Why Larger Sensors = Less Noise | How Sensor Size is Related to Image Noise
Вставка
- Опубліковано 25 лип 2024
- Now Available!
MAVEN FILTERS - Color-Coded Magnetic Photography Filters
(Circular Polarizers, UV, ND Filters, Step-Up Rings, and more)
www.mavenfilters.com
****************************************************************
TL/DW - Larger format cameras have less noise because they have larger diameter openings per equivalent f stop of smaller sensor lenses. Either you believe larger openings let in more light or you do not. Enjoy!
Here are some great articles I used in preparing this video. (Side note, there are actually a few articles out there from respected sources who get this wrong, so buyer beware!)
Roger Clark of Clark Vision has an amazing article explaining it, some of his writing can be technically deep, but its all here. GREAT info: www.clarkvision.com/articles/d...
Clark Vision Part 2: www.clarkvision.com/imagedetai...
Richard Butler's Easy to Read Article on DP Review Article: ua-cam.com/users/redirect?redi...
Mark Levoys PDF Presentation: drive.google.com/file/d/0B6w2...
Summary On Wikipedia: en.wikipedia.org/wiki/Image_n...
SNR Equation, Shot Noise Follows Poisson Distribution: en.m.wikipedia.org/wiki/Shot_...
More on Other Sensor Noise: en.m.wikipedia.org/wiki/Image...
Another good youtube video on it: • PIXEL SIZE doesn't aff...
If you enjoyed the way I presented this video, I might be able to help you learn your camera!
Michael's training videos:
canontrainingvideo.com
Michael's "Maven Straps" here:
SKINNY
Black
www.amazon.com/dp/B0756V2KP1?...
Purple
www.amazon.com/dp/B07MLXY3ZB?...
Blue:
www.amazon.com/dp/B07M7G7X5B?...
WIDE
Black:
www.amazon.com/dp/B0756T8LRS?...
Blue:
www.amazon.com/dp/B07D72VLZ7?...
Army Green:
www.amazon.com/dp/B07HM2TCJ9?...
Orange:
www.amazon.com/dp/B07HLX9K1C?...
Purple
www.amazon.com/dp/B07MVXKCHW?...
Desert Camo:
www.amazon.com/dp/B07HM294R2?...
Thank you all for the civil discussion, even if we disagree! It’s been amazing!
And it’s also ok that many of you have started deleting your own comments. No worries!
i might seens something in other videos
for example in christopher frost videos he make a test of lens
this test the lens on a m43 sensor and on a fullframe sensor (he use aps-c but the point being the same i stay on m43)
what there often is to see is that some lenses dont produce higher resolutions (or pixels) on the m43´s sensor
in christophers videos you can see that the pixel density or sensels are more dense on the m43´s sensor
that being sayed the lens produced less pixels on the m43´s sensor
now taking a road through videos on youtube some m43´s lenses produce more pixels on a smaller density of sensels
in this case the video or picture as better on the m43´s camera as on the fullframe camera
that being fixed if the fullframe lens produse enough resolution
now going back that fullframe cameras haveing less density of sensels (they kinda different for a smartphone the density was 1200 nm (1,2 microns) and on a medium format camera it was 5000 nm (5 microns)) (size of light 300-900 nm)
1 problem was a lot of cameras dont use the full sensor and just cropping in the sensor so they have 1080p
that resultet in better image quality for the m43´s sensor again
2 problem was because of the less dense of sensels on of fullframe sensor
now having both problems (less resolutions per glass area, and less sensel density for fullframe sensors)
are im right or did i mess something up?
but there are meanings the density doesnt do anything:
ua-cam.com/video/_KYvp8PrCFc/v-deo.html
i think the measurement is correct in the way tony did this, but maybe some factors/variables are missing
and shouldnt fullframe cameras make equal high resolution glasses and creating more density for sensels to fix up this problems?
As per, brilliant work. People really don’t enjoy thinking this through, let alone accepting it when they consider their own gear. Best explanation I’ve seen so far, keep up the great work
Interesting video. I agree with what you are saying here. My background is in audio engineering, and SNR was super important to the quality of recordings we could produce. This was especially true with analogue recording, which inherently had more noise than digital. No discussion of SNR is complete without looking at electronic amplification - that is what is really causing the noise in your images. So in line with what you're saying, let's say we have two 24mp sensors - one micro four thirds and the other full frame. Since the ff sensor has four times the surface area to capture an identical field of view for the same amount of pixels, four times more photons are hitting the sensor. Of course this will require much less electronic amplification (NOISE) to get identical exposures because the SIGNAL is stronger. Getting back to my analogue audio roots, that's why 24 tracks sounds better on 2" tape versus 24 on 1" tape, and why 16 tracks on 2" tape machines are the Holy Grail of analogue sound recording.
So for sure full frame will have less noise, especially with weak signals (low light). Done. let's go shoot some pictures, with whatever camera you happen to have! : )
Thank you for the great comment
If I may add a correction: "Since the ff sensor has four times the surface area to capture an identical field of view for the same amount of pixels, four times more photons are hitting the sensor" Yes but this is at the sensel level. As you well said, if we have the same amount of total pixels, then it means the individual sensel is larger, thus requiring less amplification thus less noise. If you work in audio engineering (as I have) you know that to record a clean signal free of noise, what matters is not the number of microphones, or its ability to take in louder signals, but actually the SNR of the microphone / preamp couple.
Damn..You should be a bloody good teacher...
@@Innovate-pq9ci, you're right. Bigger pixel less amplification. So even less noise!!!
Summer2015 thanks. Larger sensors are more efficient thus increased resolution or increased noise / dynamic range performance. Usually that is thanks to larger pixels yes.
Thank you. I have been looking for an explanation like this. You actually isolated everything but sensor size and cut through the confusion.
Awesome explanation.. I can see how much effort you put into this.. 👍👍👍
Great video Michael. The noise you were talking about is called Shott noise which follows a Possion distribution, and is the predominant noise in the well exposed regions of you photo. This is why you could shoot ISO3200 in bright light and it still appears very clean.The usual noise we are worried about in low-light conditions or in shadow regions is mostly due to read and thermal noise and usually read noise predominates unless you are doing very long exposures. The peak SNR of a sensor is usually defined as ~log10(full well sensel capacity/read noise) and runs around 75dB or so for best in class FF sensors. Of course in low light the sensel is not full and the SNR drops dramatically, since read noise is fixed (at a given ISO). Usually we use high ISO in low light to make up for short exposure, and this is why ISO3200 looks a lot noisier than ISO100, a lot less photons collected for the same read noise.
Man thank you so much. Wonderful feedback.
Wayne, your formula is very helpful and I want to understand it. I’m trying to wrap my mind around calculating total sensor dB say for the example I used in the video, lets say a sensor captures 216,000,000 photons, with a SNR of 14,696. Would it be log10 (14,696/read noise)?
Hey, Michael - really enjoyed your video. Now, the practical question: when is big enough, good enough? For example, imagine a full frame sensor camera beside a medium format camera, both pointing at a nice landscape, each creating an essentially identical image in terms of focal length. Under what conditions will the medium format sensor camera deliver an image that is clearly visually superior to the full frame sensor?
Or to put it another way, how small can your sensor get before you kinda shake your head, step back and say, "crap. this ain't working. Need my bigger sensor camera." ....?
I'd love to see this question get answered. Commenting in case someone comes along in the future and leaves an answer, so I'll be notified.
Excellent illustrations! Really good work.
