this is shown in the image. There was no mention of settings used for this image, so if it was at ISO100 it would be relatively clean for both camera models, even with a huge differece in resolution.

[@niccollsvideo]

We shot those examples at 6400 ISO, which we considered a value commonly used in low light. Shutter speed was 160th at F4.

[@ChrisAlford]

Love the video. Great job comparing low light in high vs low megapixel sensors. Using prints to show the differences in sensors was really effective.

[@Bayonet1809]

Thank goodness you are publicising this fact. So many people, even camera reviewers for popular outlets, get this wrong.

[@oscarshen6855]

What they trying to say is high reselution sensors has more but smaller noise, when you are viewing at the same level of magnification, the noise is less distracting.

[@patrickchase5614]

Or alternatively you can downscale the high-res image to the same resolution as the lower-res one, and the per-pixel noise will look more or less the same.

[@jamespeirce2582]

That’s not quite right. Roughly the same level of signal degradation is present in *both* cases, but because the noise profile is more finely rendered on the larger resolution sensor, it will also appear less noisy if a difference is noticed (in addition to resolving finer detail). As a bonus, the finer noise is also easier to clean up relative to detail.

[@patrickchase5614]

@@ElementaryWatson-123 Yes, that's absolutely right. It's only mathematically provable if the sensor has random noise with a gaussian distribution, and if the fill-factor and photonic efficiency are constant between the two sensors, such that they both capture the same number of electrons per unit area. Modern full-frame camera sensors are close enough in both respects that what you say holds true as a practical matter (fill factor is very close to 1.0 after microlensing is taken into account, and gaussian shot noise dominates).

[@patrickchase5614]

​@@jamespeirce2582 You seem to be assuming constant read noise, but in reality the dominant source in modern cameras is shot noise, which scales proportional to the square root of the expected number of electrons. The fixed read noise of modern sensors (the kind that's the same for each pixel in both cases) is on the order of 1.2 e-, and is completely swamped by shot noise. Once you take that into account and work through the math, you'll see that the only thing that matters is the number of electrons per unit area. The high-res sensor has higher standard deviation on a per-pixel basis (which is why it looks worse when you peep), but downscaling to lower resolution effectively averages multiple input pixels to each output pixel. That in turn brings the noise to the same level as just shooting at the lower resolution to begin with (with the "don't use a stupid scaler" caveat that US Grant brought up). I say this as somebody who shoots both the A7S3 and the A1. I don't really have a horse in this race.

[@jamespeirce2582]

@@patrickchase5614 I think you’re taking a simple reply and trying to read too much into it. I’m agreeing with the assessment as shown in the video and adding that noise reflected on the higher resolution sensors is easier to separate in post production from detail.

[@bobharris3002]

I expect someone has already pointed this out, but the difference with noise is noticeable at higher ISO values, not at the camera's base ISO. I have a 24 megapixel and a 45 megapixel camera, like both, but the 24 is noticeably better as the ISO values raise. Flash photography is not really an appropriate test. Loved the images though.

[@idolog]

There was a lot of hype on Sony A7S3s low light, in my tests I have found out that Panasonic S5 beats it and even Canon R6 was very close.

[@clifftotten7609]

The S1/S1H and S5 use the same Sony IMX410 variant sensor. Yeah,.....below 12,800 ISO?...yeah that 24mp IMX410 is WAY better than the A7S-III. But,....when you hit the S-III's "second stage" preamp? Well?....yeah,...that is when the table gets flipped over. The S-III takes off like a rocket and passes the IMX410. (Vlog and Slog-3 dual gain preamp stages)

[@aarontalib6932]

My only issue with this video is that in my experience with camera retailing. Low light performance all most never comes up with it comes to taking stills. Its all most always a video related benefit.

[@LEXPIX]

Well explained. More is better, since you can always downsize a higher MP image to hide even more noise. I agree when it comes to video is where the faster readout of less MP is advantageous.

[@RicochetForce]

Exactly. Folks like The Hybrid Shooter noted this about photos from the A7R3/4, downsampling to 24MP leads to major noise compression and a cleaner image than a native 24MP file.

[@SCrowley91]

This same issue is present with most dynamic range measurements. Measuring per pixel only makes sense if you're cropping 1:1. This is why ASC members say the Alexa LF is better in the shadows/low light; it has more pixels to average out. Sadly, CineD's tests don't offer any normalized option for dynamic range (scaling everything to 1080p or such). They did, briefly, and some cameras gained half a stop or more in dynamic range because of it. I really wish they'd start showing that again. It can really be misleading, especially with 8k+ cameras.

[@davidspearman3939]

Why did you test a comfortable 6400 and not 64000? I think DPreview missed the true strength of the low megapixel camera. It does much better in low light with higher ISO values. Not simple controlled environments with low ISO like you showed her. Somebody please correct me, sincerely!

[@TechnoBabble]

The A7S III and A7R IV are about equal until you get above 51200. So if for whatever reason you are constantly shooting at 102400 and up, go ahead and get the lower res sensor.

[@bozmundarts2614]

@@TechnoBabble or an a7iii, maybe an a7riii, or a7c why not, i think we can agree the 12mp of the a7siii does its best job for video... specially with that high dual native iso... otherwise i think 24/42 are still plenty, and yet for some people that arent hard working professionals is a bit hard handling those file sizes of 45/50/60 reason why 24mp is such a sweet spot for everyone and anything...

[@Quoutub]

You are right, though note that the misleading thing is also that Amazon's Dpreview compares different cameras, which might use a different technology and have a different amount of read noise per pixel. Take a photo of a distant object with a 500mm f/20 lens and with a 50mm f/2 lens (same exposure time) in low-light conditions and the distant object will have better colors with the 50mm f/2 lens, even though the 50mm f/2 lens captures the same amount of light from the object. The reason is that the distant object consists of 100 times as many pixels with the 500mm lens. The same amount of light was distributed over 100 times as many pixels. This leads to much more read noise per object.

[@firstnamelastname8197]

My personal experience with a7s series lower-res sensor 135 cameras showed me that I could use really fast shutter doing street photography in almost dark places at night… and that’s it for my personal taste. To cracking up higher-res camera’s most potential, it’s better to use higher quality glasses. Personally I’m satisfied with around 24 Megs on 135 or smaller sensor cameras, but it doesn’t stop me from appreciating my iPhone’s and a7s’s 12MP. My take for all those is that you have to sort out your mind choosing the right gear for your shot to get the most out, appreciate what you have while pushing your photography further.

[@dominic-ryan]

Great to see this topic touched on, also well done on pointing out use case scenarios. Others have pointed out that ISO 6400 probably isn't that high for a modern full frame camera. What happens when you push the ISO above 25600 which does happen for low light video where there is far less latitude for adjusting shutter speed to gather more light?

[@andersistbesser]

Why would anyone ever use 25600? Its unrealistic

[@dominic-ryan]

@@andersistbesser Niche requirement, but a lot of people with interest in the subject matter of this video. Anyone shooting in dark environments without controlled lighting, e.g landscapes under moonlight, caves, forrest floors with heavy/dens overhead vegetation, etc. The HGC of the a7s III doesn't kick in until 12800 for slog 2/3, so shooting video at 1 stop over this isn't uncommon for very low light scenarios.

[@TechnoBabble]

@@dominic-ryan Shooting video is completely unrelated to this test, S-Log2/3 uses a different ISO range as it's meant to be exposed differently from a "standard" profile. When just shooting stills the second native ISO on the A7S III is 1600, meaning it was already using it's high gain circuit in this test.

[@pieterdhaeze]

Great video. Only one remark. Did you look at the prints from the right viewing distance, say 2x the longest side? On a table with your nose at the print you still are pixelpeeping ;-) Differences in noise and detail will then be even less apparent than from close-up. What do you think?

[@markus8282]

No one keeps looking at fine art prints from that far away after getting interested into it. Same is valid for paintings. Just visit an art galery and check for yourself if you don't believe me. It's only true for advertisements, billboards etc.

