Peltier Cooling for Canon EOS 350D

This page describes a modification I performed on my own camera. If you attempt this (or anything similar) yourself then you do so entirely at your own risk.

Summary of Results:

This modification was carried out with easily available materials.

Some 22mm copper pipe was hammered flat - then filed and whetstoned flatter for better thermal contact with the back of the CCD and with the Peltier. A standard CPU heatsink/fan was used. The Peltier chip itself was salvaged from a Maplin Drinks cooler but it's a TEC1-12703 and is readily available on eBay, for instance. It consumes 2.5A at 12V. The mod increase the weight of the camera from 600g to 850g.

Here are a couple of images before and after the insulation was added:

The graph below shows the standard deviation of noise in successive 5 minute dark frames done at room temperature (21C). The Peltier was switched on immediately after frame 21.

The image showing frame comparisons also shows a 1/4000sec bias frame.

The final stage will be to seal the camera body in a plastic bag with desiccant before using it outside - I hope that will reduce potential dew problems.

Further details:

Here are some further details on the procedure used.

This first image shows the rear of the camera stripped down to gain access to the CCD (shown).
To do this, I followed instructions available on the web such as:
Ash's Modified Rebel XT Page

I have also removed the left hand copper clip so the copper "pipe" will be able to lie flat on the back of the CCD and I've removed the mini-circuit board on the left that holds the sockets for remote control and video out.

This shows the sockets desoldered from the mini-board:

A couple of wires are soldered on in the position of the remote control socket so a new one can be constructed later, outside the camera body:

Here it is back in position. The CCD has thermal paste applied before assembly.

Here are the notches cut in the flattened copper pipe to allow it to pass throught the left side of the camera and to allow the circuit board and the rear cover to be replaced. (The top right notch was cut in error!)

The copper "pipe" is held in place with tape before assembly.

Here is the camera almost assembled. The circuit board actually holds the copper "pipe" quite tightly.
The copper tube is 2mm thick when flattened. This is very slightly fatter than the gap between the back of the CCD and the circuit board above - so that board holds it in place.

The rear cover is now in position. When the insulation is added later, it is impossible to access the digital (USB) port so I've left this cable permanently attached (for PC connection).

Here is the side view before cutting a slot in the side panel so it can slide over the copper.

Now the peltier, heatsink and fan are tightly wired in place:

Finally, pipe insulation and a remote control socket is added:

Future Improvements:

To improve this further I intend to cut a strip fom 2mm copper plate (rather than flattening a pipe!)
Being completely flat this should improve the thermal coupling between the copper and the Peltier and CCD.
Also I'll seal the electronics close to the CCD in silicone or hot glue or similar to reduce the condensation risk.

Update: 29 Nov 2009

I performed the MkII mod yesterday.

Here's the copper strip I soldered together and cut to shape:

I've made the strip wider than the previous piece of copper pipe.
That and the double thickness means it transfers the heat much better. Since it is completely flat it gives excellent thermal contact.

I had to then do a bit of extra shaping so it would not collide with other components:

In position in camera:

Here are two 15minute subs before and after cooling (same scaling on both) - all done at room temperature.

The cooling is so dramatic that I can't meaningfully measure the noise reduction - I had to 60min subs to measure it. It's well over 30C of cooling. The "after" picture is almost indistinguishable from a bias frame except in the amp glow areas. I really need to do an amp-off mod as well to get rid of it.

Later calculations showed that the standard deviation of the pixels in the difference image between the two 60minute subs is 8.9ADU.
Compare this with a standard devation of 6.9ADU for a diffence image between two 1/4000s bias frames.