Kudos for trying to explain SNR on UA-cam to geeky photographers. I can imagine that this video will not get you a lot of views.
Even thought I studied SNR in school (I did electrical engineering) and have been a photographer for more than 10 years, it was difficult to follow your explanation. But I like the effort you put into it. Keep up the good work!
I couldn't find any good videos on the explanation, so this is my best effort!
Michael The Maven I know man. I work as a lecturer and making people understand things is hard.
For example, I didn’t get the Texas and your thumb part of the video.
Also the pixel vs Voxel thing. I see that your terminology is more accurate, but I would have mentioned it and kept using pixel in the rest of the video to keep things comprehensible. All manufacturers call their camera 10 megapixel or whatever... so suddenly calling it voxel will confuse people.
Then you scale the sensor to twice the height and width but you also make it a 4 voxel sensor. So suddenly you changed two things and I don’t know is the improved SNR the result of the size increase or the voxel increase??
Then you bring in lenses and their effect on things and then I’m completely lost.
I recommend you make more shorter videos and tackle each variable separately. Then maybe one final video referring to the other videos and combining the effect of all the variables together.
That is my 2 scents. I don’t make UA-cam videos and don’t teach very technical stuff. So you probably know more than me. Just giving feedback.
Samtagri I appreciate the feedback. This was as simple as I could make it- sorry I failed to connect with you
I've used a GH4 for 6 years. I've never gotten noise like your image on the left. When it came out it was an astonishing camera. The first to record 4k to an SD card. I have a full frame camera today, but am still using the GH4 also. It's still an astonishing camera.
I started sharing the articles several months ago.
Below I have two articles. I have summarized information from the Stanford article so nearly all credit should be given for the below summary to them. I’ve added my own explanations to some parts however.
The number of photons arriving on a pixel for any given exposure is governed by the Poisson Distribution. From this we get a mean pixel value and a standard deviation pixel value.
The signal to noise ratio or
SNR = (mean pixel value)/(standard deviation of pixel value)
SNR(dB) = 20*log10*(mean /sigma )
Therefore, doubling the width and height of a pixel increases its area by 4×, hence # of photons by 4×, hence SNR by 2× or +6 dB
Not all photons will produce an electron. Quantum Efficiency is a number we can calculate as
Q.E = #electrons/#of photons
For a typical digital camera this will be less than 50%. For back side illuminated sensors this will be greater since the wiring doesn’t obstruct photons. Larger pixels allow for longer measurements of small currents to be made and therefore a more accurate measurement.
It is reasonable to say that a full frame sensor with let’s say 24 megapixels will generate a larger signal to noise ratio than that of a crop sensor with 24 megapixels and thus less noise. This is assuming that the tech in the signal chain after the sensor i.e Amplifier, ADC etc is the same and therefore the noise contributions from the rest of the signal chain are the same. This is of course assuming the photo sites on the 24 megapixel camera are larger which would naturally be the case since they would be expanded to fit the larger surface area of a full frame sensor vs the crop sensor.
Pre gain stage before the ADC is typically an analog gain amplifier. If I assumed an inverting amplifier the gain equation is:
Vout=-Rf/Rin*Vin
Vin would be the signal of interest plus the inherent noise at the input of the amplifier pins (1/f noise, broadband noise etc). This would be where the base iso would come from and the noise would vary depending on the amplifier of choice.
The ratio of resistors chosen would determine the gain of the signal Vin and maybe be a “standard gain” whereas the quality of the amplifier would determine the noise level. One other option is for them to use a PGA (programmable gain amplifier) with values 10,100, 1000 (pre ADC).
Post ADC the signal can now be infinitely gained up digitally therefore there’s no limit theoretically.
Citations and References:
pdfs.semanticscholar.org/e0af/181fd692ef06b7a114948164c399b8869be0.pdf
graphics.stanford.edu/courses/cs448a-10/sensors-noise-14jan10-opt.pdf
Your explanation is good, especially light/proton and noise effect. Thanks.
great Maven never seen the noise issue in this perspective nice effort in explaining us. i like the way you kept it technical and simple at the same time.
so in layman terms crop sensor is 8-inch pizza and full frame 12-inch, more the surface area more cheese :> :)
Thanks Michael. If my understanding is correct, a 36MP (or so) sensor has similar noise to signal ratio as 16MP M34 assuming both uses similar f stop. However, the larger image still looks cleaner when scaled down since the noise patterns also get smaller.
It depends mostly on the total number of photons delivered.
The low light king for video is still the old Sony A7S II, with its full frame sensor and less, but larger pixels. But technology growth is letting a sensor of the same size with more pixels be almost just as good in low light. So, a smaller high quality sensor that is designed well can indeed outperform a larger sensor than is of inferior technology and quality.
There is also noise induced from the measuring circuits, so a larger pixel is beneficial, since it will acquire more light and thus more signal and higher signal to noise ratio
Great video, Michael. One thing I wish you had mentioned, though: it is not a "free lunch". Using a larger sensor, you can only boost the SNR of an image if you also reduce the depth of field. In many cases, photographers want both a shallow depth of field and minimum noise - in those cases, larger sensors are excellent.
However, if a photographer wants to maintain the same depth of field, a full-frame sensor would not produce less noise.
This tradeoff is never talked about, and I'm not sure why. It's an important point. Thoughts?
There are so many variables & different directions the discussion can go that I think it’s important to dial in just the difference between a large and small sensor in regards to image noise.
It seems that this simple concept alone is almost impossible for some people to understand so I don’t think complicating it with these other valid and important aspects would help, at least yet.
Sure we could talk about maintaining depth of field and things of that nature but if they don’t understand that an increase in sensor area does really correlate with less noise how would they wrap their mind around the rest?
If course you can. You can just increase your exposure thru a longer shutter speed. Aperture/Fstop is just one side of the exposure triangle
Great video Michael, I still find myself coming back to it when I'm needing a refresher.
You mention pixel pitch as a factor (just not an important one). In what scenario does this become important? I imagine anyone wanting to do long exposure time-lapse photography would be best suited with a lower MP full frame BSI sensor camera to minimise the impact of read and thermal noise?
Wow... how much trouble people have grasping this concept is amazing!
It really is! This is why I thought I would take a shot at it. Most photographers understand larger openings let in more photons, however, when we start talking about sensor size behind them...its a problem apparently. It's truly a fascinating anomaly of reasoning!
Nice one ! what about Sharpness vs noise : lets compare a 24MP 35mm sensor with a 45MP 35mm and see where the noise breaks apart the detail when cranking up the ISO (with controlled light) and compare the results. The question is : at what breaking point the 45MP 35mm sensor becomes equal or worse in terms of details compared to the 24MP sensor when cranking up the ISO
I do agree that larger sensors have larger signal to noise ratio. However my iphone blows away my aps-c camera with kit lens in low light photography, thanks to the benefits of computational photography. Iphone is able to take many pictures and compose them is very sharp photos, while my camera is struggling at 6400 iso and is more noisy
I wanted to upgrade from my Sony RX100VA to a Sony A6600 but now thinking to go for the Sony A7III .. (especially for the low light performance..) ..thanx Michael.. :)
finally a clear explanation of the subject, thanks
Light! The answer is always light! Thats why I want a medium format 12mpx astro body!
Thanks for this great explanation
Hello Michael. So, considering everything equal, the noise level on an APS-C with a lens of f/1.4 will be equal to the noise level on a full-frame with a lens of f/2.1??? Because f/1.4 on APS-C has the same opening diameter of full-frame f/2.1
If yes, then wouldn't it be better to use APS-C with f/1.4 rather than using Full-frame with f/2.1 because APS-C cameras are lighter and cheaper (If considering only the noise level)?
Hmm... So my a99ii isn't doing so well compared to xt2 in terms of noise... Even though it's a 40mb full frame. Can 1/3 stop light loss from the slt cause so much loss of signal?
Excellent explanation!
Well explained!!
I had heard it from other sources, but only you could say it in a convincing and definitive way. Still some videos show that, for example, a 80D and a G7 Mark II, with the same settings and equivalent 1,8 lenses for each one, will produce practically the same results, so I had the impression that lenses had a way to focus the light in a more intense way into smaller sensors, specially because an Iphone, for example, can still produce quite good pictures and the difference in sensor size is huge. Maybe some day you can disprove the 80D x G7 dilema, or the T7i x G7 one with the detailed results you always present, showing that compact cameras with one inch sensors doesn't really go toe to toe with APS-C, even when their aperture make them look like they gather a lot of light because nobody does the conversion.
In the presence of large amounts of light, the results will be more comparable. In low light is where it becomes a problem. There are some other variables that come into play, but # of photons are the most important.
It means that we can take advantage getting less noise with the full frame of the Panasonic S1 even when was cropped to Micro Four Third or APS-C in that big sensor? Thank you for this video!
According to the video: the contrary - even with a full frame sensor if you shoot in crop mode it's gonna a produce more noise. It's not that I agree with that
Why is the noise the square root of the number of photons ? just wondering. In music sampling, the noise is due to the sample being ever so slightly at a different level to the signal because it has to go to the nearest number eg 16 bit / 8 bit thats closest, so you end up with a bit of hiss caused by the very slight difference. Would it be reasoablle to say its like turning up the volume on a quiet recording, you end up amplifying the electronic noise as well as the signal, because the signal is not that much greater than the noise ?
You really have a nice way of presenting info! I bought your DVDs a few years ago, but my house got broken into and they were stolen. (Sad face). It all makes sensel the way you explain it. I think people just take it personally like someone is saying their camera is junk. "What you talking about bruddah? My camera works fine." It is always about the light, but just because bigger is better S/N, doesn't mean you can't still take great pictures. The movie Unsane was shot entirely with an iPhone. So MFT is gigantic compared to that.
It’s true you can take great images with any camera if you know what you are doing. In low light, larger aperture lenses will make a huge difference. Sorry to hear about your stolen dvds - contact our customer support and we can assist you with download replacements
Interesting video for me becz was planning to buy full frame sensors near future
I was hoping your were going to explain the ideal photon counts to get max SNR from currents and silicon based sensors.
In an ideal world, all of a the light from a FF 1.8 lens could be focused onto a phone sensor and get same amount of noise as FF sensor. And reverse, spread the light from a phone lens all over a FF sensor and the SNR would be the same as a phone. Thus, sensor size doesn't play a role on SNR. A slightly mis-leading title, but I get your point without explaining equivalence.
Hello Michael.
Seems to be you're a great Maths student at your era.
Can you also cover a topic about Lenses.
Brand name lens vs 3rd party lens.
Also another topic using Full frame lenses using on Crop body.
This one? Full Frame Lens Sharpness on Crop Sensor Bodies | Resolving Power of Lens vs Sensor ua-cam.com/video/rvEe_kqFp14/v-deo.html
I have a question. The FF 50mm 2.0 lens set for the GH5 body gives an effective focal length of 100mm. But what about the aperture? Is it still 2.0 or 4.0 (we multiply x2 as at focal length?)
The exposure is still f 2.0, but if you are talking about the DOF that is an equivalent of 2x, or 4.0. Exposure is the same
06:36 Wouldn't thr SNR still be 3 for each sensel ? I'm confused, surely its the noise of the individual sensel that matters ? A bigger sensor eg 20 MP full frame compared to a Micro 4/3 20MP 20 MP, yes I think the individual photosites will be larger, so there will be a noise floor, but the amount of signal on a big photosite will be much more , I guess ?, so much better SNR, is that right ?
What I find interesting is that comparing pics from both my A7III and EOS R.
I take pics of cars and I always have an issue with noise at 100 iso on the EOS R in post with minor tweaks. Colors can fall apart fast.
I don’t have this problem with my A7III. Colors and noise at 100 iso are perfect.
Why is this?
I think it's because the photosites on the Sony A7III are fewer, but larger (24.2MP at a size of 5.93 microns each), therefore absorb more light, which will make the noise less noticeable, whereas the EOS R's sensor may have smaller, yet more photosites (30.4MP at a size of 5.36 microns each), therefore will be susceptible to have more noise.
That's just a guess, but it makes logical sense. However, maybe the Sony A7III pixel bins, creating even larger photosites, whereas the EOS R may not pixel bin. Also, maybe the lens has something to do with it.
I'm not sure regarding the color science. I guess Sony has done a good job figuring out white balance and such to make color look as natural as possible, but I really don't know how those technicalities work exactly.
A7iii has an ISO invariant sensor. The EOS R doesnt. That makes the A7iii have much better dynamic range than the R at ISO 100.
I think I understand this, just wanted to apply the logic to my particular camera - I've got a FF Canon 5DSr, which has a 50mp sensor, compared to a FF Canon 5D Mk4 which has a 30mp sensor. I've been told the Canon 5DSr does not handle noise terribly well, but this is because the pixel pitch is much smaller to cram 50mp into the same space that you would fit 30mp, isn't it? So the logic would be same number of photons striking both sensors, but because the pixls are smaller on the 5DSr, the signal to noise ratio for each individual pixl is higher? Or have I just made an erroneous leap of faith there? If so, that would mean, when you crop down the photo from a 5DSr you get more detail than you would from a 5D Mk4, but you also get more noise Is that right? Sorry if I'm completely off base, I have an artists brain rather than a scientists. Good job I wasn't tasked with understanding quantum theory :)
Ken McDougall It’s pretty good! There are many variables to total image noise, and pixel pitch does come into play, just not as much as actual light. When the 5dsr has enough light, read noise and sensor design related noise aren’t as noticeable. There are many variables that can create noise, but in the end, light is King, so when you have enough, it’s much harder to notice. The 5dsr is amazing for detail, but my understanding is its design isn’t the most efficient. Look at the Z6 abd z7 both ff Nikon cameras, very similar in design and software and their noise performances are nearly the same. Thank you for the leap of faith!
@@MMaven Thanks for responding - think you're right about the 5DSr, quite sensitive to camera shake, and you know they knew it would be because of the modified mirror flip design, and all the extra screws in the baseplate to keep it steady. But I practiced my shooting technique and if you're careful you get something that's absolutely bang in focus and then it's outstanding. That also means that megapixl chasers buy it, get disappointed with the camera shake issue, and quickly resell it - so you can pick up really good used ones with very low shutter counts (how I got mine), can't do that as easily with a 5D Mk4!
Thank you for another clear and concise video making sense of something that often seems infernally complicated whenever anybody else talks about it - quite a knack you have for that!
Ken McDougall thank you Ken! I am doing my best to keep it all easy to u derstand
Assuming the two cameras would be the same generation if you would down-sample the 50mp to 30mp they should be equally noisy or print the same size 😉
@@Photo-ms1mo You may be right but I think they have different design sensors, the 50mp has more pixls crammed into the same area as the 30mp. I'm going on 'gut' rather than 'knowledge' here :) I think that's what Michael is referring to when he mentions pixl pitch. The 5D Mk4 has a much bigger usable ISO range than the 5DSr
Very technical explanation and totally make sense. I wonder how camera manufacture increase SNR with same size sensor with every iteration? For example, canon 1Ds mark II (16.7 mpix) vs 1DX (18 mpix). They are roughly similar in term of resolution, but the 1DX have better SNR.