[@pieterdhaeze]

@@markus8282 True, but sometimes it is not possible to get close. But you are right, if you can, you do. 😀

[@Maxime-ho9iv]

The print is only an example. The argument presented here holds true whatever the viewing distance is and even on screen. The argument is not that prints somehow mask the difference. It is a downscaling topic, the print is just a way to downscale the images to the same resolution (i.e. the same physical size).

[@DustinBKerensky97]

Keyboard Photog: "This camera is no good because of it's low DXO rating, low light performance, and how grainy it is when you pixel peep. You're not a professional photographer if you use this." Photogs: "What photographer do you admire the most?" Keyboard Photog: "Henri Cartier-Bresson. His grainy, blurry, picture of 'Man Jumping the Puddle' that has no detail in the shadows is a masterpiece!"

[@nateo200]

You guys have both integrity and balls bigger than most! Nice short and sweet video, I have seen longer videos and articles that aren't as fun.

[@Soundwave857]

what about dynamic range and colors at different and extreme iso? 6400 is pretty easy for modern cameras ngl. Im also interested in 61mp fullframe vs 24mp apsc vs 24mp fullframe.

[@darkvadorus31]

yep, even with an aps-c or m4/3 you can easily use 6400 with no big issue

[@mikkelstormhansen9733]

@@darkvadorus31 Depends on the camera, i wouldnt use anything above iso 3200 on my 77d with the 24mp sensof

[@trym2121]

61MP full frame is pretty much the same as 26MP apsc so there is no need for comparison. Just crop it and you get 26MP apsc. Compare directly to 24MP for less work to be done

[@Soundwave857]

@@trym2121 that's why I want it compared

[@burritobrosvideos8060]

There will be no difference. This myth is busted, time to move on.

[@exploresouthwest]

Now do a video on "35mm Equivalent" specs when using a FF lens on a crop sensor. Specifically how much bright the image is at "equivalent" F-stops.

[@TangerineTux]

That’s why there are equivalent ISO settings too.

[@be_bigger_today]

Chris - your daughter is adorable - and a good sport being your subject every time you test a camera or lens. Hope you frame those prints and hang them in your home! 👍

[@Durio_zibethinus]

It's Jordan's daughter, but yeah she is adorable 😄

[@thierryhoornaert9950]

Thank you for this eye-opening post. Is it also true for not-backlit cameras like the Canon EOS 90D vs comparable APS-C?

[@tushardani]

Watch for Jared Pollen's reaction to this one 😎

[@joezuu]

Is he the one who wrote the book, "The Frotographer's dilemma"?

[@Tzadeck]

Fro Don't Know Photos

[@DustinBKerensky97]

The cameras used in this don't focus fast enough, the FPM of the bodies is too low, and the f/stop of the lens is too small to even be considered a valid test for real professionals like him.-- Now let me pause for a second to tell you about his latest $100 image filters you can buy! --Maybe some photographers just getting into photography to look cool with a camera around their necks might find it interesting, and Sony and Nikon pander to lowest common denominator consumer. But as a professional he just doesn't get why people would stoop to using equipment of this low of caliber.

[@kashima3626]

Actually he is a bad photographer

[@Maxime-ho9iv]

There is no « reaction » to have. This is very basic stuff.

[@angliski911]

Chris, this is not a fair comparison. You should have compared RAW images as out of camera jpegs could be compromised with the in-camera converting software...

[@Leptospirosi]

8:25 you clearly can see the problem there: if you set the picture to 1:1 in term of pixel density you see the low res camera has WAY less noise then the high MP one: this is phisics and every single photodiode on the camera collects more light at a given size from the lens. But noise is a component of "grain" and the more density you have, the less noise becomes apparent showing the same scaled detail. It is the reason because looking at a screen monitor with a magnifying glass shows you the elementary component of light breaking down the image to pixels. In modern cameras with such a high reserve of MP embedded, noise reduction and light sensitivity, pixel density is not a thing when you print, turning a digital picture into an analog one.

[@dclark92064]

WRONG!!!

[@burritobrosvideos8060]

Well, youre wrong as hell haha. If you down sample the larger file to the size of the smaller one you will have less noise and more detail. Still, there is no reason to do that.

