RustyBug
Offline
• • • • • • •
Upload & Sell: On
|
Tony Ross wrote:
When we render our finished image on a printer, some of the printers accept onty 8 bit data, so what matters is how accurately we can achieve those 8 bits. Some have an extended gamut through the use of additional inks, but if the data we are feeding them is only 8 bit, we can still only realise 8 bits of colour.
When we render our finished image on a screen, some screens can only show 8 bits of colour. Some can show 10 bits of colour (maybe even 12 bits of colour) - again, our output is limited.
What might limit our ability to product those 8 or 10 (or 12) bits of colour for our output?
If our image was captured at high ISO, then we might already be working with a limited dynamic range, and it doesn't matter if our output delivery is capable of 8 bit output if we're limited to 6 stops of dynamic range due to high ISO.
Or if we underexposed our image by 5 stops, even under conditions when we should have 11 stops of dynamic range - we'll only have 6 stops of image.
Or if we have a significant amount of nose - that's encompassed in the definition of dynamic range, of course.
When does 16 bit make sense? When the electron wells have sufficient capacity to make it worthwhile to measure the values to 16 bit precision. 14 bit is generally regarded as ample for full frame sensors. Heck, it wasn't until around 2007 that we got 14 bit on full frame - cameras like the original Canon 5D were 12 bit, and some cameras still operated in 12 bit mode for quite a while after that.
With the push to high megapixel counts, we get smaller pixels, with lower capacities. If you really want 16 bit, you need fewer pixels and bigger capacities - start demanding an 8 megapixel sensor if you want 16 bit :-)
...Show more →
+1 that as long as your info is north of your output threshold (i.e. you're supplying 10 bits of data-continuity to an 8 bit device, etc.), your 8 bit device isn't going to render any differently than if you are supplying 12 bit, 14 bit or 16 bits of precision, to it.
That's why the point of "heavy math" can create situations where the tonality continuity is 8 bits or less of precision. At that point, the difference in starting at 8 / 12 / 14 / 16 can influence whether or not you remain to have more than 8 bits of precision continuity. Exponential operations like Curves, multiplied by Contrast, Sharpening and other operations can stack and build to erode the continuity of mathematical precision. 16 bit simply gives you the more wiggle room for the math, than starting at 14 bit does.
To be clear, I'm not advocating that everyone needs 16 bit. Most don't. I'm only saying that there is a difference (vs. marketing goo only), and it isn't until you are cranking on the numbers that the difference does (potentially) present itself. My point is in understanding that the math has to be heavy for such difference to reveal itself, beyond the threshold of our output devices (screens, printers, etc.).
That said, if an image math isn't going to be cranked on ... yup, the difference in 16 bit vs. 14 bit, the difference will be "lost" on the limits of an 8 or 10 bit device, as it can only function to the lower level of precision, so long as there haven't been mathematical "gaps" that breach the lower bit precision (i.e. causing banding, etc.).
|
