Pixel Density vs Noise: Do Higher Megapixels make Noisier Pictures?
Talent matters: Model shooting with an eight(!) year old Sony A-mount camera at SonyAlphaForum.
Sony Showdown: A7II vs. A7R vs. A7S (Adorama Blog).
Optical Advantages of Curved Image Sensor (Image Sensor World).
Pentax FA 43mm F1.9 on Sony A7 (FocusPeaking).
Sony to Sponsor PAGB Inter Federation Cup (Photographyblog).
IISW 2015 Program: All-Pixel PDAF, Global Shutter Stacked Sensors, and More (Image Sensor World).
Sony A7II + Minolta APO 100-300mm (SonyAlphaForum).
Animation to show how CMOS and CCD image sensors work. This animation was used in a trial against Canon Inc (Gizmodo).
Niklas: “I saw that you posted some videos and just wanted to share a video I shot entirely on the RX10 compact camera. t’s a great camera and I just wanted to prove that you don’t need an a7s to get a cinematic look. https://vimeo.com/98879187. Swedish midsummer in the archipelago, shot entirely on the rx10, enjoy.“
136 Responses
Uggg… I will never get those minutes back.
“If all sensors were the same size then we can see that … ” Bro. You lost me right there. Everything you said after that is already based on a flawed premise.
Its not flawed, he wanted to say that if, lets say A6000 was a full frame with unaltered pixel density, it would have over 50 mpix, and the level of noise would still drop down because of larger sensor area.
Good point. For images of the same final size, the differences are visibly irrelevant (as his samples show).
Same final (small) size.
This only applies to people who buy added MP without need. Whereas the rest of the world, expects to buy higher res. camera’s with the intent of using it. – To which I’d add, I have to agree with TheloniousGonzo. The entire presentation is based on flawed reasoning. Not to mention that the author seems completely oblivious to the reasons sites such as; Imaging Ready, DXO and DPReview and most every other, presents 1:1 images to show a camera performance rather than cramming everything at the same output size.
And so, in short, this only works for the noob who purchases high MP camera’s without just cause.
So you assume that the result would be different if the test was done with upscaling to the MP count of the high MP camera?
It wouldn’t. Upscaling increases visible noise at the same pixel magnification, so the relative noise difference between the cameras will be unchanged.
It can be added that the missing resolution of the low MP camera would be even more obvious, so this change of the test would actually favour the high MP camera.
Its like he had a point he wanted to make so he just ignored the stuff that didn’t help his point, assumed stuff that would, and then extrapolated the rest. His reasoning was so biased that I couldn’t take any of it seriously.
Well, that was clear from the first release tests. A7S is video monster though… Still, he doesnt explain why A7S has clearly better images on insane ISO levels (above 100000) than any other downscaled high megapixel camera, or to say, any other camera.
Good point. In fact, while he admits briefly that the A7S is flipping superior for video, in one of his blogs, he avoids the extremely high ISO comparison. I would like to see that test.
We need good counting statistics.
I thought D4s is what nikon will tell people to buy if they want a low light camera.
This guy is always such a tool.
There is a relationship between pixel density and noise, amongst sensors of equivalent tech, regardless of scaling. That’s why the A7S just happens to be 12MP and destroys everything else on high ISO tests. Do you think Sony just made it 12MP for a laugh?
Downsampling (AKA averaging noise across multiple pixes) does *help* with ISO performance of high MP cameras downsampled, but downsampling 3 pixels to 1 absolutely does not replicate the high ISO performance of a single pixel gathering all of that light, one of the biggest reasons for this is that it’s a bayer array, so you’re averaging different coloured pixels together.
Of course, if you do what Tony did and ignore sensor tech and just average *all* of the cameras from 2014 together regardless of manufacturer, scaled up or down to an arbitrary resolution, it gives the illusion that there’s no relationship. But then you’re completely ignoring the very important factor fact that there’s different sensor tech in a Nikon J1 vs a GX7 vs a D810 vs an A7s, vs a D3300 vs a 7D which is completely confounding your little experiment.
The downsampling shouldn’t work nearly as well as it does. But at least with the algorithms in Lightroom, it’s within the margin of “very difficult to tell the difference” in the kinds of photos shown in the video.
Downsampling or better call it downsizing does help to improve noise only if noise is not too high. So his logic is right for low noise situation as he showed the Dxo mark iso which is again measured for low noise.
But I agree with you that at high noise the pixels are already badly corrupted that you can not recover them with averaging. That is where innovative sensor design of A7s comes to play that removes the noise even at extreme low light at the first place where the image cannot be recovered by simple averaging had it been shot with a high pixel density sensor.
A7s don’t destroy any other full frame 12 Mp camera with its High ISO. Yes, it’s excellent for movie mode, but not for photography – D700 or D3s are at least as good.
The only reason the A7S is 12MP is so that Sony can sell you a 16MP Version in the future. After that sell you a 24MP version of it with roughly the same results. After that sell you a 36MP version of it with roughly the same results. Don’t forget all of the tweaks and form factor issues/ect. they “Improve” along the way with at least 2 versions staying the same MP. SO We will be on AT LEAST the A7S Mark6 before they produce a camera that produces high iso image quality on a 36MP camera so they make A LOT of money along the way when they could already produce it now if they wanted to.
This video has a lot of misinformation in it. It will take so much work to now un-learn all the folks who watched it. Glossing over nuanced issues with such broad generalizations to the point where the things you say are dead wrong on their face means it’s time to start over.
EDIT: It may be better to say it’s very misleading and the examples shown are not designed to prove the points he suggests they prove.
I actually bought this guy’s book awhile back. While reading it, I learned stuff, but at the same time I thought alot of the things he said in the book were incorrect, or based on personal opinion. So now I just take what this guy say with a grain of salt.
Well, why don’t you write a blog post on it. You write some good stuff- you can tackle TN head on.
I would much rather see Sony debunk that myth by making high MP cameras with low noise for a change!!!
I would definately buy that. I want the A7s, but the low megapixel tells me I want the A7r. RIght now I compromise with A7. I wouldn’t mind bu ying A72 just for the IBIS, but I wanna see whats coming out!
Have you heard of the A7R? It’s not quite at the level of the best Nikons, but it’s pretty damn good. Better than the A7 or A7II (unless you’re looking at pixel-level noise, which would be silly).
What I mean is affordable cameras. Sony still cannot compete when it comes to noise levels with other manufacturers’ offerings at the consumer to prosumer level. That is a serious shortcoming IMHO.
The A7R is pretty affordable for full frame, but there’s also the A6000. It’s near the top in resolution and noise as APS-C goes, and a lot cheaper than most.
I don’t want to see this dude, show me new Alpha cameras!!! It’s already the 15th in Australia…..
It’s not just the 15th, it’s 4p.m. in Sydney, I’m not feeling good about a launch!
The official public launch will be on May 20-21!
Woohoo! What SR is that? 5?
Public launch for what too, A7rII and A6100?
Really? Wheres the info for that? Its not on the front page yet.
I wonder why he didn’t compare the A7r to the A7s for photo quality…One thing he didn’t mention is that every camera manufacturer records RAW files differently. Nikon’s D810 NEF files are lossless while the Sony a7 series cameras record in an ARW lossy file. Meaning the Sony a7 cameras have more compression and less color information even though they are 14 bit files. In most daylight scenarios it’s a negligible difference but in low light or high contrast photography it’s definitely a handicap as the ARW files display more noise and color shifting artifacts.
Here’s a link for an illustration of the issue:
http://www.rawdigger.com/howtouse/sony-craw-arw2-posterization-detection
I thought he just compare between A7 cameras, so all will suffer the compression issue
I love my simple a900. My clients love it more.
Let’s stop with the uber duber super specs. I have magazine covers and fashion editorials published with the old 3mp Fuji S1. We reached the commercial quality peak years ago.
I’d be happier if ergonomics evolved as fast as sensor tech. Sony Emount is bottom of the totem pole for ergonomics… like underground. Start by removing those antiquated strap lugs that stick out like little knives. Then return to Minolta vertical grip, instead of current that allows my palm to engage VG shutter while in horizontal mode. Just awful ergonomics Sony Emount. Take a cue from Amount 25 year legacy that’s evolved to perfection.
Hmm, I like the A7/Nikon D800++ strap lugs, so what strap lugs would you recommend? The one like Leica T? now that’s unique!
Actually the strap lugs on original Nex 5 were much better for me. They were positioned so that the lens would always hang down when the camera is on your shoulder.
Strap lugs stick out because of the way Sony engineers tried to make camera smaller — it you disassemble camera, you’ll see.
SAR, please please please do not give this guy anymore publicity / click rates than he deserves. He’s already poisoned a heck lot of impressionable minds with his nonsense on equivalence, and now this…
Haha, you must be joking right?
The a7s is 12mp is because to avoid having the need to do line skipping in video
They can use the 36mp from his example for 8k though.
Ah yes, the 8k that everyone uses as a standard as of now, right? 4k is so 2014!
You need a 43+ Mpix 3:2 sensor, to be able to shoot 8k in 16:9. So until Sony releases a Canon 5Dsr competitor, can’t even speculate about 8k.
obsiously at some point Sony will implement full sensor readout in higher MP sensors
Great video for advertise to buy a Nikon D810 instead of the sony a7s. He said- he found not one picture that is in the a7s better than in the d810 (after re scaled).
Yes, higher mega pixels do make nosier pic… provided sensors fabrication technology is the same, the DSP behind is the same. But technology progress, there will be better material, better fabrication technology, better DSP, etc. to negate the nosier smaller pixel pitch.
I find the A7r noise is same as Nikon D7100.
Your findings are wrong. Signal-to-noise can be measured objectively, and the A7R is far superior to the D7100, as should be expected given its much larger sensor.
What matters most is how big sensor and how fancy A/D converters you can afford.
Comparing equal sized sensors, it is the type AD converters used that sets the high-resolution vs low-noise cameras apart, as the former has many small, simple and efficient converters whereas the latter a few big, complex and hot ones (with better pre-amplification).
Urrghhh… Tony Northrup… He has some weird televangelist quality that freaks me out. Urrgh…
I was going to say seeing him somewhere always creeps me out, but yes.
This CLAWN again!!!
SAR its becoming a fan of this guy?
i remember him for his stupid video on “equivalence”
MMMMM What i hate more?
– the big yellow N in the wall
– that he thinks 6400 its high iso (were are not in 2004 anymore)
– that he contradicted himself from the equivalence video
– that he cannot buy a damn “just for men”
– that he try to sell me his stuff at the end of the video
– his irish yellow/black shirt (seems this irish guy like more yellow than green)
im start getting green, oh better not calling the vanished hulk guy lol
No fan. Just reporting what others say about that matter. My position is neutral in Sony Tidbits posts. You discuss 🙂
Then you shouldn’t use a headline like “Tony Northrup debunks…” but say “Tony Northrup’s take on…” because now it looks like you agree with him and then you are not neutral anymore.
Totally agree.
tbh, he is just trying to hustle for living like every other youtube photographer, only that his book is alot cheaper and better written than, ahem, say FroKnowsPhoto…
What the hell…
Tony again! Nooooooooooooooo…..
I watched only one of his videos and couldn’t go after minute 2.
Please, Andrea, don’t put him up again.
It’s the hair isn’t it.
it’s the hair, the creepy smile and that voice…
http://img3.wikia.nocookie.net/__cb20121120073933/glee/images/8/8f/Stare-wtf-creepy-smile-8e6b42f1644ecb1327dc03ab345e618b-1.gif
He talks about density of pixels, but he doesn’t say that on a full frame sensors (at same MP count) the pixels are bigger so they gather more light so there’s less noise.
Then he brings up DxO (why????? I hate that site) and he says that if we wanna know the real perceived MP of the camera, we have to add a lens.
Maybe he should add that lenses have different building quality, therefore they perform different therefore the perceived MP varies.
It’s like saying that any old 50mm lens attached to a Nikon D810 or A7R, will give the same result of am Otus 50. C’mon, Tony, what the hell are you saying?!
What a shock! How can one make excellent photos with A300 or whatever 8 years old camera?
Sony early DSLR cameras were EXCELLENT, I think the greatest time was when the line A200-A300-A350-A700-A900 conquered Nikon and Canon, winning up to 30-40% of some DSLR markets. With built-in stabilization and great A-mount optics they offered the best quality/price factor.
One can still buy any of these and enjoy. They are definitely better to handle than comparable E-mount cameras and have superb color balance, if one can live without built-in video mode and crazy high ISO.
I don’t think we need to be so insulting towards Tony. He seems sincere to me.
But he is way wrong here. Saying that the A7ii noise performance is superior to the the A7s is something that I (and many others) can verify is nowhere near true. Even when images from both cameras are downscaled to the same web size, they are dramatically different. There is nothing close to the A7s’ noise perfomance. I’ve been using my A7s daily since I got it last July; upgrading from the A7. Even the mighty $6000 D4s’ noise performance is not on the same level as the A7s.
I can see how Tony’s videos can polarize people, especially after his older one on equivalence (which, as “hatetony” says below, contradicts his point about noise), but I second your appeal to tone down the replies.
One question: in moving from the A7 to the A7s do you ever miss the additional detail at low ISO?
Post proof then. Better : send them to him.
I shot the linked album almost entirely ABOVE 10,000 ISO with a Sony A7s. Most of these shots were actually shot above 20,000 ISO. Some up as much as 80,000 ISO
My A7 couldn’t come close to this performance. Neither can the A7r. I mostly shoot in very dark environments without flash. The A7s finally allows us to clearly see the world in the dark that all previous cameras were unable to capture without additional lighting. Allowing me to capture my favorite shots yet inside a very dark club without using flash.
(ISO sample images are in the first few comments after the post)
https://www.facebook.com/groups/SonyAlphaTalk/permalink/339434052913711/
A7s + Voigtlander 35mm f1.2
20,000 ISO
A7s + Voigtlander 35mm f1.2
40,000 ISO
A7s + Voigtlander 35mm f1.2
51,200 ISO
A7s + Voigtlander 35mm f1.2
80,000 ISO
A7s + Voigtlander 35mm f1.2
51,200 ISO
AND….. this is the image quality I was getting from the A7 when I pushed it all the way to 25,600 ISO. Also with the Voigtlander 35mm f1.2
The difference in the noise performance low light capability at high ISO levels between the A7 and the A7s is night and day.
What does it say, there is no pairwise comparison.
You converted your samples to BW to hid noise lol.
So then these are examples of noise performance in black and white. If you’re able to “hide noise” in post, when shooting above 20,000 ISO (!), that speaks very well of the sensor.
Also, these weren’t exactly converted to b&w, I shot them as jpgs in b&w. I never saved the color information, because yes high ISO color noise disturbs me, even though the A7s is still better at that than any other camera I’ve ever used.
Try getting shots of this quality in a very dark club at these ISO levels with any other camera.
So there is no ISO wheres A7S pictures are good and A7R’s are crap, is just at very high ISOs A7S pictures somewhat less crappy.
Lastly, do not click the images to open them in higher res. It’s not accurate. These were exported at 1200px on the long edge, but for some reason they open up in a new window much larger than that, introducing a lot of distortion noise that was never there.
They’re having an offly close conversation in this one.
Nah, it’s okay if it’s done for the camera. Aberration + camera = art
how dark is the room, why do you need to use 80,000 ISO?
I dig it.
Dude. Your noisy low res images are proof of nothing.
Take both of your cameras. Shoot at same iso, same lens, in raw. Downscale the A7 images at 12MP. Change nothing else. Post result.
I give your prop for having balls and posting your own content.
I don’t say you are wrong. But Tony shows proofs of what he is saying. Do the exact same thing as he did and show us the results.
The A7S comes into its own at very high ISOs (as the DXOmark results show). The really big distances occur beyond ISO6400. The gap really opens up at 12,800 ISO and beyond. It’s clear from quite a lot of analysis that Sony designed the sensor to have exceptionally low read noise at these extremely high ISOs. That means the dynamic range is much better at very high ISOs (perhaps as much as 2 stops towards the top end). The A7S sensor appears to have high ISO read noise of perhaps 1/3 of an electron charge, which is remarkable. (Unfortunately that isn’t echoed at low ISO where the noise floor is rather higher than cameras like the A7R, at least at equalised output, which is why the low ISO DR on the A7S is not of the best).
Also, his methodology of examining noise by photographing a white wall is highly misleading. What is seen there is largely shot noise, almost entirely defined by the sensor size, and not the characteristics of the photo-sensors for roughly equivalent quantum efficiency (QE). What really defines high ISO performance is the noise floor. That’s the noise in the dark areas of a photograph, as that is dominated by noise from the sensor, and not that inherent in the nature of light.
Also worth noting is that there is one inherent noise advantage to larger pixel cameras. That is, that whilst down-sizing shot noise (what is largely the mid-grey SNR number) produces exactly the same result, it is not true for the electronic noise introduced by by the sensor. Then multiple sets of read noise get combined (in quadrature). So the difference between the “large pixel count” and the “low pixel count” sensors will be seen with a lower DR at higher ISOs. Which is what we see on the A7S of course (albeit there are other design optimisations in play, so the issue isn’t just that).
This is not to say he’s entirely wrong. For very many photographers, the higher pixel camera is a better choice, as the A7S offers little improvement at typical ISOs and doesn’t give the same detail capture and ability to crop.
So, if you want the ultimate in really high ISO cameras, then the A7S offers a very significant advantage. However, it won’t suit everybody, and that video fails to point out where the real differences are.
nb. all comments are about still image. The A7S video capability is another thing altogether, largely because it allows for “full frame” capture without line-skipping so more of the sensor is used. Most higher MP sensors in still cameras throw away much of the information that arises at the sensor.
I agree, there is no image quality benefit to the A7s at low ISO levels. At high ISO levels the A7s is not just a leap in technology, this is some crazy magic voodoo stuff. I don’t know how Sony made this happen, but the A7s high ISO results consistently blow me away, night and day above all other cameras. It’s forever now between me and that sensor, until we get IBIS in an A7Sii 😉
This is the magic (or at least the result of it). Electronic noise of less than one electron at high ISOs.
http://www.sensorgen.info/SonyA7S.html
Given the EQ of the sensor, then that’s the equivalent of being wrong by about 1 photon.
In contrast, this is the Nikon D810
http://www.sensorgen.info/NikonD810.html
At higher ISO the read noise is around 1.7 electron charges (maybe 3-4 photons “wrong”), and that’s against a signal level one third as high (as each pixel is one third the area). So the DR (per pixels) is going to be about 9 times higher. Down sample, and that improves, but will still have a DR about 5-6 times worse than the A7S, or in the region of 2.5 stops. Which is about what DXOmark shows at the very highest ISOs. The 6.3EV of DR that Nikon gets at 51,200 ISO (actually abut 36,000 measured) is what the Sony achieves at about 200,000 ISO (doing a bit of interpolation between points). That’s around 2.5 stops, or a difference of approx 5.6:1 (2^2.5).
Meanwhile, on the 18% grey test (roughly like the white wall “experiment”), there’s much less different as read noise is a relatively small part of the whole.
So for those interested, the most important thing about a high ISO camera is how well it maintains DR. For equivalent size sensors, the 18% SNR is going to be rather similar provided EQ is about the same (modern sensors are reasonably close on that). Note that the exception is the SLT cameras as the translucent mirror filters out about 1/3 of the light which reduces the effective EQ by that much (so compare an SLT and non-SLT camera of the same sensor size, and there will be a significant difference in 18% SNR.
Yeah, Heaven forbid Sony trying to make a profit, eh? They should make cameras for free!
Well I always see the Tony video. They are in my opinion ones of the best… [Who said Chelsea is pretty ?] …
About this video, yes he’s wrong about ISO 6400 saying it is high ISO.
But think about this :
A 36Mpx (I don’t care if the RAW is lossless or not) take a shoot. It “absorb” light, and make a picture. When downscalled, we all know that the noise will be less visible since the software will make average pixel color.
In every case the same amount of Light is “absorbed” by the 12Mpx and the 36Mpx. The picture should be almost the same at 12Mpx AND low ISO (under 1600).
The power of A7s and the 12Mpx sensor is that, it can take shoot with ISO 492 000 with high dynamic range… higher than any other camera. It will be easilly seen at ISO 51 000. In fact, A7R/D810 will have a crappy dynamic range at this ISO. Downscalling the 36Mpx picture will result of crappy image with low noise but color that are messed up etc…
I see lots of ppl that bought the A7s are complaining… but did they bought the camera to shoot 12Mpx in ISO100 ? ? ? They did for going further than ISO 25 000 and they did well.
I like the way this guy is allways finding a way to show that Sony is not as good as other brands. He forget in this video that you can get a lot better low light video from the A7s then from any other camera in this class.
I like the way he uses the DXO findings just to his liking, telling everyone that a given nmber must be calculated in a way that Nikon or Canon is looking better.
I like the way how in the end he always manage to make the video into an advertising camaign for his books.
So This video is the same old, the same old…. A boring TN video…
If I have a cheeseburger in one hand and a chicken sandwich in the other, and I take out the beef in the cheeseburger and replace it with chicken, the cheeseburger is just as healthy! See that.
I’ve watched his videos for a long time. Most times I thought he makes valid points and I liked the way he showes relevant proof and examples. However, this time I felt like he is not like he used to be. I’m not sure if he was kinda in hurry, or he is just another c/n advocate, or maybe he just made a mistake. Anyways, I’d like to say this test is flawed. First off, he didn’t consider effective ISO. I mean, ISO 6400 of camera A is not necessarily the same as ISO 6400 of camera B. So something like one third stop of difference may be caused by different effective ISO. Second, he didn’t compare 5-digit and 6-digit ISO saying that some cameras don’t have that exact ISO besides A7S. However I remember he did a video capability comparison using Virtual ISO of ONE MILLION (seven digits) not so long ago. Many prior reviews showed that A7S really shines when it comes to high ISO, e.g., 12800 and above. I wonder if he never read those reviews.
This guy always makes my head hurt 🙁
Bigger pixels mean more light / pixel. Noise has nothing to do with pixel size, but signal has, so SNR can be higher with bigger pixels. The reason we can push resolution higher and higher is because we decrease noise with each generation, so SNR won’t suffer.
Yeah, it’s supposed to hurt when your beliefs are questioned. It what you wrote were true, someone would made e.g. 2 megapixel APS-C sensor for very high ISOs too. They don’t do it — because there will be no major improvement.
They do it for security cameras. For photography 2MP would be useless.
Btw, here’s your 2MP low light camera:
https://www.youtube.com/watch?v=eHnUtmenGQ0
2 MP is perfectly enough for web and much of photojournalism.
So does this 2 mp sensor show any improvement in high ISO?
Did you watch the video? Looks better than any DSLR currently on the market.
It looks better than any DSLR in the market, yes, but this is not in market….so, in the end, it looks like nothing any of us can use.
>but this is not in market
And then what?
The ISO in the video is approx 819200. The 2MP cam was a test camera, but the organic, the apcs and the quantumfilm sensor can good possibly use this high iso with 12MP and Global Shutter and less noise. Also you don’t must wait for this test camera, but rather for a 12MP camera with a new sensor technology.
2 MP means color fidelity and dynamic range are slightly less and there’s no room to crop. Also keep in mind that Bayer filters are a thing, so that 2 MP isn’t really 2 MP qualitatively.
botttom line, it would not create major improvement in high ISO
Someone knows a free software that allows you to downsample/upsample images?
Most of commenters here are ignorant of that MOST OF NOISE comes from photon shot noise, not pixel architecture.
Unless we are talking about ultra-high ISOs where pictures are unusable anyway, lower MP counts don’t provide any improvement in SNR. It’s not like A7S 3200 ISO looks like A7R 1000 ISO.
Lower mp sensors always have benefit of faster readout, through.
He’s talking about measuring noise in print, higher mpx getting downsized will always result in cleaner file because of combining pixels. So, in essence he is saying whether its A7S or D810, it makes little difference on an 8 by 10 print. Noise only makes a difference if we pixel peep and view files at 1:1 for these two cameras. Simple idea guys, no need to get worked up.
First of all, he’s saying a lot more than that. He’s reaching conclusions based on flawed methods. Secondly, “little difference at 8×10 print” is true at relatively low ISOs, let’s see what he gets out of an ISO 102,400 photo on both… er wait, the D810 doesn’t go that thigh, make it 51,200.
Why on earth do you need 102,400 ISO? You know everything above native is just software increases right? Read up: http://blog.robertscamera.com/2012/06/iso-native-vs-extended/
Yeah, and? 100-102,400 is native on the A7s, outside of that is extended.
Keep flaunting that awesome 102400 native iso figure, keep shooting with it, its going to make you the best photographer in the world!
Lol, we’re in a technical discussion of high ISO performance, don’t get butthurt and start talking about how it makes no difference!
Everybody is knocking on this guy like SAR is promoting him. Yet, no one said anything about the angry photographer being introduced on here yesterday. That guy is hilarious!…and probably nuts!
If you slow a Ferrari down to 20mph (downscaling), you can be moving at the same speed as every other car ever produced. Just kidding! I got what Tony was trying to prove, but why compare images at ISO6400? You don’t test the King of Low Light at ISO6400; there are plenty of other lowly cameras for that particular test. Just as you won’t say, “Hey, let’s drive this Ferrari at 60mph, let’s see what this baby can do!” 🙂
“8 year old camera……” What ?! With a press release from Sony jan 2008 don´t give 8 year, does it ?
Wow, I read through all the comments thinking somewhere in here someone is likely to have posted a link to sample images shot on an A7R and A7S that debunk Tony’s video.
But I couldn’t find such a link.
I used to be a Tony-hater, because something about the way he explains stuff always seems a LITTLE OFF. But now I focus on his testing methodology and the results, rather than his summary and rationalizations, and find his videos very useful.
From this video I learned that for images at normal sizes, the two Sony cameras end up with the same amount of noise visible. Compare video output, they are very different. Crop a whole bunch or enlarge to poster size, and they are probably very different.
What I’d like to see is a super high ISO comparison — in a range beyond what the A7R is capable of (i.e., just push it in PP). I’d guess that the A7S wins, but the only examples I can find online at at 1/1 pixels, which is not the same test.
Here you go:
http://www.dpreview.com/forums/post/54031579
Thanks for posting. At 25k ISO, it’s about the same. At 50k a sharp eye can spot a few details. At 100k, it’s getting obvious, though within the realm of correction in PP. And at 200k ISO it’s probably game over.
So, that test at DPR basically confirms Tony’s testing results but with the additional data points shows that Tony’s conclusions don’t hold up above 50k ISO — at least with the lens used.
http://horshack.smugmug.com/photos/i-mTFMHrZ/0/O/i-mTFMHrZ.jpg
http://horshack.smugmug.com/photos/i-7WKVkxB/0/O/i-7WKVkxB.jpg
The differences between the A7r and A7s start to become visible in the shadows at ISO 12,800. As you increase ISO further what were the shadows become the midtones, making the differences more visible across the full tonal range.
Okay, clearly my eyes are not as good as yours. In those two samples, I am not seeing the noise increase that early.
Looking at 5Ds files, I would say its actually pretty ok.
You are right Rhys, I’d like to see the comparison at 25600 52000 iso, this is an unfair test.
Just posted this as comment in Tony’s video:
——-
For equivalent-generation technology sensors, a lower density sensor will have better High ISO Dynamic Range vs a higher density sensor. This translates into lower shadow noise at high ISO, which will be most visible in images dominated by dark tones such as night scenes. The amount of noise in the midtones and highlights will be the same. As you move up to ultra High ISOs, what were the shadows at lower ISOs become the midtones at high ISOs, and the overall noise advantage of the lower density sensor becomes more obvious. Here is an A7s vs A7r comparison I did that helps demonstrate this, using the same absolute exposures, lighting, and lens, and normalized to the same resolution:
http://www.dpreview.com/forums/post/54031579
Here is the technical explanation behind this:
http://www.dpreview.com/forums/post/55278918
http://www.dpreview.com/forums/post/55280773
Lastly, for tripod-mounted High ISO comparisons, a higher density sensor will have an unfair testing advantage over a lower density sensor owing to the nature of real-world, shutter-speed limited shooting scenarios. I discuss this in more detail here:
http://www.dpreview.com/forums/post/54024514
Thanks. That’s a very interesting comparison. What I see is that up to 25,600 the A7R is better then the A7S in highlights, but the A7S is winning in shadows.
Yup its not an accident Tony shot only white walls and well lighted floors
I am sure he would make a follow up video to address the legitimate comments. Let’s wait till then.
OK, the A6000 is indeed up there with the best. I was thinking of SLTs versus SLRs more than mirrorless cameras.
When you have to ignore trusted tests and make your own to prove your point.. you pretty much failed. Scaling etc does level the noise quite a bit. But sensor pixel are more than.. light gathering surface. They are circuits.. more pixels more circuits. Smaller pixels less signal vs noise in circuits.. When they create noise free circuits and sensors that don’t have random noise this all might be true. But until then given the same generation and same basic technology smaller pixels have lower SNR thus more noise.. Which is why the D810 stops as ISO 51600 and A72 goes to 400K .. Because as the signal goes down the smaller pixels fall behind faster.
Sorry to be picky, but Mr. Northup is a little to the party! It has been well established for some years now that more pixels does not necessarily mean more noise.
It’s just that some will always refuse to believe!
This is a lecture well targeted at Sony users since it is using technical lectures to illustrate the difference. I think the advantage of FF is color depth and dynamic range, especially under low-light condition. But this is adv over APSC. He needs to bash Nikon D810. So the only thing to pick about is noise. However, if you ask me to do it, I will simply illustrate how A7s could change people’s experience of photography. Fujifilm did a great job here.
A7S is centainly nowhere like the rightmost one.
I want googleglass-like camera
I want an A7s without viewfinder (but with IBIS) that tethers responsively wirelessly to a google glass like evf. That would be the ultimate!
haha, what I really want is direct reading out images from my eyes.
You are missing a picture all the way on the right. It must have cut off. It would have been a 8×9 LF field camera.
This lecture is useless to me. I look at this issue in term of photon counting. If you get larger detector size, typically you get better counting statistics. But the type of detectors and function you applied to analysis the data can yield a big difference. So scientifically, this lecture basically tells nothing. But commercially, it busted a myth.
I stopped watching and disliked around the “Sony lenses tend not to be as sharp” spot, which as most of us here know is absolute BS. He’s also forgetting when showing his “No correlation based on pixel pitch” graph to compensate for 1 stop of light when comparing APS-C to Full Frame since the surface area of APS-C is half the size, ISO 100 on crop is equivalent to ISO 200 on Full Frame – adjust for that and things may look more level.. and there’s also something to be said for sensors based on technology generations.
What annoys me the most is all the people that will watch this drivel and be driven back to “Nikon lenses are better than Sony’s” when the sharpness of the new FE lenses (granted, maybe not 24-70/4 so much) are incredible – even Tony himself said the Zeiss 55/1.8 is the sharpest autofocus standard prime on the market right now.
First of all: the popularity of Northup among Sony fans is unbelievable.
He’s just another blogger. Why do you care what he says? 😮
Second: Nikon lenses ARE better than Sony E-mount for at least 2 reasons.
a) they exist
b) they are usually sharper with less distortion.
