CO2 DSLR Cooler for Astrophotography

Using liquified pressurized carbon dioxide gas to decrease camera sensor thermal noise.


I think the usual noise referred to by photographers is actually ‘quantum mottle’, the spattered look of a photo when there are not enough photons arriving on the sensor to get an adequate coverage of the subject. It’s kind of like too light a coat when spray painting, giving the subject a speckled look. This noise in photography is decreased by having a more sensitive detector, better glass and coatings, greater aperture, longer exposure, etc.

In astrophotography, quantum mottle is a big issue too, accounting for the huge mirrors and very long exposures. But the long exposures also reveal a couple of other sources of noise, the biggest being thermal noise.

The longer the sensor is left reading photons, the more the little electron ‘well’ of each pixel is filled with thermal noise. The average amount of thermal noise per pixel is exponentially related to the temperature and linearly related to the exposure length. In addition, each little well accepts a varying amount of the thermal noise compared to its neighbors, so subtracting ‘dark frames’ from the exposed frames is commonly done to level the amount of background thermal noise.

But just like the arrival of individual light photons is somewhat random, so too is the arrival of the thermal electron. Thermal noise variance is proportional to something like the square root of the total amount of thermal noise. So not only does the total noise increase with temperature, but also the variability or randomness of the background noise also increases with temperature. This variance seriously limits the ability to remove the noise from the image later by dark frame subtraction. Astrophotographers usually average together a large number of dark exposures to enable subtracting more closely the right overall amount of thermal noise, but the variance in thermal noise on the light frame exposures keeps this from working very well for the dimmest areas of the subject.

So when an interesting photon from a galaxy or nebula is arriving only every few seconds or so, keeping thermal noise to a minimum is paramount to keep that photon from being lost in the soup of background thermal noise.

Many use dedicated astrophotgraphy cameras that have built in Peltier coolers, or use a refrigerator type arrangement over a DSLR to keep the sensor cool. Unfortunately, the astro cameras with a DSLR-size sensor are more expensive, and I have no room with my telescope mount for a box around the camera. Plus, I like to experiment.

Typical DSLR astrophotography exposure time for a single frame is 3 to 10 minutes, with as many exposures done as possible. The exposures are then stacked. Typically, accumulated exposure time is 4 to 8 hours for one subject. The camera ordinarily heats up inside to about 20 degrees over ambient during such long exposures.

Internal USB Temperature Probe

I installed a USB temperature probe inside the camera so I can monitor the temperature inside the camera. The particular model I used, the "TEMPerNTC" has a temperature sensor inside the USB plug, also has a remote wired probe, and comes with some drivers and software. The device montiors the voltage drop across a thermister at each location, and reports a voltage to the software. The device I think is actually four devices: thermistor sensors, a two channel analog to digital converter with serial output, and a serial to USB adapter. So the software includes a driver for the AD converter, and for the serial to USB adapter, as well as the reader software. A newer version is available, but may not have third party software available for it yet.

The drivers work well enough for the USB temperature probe, but the included reader software for Windows is terrrible. Fortunately, others have written some third party software for this particular device that works great. Albert Huntington put together some software that allows calibration, email notifications, network access . . . amazing!

Removing the stainless steel cup from the outside of the external temperature probe was difficult, requiring a fiber cut-off wheel on my Dremel tool, safety glasses, and a dust mask. The thermister is apparently coated with epoxy, then epoxied into the metal cup, and is difficult to extract. I went ahead and installed a longer cable, but may not have been necessary, as my USB hub is mounted on the telescope tripod.

Since I changed the cable, I went ahead and calibrated the USB temperature sensor using the third party software. I did the calibration at only one point (freezing), using a glass of ice water, after first sealing the sensor again with epoxy. The software has the ability to modify the voltage/temperature slope also, and can then accomodate calibration at two points (like freezing and boiling of water).

The photo above is the sensor assembly for my Canon XSi/450D DSLR camera, and includes a stainless frame, the sensor, overlying processing circuit board, and a covering stainless sheild. Gary Honis has an excellent online guide for replacing the IR cut filter in a Canon XSi that shows very well the steps involved in exposing the sensor assembly. I glued the thermistor and cables to the DSLR sensor mount stainless steel frame with some cyanocrylate glue. I initially insulated the solder joints with some epoxy, but later revised this and used a very short piece of shrink tubing, with hole on the side to expose the glass thermister to the frame. This picture was taken after removing the liquid CO2 expansion chamber, but before installing the CO2 gas tubing. I had to remove the temp probe temporarily, but cutting the CA from the metal with a blade was very easy.