[@peoplez129]

There's 2 major benefits of lower resolution sensors: 1. It's not about noise, it's about color fidelity. More samples of light = more accurate color (and that especially matters under different white balance settings). White balance only works soo much, because if a spectrum of light isn't in the scene to be measured by the camera, you simply don't have the readings to create a proper white balanced image or enough data to push it very far. This is especially true under mixed lighting or modern LED bulbs which can have very wonky spectrum imbalances coming out of them. Yes, colors can largely be subjective or close enough, and can be manipulated, but a good starting point isn't a bad thing. Camera's also perform differently under different white balances. Let's just put it this way, the original 5D has a better color rendering index than an A7III, and especially so under wonky white balances. All camera sensors perform their best under daylight. But when the lighting isn't daylight, things can drop off dramatically when it comes to color readings from the sensor. It's easier to sift out color you don't want in a bad white balanced scene, if you at least have a decent enough reading of the spectrums you do want to express. 2. Diffractions hard limits are mathematically dependent upon pixel size. APSC's are diffraction limited at f/6-8. A 20MP FF would be hard limited at f/12. Only a 12MP FF sensor can actually go to f/16 before diffraction sets in. Of course, extra resolution may largely negate this advantage. But in certain situations, something off in the distance could be sharp on a 12MP FF sensor at f/16, compared to the same thing being soft on a 50MP FF sensor at f/12. One thing that can be noticed between the 61MP and 12MP, is even though the 61MP is sharper, it's actually doing poorer with micro color contrasts. If you notice at 6:15, her freckles are more pronounced on the 12MP. Some of them are even gone almost entirely in the 61MP image. This again goes back to more samples of light = more accurate color. With higher resolution you can get more sharpness in details, but with bigger pixels you get better small color detail. 12MP also is a lot cleaner to remove color noise from, because those shifts are happening over larger areas. Part of it has to do with processing. Color noise removal effectively averages pixels. The more accurate reading you have to begin with, the better the average. You can even see it in her skin color at 6:15. Her skin is slightly more pale and less saturated, with less warmth. Why is this? Because warm spectrums are actually the largest wavelengths of light. Meaning you're always going to gather less samples of warm light than cool light over the same given area. So the bigger the pixel = the more samples of warm spectrum. It might seem trivial, but it's not. Warm light is something like 700nm. A 12MP sensors pixel might be 8000nm across. So technically it can only capture 11 photons of light side by side. Anything outside of that width gets cut off. Compare that to a sensor with only 4000nm pixels, and you're only capturing 5 photons side by side of warm light to sample, that's even less than half. We can make up for this with better signal processing of the lesser light we do gather, but that can only do soo much. Of course, photons don't flow in as neatly as side by side, but regardless of where they come in, a larger area for them to come in means more samples in that specific pixel. You could say that different pixels themselves are samples too, so 4 pixels vs 1 pixel, can create more micro samples, which does help with sharpness in the form of contrasting light/dark, but does not help with color measurements. Even if you average those 4 pixels down to 1 pixel, that average will simply not be as accurate as that single pixel measurement over the same area. A large part of designing higher megapixel sensors, means weakening the color filter arrays on the sensor, to let in more light, otherwise they would be even more noisy. This also affects color fidelity, as it allows light from other spectrums to bleed more into the wrong pixels. So red pixels get more green and blue light leaking in, shifting their final brightness value. It wouldn't be soo bad if that was all that happened, but you get the same across every pixel too. So green pixels get more blue and red, blue pixels get more green and red, throwing the whole thing out of whack. At that point it's not longer a matter of brightness of a single color, but rather, brightness of all colors, changing the dynamic between how all the colors are expressed. And since it does it to varying degree's, it's impossible to correct for. Which means it's not as simple as merely lowering the brightness value for each color, because the brightness value might be 12 points too much on the red pixel, 30 points too much on the green pixel, and 45 too much on the blue pixel, making for a completely different final color reading when combined. Not to mention that's on top of a Bayer configuration to begin with, which already has the drawback of using demosaicing to determine pixel color, while demosaicing itself adds noise just from the mathematical process of demosaicing. Can a person not care about all of this and just shoot? Yes. But there is a tangible difference that can give images an extra quality to them if someone has a keen eye to appreciate it or use it to their advantage to express that extra quality, even if the viewer themselves can't put their finger on what that extra quality is, it can be noticed.

