A77 Noise Analysis

Started Sep 12, 2011 | Discussions thread
ForumParentFirstPreviousNextNext unread
Flat view
RicksAstro
Veteran MemberPosts: 3,422Gear list
Like?
A77 Noise Analysis
Sep 12, 2011

I was frustrated trying to glean an idea of how good or bad the noise characteristics are for the A77 in RAW. I downloaded Raw Therapee which handles the A77 files well. I believe it's based on DCRAW, to it is a very neutral RAW processor. I'm figuring the basic electronics in the pre-prod camera won't significantly change, so the RAW files should be quite similar. Although I'd hope their jpg engine will improve with the production firmware.

I then downloaded a boatload of raw files from ImagingResource...their test scene with the bottles and such. I downloaded all the 16mpix chipped cameras (D7000, A55, A580, K5, 5N) as well as the A77 and a few other cameras (60D and G3). I used an Astronomy program which measures pixel standard deviation over an area and chose the dark grey (not black) square on the test chart at the bottom middle and measured the standard deviation of that square, simulating medium shadow in an image. I loaded all these values for each camera and each ISO into a spreadsheet, and indeed the A77 came out the worst by a bit.

But of course this is an unfair comparison, since this measures noise at the pixel level, and the A77's pixels are much smaller. So I resized (using CS5 bicubic) the A77 image to the exact size as the other Sony 16 mpix images. The result shows its performance is virtually identical to the A55's, which makes perfect sense since it also has the mirror in it's path and has the same amount of light hitting the overall sensor.

The other 16mpix cameras without the translucent mirror are remarkably similar, with the 5N being the least noisy sensor. You'll notice the dramatic effect of the k5's RAW noise reduction at 3200 and above.

So, basically, if you want to take advantage of the higher resolution of the 24mpix sensor, you'll get a little more noise at the pixel level. But the more megapixels gives you that choice. This should help dispel the more mpix=more noise in an image myth. Yes, there's more noise at the pixel level, but your image should look noisier.

This little exercise has calmed me a bit...I'm keeping my A77 pre-order. I still have my 5N if I really need low light performance, but for outdoor sports shots, the A77 should do perfectly well!

I will also presume from this the NEX7 will perform similarly to the other 16mpix cameras when sized the same, although perhaps not to the 5N level.

I didn't plot the 60D and G3, since these weren't apples to apples as much as the others and just confuse the chart. I left the data there for your perusal.

The vertical scale is raw standard deviation for that square.

-- hide signature --
ForumParentFirstPreviousNextNext unread
Flat view
Post (hide subjects)Posted by
ForumParentFirstPreviousNextNext unread
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow