WOW what a game changer A7s

Started Apr 6, 2014 | Discussions thread
Shop cameras & lenses ▾
Veteran MemberPosts: 3,201
doubt Sony lies, claiming better high ISO performance
In reply to Ontario Gone, Apr 7, 2014

Let's keep in mind as others have pointed out, that information-free A7s noise predictions at this point are inherently pointless. Since we're all going to get the real data within a few months.

Ontario: are suggesting that the lower resolution will specifically make noise performance better. It won't.

This is a sort of silly prediction. Firstly since there are reasons why larger pixels on a same-size sensor, might make it possible to lower more than one kind of noise. That is at a given display size, for a given level of ambient light, a given lens and aperture, and a given exposure time. Boring details below.

In contrast, there are no reasons why smaller pixels will decrease noise for a given sensor readout after a given exposure time. Please let me know if this is not true, am eager to learn.

Secondly, Sony has claimed better high ISO performance for this new camera (presumably compared to other same-sensor-size Sony cameras available for sale before May 2014), and it is unlikely that a zillion-dollar-brand company like Sony would say something so easily and soon proven to be a lie. Sony has lots of brand perception to lose, in both short and long run.

Ontario: Noise performance is about total light gathered,

Well that is surely the most important factor, but it is a touch misleading to imply it's the only factor. For example you surely do not guess that all cameras with the same sensor size show the same total noise performance.

and the A7/r and A7s gather the same amount of light.

Indeed all cameras without translucent mirrors with the same sensor size "gather the same amount of light". For example the newest Nikon D4, an older Sony A680, etc. Would you hazard a guess that all those cameras have the same low light noise performance as the new Sony A7s.

Perhaps with a newer sensor they may have tweaked something minor, but it will be just that, minor...

I agree that every individual aspect of operation that improves with lower-resolution cameras is individually minor, but when all things are considered ranging from lower sensor read noise on through to smaller raw files and quicker file write times, I do not share your prediction that the total difference for ordinary still photography in low light (even in the narrow issue of high ISO noise levels) will be minor with the A7s vs A7/r.

Let's discuss some of the things that can contribute to lower noise when you move to larger pixel sizes. (To rebut this you will of course have to tell us about some noise contributors that go down with smaller pixel sizes.)

  1. There is such a thing as sensor "read noise". It is no great stretch of imagination that when you do less reading of fewer pixels, you're gonna get less read noise.
    To be a touch more scientific and less concise, we can point out that if you have a read noise of plus or minus 3 electrons stored per pixel well, and you are reading storage areas of pixel wells that are holding 100 electrons, you have 3% noise. But if you have triple-size pixel wells holding 300 electrons for the same ambient light and exposure, now you have a (3 times less obtrusive) 1% read noise. This in fact understates the case, because when you read pixels out at a 3 times lower rate, the machinery that operates so much more slowly inherently does so less noisily.
    Now if you contemplate pixel-binning 3 smaller pixels of say a 36 megapixel A7r camera, it is an unfortunate fact of life that you "only" get a noise improvement of the square root of 3. Call it an improvement of about 1.7x. Not the 3x read noise improvement you could'a gotten from the 3-times-lower-res sensor.
  2. There are non-light-gathering electronic and optical areas on the top of sensors that become percentage-wise less obtrusive as the pixel well sizes grow. For example let's say you had a 1 billion pixel 24x36mm sensor. That sensor would have so much of its top surface taken up by the non-productive edges of microlenses and other various kinds of partitions between adjacent pixel wells that it could not possibly capture as many photons as efficiently--as a single-pixel huge photodetector that had all of its surface dedicated to light gathering, with no interconnects or microlens edges at all.
  3. There's got to be less heat generated by lower-megapixel sensors and their readout machinery, which we can describe as leading to less thermal noise. There are fewer analog amplifiers, the read out rates can be slower for a given taken-frame-rate, etc.
  4. There is certainly less processing for the rest of the camera CPU chain to do with fewer-pixel images, which at a given frame capture rate is going to lead to a literally cooler camera and less thermal noise.
  5. One can even imagine feeling freer to design or choose lenses with better large-structure contrast but poorer finest resolution (exactly consider the newest Sigma 50/1.4 Art lens compared to the Zeiss Otus 55/1.4), for the lower-resolution cameras. And the better your raw image contrast, the less contrast enhancement one needs in post-processing. And the less post-processing contrast boost you need do to an image, it is surely familiar to many of us that the less noise you will see.

Having said all this, even cheap APS-C cameras don't do so bad in the noise department these days.
Ontario why not show us some of your own work, that shows your experience with low light?

Reply   Reply with quote   Complain
Post (hide subjects)Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow