How are camera sensors still improving? | Upscaled
Вставка
- Опубліковано 9 лип 2024
- I’ve been trying to decide on a new camera for myself, and it’s easy to get lost in the specs. People argue for different brands, some folks insist only lenses matter, and I guarantee you someone will comment here that it’s just the light and the scene that matters and you shouldn’t care about gear, and that person is insufferable. So on my quest for a new camera, I set out to learn how sensors really work, how they've changed, and where they’re going.
Check out our sources here
docs.google.com/document/d/e/...
Get More Engadget:
• Like us on Facebook: / engadget
• Follow us on Twitter: / engadget
• Follow us on Instagram: / engadget
• Read more: www.engadget.com - Наука та технологія
With the intensity my dude is speaking, I fear about the veins in his neck.
Or his hands flying clean off 👋
Lol, for real
Yeah. Pent up energy released all over camera sensors!
Lol
This is a pretty nice explainer. I would like to correct a few things in the video.
1.
Smaller pixels have higher noise and lower dynamic range, but when well scaled the noise and dynamic range PER UNIT AREA remains constant.
If you take a pixel and sub-divide it in to four smaller pixels, with careful manufacturing process controls each small pixel will have 1/4 of the sensitivity and 1/4 of the full well capacity (FWC) of the large pixels. But each smaller pixel will only see 1/4 of the photon flux (smaller buckets left out in the rain), so they can capture precisely the same scene dynamic range as our larger pixel. Meanwhile the noise of each pixel has increased by a factor of two, but when averaged together you get a SQRT(4) = 2 times reduction in the noise.
This is the key point, for a fixed viewing condition (the same screen, or a print of the same size), the performance remains constant, so long as the scaling hasn't pushed to a point where the drop in sensitivity and FWC are worse than linear. Sure, if you zoom to 100% in Photoshop the smaller pixels will look noisier, but you're zoomed in twice as far as in the larger pixel case in real terms.
This is why half toning works, and why an inkjet printer can make photo-quality prints using tiny, discrete dots of ink. When viewed under conditions beyond the limits of human visual acuity, the dots merge together to form the illusion of low noise continuous tone.
What does improve noise and dynamic range is a larger sensor.
2. Gallium Arsenide sensors (or more accurately InGaAs) are used for medium wave infrared imaging, not the long wave thermal infrared imaging showed in the video. That generally uses Vanadium Oxide micro-bolometers, which are an array of small resistors that change in response to temperature. This imaging is in the 8-12µm band (below that, down to ~5µm there's absorption by water vapor in the atmosphere).
3. In BSI the silicon is thinned down so that visible light can pass through it and reach the photodiode. The pixel wiring does not pass through the silicon as described. At the edges of the sensor connections through the silicon are indeed needed to provide bond pads so that the sensor can be wired to the package or printed circuit board (in the case of chip-on-board).
4. [Not a correction, an addition] You accurately described stacking, how it provide real estate for better sensor circuitry and DRAM. Another thing it can enable through hybrid stacking with close to pixel scale connections is global shutter functionality. At the extreme, a small cluster of pixels can share an analog to digital converter on the ASIC layer. By quickly converting each of these in turn, a near global shutter functionality can be achieved. Even without that extreme it's also possible to put the additional analog circuitry required for global shutter pixels on this ASIC layer and do a better job at isolating them from the light than in a monolithic global shutter sensor. This can lead to higher shutter efficiency.
Overall I enjoyed this video and applaud the job you did in making this understandable to a mass audience. Personally I found the overly-energetic hand waving to be annoying. I know this is what people are told to do in videos to be more engaging, but there's a point where it becomes distracting and comes over as inauthentic (fake).
The problem is, it's not exactly linear. Light itself has different wavelengths that are a specific size of nanometers across. Blue light is "smaller" than warm light, about half the "size". While the ratio of light captured will remain linear, the difference in pixel size can mean the amount of not just light, but a specific wavelength, is halved or less, while the efficiency of solar panels (which is what a camera pixel essentially is) have not doubled in even 10 years, which means it's impossible for sensors to perform just as good with half the light.
Sensors actually cheat a bit to reduce noise, by making color filter arrays more transparent. So your red pixels get more blue and green light bleeding in, and so on. It increases signal, but at the expense of color accuracy/fidelity. They then subtract the ratios of expected light bleed in camera, shifting the pixel to be brighter or darker based on the ratio of light that is factory measured to bleed through into the wrong colored pixel. The problem with this is it's impossible to account for scene characteristics.
For example, you may have a red pixel that lets in 20% too much blue and green light, making that red pixel much brighter. But if what you're capturing on that pixel is green in the scene, it's going to push through a LOT more green light, and over brightening the red pixel even if it is ultimately converted to green through demosaicing, it will still be a less than accurate final result. And since it is a red pixel capturing what will ultimately be turned into a green pixel in the final image, all you have is the green light bleed mixed with the red and blue light bleed, meaning you have no way of sifting out that green measurement from the rest, you simply have to take them all.
It's kind of like having a glass of milk, pouring 20% soda in it, and 30% orange juice......and labeling it just "milk", when in reality it looks nothing like a glass of milk anymore. It'll be just an orangey brown sludge instead. And that's how inaccurate modern camera's are. Camera's made in 2005 are actually measured to have a higher color rendering index than even the best of the best today. This is because in the pursuit to increase megapixels, compromises had to be made to the CFA, because we've already pretty much reached the limits of what silicon can do. And if you think of it in terms of solar panels, that's where you see the truth, because again, camera sensors are just solar panels that generate a charge from light. How big is a 1000w solar panel today compared to 10 or even 20 years ago? Little to no difference, because we've at best only squeezed like 15% extra efficiency out of solar panels in the last 20 years.
So when you double the megapixels, you halve the samples of light, without increasing efficiency by double. That's why CFA's are actually the main driving force behind megapixel increases. But that can only go soo far, because you can only do soo many tricks before you let in soo much light bleed, you effectively have no way to distinguish color well.
@@peoplez129 linear scaling of pixel performance is hard, but so long as not pushed too far and kept well above the diffraction limit for red light (for visual imaging we need up to 750nm) the sensitivity and FWC scaling has been remarkably linear. I haven't looked at the latest really tiny pixels, like Samsung's 0.64µm (which is ridiculous), but I have data for larger pixels that show linear scaling.
The color crosstalk you refer to does not impact color reproduction. It impacts chroma noise. The color correction matrices subtract this crosstalk and if correctly computed for different illuminants (different light color temperatures) and interpolated, produces very low delta-E/delta-C color reproduction results. What does happen though is noise gets amplified by the correction, but that doesn't shift the mean patch color reproduction score. Another downside is high crosstalk reduces the pixel MTF, resulting in camera resolution loss (resolution in terms of linear resolving power, not pixel counts).
Yes, what I wrote was the first-order explanation, and you are correct that there are a lot of second order effects that come in to play when scaling pixels, including effects from the CFA, microlenses, tungsten grid, oxide/nitride dielectrics, and deep trench isolation. At my last company the CFA team were constantly running large numbers of wafers to tweak parameters, try new CFA materials (well, less frequently), in order to improve performance. It cost a lot of money!
Never the less, it's wrong to make a blanket statement that bigger pixels are better when you normalize to fixed viewing conditions (instead of looking at per pixel SNR). DPreview even ran a feature on this recently. [I didn't watch the video, because I've been involved in this kind of testing in my professional capacity.] I even have experience with a smaller pixel that had substantially better FWC per square micron than the previous, 2x larger pixel because it was a popular pixel that had been improved over multiple generations vs. one that had not.
@@JonS Go look it up, the original Canon 5D has a better color rendering index rating than the Sony A7III. Two camera's, more than a decade apart, and that's despite the extra noise the older camera has. And CRI doesn't just matter for normal conditions, but also difficult conditions, such as mixed and artificial lighting. The CRI's drop drastically under less than perfect daylight. So what you end up with is a camera more capable of resolving color in these lighting scenarios.
For example, let's say you have a really warm light, but it's a weird CFL or LED that also has a lot of green. What little light is reflecting off of the scene, is going to be limited by the available wavelengths of light. So if your light source doesn't have much of the wavelengths objects would reflect, you end up with light being absorbed by that object, resulting in less samples of the color of light that object reflects. Lower CRI = less ability to see colors. So it's a lowlight issue, but also a color issue. The better CRI camera will see more of the actual color in the scene. SNR might be a thing, but there is something outside of that. Modern camera's boost SNR by making CFA's less discriminate. So it's a bit of a false SNR.
Just like with raytracing, the more samples of light bounces, the more realistic a 3D model looks. So SNR is only part of the issue, it's just comparing noise to signal, with a compromised regard for the fidelity of that signal. strength =/= fidelity.
Since bigger pixels can capture more warm light, that's pretty significant. Since warm light is the least energetic, capturing more of it is a good thing toward getting an accurate signal. It also means that there's less "noise" with warm wavelength capture because when you filter for CFA light bleed (which is what EVERY camera secretly does before the RAW file is even written) you don't have to filter as much. Compare that to more energetic and more numerous blue light because of its smaller wavelength, and blue light is a big issue when it comes to CFA light bleed. It can much more dramatically bleed through the CFA's.
And since the scene you're shooting is dependent upon many factors, the camera has no knowledge of this. So it can filter out baselevel light bleed, but it doesn't filter out of something is extra blue in a scene, while striking a blue or green pixel. If that green pixel has extra light bleed on the blue spectrum, and what's being captured is blue, you end up with lots of light being added to the pixels luminance value, while also doing it on a pixel not meant for that color. Since you don't know exactly how much blue brightness was actually there compared to other colors (and even if you could, you'd end up with a very noisy signal because you'd only have the blue light bleed to make a signal from), so what's happening is fidelity is being reduce in exchange for signal. You still get an image of course, and color can be adjusted/subjective, but those camera's people usually praise as being able to just spit out a JPEG in camera with zero to no edits, tend to also be high CRI camera's. The colors are better by default. Garbage in, garbage out.
A lot of people talk badly about Sony's skin tones. And the reason has less to do with color science, and more to do with CFA's. The skin tones suck because the CFA is bleeding extra light into pixels. So you end up with hue shifts. And since we're talking about 3 pixel spectrums, it's a bit of a jumble. You're not just getting a little extra green or red on the blue, you're also getting extra on all the other pixels too, which ultimately causes false variations in pixel luminance, which doesn't just affect fidelity, but also amplifies noise.
@@peoplez129 There are many, many factors in a camera's color reproduction. I don't really need to look up SLR camera results from two different companies using different image pipelines and tuning preferences, as it won't tell me what's going on at the pixel level. For that I can rely on my own industry experience having worked on both CMOS image sensors and cameras for over 25 years. This is what I do for a living.
Yes, "less samples of light" means more color error at a particular pixel, but the white balance algorithm is looking at resolution invariant color statistics, in other words the mean color for a large patch of pixels. So if you're averaging looking at patch of 100x100 large pixels, or 200x200 half-sized pixels, the result is the same. So long as those half-size pixels are scaled linearly.
There are a lot of myths around CFAs and color reproduction. At Aptina we developed a RYYB pixels. The yellow was a weak yellow (so this was part of a RCCB development). We proved both through analysis and real world testing that we could get the same color reproduction as a traditional Bayer.
@@JonS The CRI indexes I speak of were 3rd party standardized testing, not some manufacturer claim. If it can have a better CRI index despite worse noise, then that's orders of magnitude better, because SNR itself is not just dealing in double digit percentages, but just a few single secondary numbers of difference in SNR ratio means doubling of noise. Which is just a testament to how good CFA's used to be made. Camera's with newer CFA's only have decent looking colors because they technically come with the saturation cranked, baked into the RAW file itself. People like to think RAW files are RAW files, but in reality they're heavily manipulated in camera before they're written as RAW files. I mean just look at Sony's star eater problem, where even in the RAW files, astrophotography results in stars effectively being subtracted out of the final RAW image as noise.
What a quality video. Getting down into the device level physics and building up from there with real-world context. Very well done for only 13 minutes, thank you.
THIS is why I subscribed to Engadget. REAL TECH JOURNALISM!
Same here
Only this guy*
only for UPSCALE
Exactly! How long have you been watching them?
My man is back. 😎
Get us some more processor and storage stuff too bro.
That’s the first time I actually understood how a camera sensor works. Thank you!
This presenter, and this series, is SO GOOD in every single way! I always learn something new, even if I _feel_ like I know everything about the subject at hand!
He way over does the enthusiasm routine.
I've owned both the a1 and r5 and I stuck with the a1 because if you shoot high bursts the workflow to quickly pic the perfect shot and delete the rest in camera makes a huge difference in storage. The a1 to me is just easier to get the pics I want the fastest way however the r5 is obviously no slouch. Good luck in choosing
I’m still sticking with Fuji. The only thing they need to improve on is focus speed and accuracy. I like these cameras because they’re smaller and lighter and APC takes good quality images.
To me it’s not all about the sensor, it’s about how you see your subject. People can talk on all the specs they want, but in the end it’s what you can do on the other side of the camera that counts.
@Ecko Eckard
Obviously you’re mistaken, as other people agree with me.
@@benjones8977 I agree with your first comment but it doesn't mean that it's true for everyone and everytime, there are alot of people who'll disagree as well but that's fine, I buy cameras for myself not for others.
@@reenachauriya4931
I was just giving a smart come back, nothing personal. 😘
Ironically I left my X-H1 for a Nikon Z6. Couldn't be more happier with the change
@@chickenpasta7359 why is that? My big worry about Nikon z system is the size of the lenses - seems like they've made no effort to keep them compact. Sure, light gathering is better on FX, but at the cost of a lot of weight. even Canon made an effort. What are your thoughts? perhaps Z6 is just newer and faster/ better IBIS?
How and why is this video so much better than every single video mainstream tech journalists have ever made? The research is so deep and well understood, the video was a large chunk of a degree semester (speaking from experience) delivered in 15 mins
I’ve recently bought an a7R III and I absolutely love it, I would go with the a1
0:22 WHOA I was not expecting a Foil Arms & Hog cameo in an Engadget video!
Same
Now this is what I call good content !
Keep up the good work guys !
This is super informative, thank you and your team for making it!
this is exactly how video should be made, bottom up so you don't get caught in the marketing and buy stuff that is actually meaningful for your task. Well done
Concise but compete. Thank you for making this content.
I love how in depth this is
This is a really well done video. Learned a lot about the inner workings of sensor tech.
Let's also not forget that we currently just filter away 2/3 of the light with Bayer filters. Doing something more clever there (say Foveon or microprisms) would instantly net you a threefold increase in photons hitting your sensor.
Cool idea!!!
The challenge is finding the something clever that doesn't introduce its own problems (ie: noise and color separation hurdles)
@@richardbutler8532 sure, I'm just saying before we admit defeat to physics because there's no way around shot noise, let's not throw away most of the photons.
i agree only the light and scene matters. using your rain bucket example, if i can multiply the rain i would gain more water in the bucket, similarly if i can add more light into the scene, the noise would not be noticeable. cameras have gotten really good so much that cameras can see more in the dark than many people. so at this stage its far reaching to worry about noise. but definitely camera progress is good. i'm generally more impressed about phone cameras and how they produce great image comparable to big sensor cameras. thats where the real innovation is. that's why people rather spend $1000 on the latest phone than $1000 on the latest camera.
i also agree the lenses matter, if you have recent apsc or full frame, with f1.4 or f1.8 lens, that really all anyone needs, regarding noise.
good video!
Amazing research! Thank you all for this video.
Great video very accurate! Blows away all the photography channels in terms of accuracy and sheer quantity of valuable information.
Seeing as these videos are so popular, it makes you wonder why no one else is trying to get a piece of the ‘really deep dive into chip/tech design explanation’ videos
this video deserve more likes, I can tell there's a a lot of researching went into making this video.
Substantial & Detailed Education To Us! Thank You.
Great explanation! That said - technology isn’t everything when choosing a camera.
Lens selection, ergonomics, and the camera “feel” is also a factor.
I would personally chose according to what camera makes you enjoy using it more - the more you enjoy it, the more you are likely to use it and get value from it. Have fun!
Yes, and not having a terrible UI on an otherwise amazing camera (I'm looking at you Sony)!
i'd rather use something i don't enjoy but get the result i want. i'm all about the result. if i was buying the camera for personal fun, then i would consider which i enjoy.
@@truthseeker6804 unless you are a professional sports/ wildlife action photographer, If you can't get the desired result with any higher end modern camera, it's on you, not the camera.
Cameras today are like phones, they are all perfectly capable, it just comes down to ui/feel preference. (Again, unless you absolutely need 250k iso or 15+fps bursts)
That's why I don't bother with digital. For some reason digital cameras are all cheap plastic unless you spend a ton. And even when you spend a ton, it's all menu diving to adjust stuff.
very nice presentation. would love to see a similar video on different lens motors being used currently
Great explanation. You’re very knowledgeable.
Congratulatios, great work. Thank you.
I love this guy. Give him more content!
This is so refreshing! Proper sciency stuff! None of this UA-cam recycled content
I just wish any major company made a CCD sensor camera now just for photography. CCD colors were something and I still love them.
This video is like an entire semester on camera sensor technology! 😍
I saw the thumbnail and i thought this might be interesting. I saw the presenter and i know it will be interesting.
Excellent summary, with the right level of detail and technical thrown in and all nicely explained and presented.
Camera suggestion - get what you need, not what you want - apart from lenses - they have settled down tech wise and hold their value better than the camera attached. Which is why camera's of old got driven by the lens and the film - today, your camera is locked into what film chip you get when you buy it. So, do you need the latest and greatest or is year gone flagship suitable for your media needs.
The was unnecessarily geeky!
Loved 🥰 it! Bravo
Literally what dances through my head and a huge part of why I enjoyed image making aside from the artistic elements.
Great video. More from this guy please.
Heyyyyyyy new upscaled!!
Outstanding content!
If anyone with a struggling budget is looking for a professional camera, I would absolutely recommend a used Canon 5D Mark II
The first things people think of when considering what factor matters most in quality images are (as was mentioned in this video) the camera, the lighting, and the sensor. But the one thing less often considered that can really make the difference is post processing skills. You can't do anything to improve your existing hardware, but you can improve your pics if you know how to do it. (Hint: shoot raw & jpg, even if you haven't acquired your skills yet). I've got some old pics from a Rebel XT (8 mp, low dynamic range compared to today's sensors), I can make them look gorgeous. But I've been post processing since 1998 or so, and I've worked hard at improving. It's a hard won skill, but to me, very much worth it. Especially when it comes to pics of little kids who are now adults, and people who've passed away. All you have are those old pics, so being able to make them look much better is a valuable skill.
This is excellent. More of this stuff, Engadget!!
Cool series though, keep that up!
I'm starting a _GoFundMe_ to buy this guy a straight-jacket.
Terrific piece though.
Excellent video.
I mean I hate to be that guy but, it really is all about the lighting haha. I really feel that camera sensors are at the point where they are so good that they far exceed most people's need or capabilities. So you're better to decide what gear you like based more on user experience than pure image quality. Image quality is basically universally amazing across all brands now so think about lenses and comfort more than the sensor. If pure image quality is all you want it's still hard to beat large format film haha. You'll just never get sensors that are 8 inches by 11 inches in size and that large of a sensor comes with so many benefits. There's a good reason large format is still the standard format for many landscape photographers.
With the right lighting setup, you can make any shot look great regardless.
While everyone has a CMOS, I have a Foveon X3 for the Sigma SDQ H. It’s a powerhouse sensor. Pros: Image quality is sharp. Color vibrancy is film like, little to no post processing added. Infrared photography is incredibly beautiful. The only cons that I can think of is the battery life, drains to 50% after 50 captures, it’s a one ISO camera, anything passed 800, the noise is visible, and photography only, no video recording.
Good work 👏👏
really good video!
“You can’t just make bigger sensors, A big slab of Silicon is expensive”
Film Photographers: “I don’t have such weaknesses”
27 seconds into the video and I already like it! So true...
Agree these are great vids. Love this series!! Light is important tho. Marques just had a car reviewer on his channel but then I tried his videos. Sorry Doug there’s shadows everywhere. Get some big cheap lights. It was bad.