It’s a good and relevant question for example BSI - backside illuminated sensors are more efficient & therefore can experience slight increases of efficiencies that reduce overall noise.
This video is the discussion of sensor size with all things being equal in relationship to shot noise.Once we can all get on the same page as far as that goes we can talk about all these other things
That is because there are multiple sources for noise, one is the random nature of photons, another one is read noise, which is caused by inefficiencies in the electronic circuitry. The first is a law of nature, the other can be, and is improved, but even if a perfect sensor could be constructed, the random nature of photons will persist. It can only be overcome by gathering more light
Richard Rönnbäck amazing comment
Wouldn't be nice if you could lower the resolution of the image capture in RAW. Meaning more sensels working together to generate each pixel. That would by a nice trade off to have on a camera.
HI, Im an APS-C user Nikon DX.
It astounds me the number of people who blog that they have changed from Canon or Nikon to Micro 4/3rds.
Thanks. I don't understand why so many people get confused by this, same exposure doesn't mean same amount of light/photons (= information). Next time I'll point them out to this video.
You get it! Exposure is brightness. Greater area also means more light.
It’s a great question. Most photographers have no trouble thinking of wider lenses letting in more light.
The moment you mention a different sized sensor is behind it...well that’s a problem for some reason!
This was very informative. However, I still have a question. If a sensor has fewer megapixels then wouldn't it perform better in low light? If a full frame sensor had 24 megapixels and a micro four thirds sensor had 12 megapixels then wouldn't they have fairly equal low light performance under the same conditions with the same aperture? The reason I ask is because the Sony a7s has a full frame with low megapixels and it is a super performer at low light.
Smaller sensors only appear to have less noise because they have fewer pixels, so the noise is less apparent. There is a DPReview video where they test an A7r against an A7s and find the noise is practically identical when the images are scaled to the same size.
Hi Michael! Great video, I’m a subscriber who previously passed over this and similar vids (diffraction, shot noise) until now as I’m struggling. I shoot an APS-C camera for wildlife in a notoriously cloudy part of the planet, PA. Noise is killing my passion but I have a tech ? and you’re the only guy to come close to scratching the surface. You’ll notice that I’m going to dance around the brand issue. If I have a 30mp APS-C + 400mm lens and I have a 24mp FF + 600mm lens I should render the same field of view. However, per your SNR formula I should get a better SNR w the full frame. I get it, but here’s my struggle: Will the larger pixels on the FF 24mp be more efficient and effectively add to my SNR? Photosite area square adds vertical growth to SNR; will pixel diameter add horizontal growth to photons accumulated on the photosite? Think of this question as comparing a D5600 @ 300mm to a D780 at 450mm. A lot of vloggers want to compare the A9 to the A7R4 w same lens in sunny FL but nobody is pushing ISO on an A6500 and an A7iii w comparable field of view. If you’ve made this video and I’ve yet to discover it plz point me there. If not, any help you can offer is greatly appreciated. Thanks!!
It is more simple than that which of the two allows for more light to hit the sensor? If signal to noise ratio is measured in terms of light which is photons the full frame sensor is going to have less noise mainly because the aperture of the lens is wider to get that full frame sensor projection
I agree mostly your points. Although I found my A9 don't have two full stops over my EM1 Mk 2(not very noticeable on iso below 800 and about 1.5 stop on high iso). I think it will be also related by the process power and the NR by the software part(even on raw file) . As I shot almost 10 yeras 4/3 system (now micro 4/3) , I understand we choose m4/3 for compactness (trade of image quality). We mostly share our work on social media and will not need to larger print. The m4/3 fits most my need. But I think Olympus need to bring more fast lenses (especially telephoto ) to support the sports camera like the the new em1X . Another than that . There's no way to fix indoor sports problem.
Hoping you will test the EM1X on the near future. Thanks for your work.
Remember its really about 3.8 something times more area all things being equal. When we start talking about different sensors designs, filters, etc, then these other variables come into play. The main goal of the video was to set the record straight when we are talking about larger format lenses delivering more light to larger formats.
@@MMaven Yes, so small format need much fast glasses to reduce the ISO performance lost. If we need to have same image quality on m4/3 compare with full frame. We need 2 stop fast f-stop lens. olympus have great IBIS on some low light situation. But it can not suit on sport photography which need fast shutter speed to freeze action .That's why i say Olympus need more fast lens to support EM1X on that area. M4/3 never can get same image quality on same ISO and same aperture setting compare large format. Especially , Olympus use same sensor on EM1X which same as 3 years old EM1mk2 .It will struggle on real world sport shooting. Because anything ISO higher than 6400 on my m4/3 camera simply unusable on my opinion.
Nicely done ~
At 11:50 the cropped part has same noise the the full frame, but since a 1/4 of the photons hit the specific part the SNR is lower. This doesn't make sense to me.
Because he's just completely wrong.. what he says makes absolutely no sense. It's like he's suggesting that len's light intensity increases with bigger sensors.. lol..
YES INCREASES because the lens size and opening increase on the ff @@bensslightnature
@@Photo-ms1mo you have no idea what youre talking about, do you even own a camera? you just say what you hear from others and follow like a freaking sheep
Thank you Michael
How do you explain the fact that Canon 80D APS-C Sensor PRODUCES LESS NOISE THAN A Canon 5D Mark III? I guess is that processor comes to the scene.
Same question
Canon has improved their 24 Mpx. With the same technology 5D would have 1 stop better iso performance, not 1/2 stop. Look at dxomark.
MichealTheTechnologist - That's your new name now.
Absolutely excellent explanation! There are always people who WANTS to misunderstand that and they will drag in all the usual arguments about exposure, even though we all can see the result of the sensor size
Wow, this is way off! The size of the sensor or surface size has little to do with SNR. It’s more to do with pixel size, the quantum efficiency of that pixel, its saturation capacity, and finally how that information is processed i.e; CCD interline, CMOS, or BSI sensors. A 9mp APS-C crop sensor will have way better higher gain (ISO) and SNR than a 40mp FF camera. Capturing more light doesn’t equate to better SNR, or better picture quality.
Now if you have a 24mp FF and a 24mp 4/3 with equal lens focal distance i.e; 25mm lens on 4/3 and 50mm lens on FF, it’s not the size of the sensor, but the size and well of the pixels. 24mp on a larger surface area mean the pixels must be larger to cover that area. Hence the better SNR on the FF.
It’s not so hard. Wider lenses let in more light, more light = more photons. More photons = greater signal. Greater signal = better SNR.
That’s all there is to it
@Richard P "You can’t just double the light and expect good things." Sure you can, especially when light is at a premium (low light situations). Sure you have shot in low light and know the difference even 1 stop makes?
This is the heart of the matter, either FF sensors receive more light or they do not.
Michael The Maven actually it’s not that simple. Photons aren’t all created equal. Photon saturation into electrons has nothing to do with the size of the sensor, but with each individual pixel well size and pixel micro lens, and filter.
NOT TRUE! The 9 mp camera have less noise but also less signal (resolution in this case) than the 40mp FF so if you downsample the 40mp to 9mp it will be much cleaner ,also often cleaner or on par with the 9mp FF sensor size.
