DP2 Merrill sensor read noise finally measured

Started Sep 5, 2012 | Discussions thread
Flat view
Aku Ankka
Aku Ankka Contributing Member • Posts: 591
DP2 Merrill sensor read noise finally measured

Hello all,

Thank's to docmaas providing me lots of test files from DP2 Merrill I managed to measure it's read noise and dynamic range at the base ISO.

I also redid my measurements for SD14 thank's to files from Rodrigo Cunha .

Thank you to the both of you for your efforts. It is much appreciated!

I had to rethink the measurement procedure in order to solve the issue of data not following proper distribution. Finally I figured that one can't measure (at least using regular methods) the color channels read noises individuallly, but that one has to calculate 'per pixel', not per photodetector. In hindsight this makes sense considering the operation of the pixel structure.

The following dynamic range figures are not normalized in any way.

First the old SD14:

Sigma reports the read noise to be under 70 electrons. This is however for a single photodetector (green) - for the whole pixel it is in the ballpark of 136e !!!

DP2 Merrill

Now I am happy to report that the sensor in the DP2 Merrill is very significantly better when it comes to the noise performance.

Using slightly different methods I got a bit different read noise figures, but it is safe to say the it is between 22.3 and 28 electrons per pixel - a massive improvement over the previous generation sensor. Regardless, this is very impressive for this kind of pixel architecture.

Due to the very large full well capacity (typical for the old Foveon as well) the "engineering" dynamic range can be upto 12.2 stops (at the pixel level) if the light is of broad spectrum, but less if dominated by a single color.

Comparing to Bayer sensor cameras is not straightforward due to very different pixel structure - the dynamic range of the Foveon depends quite a bit on the spectrum of the light, while this is not the case for Bayer sensors - in practise I'd guestimage that the Foveon DP2M DR for most scenes is somewhat under 12 stops.

Where this puts the new Foveon? The read noise is still much higher than the competition - the full well capacity is on the other hand still much higher than the competition has - this helps lots with tonality. If the light is well spread on all the channels (ie. greyish), the DR is competetive (better than Canon, worse than Sony), the more off from gray we go, the lower the DR will be. It is also good to remember that the DR of the Foveon comes more from capturing plenty of photons, while the competition gets it's DR more from minimal read noise. This means that while the engineering DR with some cameras is higher, the new Foveon sensor has very competetive DR if we set a moderately high quality standard (and use black and white output).

Please remember that the DR does not take account noise which comes from the color separation. This makes color image DR comparisons even harder against conventional sensors.

Anyhow, Sigma has done splendid job with the new sensor. It has very different characteristics from the Bayer sensors, and direct number based comparisons are difficult and doing it blindly can be misleading. At best the new sensor can be a formidable performer.

-- hide signature --


Sigma DP2 Sigma DP2 Merrill Sigma SD14
If you believe there are incorrect tags, please send us this post using our feedback form.
Flat view
Post (hide subjects) Posted by
(unknown member)
(unknown member)
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow