The Economist article on mirrorless cameras

Started 3 months ago | Discussions thread
Jerry Fusselman
Regular MemberPosts: 428Gear list
Like?
Re: Sometimes known as ISO-less processing
In reply to Erik Magnuson, 3 months ago

Erik Magnuson wrote:

Jerry Fusselman wrote:

If my prediction is correct, that is the reason not to go beyond ISO 1600 when you want the best possible image with your equipment under the circumstances.

When looking at DxO DR data, look for non-linear sections of the graph. Where the graph is linear, it makes essentially no difference what the camera ISO is actually set at as long as you process for the desired brightness (your raw converter may have limited range.) The A7R is pretty much ISO-less from ISO 100 to 6400. The Canon cameras start getting linear after ISO 1600. That's why ISO 1600 is the point where the results start to become very similar .

Ummm, not exactly. The definition of ISOless is sometimes stated as "a sensor where the read noise does not change as a function of the ISO setting," but that definition is not so useful, because measured read noise always changes somewhat as a function of the ISO setting.

What really matters is the slope, not the linearity. If the DR loses about one stop when you double the ISO, that's what should be called ISOless. Call this a unit slope of DR.

When you have a unit slope of DR, it means that there is no point to doubling the ISO, because you're going to lose that stop of dynamic range anyway, and you can probably handle the adjustment of weight better in post than an algorithm can in the camera.

My estimate of M = 1600 for the A7r is, to my amazement, demonstrated in both sources I found on the Internet. I did not know this ahead of time. You can see that the unit slope for DR begins at ISO 1600 in two reports:

http://www.dxomark.com/Cameras/Compare/Side-by-side/Sony-A7R-versus-Canon-EOS-6D-versus-Canon-EOS-5D-Mark-III___917_836_795#tabs-2

http://www.sensorgen.info/SonyA7R.html

I think you will agree with me that the surprising up turn in DR at ISO 25,600 in both references is also surely due to hidden in-camera noise reduction, not true improvement in DR.  It should be ignored.

However, when I look at both sources for the Canon 5D M3, it appears that M = 3200 may be a better estimate.  Alas, I could never get happy with my results at ISO 3200 with the Canon 5D M3, so maybe that biased my judgment, or maybe M = 3200 is just moot for my taste with that camera.  Or maybe I should improve my test methods.  I'm not really sure which is my moral for apparently being off by a stop.

But your emphasis on linearity is incorrect.  It is slope that matters.  The A7R is not really ISOless from 100 to 1600, but it is from 1600 up.  Both of my sources above show that.

-- hide signature --

Jerry Fusselman

 Jerry Fusselman's gear list:Jerry Fusselman's gear list
Canon EOS 5D Mark II Canon EOS 7D Canon EOS 5D Mark III Sony Alpha NEX-6 Canon EOS 100D +40 more
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