You can’t randomly compare lenses and decide that Sony makes sharper ones, because they have the excellent 55/1.8. Compare by type, because it limits the performance. If you do that, Nikon equivalents will usually be better optically.
And it’s not like Sony can’t make better lenses. They have money, they have Zeiss formulas. They can for sure.
But DSLR manufacturers design lenses to be good optically.
On the other hand, Sony designs lenses to be sharp and small.
“Sharp and small” will never be as sharp as it could be (not to mention distortion).
When the global hysteria about camera size ends, I’m sure Sony will give us many excellent lenses and will convert many great A-mount zooms. 🙂
And just a tiny heads-up. 55/1.8 is NOT the sharpest autofocus prime. Nikkor 200/2 is way sharper. I’m sure we could find a few sharper tele primes (85 to 400mm) from both Canon and Nikon.
+1 I ended up leaving Sony (even though I liked their cameras) because of the inconsistency of lens quality. I just couldn’t buy the lenses I needed across my required focal length and for them to be be Pro quality. You’d occasionally find a really good lens but would then be let down by the quality of another. UWA & Long lenses were not great but 24mm – 50mm were really good. Gave in in the end.
Primes in the 24-80mm range are usually the smallest by design. They often have as few as 6 elements (Sony 55/1.8 has 7 – based on fairly simple Sonnar formula).
Compare that to some of the best wide and tele primes:
Nikon 200/2: 13 elements
Canon 14/2.8: 14 elements
Even the Zeiss Otus 55mm has 10 elements, but that’s a Distagon (better distortion correction) and is f/1.4.
Sadly, Sony decided to go “retro way”.
It is true that 20-30 years ago most amateur photographers usually had just a few manual primes between 35 and 100mm. This approach is very easy to sell to many people (not just hipsters :)).
But the world has changed and things people shoot also changed.
Amateurs do sport, wildlife, macro – things that were almost reserved to pros and advanced enthusiasts back in the day.
Occasionally someone tries to offend me, because I’m using an “ancient” SLR technology. That might be true, but at least my ancient SLR works with lenses from the XXI century. 😛
That said, while I was really skeptic when the A7 came out (properly rubbish camera, I borrowed it and hated after 3 days), the A7II is acceptable. I could totally see myself switching from DSLRs, if Sony either improved the A-mount adapters OR gave Ax PDAF (and native FE-mount tele zooms).
And of course I would like the future Ax to be properly sealed, not just “tightly assembled”…
But all this wishes mean a much bigger gear – something Sony might be reluctant to offer at the moment. I can wait another 2 years, but who knows what will Nikon offer in the meantime. 🙂
It is a myth, except Canon tries to perputate that myth with the lousy 5Ds. Recent samples I saw look terrible even at Iso 400, anything above that is nearly unusable.
Most of Tony Northrup’s myth busting videos seem to completely disregard that a lot of these phrases are shorthand to describe how the market is generally organized at any single point in time.
all he is saying(it does not matter much) might be true for the mid tones and the highlights but it is not true for the shadows. there the bigger fotosite will gather more photons which converted will give more accurate signal, thus more detail in the shadows. since higher iso in digital means more amplification of the signal, this all means that having more accurate signal in the shadows allows higher iso. i’m not sure that doing averege of a few pixels after analog to digital conversion is the same as having a bigger fotosite. the only way to achive the same is to somehow make the downsamlping before analog to digital by combining the photons/signal from a few fotosites.
Anyone that’s looked closely at the A7s knows it’s ISO advantage isn’t at ISO 3600, but at much higher ISO’s, where all the others start to really fall apart, to borrow your analogy, @60 a Pinto and the Ferrari will both do well, but triple the speed to 180 mph and the Ferrari will be leagues better.
Myth? How many megapixels has A7S?
This video is frustrating because he first says “Off course there’s more noise from sensors with higher pixel density” and then he blows a lot of smoke claiming that it doesn’t matter.
Who’s he to tell if it matters or not?
If you choose a sensor with higher resolution and higher pixel density because you want better “cropability” then the noise threshold is highly interesting.
Higher pixel density DOES produce more noise than lower. He stated that himself!
If it matters or not is a completely different issue, but you are not promoting people to think about that as an issue if you mislead people from the beginning claiming that pixel density doesn’t affect noise.
In my opinion he’s full of it and not contributing to help people understand the relationship between noise and pixel density.
I think his argument is that 36MP zoomed in 1x have more noise but if you resize it down to 12MP, the noise become less obvious and should have a closer result as the image taken with the 12MP camera zoomed in 1x.