Photo2015 that’s not how it works at all! Camera pixels don’t combine into smaller electrons to form a clear pictures. There is no “down sampling” as you put it, of pixels.
If I understand the pixel pitch concept, the same pixel pitch on any sensor would be the same signal to noise ratio. But on a smaller sensor that would equate to less resolution. Correct?
Per pixel, yes! However that’s only part of the discussion. Total number of photons determines shot noise. This is the heart of the matter.
@@MMaven So a larger sensor with the same pixel pitch would not only allow for greater resolution but a stronger overall signal(photon count) equaling less noise.
SDC exactly!
@@MMaven Thanks, Michael. finding the right words for this topic is difficult even though experience tells us it's so. Keep up the good work.
Great video I made one linking your video. Stay blessed bro.
Thank you!
Thanks, cool vid, very smart content! Yet my Canon Eos-R is almost unusable in low light whereas the GH5 works.
lol dont get me started on that camera. There are many variables, which is why I only focus on sensor size to image noise in this video. The sensor in the EOS R has many known issues. Thats a big part of the problem.
Michael The Maven I wish there was a way for you to speak the truth without being hated. Companies hearing the truth from guys like you is what makes them build better products.
It stinks people’s feelings get hurt and we have to dance around to not piss people off.
Michael The Maven, I am a little confused by your statement in this video concerning the difference in SNR in low light between sensor sizes. Yes I understand that the FF lens/sensor would get better image quality for the same framing, but confused about the "2 stop" difference in image quality... Wouldn't there only be a 1 stop difference considering the square root of 864 is 29.3 while the square root of 225 is 15, which is roughly half of 29.3?
Also, are you comparing the performance of a FF camera and MFT camera in low light conditions at each cameras base ISO, or at a high ISO?
I believe this would make a difference since FF cameras typically have lower base ISO's since there pixels have a larger deep well capacity but conversely would need longer to saturate. As for high ISO's wouldn't the square root of the average area (or photons) a sensor collects be the difference in SNR?
I hope you can find the time to answer. FYI, I really appreciate your no nonsense reviews and informative videos.
Also what do you think about the website "photons to photos"? www.photonstophotos.net/
Looking at sensor performance between FF and MFT camera formats on this website there appears to be a 2 stop difference at base ISO's but only a 1 stop difference at equal high ISO's.
Where did I specifically say a 2 stop difference ? I don't recall saying that on the video. Thanks!
Great website BTW. Thank you!
@@MMaven You are correct, you did not specify a 2 stop difference in this video. However, it could be easily misinterpreted as a 2 stop difference by your audience when you talk about a FF sensor collecting 4 times as many photons compared to a MFT sensor, if they mistaken SNR as having a 1:1 relationship between area. I believe it would have been beneficial to clarify this common misinterpretation.
Thanks for your response. I am glad you enjoy photons to photos, there is a lot interesting info to geek out about on that website :)
If i have a full frame camera and cover half of the sensor, and take a picture with it, will it produce more noise now that the sensor is half the size? This is a nice experiment to do. Coz im wondering, cellphone cameras that are coming out now are becoming good in lowlight.
Same noise on half the image. All you are doing is blocking photons from hitting the other half, so in this case it’s proportional
@@MMaven Tnx.
@@MMaven Which means that SNR decreased and noise stayed the same?
@@MMaven So the noise should be proportional as well according to the video. Why it doesn't?
Ok, great. But how exactly do you measure photons, MtM? Besides camera sensors, what other instruments are there for that particular job? Having said that, how do we know we are actually measuring photons?
I like that!! how do we know that you are writing this comment, not someone else?
@@inseries5494 Kind of lost me there, my friend. I asked a very simple question. I'll rephrase it: how do you objectively measure photons?
They are not measuring the photons in fact. The sensors measure the emission of electrons due to the photoelectric effect (check Google if you are not familiar with this term)
Thank you @@kalinevtimov4298 . The video showed otherwise, hence my original question.
@@EduardoPortasRuiz Yes in the video he has oversimplified the things a bit
I'm still confused. If it's "Photons per sensal," Then everything being equal--image SNR should be the same for full-frame or APS-C.
Only if noise is linearly proportional, which it isn’t. Shot noise follows a predictable equation for noise...as you increase the total #of photons, the relative noise decreases.
But, is it not the same number of photons hitting the single sensil, no matter what total size of the sensor? ( My head hurts and I smell smoke.)
Willie Edwards Are there more total photons hitting Texas than my pinky? If you can believe that, you are starting to understand.
Michael The Maven then the explanation is inefficient. Because then that would imply that noise generation is not per unit area, but per total area... Or that light intensity is the proper term and not light amount.
Michael The Maven... When we say amount, people imagine that it's like taking a lbs of sand and spreading it evenly across the surface... But intensity is something completely different. F STOP terminology is then an improper measurement number because it implies that the same amount of light is hitting the sensor, when it very well may be that the larger iris of the FF lens allows for a higher intensity of light even if the exposure is the same.
Still complicated and confused! How about smaller sensor with larger lens, would it be less noisy?
If a speed booster is used (aiming the image circle at the sensor only)...yes!
let me ask u a question. Does bigger microphone produce less audio noise than smaller one?
All microphones have variation. Bigger factor is distance to subject
Larger sensor = More bragging rights.
I take you are a noob...
mind blowing teaching
Thank you
I appreciate how you use the SNR equations to demonstrate your point: quite easy to see. However, I disagree with the idea that larger openings let in more light: The amount of light or photons reaching the various size sensors should be the same. Larger format cameras have a larger sensor than that of DSLR camera. Assuming that both a larger format camera and a DSLR camera have 50 megapixels, the pixels in the larger format camera are bigger than that of the DSLR camera, because the larger format camera have a larger sensor. Since the pixels in the larger format camera are larger, they are able to collect more photons per pixel unit. If we carry this idea and apply it to the SNR equations, we can see that signal to noise ratio is higher in the larger format camera vs. that of the DSLR camera.
Chris Tams if you disagree that “larger openings let in more light” than wider apertures do not either.
@@MMaven Then let's do an experiment. Let keep the ISO the same: at 1600: also, both camera sensors are the same. You are using a large format camera, and I am using my DSLR camera to shoot a darker scene. We are both shooting in manual mode, and I am assuming that the exposure metering on both cameras are the same: Since ISO is constant at 1600, so to get the same exposure value on both cameras, we each need to adjust slightly the shutter speed. Even if your large format camera allows more light in than mine, the shutter time for you is shorter than mine, to get the same exposure. Since my DSLR allows less light in, my shutter time is longer than yours. The interesting question is, "Do your pixels receive more photons than mine?"
@@MMaven It is not the case that I don't believe that a larger aperture allows for more light to come in, but, rather, I disagree with the statement that "Larger format cameras have less noise because they have larger diameter openings per equivalent f stop of smaller sensor lenses."
@@christams8863 They are physically larger openings on Larger format cameras. Either you believe there is more light in larger openings, or you do not. While the f-stop, or exposure can be exactly the same, the SNR is better when more light is coming in, and therefore, there is less noise. If you cannot accept larger openings let in more total light, you wont be able to accept this principle, but thankfully, larger openings letting in more light is an easy concept to grasp. best wishes.
why does my Fuji gfx 50r have so good image quallity and nois ratio vs my friend.s canon eos50r his canon is 50mp mine is 51.4mp practicly the same ammount of MP but my gfx blows the canon out of the wather on nois and image quality colors and everything else
Did you watch the video? :)
great video to present this theory... but i want to see it tested to see if your hypothesis is correct
It’s easy to test. Either larger apertures let in more light, or they do not
If a manufacturer made a Micro 4/3 sensor of very low pixel density, let's say 8MP for example, wouldn't it have about the same noise performance as every "sensel" would be bigger?
It’s not the size of the sensel that’s doing it...it’s the wider opening lens!
So just putting the FF Lens on the micro 4/3 should solve the problem in your opinion?@@MMaven😉
Photo2015 with a speed booster? Absolutely. The concept of image circle was covered. Total area matters
Very interesting ... well explained. I wonder if Tony Northrup has seen your video after coming out with his. You focused less on exposure and more on noise level. I am thankful that I watched your video, because you confirmed what I had always believed. KEEP MAKING THESE INFORMATIVE VIDEOS! It helps camera enthusiasts like me understand how sensors work. By the way, you do really good work! Thank you!!!!
Theres a lot of confusing information out there, so I thought I would take a stab at it. Knowing I was able to help, even just you made it worth it! thank you!
Fundamentally, the light gathering power of an optical instrument is determined by the area of its entrance aperture, which has nothing to do with the area of the sensor. For an imaging instrument, such as a camera, the depth of field is also determined by the entrance aperture area - again independent of the sensor area. For a fixed entrance aperture area - for a fixed depth of field - a small sensor camera will be operating at a numerically smaller f-stop and iso than a large sensor camera. This is why small sensor cameras have very similar low light performance to large sensor cameras in many if not most practical situations. Large sensor cameras only win when depth of field is not an issue- e.g, Astro-photography - and when bulky lenses with large entrance apertures are practicable.
Except that making a very small f-stop lens can sometime be extremely hard to manufacture. For example, an image shot with a f1.4 lens in a full frame would need a lens with something like a f0.9 in APSC sensor cameras to achieve the same depth of field, which is a lens very rarely available for regular consumer.
But in support of your point in one case, if you do can achieve the same depth of field, it's even possible that the smaller sensor one have actually lower noise than the larger sensor due to the reduced read noise (assuming same manufacturing quality)
dude, are you sure on this one? looks not right to me.
your explanation is true if each pixel is capturing entire image, but as long as each pixel captures different parts of image, it all comes down to pixel pitch only, and individual photosite does not care what other photosite capture.
snr is not only the hardware thing like in analog, snr in digital is affected by software part too.
Either larger opening lenses let in more light or they do not...it’s basic physics, most photographers understand it.
Those larger openings project “more light” onto larger format sensors.
Even a speed booster is proof there is more light from FF Lenses to MFT
So...Yep! I’m sure on this one.
@@MMaven lenses have nothing to do with this, my phone has f1.8 and this is related only to exposure, not to lowlight capabilities of sensor
you can say that smaller pixels gather less light, but pixel on the right side does not care what pixel on the left side sees
Vadim Om and can you use your 1.8 smart phone lens on a FF Sensor first get 1.8? This is what you are implying
Vadim Om lenses have everything to do with this. If you don’t realize that you won’t understand this
@@MMaven this is obvious dude, and it's not related to the sensor noise
The difference between you and Ken is, your smart and Ken just thinks he’s smart.
Good video.
Facts matter,....not what people believe (doxa). and only facts matter.
Oh Leonard, what an only emotional and irrational argument!
The old Romans called that argumentum ad personam which often occured and occurs when rational arguments - argumenta ad rem - are missing!
Theoria Apophasis Problem is the facts aren’t on your side Ken. m.dpreview.com/articles/5365920428/the-effect-of-pixel-and-sensor-sizes-on-noise
Ken is right and you are wrong
@@bellmountaincastle Let me guess another Fuji shooter lol
You mean Microstrip antenna arrays (or printed antenna array) which are used typically at microwave frequencies in telecommunications? This exist...
Hi @michael! (Sorry for my English)
I love your video and your explanation about snr... and I’m agree with you. Photography is my hobby but I work with medical images and I have the same problem with medical images than photography.
Less photons = more noise
Very difficult))
On high ISO with same speed of lens and shutter bigger sensor give lower noise. For compensate small sensor, you can use lower shutter speed or take lens with more speed
For 4/3 and FF with same shutter speed and noise level will be equivalent - 50mm/2 ISO400 = 100mm/4 ISO800. Just multiply by the "crop factor"
On low ISO (100-400) noise level must be the same
But then you overexpose...
@@jayzn1931 No, on the sensor will be same lite spot
And why has a Crop-Sensor Nikon D500 much less noise than a Full Frame Canon 5DSR??? Because Noise has nothing to do with Sensor Size. Its related to SNR and this is usually related to the pixel pitch and the used technology. But not to the Sensor Size... However it is often the case that bigger Sensors also do have a lower pixel pitch and therefore they also do have less noise, but the Sensor Size ist however not the reason for it.
100% correct
a thing we in the astro field call quantum efficiency
Please refer to the articles provided in the description. Pretty good science in there. Roger has a PhD from MIT with over 300 published papers, and well over 22,000 citations. He knows his stuff, so...you might want to actually read my basis for this video. Thanks!
Well that's because the 5DSr has a shitty sensor with lot of read noise. Compare the D500 to the D850 from the same generation: identical pixel size yet the D850 has 1 stop better low light performance
What if both Nikon D500 and Canon 5DSR both have the same pixel pitch? Will they both have the same noise? I Don't think so! Full Frame wins LOL
If you had such an unstoppable desire to build a lens that could fit on a m4/3 mount, project an image circle with the diameter of the diagonal of a m4/3 sensor, and deliver as much light as an ordinary lens designed for a large film camera, you could have less noise than you would on a 35mm camera shooting at high iso. Exposure would be different so there's that. Shot noise isn't even a property of the sensor as i understand it. So it really boils down to manufacturing and design standards for whatever system youre using. But on the DSLR and Mirrorless market, i dont see those monster lenses. There are speedboosters out there tho that will do this to some extent. So noise depends on the whole stuff, and for standard DSLRs and standard lenses, the whole stuff that makes a FF camera will produce less noise than the whole stuff that makes a m4/3 camera for the same image.
All accurate! I think you get it. The secret sauce is how much light the lens is delivering...not the size of the sensor itself! That’s the key!
Good Morning Sir!
To conclude the discussion: if the size of the sensor would be significant then pixel binning to increase the signal-to-noise ratio and thus to decrease noise would not work, but it does!
Binning happens after the signal is captured
@@MMaven Binning results in less noise because the deviation is statistically reduced by taking the mean (about square root of number of pixels). Having e.g. four small pixel and combining them will result in nearly the same result as one pixel with the size of 4!
Outstanding explanation!!!
A larger sensor DOES NOT collect more light - a larger APERTURE does... - having a larger sensor behind an aperture does not magically make more photons enter the aperture, instead the only thing that changes is the crop factor.
If you want to take and equally zoomed image (field of view) you need more focal length to fill the bigger sensor, than you would need to fill a smaller sensor with the same image, in that case the SAME light would simply get spread out more along the focal length and would hit the larger sensor less concentrated, where as the smaller sensor would receive the same light, less spread out and more concentrated. - Number of photons would allways stay the same, because the sensor only records the photons captured by the aperture.
Everybody seems to think that a larger sensor collects more light, but that is what the aperture does... - you are confusing the sensor with the light collecting aperture.
Your first paragraph is literally agreeing with my video. Second paragraph, not really sure what you are saying.
@@MMaven Ill admit i didnt watch your entire video before commenting, but now i have and it seems your explaination still misses the point a little bit. Correct me if im wrong.
In a nut shell only aperture collects light - a bigger sensor simply allows manufacturers to construct bigger aperture lenses, keeping the f-number the same, and f-number is the limiting spec when designing lenses.
To fill a bigger sensor with the same field of view image you need a longer focal length than you would need for a small sensor. - having a longer focal length now allows you to increase the aperture and get the same f-number as the shorter focal length, smaller aperture lens - and with a larger aperture you collect more light.
So its really not about the surface area of the sensor "collecting" more light, since you could just focus that same light on a smaller sensor in higher concentration, but about lens design.
Thank you for your video.
Jeffrey Lebowski doesn’t miss the point. The title of the video is why larger sensors have less noise. It’s because lenses required to cover them entirely are wider apertures. That’s it.
I’m still not sure what you are trying to say about the relationship of focal length to apertures to sensor sizes.
@@MMaven I see, i agree your video says alot of correct things, maybe im just splitting hairs here.
The relationship between sensor size and focal length is that to take the same image - same field of view - with a larger sensor, you need a light collecting element (lens) with a longer focal length (the image has to be allowed to spread out over a longer distance to cover the larger sensor) - this is implicit in your video, where you show three different lenses with the same f-numbers but differing apertures (diameter of the opening). - those lenses have differing focal lengths. (crop factor not taken into account, because we're talking about actual physical focal length).
You dont need a larger aperture to cover a larger sensor, as you state, just a longer focal length. But a longer focal length ALLOWS you to increase the aperture, because people know, how to make a lens with a certain f-ratio, - angle of the cone of light being focused.
So if you want to improve your S/N ratio you start by increasing your aperture (diameter of the opening) for the reasons, that you have explained in the video, then you realize that this will lead to a longer focal length, because you are limited in your lens design by an f-ratio, then realize this will lead to a larger image/projection, and only then build a bigger sensor to capture the larger image, because you want to take images with the same field of view and not be zoomed in by cropping the image with a smaller sensor.
So a larger sensor is not intrinsicly better or the starting point, when it comes to shot noise, instead it is required to acommodate a larger aperture lens, because of increased focal length...
I hope this is understandable, i only just understood this my self, partially thanks to your video.
Let me know if this was helpful or interesting to you or not.
I really enjoyed your video, thx.
Isn’t the noise just more apparent on a Mico four thirds camera because you must enlarge an image 4 times larger than full frame making noise 4 times larger?
20mp GH5 vs 10mp GH5S vs 24mp Z6 vs 46mp Z7 This should put to bed any theories that pixel pitch plays any MEANINGFUL role in noise and that sensor size is key! www.dpreview.com/reviews/image-comparison?attr18=daylight&attr13_0=panasonic_dcgh5s&attr13_1=nikon_z7&attr13_2=panasonic_dcgh5&attr13_3=nikon_z6&attr15_0=raw&attr15_1=raw&attr15_2=raw&attr15_3=raw&attr16_0=6400&attr16_1=6400&attr16_2=6400&attr16_3=6400&normalization=print&widget=1&x=0.18771705871298186&y=0.5071292281006071
That's exactly the same thing author presented, just in different words...
What your saying is simply NOT correct. A M43 sensor using lens with same "angle of view" and with the same aperture collects 1/4 of the light of the same FF system.
@Richard P so basicly, I will get more noise on my D500 than my D3 because my D500 is crop and D3 full frame ?
@@robert7622 The Z6 has the least noise, followed by the Z7!
Good stuff! Thanks.
Smaller Sensor = Smaller Camera for the win! I'll tolerate a degradation of the image for the increase in portability of M43. My personal preference.
Not that much of a correlation actually, the real difference is the lens. An equivalent aperture on fullframe is just as big as a comparable M43 lens, and also camera bodies are pretty much the same size nowadays.
@@v0ldy54 I can fit my Olympus EM10 with a Panasonic 20mm/1.7 pancake lens in my coat pocket. I want to see you do that with a FF.
@@bobk4438 A7iii smaller then then the best MFT lol
@@robert7622 Until you mount a lens on it :D
@@bobk4438 camerasize.com/compact/#487.394,521.94,ha,t
Nicely done.
Software always improving the speed of signal
Great explanation
Better lighting = Less Noise
I guess objectives and whatever also have something to do with noise.
Of course!
Great Video, man. u deserve more subs and likes. AND i helped u with 1sub and 1like. hope u make more such contents.
James Frank Anderson thank you!
I think your explanation only holds up (while using a biased photons model) if you put no lens in front of the sensor and if you assume every sensor processor is the same regardless of sensor type or size.
Your illustration at 6:07 about the sensels is incorrect. When you multiple the number of sensels, you also need to divide by that number somewhere else in your equation. I'll leave it to you to figure out where.
Ok, I can't help myself. I'll correct you here. In the example with 4 sensels you can't merely sum the light intensity and ignore the number of sensels(of which individually are responsible for the SNR value). You must divide. And hence: sqrt(36/4) = the expected value of 3. And hence, assuming the same level of technology (individual sensel SNR ratio), the overall image noise level remains the same as long as the sensel size remains the same.
And thus:
Given individual sensel SNR of 3
And 24 million sensels:
(9*24,000,000) = 216,000,000
Sqrt(216,000,000/24,000,000) = 3
Properly correcting you 24mp camera example, and crushing the incorrect math.
Also, it would be better to think in terms of light intensity per square millimeter instead of a count of photons, but this is irrelevant to the math, and it's safe to be plugging in 9 to represent the intensity level.
So you believe 36 photons are not a greater signal than 9? This is what you are suggesting. Noise is calculated by taking the square root of the number of photons. The illustration shows exposure can stay consistent, while the signal increases. Even if you used a single sensel, 4 times the size or even the same size, as you are suggesting...the math would still be the same! Either 36 photons are a stronger signal, or they are not. 36 photons are greater than 9!
Also, do I really need to point out the ludicrously high SNR value of 14,697:1 from your 24mp example? You disproved your own math here with such an insane number.
3:10 wait a second... Your Pinky isn't square!
That is not true!
Firstly the whole image is not the signal!
Secondly it is not the sensor it self that gathers photons - the pixels INDIVIDUALLY gather them! Due to quality improvements the main part of noise nowerdays stems from light noise that is the indeterminism of a photon due to quantum mechanis. The less photons the more the variation of gathered photons and the more noise!
Sony offers different full frame cameras with different noise characteristics. The A7S2 is dedicated for low light with excellent noise behavior has a pixel area of 69,22µm2 (12MP), on the contrary the Sony A7R3 with much less noise tolerance has an area of 20,25µm2 (42MP)!
Sorry my friend- the whole image counts as signal. Please refer to any of the many articles I have provided in the links. Thanks!
@@MMaven Firstly You did not even touch the more important second part dealing with pixel area and photon physics!
But the noise is not equally distributed over the image it varies with the brightness of the different parts, so there mustbe different signal-to-noise ratios in one pictures and that was it I wanted to express!
@@MMaven And to continue: the f-stop measure normalizes the amount of light in respect to sensor size in coloboration with the crop equivilance - the same f-stop means the same amount of light per unit area. Otherwise - as Ken Wheeler often sarcastically stated - all light meters would have the sensor or film size as an entry parameter!
@@WMedl I never disagreed with any of that, I actually demonstrated this (several times) in the video. So, no it doesn't sound like you are open minded to the truth. What you and Ken cannot accept is total signal matters. It really does! Had you read any of the articles with a sincere heart you would see it. Good luck to you!
