Dark frame Analysis of CDS-5D. by Brice Grassin,

This analysis report was based on the dark frame images of cooled 5D markIII and these were taken at the temperature range +20C ~ -30C .
Below linked images are the unscaled original dark images.


Below is the results of analysis by Brice Grassin,( contact@bgimage.fr )

First, let’s start to see histograms about pixel population around the default Canon 5D mkIII’s pedestal (sort 20°C -> -30°C):


-> Nothing unusual; the shape is pretty good at low temperature with a population quasi inside the offset +/-128 ADU.

High frequency waves in the population are typical of the FPN (fixed pattern noise), present at all temperatures on 5D cameras.


Let’s see now closer, with a deeper analyse and especially how main noises are affected by the low temperatures:

-> Standard deviation of the dark frames (in blue) are very well reduced, even when temperatures are not extremes. Between 0°C and -10°C it seems to have a pivot point.

Beside, RON’s sigma (in grey) is down too with low temperatures (unexpected!). This fact could be interesting in some configurations, when RON (read on noise) is a limiting factor.


After noises, let’s take a look on how the cooling device is efficient:


-> The cooling system working very well in fact! >90% of efficiency is reach before -10°C.

Noise reduction is well link with the cooling efficiency and the amount of reduction is very significant just from 0°C.


Beside, under -10°C the system looks to saturate; gain is <2% from -10°C to -30°C.

This fact tells us that the 5D MkIII CMOS can’t do very much in question of noise. Mostly when we know that the sensor can not go under -48°C because of micro-lens.


And for the end, let’s check how the cooling device helps in dynamic range:

-> Of course we are speaking about dynamic range in a single frame here, where the theoretical maximum is around 9.6 stops for the 5D MkIII @ISO 3200.

But this value is for a very short exposure time, far of the 120s here.

So this is very interesting _in absolute_ to see that the dynamic range reach 9.3 stops @ -30°C. And it represents a gain close to 24% from the same exposure time @ 20°C.

For sure, almost 2 stops of gain per frame can be significant after compositing 10 or 20 frames in deepsky imaging.


Well, I hope you find all of this interesting. I think you will take time to read all those graphics.

Of course there is a lot of things to say about all these datas if we take time to make links and afterthought. It is a work through the time and with experimentations to measure gains.


Brice Grassin,
BGImage • Imaging solutions & consulting (France)