Too many megapixels? Six is enough for an 8x10.

Started Apr 26, 2009 | Discussions thread
Daniel Browning Senior Member • Posts: 1,058
Dynamic range, 1:1 S/N, and charts

Ergo607 wrote:

All I can say is that I already have problems with the assessment of
the LX3 having 10.7 stops DR (it's much lower in my experience), let
alone having it some theoretical 12.3 stops...

This is essentially because you and I are talking about a different "standard" of what's acceptable for noise. Let me explain.

Raw files have no such thing as "black". It just goes from white to noise. So to make an image, on must choose where to place black. It could be placed 7 stops below clipping, and that would result in 7 stops of dynamic range. The more stops down you go, the more noisy the image looks. On the LX3, 8.5 stops down from clipping, the noise is pretty bad. that's probably about the spot that most people would place black on a raw conversion. Furthermore, a curve will be used to crush the lower stops (from 6.0 to 8.5, say) and make what noise there is even less visible.

There may be plenty of visible detail if you go further. At 10.7 stops, the noise is in equal competition with signal (1:1 S/N). This is the point at which it is convention to draw the line for dynamic range. But it's just convention, it's possible to go even further. Since S/N doubles for every four-fold decrease in resolution (0.5X spatial frequency), greater dynamic range can be used for images with lower spatial frequencies. That is, using 10.7 stops dynamic range for a 30x20 print might be too noisy for the taste of most folks. But in a wallet size print it looks fine, perhaps even 11.7 stops would look noise-free. Since the spatial frequency is lower, so too noise is lower. (This is a linear-raw-only advantage. Nonlinear formats such as film have the respones curve built into the capture medium, so it's not possible to trade resolution for noise.)

You probably use something higher than 1:1 for the lower bound of your dynamic range, depending on how much noise you can tolerate in your shadows. Since everyone has a different sense of taste and preference with regard to shadow noise, it's common to use the 1:1 convention when a monolithic number is required.

However, a monolithic "dynamic range in stops" number is not enough to tell the whole story of course, just as MTF-50 cannot tell the entire story of a lens. It's highly useful to plot S/N over the entire exposure range to see how dynamic range changes for a given lower bound in S/N. That is similar to how it's useful to plot MTF over the entire Image Height and/or spatial frequency.

Look at Emil Martinec's S/N plot of the LX3 from the following post, for example:

http://forums.dpreview.com/forums/read.asp?forum=1033&message=29866728

[This has nothing to do with pixel size, by the way, just trying to share information about dynamic range.]

1:1 S/N is bold horizontal line. The blue line intersects 1:1 at 1.5. 12 minus 1.5 = 10.5, which is about the same as 10.7 above. But what if a photographer can't stand the sight of any noise in their shadows, and they prefer an minimum of 8:1 S/N for their darkest shadows? That would be a S/N of 3 stops (2:1 -> 4:1 -> 8:1). At this level, the blue line is at 5.5. 12-5.5 = 6.5. So Dynamic range drops to just 6.5 stops. Other cameras and sensor sizes plot out differently, so the rate of change for dynamic range with exposure can be different.

I hope that helps.
--
Daniel

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