Amazing lens, incredible sharpness

Anders W wrote:

Great Bustard wrote:

Anders W wrote:

Great Bustard wrote:

thk0 wrote:

Ah. I see I misread your post. You are only making a point about total light gathered, not how it might translate into IQ.

Actually, I'm saying equivalent lenses will put the same total amount of light on the sensor for a given shutter speed, and that this will result in the same noise for equally efficient sensors.

I was questioning whether eg 4x more total light (ff at same f-stop) gives exactly 4x less noise.

4x as much (2 stops more) light will result in 1/2 as much (1 stop less) photon noise. The other player is read noise (the additional noise added by the sensor and supporting hardware).

The read noise becomes dominant only in the portions of the photo made with very little light -- most of the noise in the photo is dominated by photon noise.

True. But a) the portions of the photo made with very little light are those where quality shortcomings are most visible...

Only if they are pushed way up, such as with a very strong tone curve or very high ISOs.

I rarely shoot at very high ISOs but I often push shadows, sometimes quite a bit, at lower ISOs. And in both cases (high ISOs or lower ISOs with some shadow pushing), that's precisely when the shortcomings become most visible.

...and b) the lower the overall light level, the greater the portion of the photo where read noise makes its presence known. Both consideration are of considerable importance, in my opinion.

Let's quantify that. The EM5 has a QE of 53% and a read noise of around 2.5 electrons per pixel at the higher ISOs, which means the read noise matters more than photon noise for signals less than 12.5 photons per pixel, which is 6 stops below full "saturation" at ISO 6400, or 2.3 stops above the noise floor.

So, yes, read noise most certainly matters more and more as the light dims, but the photon noise is still dominant for most of the photo until you get to outrageously low levels of light (e.g. much lower light than the conditions you would use ISO 6400 for).

That's not to say that read noise is irrelevant -- perish the thought! -- but it is to say that it is of secondary, not primary, importance for the vast majority of lighting situations.

What does it matter that photon noise is dominant in most of the picture for most pictures shot if it's in the deep shadows that the shortcomings become visible. I can't remember that I have ever been bothered by highlight noise and only sometimes by midtone noise (and then usually at higher ISOs). However, I remember a lot of occasions where I have been troubled by shadow noise, at base ISO and higher ISO alike.

And apropos of sensor development: Note that the QE of the E-M5 sensor exceeds that of the old 12 MP MFT sensor by about 20 percent (about 10 percentage points). The old has a QE of 40+ percent according to Sensorgen (E-P3, E-PL1; G1 excepted) and the new is 50+ percent. That's about 1/4 EV better as far as photon noise/max SNR is concerned. However, in terms of DR (where read noise weighs in more heavily) as well as in the perception of most people (based on the reports I see on this board), the new sensor is about 2 EV better. Why is that?

In fact, dare I say it, read noise is all but irrelevant for modern sensors at and below ISO 1600, and likely a couple of stops lower light still.

You dare say it, but be prepared for a few objections.

What you say is all part of a thread I've been thinking of starting in the Open Forum on DR.  Give me a little while to start the thread, and I'll PM you a link to it.