The SIGMA DP1 is the most fun $28 ever bought me.

allineedislight wrote:

It is funny and a little bit astonishing that an AFE was included in the Quattro cameras, whereas you deem this setup as 'unpopular'... You mentioned a drawback attached to the use of an AFE, but there must also probably be advantages, otherwise people would not be using them at all?

The reason for introducing AFE was to achieve better high-iso images. It is understandable that Sigma tried to expand the high-iso capabilities with the DP1x and DP2x to remain competitive with other manufacturers.

On the other hand, AFE in the DP1x and DP2x causes two severe problems which are the reasons why these models are comparatively unpopular:

First, with AFE it is easier to clip highlights, and it is more difficult to recover clipped highlights in SPP using the raw files. The smooth highlight roll-off of the non-AFE models is probably one major reason why the classic Foveon looks so film-like (the other one is the ability to accurately model local tonality changes with the foveon chip).

Here is an example image from http://www.mittenkontakt.de/2011/08/sigma-dp2x-versus-dp2s-teil-2/ that shows the issue (look at the highlights at the stones):

...

DP2x - troubles with highlights (source: http://www.mittenkontakt.de )

Second, AFE leads to banding in ISO 1600 and 3200 that is very visible in black and white. In contrast, the DP2s shows no such banding and in fact can be used at ISO 1600 and 3200 - with SPP version 3.5.2 it even looks very good in b&w (a bit like Tri X film). So if you like to use your DP2 also at high iso in b&w you should look for a DP2s rather than for a DP2x.

Here is an example of the banding of the DP2x (using the DP2x example raw files of "mittenkontakt" provided here: http://www.mittenkontakt.de/2011/08/sigma-dp2x-versus-dp2s-teil-3-high-iso/ )

...ing at iso 3200 - DP2x, developed with SPP 3.5.2

Again, thank you very, very much for your insightful and very useful explanations.

I think I was completely wrong to assume that a hardware-embedded AFE was totally equivalent to a post-operation carried out in software. The answer probably lies in the indications given by Ted about the strength of the signal flowing from the sensor to the AFE and then to the ADC: without the AFE, some kind of underflow must occur, resulting in information loss that no post-treatment can ever revert.

And on the other hand, I understand now what are the drawbacks of using an AFE.

Again, thanks to all for taking the time to explain the technical inners of the Foveon world to newbies, it is much appreciated.