This night I dreamt about zero cloud forecast for the next 3 nights. Unfortunately it has nothing in common with reality – the weather is as it is and there is no single symptom telling it will improve soon. So I continue preparations for the astrophoto season 🙂
Insprired by this link I decided to add passive cooling to my QHY5 guider camera. The implementation is very straightforward assuming your QHY5 camera has already cut out piece of the board under the CCD chip. Then all you need to do is to place there a piece of aluminum (or copper) that will fill the space between the chip and the camera case. The PCB cut out is 9x10mm large, so you need to choose the proper cold finger size. In my camera the distance between chip and camera case is 15.3mm, so I choose aluminum rod 8mm diameter and 15.3mm length. I drilled a whole at one end for 3mm bolt and fixed the cold finger to camera case. Both cold finger surfaces are covered with thermal grease. 

qhy5finger

One important thing is that cold finger surfaces need to be flat, so it will contact with case and CCD chip with the whole surface area. And that is all – camera was assembled and I performed a few tests. Three series of dark frames were taken with and without cold finger. Exposure times were 3, 5 and 30 seconds, and each serie lasted for 5 minutes. There was 15 minutes break between each serie to cool down the camera. Ambient temperature was 21C. Here are sample picture crops from the images taken at the end of each serie (click to enlarge):

qhy53s
qhy5-5
qhy5-30
The images were stretched to the first 20% of the histogram. The cold finger work is clearly visible. In the table below there are a few numbers calculated in Maxim from the camera image 300x300px selected crop. 

Exp. time Std dev. HOT Std dev. COLD Hot px val HOT Hot px val COLD
3s 1.18 0.65 38 17
5s 1.89 0.89 62 30
30s 10.5 4.39 204 92
 
It’s quite easy to estimate that noise was decreased by the factor of about 2.1 that corresponds to the chip temperature drop of about 7C (for CCD chips the temperature noise doubles for every 6C temperature increase). Taking into account our investment for this mod the benefit is quite impressive. 
The last picture is the graph of single selected hot pixel value growth in exposure series for 5s exposure with ant without cold finger mod. Horizontal axis represents single exposures, so each value means 5 seconds (plus download time).
qhy5-graph
X axis indicates sample number – in our case it is each 5 second exposure. For non cooled camera we can see that hot pixel value increases during this 5 minutes time, but even at the end of this period it is not yet flat. This growth indicates increase of CCD chip temperature during session until the moment it reaches kind of thermal equilibrium with the environment. For the “cold finger” version we can easily see that hot pixel value is lower, and equilibrium is reached much faster – after one minute (12 samples) the plot is reasonably flat. This is quite important information however, especially for non cooled QHY5 cameras – if you want to take dark frames for calibration do not do this at the session start. Take at least 2-3 minutes of test exposures, then take darks when the camera warms up and reaches equilibrium with ambient.
Cold finger solution seems to work fine and I have high hopes it will improve stability of guiding process, especially facing the fact that hot pixels are the major pain of QHY5 camera. It probably also benefit when using this camera for making planetary exposures.