@@WMedl Sure I did. Man, thats too bad. I thought you were sincere. Good luck!
I don't disagree with anything technical you have said but rather I look at the significance or relevance of the science. I need to lose about 30-40lbs, if I lose 3 it is MEASURABLE but not SIGNIFICANT or even DETECTABLE. WHILE larger sensor cameras generally will have measurable improvements it cannot always be attributed to the larger sensor, I think it is fair to say in most cases the larger sensor cameras also likely have better processors, but not always. The XT-3 likely has better processing than a 6Dmk2. There are so many variables in different model cameras and different generation cameras it is inaccurate to try and make generalizations. The best comparison is to directly use 2 specific cameras and compare the results in the same conditions with Photographers of the same skill level. Even in the case of a camera having a measurable advantage, the results will likely be UNDETECTABLE to all but those with a trained eye
I appreciate the comment. It’s more about the total number of photons hitting the sensor. We see this with ff speed boosters on MFT cameras, as that light is concentrated down, the actual exposure goes up.
Larger format lenses allow more light to enter per equivalent f stop.
Coming from radar technology, the SNR of a sensor array (SNRarray) is calculated by SNRarray = 10log(M) +SNRsensor with M being the numbers of single antennas / sensors and SNRsensor being the SNR of the single antenna / sensor. The array gain (G) is defined as the ratio of the array SNR and the individual sensor SNR and measures the enhancement when using M sensors rather than a single sensor: G = SNRarray / SNRsensor. Both array gain and sensor array SNR do not (!) consider the size of the complete sensor itself. Both formulas can be applied also on imaging sensors as you just deal with electromagnetic waves of different wave lengths. Hence, Larger Sensor = Less Noise as general rule is non-sense. I am sorry to say that - for all folks who still want to believe that. Also, stating WIKIPEDIA as a source is everything else than science based argumentation, it is in fact a pseudo-science.
That’s funny. I don’t even know anything close to you as antennas, but it sounds like you are saying antenna sizes don’t matter. Is that accurate? If so why not use microscopic antennas?
@@MMaven Radar antennas have to deal with emitting a signal as well as receiving it. On the emission side of the house, a larger antenna allows you to have a smaller beamwidth (i.e. more angular resolution) which is hugely important if you want to know where in space the thing you're detecting is.
Wikipedia is only as good as the sources section backing the article up. ;)
Can you please point to all the smaller sensor cameras out there that DONT follow this General rule? There must be dozens to back up your "but radar" analogy
@@robert7622 Look, this is a pointless discussion. The theory (and implications) is just a fact, you may not like it but it is a fact. It is not about radar / antennas or whatever. It is physics. I have really no intention to step into a discussion or fight about what people want or don't want to believe. And this discussion about FF vs APS-C is such a useless fight of believes.
It doesn't make sense to me when at 6:50 you add total number of photons of 4 sensels. Each sensel makes 1 measurement, and 4 sensels make 4 measurements. If there was only one 4x bigger sensel, then we would have 1 measurement, and only then I would agree with the way you calculate SNR=6.
In other words, the sample increased 4x, but our needs increased 4x as well, meaning the sample is just as poor.
As I see it, each pixel has to make indepentent measurements. Therefore it only matters many photons hit the specific pixel. Average SNR would make sense. I agree with SNR=sum(sqtr(N))/#pixels and not with SNR=sqrt(sum(N))
You are making the assumption that noise also scales linearly in proportion with surface area. It doesn't. Noise increases, yes, but as more photons are added the SNR is more favorable to the signal. This was demonstrated in the video. The More total photons = better SNR
@@MMaven Lets say I deactivate the left half of a sensor. Do I get same noise but less photons and less SNR? (great videos though good job)
@@MichaelM1991 nice point there on deactivating on half of the sensor. If i cover half of the sensor in a FF camera, will it produce more noise coz its now smaller???
No, it won't, but that's not the point. The point is, let's say you're trying to take a picture of a chair. After you deactivate half the sensor, you won't be able to see all of the chair in the frame. You'll be forced to back away, and take another picture. This second picture will have more noise, yes.
@@ParanoidAandroid By simplifying things to understand them, we assume there are no darker or lighter areas in the pictures we photograph. Therefore going 2 steps backwards, I dont see how the image gets darker, or the number of photons going through the lens decreasing, or noise increasing.
Hmmm....so those Photons are the reason why. Theoretical particles were the answer all along I guess.😋
theoretical parcticles that have been observed to trasmitt electro magnetic energy
Short version: Pixel pitch matters, not overall sensor size. You know those backup cameras used in cars? They use small sensors with massive individual elements. They have great low light performance, but really low resolution; not my first choice to take good portrait photos with them but they do allow you to see in the dark. If sensor size actually drove low noise performance, they'd need to be massive, but they aren't.
Long version: Far more goes into calculating the SNR of a sensor than just the number of photons collected. Each element in the sensor introduces some amount of noise regardless of the signal. So, the more elements you have, the more noise you have, which means that higher resolution sensors by default have to deal with more noise, al other things being equal. Ouch. If you want to improve SNR, you need to increase your signal without increasing the noise added by the individual elements, or keep your signal the same while decreasing the number of elements in order to reduce noise.
Case 1 is simple: increase the size of the sensor while keeping the resolution the same. This is what you see with the current crop (no pun intended) of digital cameras on the photography market: lots of sensors with about the same resolution (~24mp) of different sizes. In this case, yes, larger sensors will have better SNR because they have fewer elements per unit area of sensor (i.e. lower pixel pitch).
Case 2 is less common in the photography market: keep the size of the sensor the same while decreasing the resolution. This is more common in other areas of digital imaging, especially IR and microwave (I've been a radar engineer for over 10 years), and it has the same effect on low noise performance as the larger sensors in digital cameras: it decreases the number of elements per unit area of the sensor.
Bigger sensors in photographic cameras currently on the market only have better low noise performance because of their lower pixel pitch. If a camera manufacturer really wanted to, there's nothing stopping them (at least from a physics/engineering standpoint) from releasing lower resolution, smaller sensor cameras with excellent low noise performance.
Im afraid you have been given some bad information. If pixel pitch matters so much, why would it not matter when talking about increasing sensor size? You are referring to the same physics. Sure there are some technologies out there that are great, but all things being equal, size really matters!
@@MMaven But that's just it: it does matter. I gave you an example in my comment above. As for why camera manufacturers don't talk about it, why would they? Resolution is an easy number to talk about and a larger number is always seen as better, but with pixel pitch it's far less clear how to market it. Does a consumer want a higher pitch or a lower pitch? Do our top end cameras always have one or the other? Etc.
Engineering often makes for bad marketing. ;)
This all boils down to a logical fallacy that I was guilty of making for a long time. When someone says, "Okay, let's change the sensor size but keep everything else the same", that's an inherently (if unintentionally) misleading statement. Why? Because if you increase the sensor size and keep the resolution constant, the pixel pitch drops and that impacts subsequent discussion whether you're aware of it or not.
Benjamin Cushwa sorry brother, Sensor size matters. Please refer to the links in the description.
@leicanoct Sort of. There are limits to how small we can currently engineer individual sensor elements (and circuit components in general). We can't make 50MP crop sensors because we can't make the individual elements that small (at least not affordably). But full frame? 50MP on full frame we can do.
Benjamin Cushwa I think it would be helpful for you to refer to the articles in the comments and then come back and watch the video again.
This only works when the small and big sensor have the same pixel count