Great video. I think both the Sony and Canon are great cameras, can't go wrong with either one
I really enjoy Fire Marshall Bill’s tech videos.
Great video! So much info in such a hurry, each of these videos are an hour session in covered in just a few minutes. Regarding DGO; Dual gain output you have on Arri (ON Semiconductor) as you mention, also BMPCC Original (Fairchild). (All way ahead of Canon as you mentioned). Sony also has a Dual Gain Output under the Starvis 2 security sensor branding. So eh, basically all the sensor design houses have DGO. Would be interesting to be a fly on the wall over at Sony and learning their reasons not to go DGO on their full frame sensors. Maybe DGO matters more to small sensor designs or is just easier to produce on smaller chips.
Useful video. Although he speaks very fast,I learned what the BSI-CMOS sensor means in my Nikon Coolpix P1000 camera. Also it gave me an explanation of the difference with my old Sony DSC-R1 camera. I was wondering why both cameras performs well at about ISO800 maximum. Cause my old camera has a much larger sensor (APS-C). Maybe the explanation is the BSI difference and better chip for the Nikon. It also triggered me to use the burst mode in my Nikon to see if it will perform better in low light. I hope that I can shoot RAW and burst mode simultanously,but all this I'll check it in the manual. Thank you Engadget.
I love your Upscaled series, so thank you again. I also get that bayered CMOS' still get the love because of the dynamic range. Wouldn't a multi-layered CMOS get better detail and color with just less light on the censor like the old Sony F35 and Foveon X3 sensor (yeah, I'm old). It emulates film and has an upscaleability I've not seen since REALLY good film scans. Again, thank you for your series.
I don’t know. They keep saying only use foveon from 100-200 or It will be noisy
This explains the the Canon 6D 1st gen is still one of the best choices for astrophotography. Improvements in the last few years hasn't been all that important for a single slow exposure
That camera is still a king to me. I have a Sony a7iii as well and still am disappointed to this day of buying it as an upgrade upon release
I love this guy!
R5 for sure, love it
Enjoyed the video. I recently bought an R6 and honestly for what I do it works well. If I ever do need more megapixels for cropping I have gotten good results from Topaz Gigapixel and I don’t have to store a ton of huge images. I do a fair amount of bird photography and the eye AF is magical. I paid $2400 for the R6 and the R5 is $3800 and the A1 is $6500. I prefer the Canon menu system but I love Sony’s 200-600 lens. Unfortunately to get great AF performance I would have to buy the A1 so the nice price of the lens is overwhelmed by the price of the body. I will stick with the R6 for a while. I was a wedding photographer in a previous life and this is the perfect camera for that. I took family picks in Maui last week at night with only incidental lighting with the cheap RF 1.8 50mm. Shots at 25600 ISO were nice. Even a few at max ISO were useable with some noise reduction. None were out of focus. I am far from a Canon fanboy as a mostly Sony and MFT shooter in the last decade but the R6 has made me a believer.
Second hand A9's go for just over £2000, which is a pretty good deal giving a useable 20fps.
the charisma in yout top half got me subd
8:33 computational photography is one but another reason that smartphone photos improved is simply because the sensor size got bigger. iPhone 13 pro’s sensor size is about 4 times as large physically as the sensor on iPhone 5. Also the benefit of BSI on low light performance is actually very limited on larger sensors such as APS-C and full frame. Fujifilm used BSI sensor in their X-T3 however the truth is noise at high ISO actually increased slightly (not saying BSI is the culprit here, there are other reasons for this) compared to X-T2 which used a non BSI sensor.
This man is amazing.
I love my A7C mirrorless, but it's ridiculously hard to keep dust off the sensor unless you never change lenses.
I love this type in depth explanation. But I'll never forget the best camera is the one that you have ON YOU.
Great explanation. Could you please stop with the subtle/not so subtle zooms and pans every time dude is on camera? I understand that changing the focal length is a good attention technique, but this is a little much.
Especially given his constant, rapid movement already attracts attention.
Lighting is really important, but a good camera look good in most lighting
I'm curious to see what can that Phase One can do! 🎥
Wish you would have covered pixel binning/charge combining and the increasingly smaller pixel sizes in the latest 108mp and even 200mp phone camera sensors of the last couple years by Samsung and Sony. I still don't know if it's a gimmick or if the whole "we can combine 9 pixels into one!" is a load of crap or not. They keep on making increasingly more and more of them and place them on flagship phones so it's not going away.
It's not this tech has been around for a couple decades.
I would watch this man explain anything.
Definitely go for the Phase... or for a more ludicrous/insane one, the Alpa 12C I think? :)
You forgot Panasonic’s fabled organic photodiode on CMOS sensor which could boost dynamic range tremendously. It has challenges though.
Watching this while taking back shots 😎
Improving fast focus mechanism on larger sensors can also be next, like some kind of focus tunable lenses (liquid lens) paired with a curved IMAX size sensor .
How about making pixels which when full automatically empty out and get back to collecting and add the collected value to the end value?
Then the end values of each pixel are read out and converted into an image.
This would need to take into account the drain speed of each pixel though...
But in total this workaround could result in basically an infinite bucket size. (of course not literally infinite lol)
Yep, that concept has also been known for a while, but no such sensors yet. Another alternative is making sensors that fill their pixels full well depth nonlinearly, many do that already.
This can be achieved with software actually. First you shoot an image at low exposure, thus the buckets are unlikely to be filled up, then you shoot it at a high exposure, to fill in the noisy blacks. Then you stack the 2 together. You can do that right now with many DSLRs with multi exposure shots (and stack them in post).
@@ginsederp yes, but an improvement to get this with a single take would still be an improvement
At 4:52 I just want to make a minor correction, you do not boost the ISO when brightening a scene, in fact the ISO is just a stored value in your raw image. Compression is a major factor (and yes raw images are still compressed) but taking a underexposed image at say ISO 800 and correcting it is going to be just as noisy as taking a photo at ISO 3200 that is properly exposed. If you shoot video with a Red or Arri camera you can dictate the DB/ISO of a raw photo sequence after the fact for this reason.
How the camera makes portrait effects i.e why the camera blur the out of focus i.e losing details in unfocused areas
That’s aperture and distance from subject and longer focal length
this was a very fun video, also id choose the r5 the ergonomics seem better and the body just looks nicer . the A1 seems very overkill if you're not shooting high action sports and you could use the $2.6K you have left for lenses and other goodies if you choose the R5
Quite informative video. I think future sensors will be having only Electronic Shutters. Also I would like to know how a BSI Monochrome sensor or like does a monochrome sensor need BSI ? & a BSI Foveon sensor will be :) Thanks :)
Google has shown camera sensors only matter up until a certain point. The most important thing for better photos is the computational algorithm and ai
Let me be another one of those it doesn't matter folks, but not in that same way. What i mean is, for most applications, the different cameras in the same range are so close, people looking at your image won't notice a difference. Honestly would look past the specs and for the things that many people don't consider.
1) i've been telling people for years, if you have a good friend that uses one of those systems, get the same one. the advantage is that they might be able to offer you gear specific tips or maybe even let you try out their gear before you buy that new lens.
2) how nice is it to use? Like which one is nicer to hold in the hand for a long time? which one's menus just feel more intuitive to you? etal.
3) Are there any specific types of shots you want that said camera or brand had specific equipment that make it easier to get the shot?
Let's be real its not the camera that matters, after a point it's the photographer's skills.
That's why I gave away my mirrorless to adopt a good phone camera instead. Least conspicuous setup and therefore optimal for street and documentary work.
Buy the A1, I’ve owned it for a few months and I love it.
Oh, and iron the short bro!
Its actually true that only light and background matters. ( And your art sense of course)
More often than not, lower resolution adds to the beauty.
Someone drinks too much coffe maybe? Nice video!! Thanks!
This is kinda interesting since I saw a few videos in the last weeks popping up saying that bigger pixels do not help with collecting more light and stuff like this. Which imo is complete nonsense because if this has nothing to do with it, why bother having bigger sensors and bigger pixels in the first place? 🤦🏻♂️
I have a lot of new cameras but can anyone explain why a Fujifilm from 6 years ago just takes the best pictures and video
Fujifilms engineers are also photographers, so they have a better understanding of what the end goal is than most of the engineers at other companies. I feel like the engineers at other companies treat the camera output as the equivalent of the display settings on a tv. Crank up the brightness and saturation so it catches your eye more than the other brands when looking at them in a store. People who care about what the output looks like will edit the raws, or so they think.
Fujifilm seems to put a lot of effort into making the images look good straight out of camera. They have actual useful gamma and contrast adjust settings in camera, and they do a lot of testing to make the presets looks good.
No discussion of improved Analog to digitall converters or preamplifier and dual gain/ISO. Your dynamic range IS the number of bits you get from AtoD converter. 12 bits men's maximum 12 stops, 14 bits is maximum 14 stops. Issue is larger you make AtoD converter the slower the conversion and the cost of AtoD doubles with each adiditional bit. Likewise the better the preamp is the more you have to worry about noise increasing complexity and cost.
thx bro
Something I’ve been wondering for a while but nobody has directly addressed: Can advanced enough sufficiencies in a lens make up for insufficiencies in a sensor? Could a huge lens make a tiny sensor as good as a decent lens with a decent sensor?
Yes, a larger lens will gather more light and provide a better resolution on a tiny sensor, with lower distortion, chromic aberrations, and vignette. Adding gyroscopes to the lens group to keep it stable will also enhance image stabilization and quality and sharpness. Also the larger the lens, the less noise the sensor will develop because of the additional light. And lets not forget about lens coatings, that can alter the spectrum of light that hits the sensor so that the sensor can uses a more efficient Bayor filter so more light passes through the filter that's on the sensor and further eliminating noise (don't quote me on that guess...lol). But this all depends on the focal length you choose for a lens as well and could make all the difference. If you could add a large lens to a smart phone, what would definitely be a game changer. Or perhaps, an interchangeable lens smart phone case. Just my thought!!! Great Question!!!! Cheers!!!
I just discovered your channel and I love it. Sony pioneered the mirrorless camera and they better low noise performance. I was always a Nikon fan but Sony won me over. Your decision may rest with the lenses you have.
Forget all the cameras you've been looking at. You should get a spanking new DJI Ronin 4D (6K or 8K models) and enjoy the equipment for about 3-4 years and extraxt the stills from the captured video clips while the other OEMs try to play catch up to the output of the DJI sensor. During this time DJI will obviously move forward as well, so my guess is it will take the other OEMs even longer to only barely catch up, not to mention ever again moving ahead anytime soon, though. With the Ronin, it seems still photography might just have become nothing more than a side show of ultra hi-res, stabilized, LIDAR-focussed, RAW based videography. While the shift from still photography equipment to videography focussed equipment has only just begun, the Ronin has just shifted that final transition into high-gear.
tbh im a canon shooter and i have the r6 its really good for doing basically almost any job except for long form video.
but if i had the cash on me? i'd go with the a1 really. that camera has it all for real
the megapixels, the readout speed, the dynamic range, the video specs etc. its basically the best tech this generation of consumer mirroless cameras has. idk if i should call it consumer but yeah, im not talking about something liek the phase one with its 40-50k price tag lol.
Can any explain why all, absolutely all, America tech orators raise the tone of the voice and octave or two in the middle of a sentence? or just in a sort of sinusoidal pattern in the whole presentation.
Global electronic shutter on CMOS would be the GOAT! I still have a Nikon D70s that does it on it's 6MP APS-C CCD but since then nothing has come out to match it. Sync with flash at any speed at any aperture. Not even leaf shutters would be able to do what that ancient plastic camera could do. ... Kinda sad.
If you're actually looking at the A1 maybe take a look at the Leica SL2 or SL2-S as well. In terms of native lens quality ... that's the long game.
If you're deciding for photos, it's a tough choice between the a1 and r5 and it'll come down to personal preference/workflow. But if you're going for video the a1 is the clear winner. Video auto focus is better, noise and low light performance demolishes the r5, and you get sony's s-cinetone from their cinema cameras as well as slog3. Oh and it doesn't overheat ever.
If you have the point spread function of the lens you can deconvolve the image and get a resolution better than the lens is otherwise capable of.
I'm on a waiting list for the Nikon Z9, it should basically be an improved Sony A1