Merrill and Quattro Noise Reduction

I know there is a risk to trigger annoying discussions again. Therefore, just to start with, I don't want to say that any of the sensors is superior to the other. And I own an SD1 and an sd Quattro H. So, no bias from my side.

Some of you may remember that I once complained about a certain grittiness of the Quattro files vs. a smooth rendering of the Merrill files. There was some advice from the crowd like reducing sharpness, detail and NR in SPP and doing sharpening and NR in e.g. Raw Therapee afterwards.

Analysing the sd Quattro H test pictures from DPREVIEW, I noticed the following which might be the major cause of the problem.

I look at the results in SPP at 300% (sorry for that).

Always sd Quattro H on the left and DP1 Merrill on the right (as you can also see by the difference in size). Both at ISO 100.

No NR : More noise on the Quattro, Merrill edges better defined.

NR at middle position for both (as well as sharpness = 0 for both) : sharp edges on Merrill, gritty edges on Quattro.

sd: Ch NR low, L NR at middle DP: NR at middle

sd: Ch NR at middle, L NR low, DP: NR at middle

NR at middle, sharpness at -2 for Quattro

Here is my interpretation:

SN at pixel level of the sd is lower because of the smaller pixels.

NR for the Merrill cleans up nicely without "destroying" edge detail.

NR for the Quattro leads to grittiness of the edges.

The problem seems to be with the luminance NR.

Chrominance NR doesn't have a big effect on the picture but also doesn't introduce grittiness.

Reducing sharpness makes the effect less obvious but still it is visible.

The conclusion may be that the Quattro files need a different Luminance NR algorithm.

To prove that it is not just due to the difference in noise from the RAW files, I compared the Merrill files at ISO 400 and ISO 100. The ISO 400 is for shure more noisy than the ISO 100 of the Quattro. But still, the edges are better defined after NR.

DP ISO 400 vs ISO 100: both NR at middle