POLL: Dynamic Range - Do we need more?

[Rick Knepper]

Not that I see a big advantage in noise at high ISO (low light performances) fromthe cameras well known to have good DR...

Look at dxomark.com's measurements of DR: Out of your 4 cameras, the Nikon D810 has the lowest DR at ISO 25600. And your comparison shows that the D810 also has most visible shadow noise at ISO 25600.

This is an area of semantic confusion. DR is relatively meaningless as intended at higher ISOs, because you can always use a lower ISO to get more DR with little or no increase in absolute SNR;

I very much disagree. DR at high ISO is not meaningless.

Would it give you context to learn that Mr. Sheehy is a sensor designer?

If you have a given amount of light available, and that amount is only adequate for a shot at ISO 25600 with no clipped highlights, then the DR increase from reducing the ISO cannot be utilized anyway.

Example of this:
Assume that your camera have 8 stops of DR at ISO 25600 and 10 stops of DR at ISO 6400. You take a shot at ISO 25600, using all available light and just barely avoiding to clip any highlights. You now have a photo with 8 stops of DR. Then you say: I could have more DR at a lower ISO, so you lower the ISO to 6400. Here the camera has 10 stops of DR, but available light is unable to use the top two stops of the DR range. So again you end up with 8 stops of usable DR.

In other words: A camera's DR at high ISO is a very direct indicator of a camera's ability to produce low light photos with low shadow noise.

 

[Allan Olesen]

Did you read my previous post beyond the part that you quoted above?

Yes, I did. And all the information you provided there, I already knew and agree to, with two exceptions:

1. I do not share your conclusion based on that information, and I have explained why: Measured high ISO DR says a lot about a camera's shadow noise in low light photography. I will add that even though it is derived from read noise, the DR number is much more directly usable than the read noise number.

2. I do not agree that the only DR which can be discussed in this thread is the highest available DR at any ISO. In fact, my first post was intended as a tongue-in-check hint that there is more to DR than people think when they say "We don't need more DR".

 

[Allan Olesen]

I very much disagree. DR at high ISO is not meaningless.

Would it give you context to learn that Mr. Sheehy is a sensor designer?

In technical discussions, I don't care much about CV. I care about facts.

And the facts are that the high ISO DR (as measured/normalized/reported by dxomark) has a direct and usable correlation to the shadow noise you should expect in a photo taken with a given, low amount of available light.

The rest is just words. Or semantics, to use John Sheehy's term.

 

[John Sheehy]

Would it give you context to learn that Mr. Sheehy is a sensor designer?

Fake News.

 

[John Sheehy]

Did you read my previous post beyond the part that you quoted above?

Yes, I did. And all the information you provided there, I already knew and agree to, with two exceptions:

1. I do not share your conclusion based on that information, and I have explained why: Measured high ISO DR says a lot about a camera's shadow noise in low light photography.

There are things that affect DR that do not affect absolute noise. That is the problem with your theory. Your theory has a hidden wild card: highlight headroom; which can affect DR without affecting absolute SNR.

I will add that even though it is derived from read noise, the DR number is much more directly usable than the read noise number.

I didn't say or suggest that an actual read noise number (in electrons or DN, per pixel) would be more intuitive or useful for comparative purposes. I said that the way DxO does DR is very useful precisely because they chart both DR and an inverse proxy for input-referenced read noise in the same chart, with their horizontally-offset data points, which normalize the trend lines for headroom. The trend lines do not represent DR; the dots do that. The trend lines reflect the inverse of read noise, normalized to the image, and RAW saturation values, which e- and DN charting would not do without crazy horizontal offsets because pixel size and count would further increase the need for normalization.

Your assertion that "DR" reflects noise performance at high ISO settings is unfounded. DxO's DR trend lines are NOT DR. They are inverted read noise trends; read noise normalized to both the image and RAW saturation.

1. I do not agree that the only DR which can be discussed in this thread is the highest available DR at any ISO.

I didn't say that higher-ISO setting DR can't be discussed; I said that it isn't necessarily what people think it is, and that the DR of the camera is the best it can do in a single shot. "The DR of the camera" and "the DR of the camera at various ISOs" are two completely different things. The former is what I think of when I hear "DR" without mention of a specific ISO or mention of the full set. The fact is, increased base ISO DR is a technological gift; DR at high ISOs can be increased very easily, just by using things like Canon's HTP. If Canon had only HTP and didn't offer non-HTP, all its high ISOs would have a stop more DR, but no decrease in noise; in fact, noise would increase slightly at the lower end of the ISO range.

In fact, my first post was intended as a tongue-in-check hint that there is more to DR than people think when they say "We don't need more DR".

Well of course; much of it is usually due to less added noise, which is always good unless a low bit depth can't handle the low absolute noise near black without posterization. However, DR has things that affect it other than noise.

 

[Rick Knepper]

Would it give you context to learn that Mr. Sheehy is a sensor designer?

Fake News.

So, you are in fact a sensor designer.

FYI, in the U.S., Trump has claimed "fake news" on every fact that has been reported about him, his campaign and his presidency, so now, "fake news" actually means "factual".