|
Brendan, here is a short summary of what is going on:
- The stripes are caused by PWM. The frequency of the stripes changes as the PWM changes. We went as high as 16KHz for the TEC switching frequency but short horizontal lines still formed a visible pattern in the images.
- The stripes are also caused by the cold finger left floating. Previous work with Canon 450D showed that the stripes go away once the cold finger is grounded.
- The back of our sensor is metal. However, unlike 450D and it cannot be grounded or the camera stops working.
So, there are two possible solutions:
1. Electrically connect the cold finger to the sensor.
2. Electrically insulate the cold finger from the sensor and then ground it.
For the option 1 we could not find any thermal paste that is electrically conductive. Most of them are designed for cooling CPUs and have good thermal conductivity but zero electrical conductivity.
Eventually Cedric (from the French forums) thought of using solder paste and it worked. Moreover, the camera read noise went down by about 15% as well.
However, I am not sure about the long term suitability of the solder paste.
Option 2 - I found a very thin (0.1mm thick) thermally conductive film that is also electrical insulator. I used it between the cold finger and the sensor, while also applying some Arctic Silver 5 (AS5) thermal paste on both sides. The thermal films and pads are usually used when the parts are pressed together by a large pressure and we don't have that - hence the use of AS5.
Then I could ground the cold finger without worrying that it may short the back of the sensor.
The preliminary results showed that this worked really well. I could not see the stripes at any switching frequency.
However, there will be a thermal loss because an extra interface between the sensor and the cold finger. I will need to measure this.
(Also I noticed an increase in the camera read noise but I swapped several capacitors at the same time which may be cause of this).
|