A Peculiar Effect in a SD9 vs. SD10 noise test

This may be of interest to those of us that still shoot real Sigma DSLRs ;-)

It's raining, so I decided to compare noise, SD9 vs SD10. I set a gray card under a lamp to try and get some tone grading so as to observe the onset of the inevitable blotching.



Took a shot with each camera, portrait mode to get the whole card in, deliberately de-focused to make noise effects dominant.



For same settings w/ same lens, pronounced rings appeared in the SD10 shot. Leading to speculation that the SD9 is 'better' than the SD10 in sorting out tone variation under low light conditions. Is it due to the different 'hot mirror' arrangement (SD9 is on-sensor) and/or the microlenses (none on SD9)?

--
"I keeps my licks in my head, so no-one else can play 'em"
Ted
SD9, SD10, GH1 and a lens or two

Most probably different sensor response, which causes some digitization effects while color conversion and black level subtraction.

I'd think that when you look at separate channels using RawDigger, then you can't see any gradations. Or do you?

ArvoJ said:

Most probably different sensor response, which causes some digitization effects while color conversion and black level subtraction.

I'd think that when you look at separate channels using RawDigger, then you can't see any gradations. Or do you?

Click to expand...

Good shot, Arvo!

Went to RawDigger and looked at the separate channels for each file:

In the channels for the SD10 shot, the onset of really bad noise was more sudden and occurred in different positions horizontally. These non-aligned positions caused the "ringed" appearance. The SD9 onset was more gradual and more or less in the same position horizontally.

Mystery solved.

--
"I keeps my licks in my head, so no-one else can play 'em"
Ted
SD9, SD10, GH1 and a lens or two

xpatUSA said:

ArvoJ said:

Most probably different sensor response, which causes some digitization effects while color conversion and black level subtraction.

I'd think that when you look at separate channels using RawDigger, then you can't see any gradations. Or do you?

Click to expand...

Good shot, Arvo!

Went to RawDigger and looked at the separate channels for each file:

In the channels for the SD10 shot, the onset of really bad noise was more sudden and occurred in different positions horizontally. These non-aligned positions caused the "ringed" appearance. The SD9 onset was more gradual and more or less in the same position horizontally.

Mystery solved.

Click to expand...

My first guess was that the SD10 image simply is more over exposed, probably due to the micro lenses.

Roland Karlsson said:

My first guess was that the SD10 image simply is more over exposed, probably due to the micro lenses.

Click to expand...

Probably not what was meant?

xpatUSA said:

Roland Karlsson said:

My first guess was that the SD10 image simply is more *over exposed*, probably due to the micro lenses.

Click to expand...

Probably not what was meant?

Click to expand...

For some reason I thought the detail was from the lamp above

So ... it is from the grey card and is close to the noise level.

As far as I know, the SD9 and the SD10 have the same sensor without and with micro lenses.

The posterization you see in the SD10 image is a totally mystery to me. SD10 should be much better at low levels than SD9. You say, mystery solved. Can you explain that too me?

Roland Karlsson said:

So ... it is from the grey card and is close to the noise level.

Click to expand...

The shot was taken to get a range of noise level. It was not accidentally "close to the noise level".

Roland Karlsson said:

As far as I know, the SD9 and the SD10 have the same sensor without and with micro lenses.

Click to expand...

Correct.

The SD9's Foveon F7X3-A91 has an IR filter on the sensor and no microlenses.

The SD10's Foveon F7X3-C9110 has no IR filter on the sensor but it has microlenses.

Roland Karlsson said:

The posterization you see in the SD10 image is a totally mystery to me. SD10 should be much better at low levels than SD9.

Click to expand...

I can see no posterization within the normal meaning of the word.

The F7X3--A91 has a fill factor of 54%. The F7X3--C9110 has a fill factor of 71% - slightly better?

Roland Karlsson said:

You say, mystery solved. Can you explain that too me?

Click to expand...

I can. I went to RawDigger and looked at the separate channels for each file:

In the channels for the SD10 shot, the onset of really bad noise was more sudden and occurred in different positions horizontally. These non-aligned positions caused the "ringed" appearance. The SD9 onset was more gradual and more or less in the same position horizontally.

--

"Divided by a Common Language"
Threadstopper Ted

xpatUSA said:

In the channels for the SD10 shot, the onset of really bad noise was more sudden and occurred in different positions horizontally. These non-aligned positions caused the "ringed" appearance. The SD9 onset was more gradual and more or less in the same position horizontally.

Click to expand...

OK ... any theories on why?

Roland Karlsson said:

xpatUSA said:

In the channels for the SD10 shot, the onset of really bad noise was more sudden and occurred in different positions horizontally. These non-aligned positions caused the "ringed" appearance. The SD9 onset was more gradual and more or less in the same position horizontally.

Click to expand...

OK ... any theories on why?

Click to expand...

No.