Image courtesy: Cinema5D
Cinema5D (Click here) posted a full A7s dynamic range test (in video):
Here we tested usable dynamic range of the given cameras. With 14.1 stops the usable dynamic range of the A7S comes surprisingly close to the Arri Amira with its legendary Alexa sensor (see our full review here).
This is an extremely good dynamic range rating and is fascinating considering that the A7S is available at a fraction of the Amira’s price and is also in a completely different weight and size class. While the Amira will outperform the A7S in other tests, the dynamic range is a very important attribute to consider when working with a camera.
Gizmodo posted the full A7s review and writes:
The A7s is a great demonstration of technology powerful enough to make you forget about its flaws. The Panasonic GH4 has more tools, more options, more usability—with 4K built right in, and to spectacular effect. But the micro four thirds sensor of the GH4 offers its own compromises: low light is not very good, shallow depth of field is constrained, as is field of view. But here’s the real kicker. The A7s will literally increase the possible scenarios in which you can get a good shot, by way of low light performance. No amount of 4K, 8K, or how ever many K can accomplish that.
More A7s tests:
Professional photographer/videographer sold his canon 1dc and purchased Sony a7s’s Matt Buckman’s Review of the Sony a7s: http://youtu.be/sjKbIIsLWdQ
A7s test at Fotopolis (translation here).
A7s review at DigitalCameraWorld.
Eric Lo: “I just came back from a long road trip and finished photo editing, all photos were taken with A7S(night), A7R(day) and Canon lenses (14L2, 15 Fisheye, 17TSE, 2470L2). http://bit.ly/1qLSLaH”
Sony A7s store links:
Sony A7s preorder in USA/CA at Amazon, Adorama, BHphoto, Sony USA, Sony Canada.
Sony A7s EU pages: Sony DE, UK, ES, FR, IT, NL, BE, AT, CH, SE, NO, FI, PL. And Wex UK.
0 Responses
The GH4 appears to be last…Life is tough when you’ve only got a four thirds sensor i guess…
Yes, it’s crappy.. C’mon, do you really think you can’t do anything good with a smaller sensor?
That’s not what I said….Settle petal…
Pretty sure the Blackmagic models fit in the middle somewhere, so what’s that about a smaller sensor? Plus the top model (the Arri) has the second smallest sensor of the group here.
Hmmm must just be the GH4 then…
It’s mostly S-Log vs. no S-Log.
Yeah…Your point??
In fact the GH4 does very well compared to the 5D, despite having a sensor a quarter of the area (and even worse when looking at the proportion of that sensor area which can be used for 16:9 aspect ratio video).
However, the secret is surely that both the GH4 and A7S read all the pixels within the video area, whilst the 5D cannot (neither can the A7 and A7R).
If you start doing the calculations and allow for the mapping of a 16:9 video format, the A7S has about 5 x the area of silicon from which to generate it’s video. Of course, this exactly matches the ratio of the area of each photosite (13.69 sq micron on the GH4 and 68.89 sq mm on the A7S).
So the lesson is, if you want great video performance, use a large sensor, use one which maps well to the output aspect ratio and which can read all the video sensor area for every frame.
or use 3 smaller ones… 😉
Anyone know where the A7 fits in that graph?
The a7/a99 sensor has has very strong dynamic range for stills. Unfortunately it does not feature picture profiles in video mode (and only is in Sony cameras that record in avc codec) that maximize the dynamic range such as s-log. The DR is probably around the gh4 / canon levels in video mode.
Huge 8 micron photosites. Big buckets that that can hold ALLOT of photons for each pixel.
Almost. The “buckets” actually hold electrons which are excited when a photon is absorbed (modern sensors manage this for over 50% of the photons that get through the Bayer filter layer). For the A7S that’s over 150,000 electrons (current technology allows for about 2,400 electrons per square micron).
Trappping actual photons is a much, much harder job.
But the honest and fair price of the Arri Amira is $1499…
Where in the world? I’ll buy ten, the lowest price I found was 30.000 Euros….
/sarcasm
Well, all of this photo nerd stuff is fascinating but what does it all mean? The a7S with an external $2,000 Shogun attached on some rig is a great, ‘economical’ set-up for people shooting serious video compared to older alternatives it leapfrogs but for the masses the GH4’s internal 4k makes it far more practical for the rest of us.
IMHO, Sony needs to address the GH4 more directly. Maybe the a99II? I doubt it but it would be a good place to start.