[@Skux720]

Thank you for this! Two equal-sized sensors capture the exact same amount of light regardless of how big the photosites are. There's no cheating the laws of physics.

[@FishinMagik]

Yes, but the individual pixel’s light allocation percentage is not equal.

[@bassangler73]

That depends, if its an older FSI sensor you will lose light gathering ability due to the wiring...but on modern BSI sensors that is less of an issue..

[@guillermohernandez5268]

I think the title is quite clickbait-ish since the larger photosites sensors are in fact better in low light for certain scenarios, I'm particularly salty that you guys didn't compare low light performance on video (but did it on Print, where the larger res sensor obviously has the advantage) because there the difference would have been really notable that was kind of unfair, nevertheless I'm glad you guys cleared it up at the end of the video. Also there's an economic factor for those who are entering the world of photography and videography, who will naturally have a smaller budget, and could see this as if they needed a larger res sensor when in fact it's not the case, the print quality on those 12mp shots was amazing for example, and could perfectly save a few hundred bucks for someone who's a beginner. Other than that, great vid.

[@shounakhinge8]

I just want to know who made the thumbnail.

[@ratemo1797]

It's "really good"😂👌🏾

[@shounakhinge8]

@@ratemo1797 yes yes, of course 😬

[@niccollsvideo]

It's vintage! Like double dragon vintage. Ah nevermind. 😁

[@JK-photo77]

Love seeing the images printed. Great video! It really shows that a 12mp image can be printed large. I really would have liked to see super high ISO (up to 51k) printed and also with basic edits. A cheap zoom that ends at f6.3 and groups are at F11, iso 51k is a reality and We all edit - there isn't a reason not to show it. I've shot with a A7sII for the above iso 6400 scenes and it has always consistently outperformed anything I have put up against it especially after edits. The real bonus with it has been the ability to edit images on my phone and upload them quickly without issue. 12mp is so much easier and for family/friends stuff, it's a joy. I hope companies still keep a low resolution sensor in line-ups for this reason.

[@Maxime-ho9iv]

Except it’s not large by any means :)

[@charliewaterman816]

This actually brings another subject that I would love to see a video from you two on! What are, and how you should decide on the best export for sharing. Social media and printing of various sizes. And great video as usual... dangerous, but great 😅👍

[@calvinatdrifterstudio8438]

So why are Fujifilm cameras getting noisier and noisier with each new higher megapixel sensor?

[@insideshots23]

Great video guys, well explained. I'm looking forward to seeing what the up and coming GH6 can provide and then a comparison on the new m43 sensor/processor in low light vs full frame competition. 🤞

[@AprilClayton]

Do you remember which camera was used for the intro? Beautiful footage.

[@randygerber2885]

And a shoutout to your printer there in Calgary. Something us small town guys wish we had :)

[@niccollsvideo]

Resolve Photo is easily our favorite printing house on Calgary. They really work with their clients to achieve the artist's intent.

[@hdwblade]

Haven't watched the whole thing yet but before I forget: Photocyte density DOES matter when you get into high density sensors, and this is why. All sensors have pixels, and those pixels are separated by walls. The more pixels, the more of the area of the sensor is taken up by said walls, and less light collection is possible. If there are less walls of the same size, more of the sensor can accept photons, therefore giving it an innate advantage over higher density sensors. Can't unexplain physics, in this case. THAT SAID, this matters much more on smaller sensors. Like the 108mp in Samsung phones. It falls apart in low light and video even though it's larger than every other sensor on the market, because of what I said earlier. On larger sensors it doesn't seem to matter as much. But again, that makes sense considering the walls between the pixels can only be so large

[@sgtLPH]

I'll never stop to adore your honesty guys

[@StephanBuchin]

The Panasonic does a great job. Good filming, lighting and settings are undeniably playing a big part in the excellent and pleasing image quality.

[@ignorance_is_strenght]

Let’s all give a shout-out to the real star of the show: Madelyn / Maddy!

[@Agm1995gamer]

From nikon pr?

[@DustinBKerensky97]

@@Agm1995gamer Apparently. She's their lead photo portrait model.

[@sam80pr]

Continuing on my previous comment: please note that it is not only about the size of the pixels, it is about the technology that goes into each pixel and the tech that processes the data from the sensor. The electronic elements used for each pixel vary their characteristics dramatically based on the type of semiconductor used and the size. Different types of transistors have different characteristics charts. I assume ‘S’ for sensitivity is not just marketing hype?

[@kenbell8752]

Great discussion. Shout out to Kostas and Resolve! He made some incredible prints for me in the past. And good choice of location. I miss the Bow River! Now go get the waders and fly rod, and get to it.😁

[@MG99693]

I think this is missing an important component. Even after reading the article, you’re ignoring the Signal to Noise ratio difference between these sensors. Sure, shot noise and read noise in a (otherwise apples to apples) comparison of sensors with two different pixel sizes will be roughly equivalent. However, the signal in those two cases will be different. A larger pixel provides a larger area to collect photons and generate charge to be read by the chip. Ap yes, if we all agree that the noise on its own is the same between the two sensors, then the signal to noise ratio (signal/noise) will be higher for a chip with larger pixel size. There’s times where Signal to Noise ratio matters and times when it doesn’t. If you are taking an image of a signal-less area (such as the black areas in the sample photos) then there is no signal, and the apparent noise performance in the image will be driven by pure noise. However, in areas where some signal is present, signal to noise ratio matters more in the apparent noise you see in the image. So in short, this isn’t a wrong analysis they’re making and the point regarding pixel peeping to 100% with different resolutions is fair and valid. However, it is incomplete and missing an additional component when it comes to analyzing true “real world” performance of sensors.

[@silverrayleigh6284]

So what kind of photo do you want ?

[@MG99693]

@@silverrayleigh6284 most people are taking a photo of something (and therefore a signal) so I would argue signal to noise ratio is more important for the average photographer than just the overall noise. However, everyone is different and it’ll depend on the application and personal preference.

[@silverrayleigh6284]

@@MG99693 again, what kind of photo do you want ?

[@orphanuprising]

I feel like on even ground, a higher res sensor can actually produce cleaner files. If you shrunk the 61 megapixel file down to the same size as a 24 megapixel file, the noise would shrink too, making it harder to see. Or maybe I'm an idiot. Either way.

[@ElGrecoDaGeek]

What you describe is how we used to de-alias CGI rendered shots out of Microstation. For camera photos it certainly hews that affect at the cost partly of introducing false color as the noise is blended, but this is minor at lower ISO.

[@Michael-qv7pn]

I would say that definetly sensor size does make a difference but just in the extremes. A smaller resolution (bigger pixel area) haves better signal to noise ratio (because the noise is the same but the signal you get is more) Thats just physics, that is how it is You wont see any difference as long as there is enough signal (btw. doesnt really matter a lot how much iso (gain) you dial in) that the noise in relation is still very very low On top there is a lot of processing even within the raw file, a raw file isnt as "raw" as one might think, which lowers the difference even more, because in low light the camera just ditches some detail in favour of less visible noise and you wont see a big difference, if even any

[@FinalLugiaGuardian]

"High resolution photos are awesome." Is what I want to say. Yet I'm watching this video on my phone at 360p.

[@romainvictoria9274]

Thanks for this video. It's all down to picking the right mpx count for you. To get a sharp print, you're good to go if, for a giving size you get 254 to 300 dpi. That's why they're standards. So if you've got a camera in the 20 to 24 mpx. And considering some interpolation possible, you're good to go for 40x60cm Or 16x24". That's why for me, that range is the sweet spot. Good enough res for most people in most situation, managable files (FPS, buffer, post and storage) My 2 cents.