Liquid CO2 Gradient Tubing

CAUTION: HIGH PRESSURE LIQUID AND GAS CAN CAUSE RAPID SUFFOCATION, CAN INCREASE RESPIRATION AND HEART RATE, AND MAY CAUSE FROSTBITE. Avoid breathing gas. Store and use with adequate ventilation. Do not get liquid in eyes, on skin, or clothing. Cylinder temperature should not exceed 52 deg C (125 deg F). Use equipment rated for cylinder pressure (liquid CO2 vapor pressure in compressed gas cylinders is 845 psi at room temperature, with higher pressures at higher temperatures). EQUIPMENT FAILURE FOR ANY REASON AT THESE HIGH PRESSURES CAN CAUSE SERIOUS INJURY.

I accidently released a 9 gram CO2 cartrige from a pellet pistol before it was completely empty. The small remaining volume of liquid CO2 propellant sprayed across my hand, turning my palm instantly white. Fortunately, my hand recovered without frostbite, but I instantly appreciated the substantial cooling capabilies of liquid CO2. The CO2 made a phase transition from liquid to gas, and made a substantial pressure change from 845 psi to about 14.7 psi, soaking up a lot of heat in the process. Wow, instant extreme cold.

So I started experimenting with liquid CO2 as a way to cool my DSLR for astrophotography. My first thought was to simply vent the liquid through a coiled tube inside the camera. As the liquid progresses down the tube, a gradient in phase change and pressure develops, and the tubing gets cold as the CO2 drops through phases and pressure.

I started by using the same 9 gram CO2 cartriges with a paintball adapter, but this photo shows the 23 ounce piantball tank (black) I later purchased that contains a siphon tube. Many paintball tanks contain anti-siphon tubes so that they only discharge gas, but mine contains a flexible weighted-tip tube to pull liquid from the bottom rather than gas from the top. This tank also has a manual valve (black knob), and mounted on it is a female paintball ASA to female 1/8" NPT adapter (nickle plated), and a NuPro/Swagelok B-2MA4 metering valve (frosted with ice). The metering valve has a 1/8" NPT solid brass plug, drilled to pass the 1/16" tubing. The tubing is 1/16" OD, 1/32" ID Teflon (McMaster-Carr part 5239K23), rated at 400 psi, and carefully flared inside the plug to keep it in place.

So indeed the tubing gets cold in a hurry, but the valve as you can see gets much colder. After a few minutes, the valve is so cold that the seals fail, expelling CO2 liquid from around the base of the adjustment knob. Perforating the end of the tubing with multiple holes keeps the tubing from whipping, but with the valve open more to limit pressure change at the valve, liquid is discharged from the end of the tube, with the tube spitting like a viper. Well . . . time to turn off the tank valve, think a little, and maybe try something else.

An important lesson here is that the temperature change occurs exactly where the pressure changes. The best cooling system is going to be one where the pressure changes at the site needing to be cooled rather than in a controlling valve somewhere else.

I think however this could be made to work. So that the valve only decreases the pressure slightly (thus does not cool much), the tubing would have to be rated at least 845 psi, and have a much smaller inner diameter so that flow is more limited through the tube. Ordinarily this would not be a problem, but such tubing is either much stiffer flexible tubing, or metal, and could not be as easily stuffed into a camera. The nearest candidate would be similar but stiffer 1/16" outer diameter ETFE tubing with a smaller inner diameter (McMaster-Carr part 5583K42 or 5583K41).

Liquid CO2 Expansion Chamber

In a effort to push the transition in phase and pressure of the liquid CO2 as close to the camera sensor as possible, I fashioned a tiny metering valve from a small block of aluminum. I did this to have as much of the pressure drop as possible occur inside the camera. I built this from a 3/16" x 1" x 12" bar of aluminim using a hack saw, mill file, sand paper, a drill press, some drills, and a tap . . . not exactly precision milling, but effective for a prototype. Initially, I shaped the screw to fit into the flared end of the tubing inside the chamber, but could not get a consistent seal. So I installed a small plug of Delrin, just small enough in diameter to slide through the taped 4-40 threads. The plug provided a very nice seal against the flared tube end.

Here again is the paintball CO2 tank with the metering valve, this time feeding the expansion chamber. The gas exhausts along the threads. With the chamber clamped to a block of aluminum to simulate dissapation of the cold in the camera, the chamber frosts up quickly. Using the combination of adjustments from the metering valve on the tank and the valve on the expansion chamber, I seemed able to achieve consistent cooling.

So I went ahead and installed the unit in the camera, a Canon XSi/450D. The sensor is the brown ceramic with the stainless flange, mounted on the alignment pin above. Notice the space between the sensor chip and the overlying processor board. I glued the chamber to the side of the sensor frame and sheild. In retrospect, I could have added some thermal transfer compound between the two first. The chamber actually is separated from the stainless steel frame by a copper clip that holds the sensor in postion, and separated from the sensor by part of the stainless shield. The chamber is only a few millimeters from one end of the sensor though.

With the camera on my telescope, the temperature, as monitored by the USB temperature probe, plumits. Wow, success? Unfortunately, the sensor window more or less instantly frosts over. Using the metering valve adjustment knob and the expansion chamber adjustment knob, I try to throttle back the cooling to a more sustainable and reasonable level.

However, within 15 minutes, sputtering starts and dry ice snow is blowing out the bottom of my camera around the adjustment knob. I fiddle with the metering valve and expansion chamber controls, try some other things including using both a regulator and metering valve, but there is too much cold and not enough heat transfer. I cannot get the flow low enough to allow the liquid to transition to gas without freezing up the expansion chamber. Not water freezing; this is CO2 freezing, at around -80 degrees C (-112 degrees F). Because of this, the temperature in the camera swings wildly as the chamber freezes and thaws. I switch to gas through the expansion chamber for that night by inverting the CO2 tank, and manage to get a little cooling by just pushing cold CO2 gas through the expansion chamber.

Above is a "20 pound" tank of CO2, a brass CGA320 to 1/8" NPT adapter, 1/8" NPT coiled extention (black), a paintball RAP4 regulator with gauge, the metering valve, brass plug, and tubing. The next night out, I try again, but with disasterous results. The CO2 tank is at around 845 psi, and the tubing is rated at 400 psi. When the expansion chamber freezes, liquid in the tubing warms a little, and pressure in the tubing downstream of the regulator starts to rise. I had a regulator set for 400 psi, but I realize now that there is no back flow over pressure valve on the regulator. So the translucent white 1/16" tubing blew open after a couple cycles of expansion chamber freezing. When it blew, the tubing whipped a welt on my arm, and the blast left my ears ringing for quite a while. It actually woke my neighbors and had them looking around out their back door. Yes, a sub-optimal outcome.

In retrospect, I think a liquid CO2 expansion chamber could work if heat transfer was better. If the chamber mass was made larger by extending a portion over and in contact with the sensor, heat transfer could be much improved, although engineering such a miniscule precision "cold finger' on the expansion chamber would be challenging at best. Aluminum is a good heat conductor, copper the best, but stainless steel is very poor, and stainless was surrounding the expansion chamber in my implementation. And of course using tubing with a much higher burst rating is, in retrospect, obvious.

Gas CO2 Cooling

This project is turning out to be a lot more interesting than planned. Disguested with the whole liquid concept, I remove the expansion chamber from the camera. But so far, I've had some success in cooling the camera just by running gas through the camera. As CO2 vaporizes in the tank, the tank cools. As CO2 gas passes through the regulator, it cools further because of the drop in pressure. Some of the cooling because of pressure drop also occurs in the small caliber tubing, with about 120 psi on the regulator needed to drive adequate gas through the system. So the system produces cold gas. The sensor on the camera actually produces very little heat.

After punching many holes in the end of the tubing with a pin to diffuse the gas and prevent a net force on the tubing from the gas jets, I crush the tube slightly so that it fits between the sensor and the back processor board. The slightly crushed 1/16" tubing is still a snug fit. The tubing is 0.0625 inch, and the space is 0.0531 inch or 1.35 mm). The tubing passes through fortuitous gaps in the sensor pins. After bending back a corner of the shield, it all fits back together with minimal modification. I route the temperature cable and tubing out through the side of the camera, using a length of 1/16" shrink tubing as a strain relief, insulation, and to decrease light leaks. The old adjustment port on the bottom of the camera serves as an exhaust port again, preventing any pressure effects across the shutter.