Or an A6000II
The a6000 already has excellent video, just not 4k, yet. In a year 4k will be standard…
http://www.eoshd.com/content/12428/surprise-sony-alpha-a6000-video-mode-huge-improvement
Early you can just use a 4K smartphone mmmkay? It’s easy to get caught up in the MP race, I understand. But those of us who do actually get this (and the A7S in general) know what it means and what it’s for. By the way, the Alexa doesn’t do 4K either, guess people are just going to dump it for the GH4’s little onboard 4K because the internet tells us its great 😉
what do you mean by “the masses”. If it’s the general public you refer to, none of those cameras are addressed to them. In fact Sony had already made a device for basic customer who want to shoot 4K and it’s the AX100 camcorder.
If “the masses” is the enthusiast community that aren’t pros and that inhabit this website and others cam tech sites, then maybe the 7S is not for us but the GH4 is not the solution either. What would you do with your SD card full of 4K content? The GH4 is not enough you need behind a powerful computer to edit your file, a massive storage capacity and at the end how do you display your video. 4K TV/monitor?YouTube? how many of us are able to watch it comfortably (meaning a powerful enough computer to stream and decode 4k).
Like you I’m interested in 4K but actually it’s more for my personal tech curiosity than an actual need to shoot 4k. The ability to record internally 4K is a nice feature but still a pros’ one. It allow journalist to shoot on the go but it doesn’t cut the massive workflow needed behind to edit the file.
He was talking about screeens. Hardly anyone got a screen that can show 4k resolutions. I’ve got one of the more higher resolutions display on my laptop (3200×1800) but it still isn’t 4k.
Given the slow uptake of 4k displays and the hardware needed for editing, it’s far from being something for the masses.
FYI, the Red Dragon sensor should have 16+ stops of DR…
This is good news. There is a low end market with the video professionals for this cam. Awesome!
It’s funny to go and read all the comments at Cinema5D-site. So many GH4 fanboys trying to discredit the test.
But but but the GH4 does 4K, like right there in the camera… amazeballs!
It’s a fanboys site. Not much different than fanboys here.
hey sony, when is 16-35mm FE f/4 coming out? I badly need it. thx!
one can only dream. oh dear sony
Err why?
you do realise that to be able to achieve the same sensitivity level they need to keep the same photosites size on an APS-C size sensor. It means a roughly 5mpix sensor.
No, they think if it was 12MP it will be the same, because it will use the same ‘technology’ (Apparently some unexplainable thing they don’t understand) to get the same results. Sad but true, had this argument loads of times on SAR.
Or maybe they think that less than half the MP will mean less than half the price… because SAR readers are cheapskates and think the A7S is way overpriced.
6mpx will do nicely!
8 MP APSC and 12 MP FF will have same pixel size.
So it;s not 5MP — it’s 8MP, which is enough for most people and perfect for 4K video
No, it’s about 5MP. 8MP is fine for 4K if those 8MP were a 16:9 sensor, 12MP is about right for a 3:2 sensor cropped to 16:9. Where you got 8MP on an APS-C version I have no idea, maybe you think MP x 1.5 crop?
No they still can’t. It’s a smaller sensor… it doesn’t gather the same amount of light as on that is 2x the area. Forget pixel densities… old misleading news.
Unless one uses speed booster. Same amount of light, same angle of view .
When even a highly regarded, professional grade, closed system/purpose built 85 f/1.2 lens on FF can’t match a 85 f/2 lens on 645, it’s very close to being silly to expect generic, multi purpose speed boosters to be a real solution for this problem.
There’s no replacement for displacement with sensors either 😉
Regarding comparison with the GH4 dynamic range.
They made mistake. They chosen the cinema look mode. Which is the mode when the video is ready for show strate from the camera. This mode does not assume post-processing.
Panasonic as well as Sony does have the specific mode, with higher DR, assuming the post-processing. And the mode could be adjusted.
Very true!
DxOMARK test the still image DR. Still image DR does not necessarily translate to that for video. It is quite possible the need to read the sensor at the required rate involves some sacrifices, like a lower bit resolution (which will limit DR if it means quantisation is above the noise floor). Of course this is just surmising, but just transferring still measurements using RAW to video (which is pre-processed on all these cameras) is not necessarily valid.
Interestingly, Sony have been claiming 15 EV DR for the A7S in video mode, and nobody has been able to explain how this can be achieved given DxOMARK still image measures. However, that makes the same mistake of assuming that there isn’t something a little different being done for video than still.
Also, measures based on whole-sensor images downsized to 8MP are not valid when the output aspect ratio is 16:9 and the native sensor is different. In effect, the video is a rectangular “crop”. As the 3:2 format sensors fit video formats better than M43, a higher proportion of the sensor area is used by video in the former. That’s why the A7S is 12MP and the GH4 is 16MP. As both use 1:1 pixel mapping in 4K video mode, it’s the per-pixel measures that should be used, not the 8MP downsized ones. (Although the difference is relatively modest). For the A7S at least, it seems that 2K (up to 30fps) uses a 2×2:1 video mapping, which might also explain the elevated DR compared to still per pixel measures.
As for the 5D MK III, I really don’t know what the reality is there.
Hope there will be a A6000-like camera with similar spec (dynamic range, noise, codec) as A7s but smaller sensor (APS-C) and 2K video (5 Mpixels?)
Basically cut the A7s sensor to APS-C and fit into the A6000 with upgraded processor, then charge maybe $1000.
You would buy a 5 Mpix cam for 1000$?
In a video camera? Erm yes, yes many probably would.
Point taken… Nevertheless, you can get a A6000 for 600$ and it also gives you FullHD… I don’t think the 5 MP sensor would make the camera more expensive than the A6000…
HAD video quality doesn’t improve with more MP. It needs three MP. Around 9 MP is enough for 4K
Unless you have a 16:9 aspect ratio sensor you need more than 9MP. With a 3:2 resolution sensor at TV standard 4K it needs a sensor with a minimum of 3,840 x 2,560 (or 9.8 MP). If you want a sensor capable of the cinema (DCI) 4K standard (which the GH4 does and Sony may well want to have this option too). then a 3:2 sensor has to be a minimum of 4,096 x 2,730 or 11.2MP. The A7S is a little bigger (about 5% in each direction), but that extra border gives the option for electronic image stabilisation.
You can do the same exercise with the GH4 and it’s 4:3 aspect ratio sensor and it’s clear 16MP is the sensible resolution for 4K.
Why not just make it 8MP and put 4K vid? Pixel density is not that much higher. And if the dynamic range and noise are not quite as good as A7s that is fine – gives people a reason to spend a lot more if they want that last little bit of DR.
Simple. You need 12MP to get 1:1 pixel mapping as 4K video (16:9 or wider for cinema 4K) is a different aspect ratio to a 35mm sensor which is 3:2. That means that the video comes from a rectangular crop. Of your 8MP pixels, less that 6MP would be in the video area. That would require interpolation of 4K, which is not what you want for high quality.
If people want a 4K capable APS-C camera with the best quality video it will, like the A7S, require about 12MP. Performance will be better than the GH4 as that has a smaller sensor and suffers even more from the video “crop” effect, meaning it needs about 16MP to pull off the same trick.
Incidentally, 2K works well too at 12MP as that’s a 2×2 to 1 pixel mapping and allows, in principle, full colour sampling at each video pixel position.
Ultimately all this just comes down to the amount of silicon that’s used to produce the video frame. Kludges (like line skipping) are always sub-optimal. All other things being equal, the larger sensor always wins – as long as the whole silicon area in the video frame is used.
Fair enough. Then just make a 12MP aps-c sensor. As you said, it will still be far superior to any other video on more pixel dense sensors – and it will not step on the toes of the A7s – another advantage for Sony. To me a A6000s with a 12MP sensor makes perfect sense. Keep the same AF system as the A6000 – and you also have a super sports camera. Nice side effect.
Such a camera may appear. My guess is it would be designed to take on the GH4. It ought to beat it on video performance as, not only is the sensor of larger area, but it better matches the 16:9 format.
However, it would clearly need a new sensor design as it would have to (like the A7S) be able to peform full-width sensor reads at 30fps, and I doubt the existing types of sensors will do that. Of course, if they do, then the “silent shutter” mode ought to follow.
One thing to be aware of is that, in designing the A7S sensor for high sensitivity, Sony appear to have had to sacrifice DR at base ISO. The A7S has amazingly low read noise at high ISO but the read noise at base ISO is mediocre at best. Strangely, the Nikon 4S exhibits exactly the same behaviour, so it may be that there’s some inherent sacrifice to be made. In comparison, all other Sony sensors appear to be essentially “ISO-less” for the great majority of their sensitivity range. In other words, read noise remains essentially constant (measured in electrons) giving a constant noise base and DR drops linearly with increasing ISO. The A7S does not do that.
Personally, I think that making such a compromise for ultra-high ISO performance on an APS-C camera would be a bad idea. At high ISOs it wouldn’t get near the FF whilst at base ISO it would sacrifice a stop or so of DR. Leave the high-sensitivity stuff to the A7S.
Also the A7S has relatively slow still frame rates, and I’ve no idea why. Perhaps they’ve done something in the “video frame area” which has robbed resources from the rest of the still area.
Perhaps the stills frame rate of the A7s was limited because the AF system could not keep up – so Sony thought – whats the point. I mean, theoretically it should be capable of 30fps – right?
I think the sensor is only capable of reading the full width of the video area at 30fps, so it’s probably more like 20-25fps in principle. However, still way more than anything else at full resolution. Of course the A7 shutter isn’t built for that sort of frame rate either.
The time taken for the “silent shutter” to scan down the frame would given an idea of what’s possible as that will give the maximum read rate (albeit I think that’s only at 12 bit resolution).
Yeah and even at 12 bits, writing 12MP files at 12fps would require 400+ Megabytes/sec… haven’t seen SD cards that can do that yet for continuous capture. With the raw compression there should be some savings but still… if they left out 4K compressed video recording for heat issues, full res raw stills at 25 is ridiculous.
Technically it’s not 1:1 mapping going from a 8MP 16:9/12MP 3:2 sensor to 4K because interpolation is necessary.
It’s not necessary with a minor side crop (about 5%) which is almost certainly what’s happened as Sony claim “no binning”. The number of pixels across the width is slightly more than is required for even cinema 4K. If the full width of the sensor was actually used then they would have to read 5% more lines vertically which makes no sense at all.
The GH4 does exactly that. It does 1:1 4K with a slight “side crop”.
I’m not sure if we’re talking about the same thing here.
If Sony ever makes more lenses for this system and I lose my fear that this system will be dropped for the curved sensor system, then maybe I will get excited and buy into it. 🙂
From what I’ve seen, the odds of them being able to roll curved sensors out to ILCs are tiny…
Curved sensors are for fixed lens cameras not ILC systems.
I know that a lens fully optimized for a curved sensor would maximize the achievable results. But surely all lenses by their nature would benefit at the edges the problem for many FF lenses at or near wide open is field curvature. If a lens focuses at the same distance across the frame, then the ‘shape’ of the in-focus field will naturally be curved thus a curved sensor would in some way benefit every lens put on it.
Well you might hope so but it won’t work out that well. Aside from macro lenses suffering, these sensors have a higher curvature than what typical lenses need. Ever thought what too much curvature can do? 😉
Thar comparison review is a total fraud!
Panasonic GH4 has 13 DR stops measured by others and myself included: http://www.dvinfo.net/article/acquisition/panasonic-gh4-dr-and-gamma-timelapse-and-more.html
Canon 5D mark 3 does not had that much DR even in RAW.
DxoMark says A7s has 13+ stops normalized to 8 MP and 12+ at full resolution.
It is a shame that misleading reviews like this are published!
because you are a MFT fanboy?
No because I don’t like misleading reviews.
We currently work with Nikon FF (Sony sensors), Panasonic and Canon FF cameras. I have extensive knowledge about all there strengths and weaknesses. I’m not an m43 fan, actually I hate the crop factor, but Panasonic does a very good job on video side, I wish I cold say the same about Nikon and Canon. Canon was good in the past but lagged behind. I’m intrigued by Sony cameras (I personally owned Alpha 100 and 700 in the past), and A7s could have been a great camera in our arsenal if it had internal 4K recording and a smaller jello effect.
Yes you gain maybe one stop of DR of the sensor if you do pixel binning or resizing the full read out, but not much more.
How much you gain by downsizing is pure math, the conversion DxO makes from the ‘screen’ value to the 8 MP ‘print’ value is using the simple formula: DR_8MP = DR_xMP + log_2 [(x MP/8 MP)^0.5]
This maths is wrong, you gain a little bit more DR after downsizing because it separates the read noise from real data a little bit better, still you don’t get new data, this is physics.
Dxomark do not review video DR, so quoting dxomark is irrelevant. HD is 2 MP — nothing to do with normalized 8 MP raw for stills as measured by dxomark
I guess the point is, if 13 is the maximum that can be done in stills, how can it do 14 in video.
its slog2 measured at iso3200 (native). dxomark done it at iso100 (native) for still.
Tone curve, etc. should have nothing to do with real DR. ISO3200 native? A lot of sensors do 3200 “native”. I think you mean “base”… in which case that’s also incorrect as the a7s’ base is a lot lower or else something shot at iso100 would be mostly blown… but they’re not. 3200 however seems to be the ISO they choose IIRC, for the slog2 curves.
S curve?
Or none of these test are consistent, accurate or done by competent people 🙂
Eno.
How about stop bashing cinema5d anywhere you can and instead show some of your tests!. If you can’t get a life and move on.
Thanks!
Johnnie (cinema5d)
I posted tests, if you read carefully.
Is your name Adam Wilt? (http://www.dvinfo.net/article/acquisition/panasonic-gh4-dr-and-gamma-timelapse-and-more.html)
If not, please direct me to your published test.
Thanks.
Johnnie
Honestly, Johnnie?
You may well be in the right with your testing methodology, not attacking you there. But saying “unless you wrote it yourself, your data is meaningless,” is probably the most ignorant thing you can say. It’s the opposite of the basis of science.
Hi Johnnie,
I for one appreciate your hard work over at Cinema5D. But I’m only a stills shooter trying to figure-out this video minefield. Thanks for the reviews, tests and more importantly, thanks for backing up your articles with actual footage. Personally I place great faith in someones opinion when they back it up with their **given name** and **work** — good for you mate, you put your arse on the line every time you post your opinion by offering up your name and showing us how and what you’ve produced!
Cheers & thanks,
Chris
No f-king way!
Agree!
Bullsh@t
Random question, A7r has 14 stops EV as well, is that correct?
A7R and A7 both have 14 DR stops, but A7s has only 13 DR stops, that review is a bullsh@t. Although A7s has better low light capabilities, that’s for sure.
This review is specifically referring to video slog 2. Eno doesn’t know what he is talking about.
The sensor in still RAW mode has a DR of 12,76 DR stops: http://www.dxomark.com/Cameras/Compare/Side-by-side/Sony-A7S___949 , there is absolutely no whay to milk 14+ stops out of this sensor in any mode. Even after downsampling the data you do not get more than 13+ stops of real DR.
You have no clue about what is real DR an not! 14+ stops of DR is pure fable!
You are an idiot. HD is 2 MP — not 8 MP as measured by dxomark. And dxomark do not measure video DR. The readout tech, line skipping, pixel binning, codec, all things that matter can have total different result for video than for stills.
I don’t think I made anybody names. NO there is a point from where you don’t get more DR from downsampling. Tray it for yourself and then speak nonsense!
There is no such thing as “even after downsampling”, you can downsample almost limit less and thus you can push the DR almost as high as you want. Half the resolution and you gain half a stop of DR, go down a quarter and you gain one full stop, to down to 1/8, and you gain 1.5 stops, which means you get 14.26 stops DR (based on the DxO data). One eighth of 12 MP is 1.5 MP, crop from 3:2 to 16:9 and you get 1.27 MP. 1080P is 2.1 MP. Which means that downsampling to 1080p doesn’t quite explain the reported DR, of course, downsampling further to 720p would easily explain the reported DR (though I think it is unlikely that they did that).
Often video has worse image quality than a still image from the same sensor, something which is expected for cameras that skip lines but also the fast read-out out itself can degrade IQ. Getting the same IQ from a still image and a single frame extracted from video stream is thus already an achievement.
But video can do some tricks when it comes to DR. What is particular with the A7s is the extremely low read noise a higher ISO values. A video stream can in alternating fashion read out the sensor with a high and low ISO setting and splice the two together. This allows an effective DR that is larger than a still image from the same sensor could have. RED have given this mode a special name (something like DR+). It is possible that some of the high DR ‘claims’ of the A7s as a video camera are based on this.
And your benchmark is DXO? Right.
For all those pulling their hair and gnashing their teeth at this test, the writers make it clear that they are not measuring absolute dynamic range:
“Before we go into details about how we test let me explain why the Canon 5D mark III outperformed the C300 and 1DC in our test. As mentioned before we’re measuring usable dynamic range. This means we’re observing actual dynamic range relative to the noise ratio of the signal. In other words, we only measure dynamic range where the signal still upholds a certain quality which is measured in noise.”
And? It seems like a decent test of ‘usable dynamic range’.
It is lost on many of the commenters who failed to read the original article.
The A7s is the best camera on the market…Why? Because I bought one.
If that’s the case, it must be true then.
Hope you enjoy your wallet getting raped.
Hitler finds out about the Sony A7S after buying a ARRI Amira
https://www.youtube.com/watch?v=vI7xnfZdgwk&feature=youtu.be
Well, Cinema5D suceeded – in driving huge amounts of traffic to their site. If those clowns were a ‘real’ director like me, they wouldn’t be wasting time on these prosaic tests. I just sit on my fat ass all day and tell people what to do.