Michael Reichmann's take on smaller full frame Sony FE mount lenses

Started Dec 18, 2013 | Discussions thread
Anders W Forum Pro • Posts: 21,468
Re: Totally different measure

Erik Magnuson wrote:

Anders W wrote:

Why wouldn't the three measures I focus on be just as relevant for what is discussed here than the one you focus on?

One because of the derivation as opposed simply reading off of the graph.

So are you telling me that the figures in the DxO graphs are not "derived" in the sense of being computed from the raw data so as to be comparable given a certain vantage point, just like mine? And are you suggesting that using measurements that are not "derived" in the above sense would somehow be preferable? If so, on what grounds?

Two because my opinion is that most people are talking about mid-tone noise as the most significant advantage of larger sensors.

And why would selecting a measure on the ground that it is supposed to maximize the advantage of larger sensors be appropriate in the present context? That simply means that you are cherry-picking the measure to suit your point of view.

That deep into the shadows you are talking mostly about minor differences in read noise

On what empirical basis are you calling these differences minor in comparison to the other differences we are talking about?

which really doesn't depend as strongly on sensor size.

If you are right about that, the implication is simply be that the efficiency advantage of smaller sensors is even greater for read noise than for photon noise.

Most people want to know:

  • given an acceptable noise threshold (DxO uses 30db for SN 18% for low light)
  • what's the minimum "equivalent" exposure that gives the same amount noise and midtone brightness for a fixed size output.

How do you know that this is what most people want to know?

Because by numbers, at least 40% are Canon shooters and Canon sensors do poorly at base ISO DR. So I only need 10% of others to agree

So are you saying that Canon shooters are just a bunch of wishful thinkers fervently trying to rationalize their poor purchase decisions? Or that they are all OOC jpeg shooters for whom the ability to push shadows in RAW at base ISO is inconsequential?

BTW: If you look at my efficiency comparison, you will see that I did not include any Canon FF or APS-C sensor, since my ambition was to compare the best sensors in each size category. I did include some Canon cameras at smaller sensor sizes, where the sensors Canon uses (I don't know if they make them) are fully competitive.

The DR at high ISOs tells more about shadow noise, and shadow noise is what people typically finds most bothersome at higher ISOs.

It depends on what you shoot in low light. For example if you shoot stage performances, DR/shadow noise could be more important. If you shoot indoor sports (i.e. a typical high school gym), midtone noise could be more important.

The scenario to which I referred you and others for a comparison of DR versus SNR-18% was the well-lit DPR studio scene, not very different in relevant regards from what you would expect with indoor sports. Had it been something like a stage performance with large shadow areas, my point would have been driven home with even greater force.

Shadow noise can be visually minimized for some types of shots by increasing the black level

Increasing the black level just replaces one evil by another (which I personally find just as bad or worse).

or differential noise reduction (i.e. do you need detail in your shadows? You certainly need it in your mid-tones.)

Regrettably, the NR algorithms at our disposal (at least those I am familiar with) are not very good at such differential noise reduction and working around the problem by manually applying differential amounts of NR to different parts of the image is often difficult and/or time-consuming.

Furthermore, the presence of fine detail reduces the visibility of noise. This is true regardless of whether it is shadows, midtones, or highlights we are talking about. Consequently, it is sometimes a better idea to add more NR in the highlights rather than in the shadows. The image below is an example where I did exactly that. The sky, where there is hardly any detail to be preserved, was given more NR than the more detailed darker foreground.

I should add here that my point is certainly not that there is no reason to look at anything but DR. Rather, my point is that if we insist on using one single number rather than multiple measures, then DR is preferable to SNR-18%, at least for the range of sensor sizes and sensor capabilities currently at issue.

Neither of the two DR-related measures of efficiency that I use depends on the full-well capacity of the sensor. They are a matter of quantum efficiency and read noise only.

There is some strange non-linearity in the DR graphs that's not in the other graphs.

There is nothing strange about this non-linearity. It just means that for many cameras, read noise (as measured in electrons) varies across the ISO range (i.e., is larger at low than at high ISOs). Consequently, you need to assess efficiency at lower as well as higher ISOs, just as I do in my efficiency comparison.

 Anders W's gear list:Anders W's gear list
Panasonic Lumix DMC-G1 Olympus OM-D E-M5 Olympus E-M1 Panasonic Lumix G Vario 14-45mm F3.5-5.6 ASPH OIS Panasonic Lumix G Vario 7-14mm F4 ASPH +28 more
Post (hide subjects) Posted by
G L
G L
G L
G L
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow