With a dremel tool and some cutting and fitting work with a pair of pliers and tin snips, I was able to trim the copper finger and get the whole assembly to fit together better.
Before I put the parts together, I smeared a thin bead of thermal compound on the CCD assembly plate. This should help improve the transfer in the absence of the gel pad.
Seen here is a larger gel pad that I had in place originally to provide some continual connection with the LCD screen. As the original effort cooled this area, I wanted to provide something similar for testing.
I've been testing via a few standard ways:
- I do a "drop from room temperature test" where I run the camera for 1 minute exposures with a 10 second gap. This runs for 60 exposures, or a little over an hour. After the set runs, I can pull the images and fire off a bat script that pulls the EXIF temperature. This gives an idea of what the hardware in the camera considers the temperature.
- I run the above test in two environments: in the open on my desk where the ambient temperature is about 15C and in the fridge where the ambient temperature is 6 C.
- I also run a set of 20 minute darks in the fridge to give a good idea of how the camera can be kept cool over a longer period of time.
- For consistency, I've been running these tests with the Peltier device at maximum cool setting (-20C on the dial).
- The camera with a closed base and gel pads on the metal work cools better than it did before. This is not a surprise. What was surprising was how quickly it settled at thermal equilibrium in the fridge. Also the thermal equilibrium temperature as reported by EXIF was 8C. Thermal equilibrium was reached in 70 minutes in the fridge, dropping from 25C to 7C. When the sequence started, there were a few frames where the temperature went up to 26C before dropping.
- With an open base, the thermal equilibrium was reached at about the same rate: 65 minutes from 26C to 7C.
- With the closed base on the finger, the I missed the test in the fridge, however, I did do a test on the desktop. I compared the final frame for noise profile.
- For one final test, I decided to remove the thermal transfer pad on the LCD and replace it with cardboard in an attempt to insulate the finger. The goal was to help the cooler isolate and pull heat just from the inside of the camera via the copper path. I also routed the thermal probe from the middle of the heat sink plate to the cold finger just inside the camera body. Thoughts were that this would help show what a real temperature inside the camera is. When testing in the fridge, the thermal equilibrium was tested from 18C to 8C. The test did not start at a warmer temperature. Final temperature was reached in 45 minutes. I also added my amp gauge to better identify when the cooling plate reached a target temperature.
I found that the device will not cool much below -5 C. This was about a 10C drop from the interior of the fridge. This was determined by adjusting the dial to see when the cooling cycle would start pulling current. I noticed that getting to 0C was pretty quick and that the final -5C was slow.
In testing on the desktop in 15C ambient temperatures, 5C was reached within 10 minutes. 0C was reached in 20 minutes. Holding the temperature at 0C results in a draw of 2.65A (vs 3.46 A during full power). This should give a bit more life to batteries in tough situations.
Next testing to do:
- Apply a thermal transfer material between the heat sink and the copper finger. Now it's just pressed with elastic bands. I am not convinced that the thermal gel pad has much benefit. When I ran the device with no thermal load, I could feel coldness on the exposed metal plate faster than through the gel pad.
- Try routing the thermistor probe deeper into the camera or give it a secure attachment to the finger. Challenge is that it should be removable and repositionable.
- Possibly disassemble the Peltier device to remove unneeded mass. I already have a light weight heat sink for the hot side. If I can figure a way to mount it onto a stiff, insulating material (considering a wood plank now) then that will be the next step.
- Do a quick run of 10 minute at 400 ISO darks to compare vs previous library to see if this reduces thermal noise. My tests up to now indicate that thermal noise is better controlled than before.