Ysarex Loc: St. Louis

Right, and since what's being talked about here is a specific brand/type of dual gain sensor then that's not what we're talking about is it.

You're blowing smoke now and you remain wrong.

The video author specifically referred to the change in read noise in a Fuji XT-3 that we know is the result of Fuji's use of a Sony dual gain sensor. The graph you presented for the Nikon D6 does not show a "dual" change. Something else is taking place there (don't care what) that you're putting up as a smoke screen.

The Sony sensor in the Fuji XT-3 uses Aptina's DR-Pix dual capacitance switching technology which is implemented in hardware on the sensor. It is not implemented in the ADC. That is the technology that is being talked about here. Thom Hogan correctly described this and noted the Aptina patent, the Sony connection and the Fuji connection. Quote: "This dual gain approach was first realized by Aptina, then licensed from them by Sony." https://www.sansmirror.com/newsviews/2016-newsview/jan-mar-2016-newsviews-2/dual-gain-becoming-the-norm.html You can look up the Aptina DR-Pix dual capacitance tech yourself. It is not a software algorithm and you remain wrong.

Racmanaz Loc: Sunny Tucson!

How about you two agree to disagree? Otherwise, it's just going to run forever lol. By the way, I apologize for my behavior towards you, it was uncalled for.

Ysarex Loc: St. Louis

Thanks. And I'm sorry if I overreacted to the video -- I can see that I have that tendency.

As for agreeing to disagree, no. This is clear and simple and he's wrong. Yes, he'll drag it out for pages and blow smoke and try every kind of sophist BS he can dig up and then eventually resort to bullying and name calling.

selmslie Loc: Fernandina Beach, FL, USA

If the camera can do a better job with noise starting at ISO 800, why not use the same approach from base ISO through ISO 640?

If you can't answer that then you don't really understand what's going on.

Ysarex Loc: St. Louis

Because the low-light, high conversion gain mode of the Aptina DR-Pix sensor design used with higher ISO values would otherwise limit full well capacity, maximum (SNR), and total dynamic range in bright light conditions. I provided a link near the beginning of the thread: https://www.uglyhedgehog.com/t-712451-1.html#12576056 Aptina's DR-Pix design is implemented on the sensor and not implemented as a software algorithm.

selmslie Loc: Fernandina Beach, FL, USA

That would make sense if it weren't for the fact that the plot of Read Noise vs. ISO:



shows a difference of 0.75 stops between ISO 640 and 800 but the difference in DR:



is only 0.06 stops.

When you look at DR (the inverse of the Noise curve) there is very little difference at the dogleg. This explains why he had to pixel peep to see any difference.

Bringing the DR down to 10.3 from 10.36 is not much to lose sleep over and it's not very likely that the highlights would have been in much danger from 0.06 stops of excess exposure.

Ysarex Loc: St. Louis

Take it up with the sensor designers; those are the reason they listed for the rationale behind it. Shocking to think that they would have proceeded with it's manufacture without consulting you. And also got to wonder why it's been so successful.

selmslie Loc: Fernandina Beach, FL, USA

It’s because it has so little impact on the cameras’ DR. That’s why it’s so hard to demonstrate.

If you can’t demonstrate it easily it’s probably not that important.

Ysarex Loc: St. Louis

Don't you mean bringing the DR down to 10.3 from 10.68.
You noted the difference in DR between ISO 640 and 800 as 0.06 stops (increase) and then added .06 to the DR at base ISO of 10.3. You compared the wrong values. The difference you should have calculated was between the expected DR loss at ISO 800 without the capacitance switch and then with the switch. That would have been around .38 stops. Without the capacitance switch the DR at ISO 800 would have dropped to 9.18 but instead measured 9.56. The capacitance switch then added .38 stops of DR that would have been lost.

selmslie Loc: Fernandina Beach, FL, USA

Check your arithmetic. I'm talking about waht the difference would be at base ISO.

Ysarex Loc: St. Louis

You added .06 to the base ISO claiming that was the difference the capacitance switch made. You got that wrong.
If you're going to try and determine the difference the capacitance switch makes you have to do that at the same ISO value. The DR you would have had without the switch and the DR you get with it -- that's the difference the switch makes. I checked your arithmetic and you did it wrong.

selmslie Loc: Fernandina Beach, FL, USA

To be more precise, extend the numbers look like this:
That's more accurate.

But the dilemma remains. How do you demonstrate a difference in noise of between 0.4 to 0.6 stops that close to base ISO?

Try it. I'll be surprised if anyone can do it.