OM-D versus Nicon D800 - dramatic?

Started Mar 20, 2012 | Discussions thread
Great Bustard Forum Pro • Posts: 40,044
A little bit of magic.

Steen Bay wrote:

Agree with everything you said, except for your definition of "equally efficient sensors". I think it's best to be consistent and say that it's the normalized performance that matters, both when talking DR and when talking SNR.

In my post above, I explained that I defined equally efficient sensors as sensors with the same QE that have the same, or scaled, read noise / pixel and saturation / pixel. If a person wanted to define equally efficient sensors as sensors with the same QE and the same read noise / area and saturation / area, that would be fine, too.

Of course, regardless of which definition we use for sensor efficiency, the proper measure of any element of IQ for a photo is per-area as opposed to per-pixel. For example, it makes no sense to compare noise pixel-for-pixel if the photos are made from different numbers of pixels.

But the main thing to bring up is what one is using for the noise floor in the computation of DR, where DR is defined as the ratio of light levels where detail can be recorded: DR = log2 (saturation limit / noise floor).

The saturation limit in unambiguaous, but the noise floor is not. As I said, some take the read noise for the noise floor, whereas others (e.g. DxOMark) use the 100% NSR for the noise floor.

For some, detail recorded at even the 100% NSR is "useless", so what we need to do is specify the highest NSR that is "meaningful". So, let's revisit how the DR changes as a function of what we use for the noise floor:

Read Noise: 11.3 stops
100% NSR: 11.2 stops
50% NSR: 10.1 stops
25% NSR: 8.9 stops
10% NSR: 7.1 stops
5% NSR: 5.5 stops

For what it's worth, assuming that the darkest inks have a reflectivity akin to charcoal (1%), then the max continuous DR a print can display is 6.6 stops, although we can certainly compress more DR into that range, with the consequence of not all levels being distinct. Here's an excellent article on that point:

There are those that argue that the DR figures given by DxOMark do are not "photographically meaningful", but those arguing that point have no understanding, whatsoever, of what DR is and how it's measured.

The DR of the captured photo is, in a very real sense, how much lattitude you have to work the file. But if, for example, you do not consider detail that lies above the 50% NSR range to be "usable", then using the read noise or the 100% NSR for the noise floor will not give you the information you want to know.

So, the "problem" with DR measurements is not the measurements themselves, but people's failure to understand what information the measure gives. Particularly tragic is when these same people blast DxOMark's information as "magic numbers". To paraphrase Arthur C. Clarke, any sufficiently advanced subject will seem like "magic" to those that don't understand it.

This is not any different then the never ending RAW vs jpg debate. RAW, like more DR, simply gives more processing options. If you have no use for those options, then there is no advantage to RAW (more DR).

And DR, like RAW, requires the appropriate processing skills to make use of greater DR for the photos that would benefit from it.

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