On the next trial at night, the camera cools nicely, but I have a problem keeping the temperature of the camera steady, as the tiny paintball RAP4 regulator keeps freezing (with solid CO2 dry ice internally), and the regulator is operating at the very lowest end of its range.

The graph above shows a night's imaging, and illustrates a couple of the thermal issues with extended DSLR imaging. The data in the graph is a temperature log generated by Albert Huntington's software mentioned above, using the USB temperature probe in the camera and at the USB hub on the telescope tripod, and plotted in Excel.

The most important observation is the extreme heating with 'Live View'. This live view mode of the camera generates a real time movie image from the sensor, displayed on the back LCD of the camera or on an attached laptop computer (as in my case). I use this mode for collimation and polar alignment prior to imaging. I turned off live view just as I started the temperature monitoring here, and you can see that the temperature inside the camera is a whopping 35 degrees F above ambient.

Turning on the CO2 cooling hastens the cooling to ambient, but I had trouble stablizing the temperature below ambient initially. Finally, at 23:00, I manage to get it reasonably stable at about 10 degrees below ambient, and went to bed.

Unfortunately, I miscalculated, and ran out of CO2 at 02:00. Without cooling, the camera internally rises per usual to about 20 degrees F above ambient, as shown toward the end of the graph above.

I picked up a Harris two-stage regulator on Ebay to address the many problems with the smaller paintball regulator. This new regulator is huge, weighing in at 4.8 pounds, but works beautifully as shown in the graph above.

I am able to dial in the temperature, and the temperature at each pressure remained very stable. Even as the pressure in the tank decreases with cooling and depletion, the output pressure from the regulator remains extremely stable.

In the upper picture is a 20 pound cylinder of compressed liquid CO2, the Harris 9296-125-320 regulator with CGA320 input for the tank and 3/8" flared female pipe out.

The 3/8" flared male to 1/4" female NPT adapter, and 1/4" male NPT to 1/8" female NPT adapters are from my local hardware shop.

The 1/8" male NPT to 0.031" barb is McMaster-Carr part 50745k11, although probably better would have been 1/4" male NPT to 0.040" barb McMaster-Carr 50745k32 allowing me to skip one of the adapters and have a more snug fit on the tubing.

So I've been imaging for a couple of months through the fall and winter with this system, and so far it works great with a couple of caveats. It is not very effecient at cooling the camera, as most of the cold goes into cooling the tank and the regulator. The phase change is in the tank, and most of the pressure change is in the regulator. The regulator still performes perfectly through the night, but is so cold I need gloves to handle it after use. Each refill on the 20 pound tank costs about $38, and I can get about two nights imaging with one tank. I have not had a problem with icing of the sensor though, as the CO2 gas floods the inside of the camera, and the C02 is extremely dry, excluding moisture from the camera and sensor. Another problem is that I have become completely addicted to these images with low thermal noise . . . here are some samples on ZenFolio.

One subtle problem is cooling artifacts on the sensor images. The upper image is normally stretched, and the lower stretched quite a bit more in only blue. If I run the regulator above 110 psi, I can see an area with slightly more pronounced blue in the image. I think this 'blue pixie' above is from extreme local cooling at the spot where the gas exits the tube on the back of the sensor. Fortunately, the 'blue pixie' goes away at lower pressures, and is easy enough to remove with photoshop anyway.

But some way to diffuse the cooling over the entire sensor would be better. And a way to more efficiently use the CO2 liquid.

Liquid CO2 Cooling Revisited

Here is a more complete drawing. I am going to try to put a simple single stage regulator inside an expansion chamber block, and attach a 'cold finger' that will then extend over the sensor. After the above photo, I drilled the block and silver-soldered the plate to the block. I am going to try this with liquid, but if this does not work, it still should work with gas as well or better than just a tube over the sensor as above. Hmmm . . .

Clear skies,
Rob Crockett
Copyright © 2010 All Rights Reserved.