Edward Plumer: August 2013

After a number of imaging attempts this summer with the Canon 60Da, I finally decided I need to build a cooler box for the camera. The temperatures at night are regularly in the high 80's and the long exposures are very noisy.

I revisited the great articles by Gary Honis on DSLR coolers to refresh myself on design options and performance. However, all of the boxes designed in that vein are for attaching a camera to a telescope. Specifically, they rely on fastening the front surface of the box to a T-adapter of some sort. Since my primary use of a DSLR is for use with camera lens, that design was not going to work.

After much doodling, I came up with the idea of making a "clam-shell" box in which the camera is fastened to the bottom section via the tripod threads and the cooler hardware is contained in the top section. The two sections would clamp together, sealing around the base of the lens like medieval village stocks.

Camera Platform - The pictures below show the finished bottom section. The base is composed of three layers of 1/2" hardwood plywood. A Losmandy dovetail plate is fastened to the underside for attaching to the saddle-plate of my GM-8 mount. I created grooves on either side of the base for the camera USB and AC adapter wires to slot though when the top is fastened. I also mounted a TEMPerHUM sensor onto the base. This is the blue USB device shown in the image below. This allows measurements of temperature and humidity within the enclosure, important to make sure I don't hit the dew point on the inside! This particular device also integrates with Astrophotography Tool which I currently use for DSLR capture.

Another plywood section forms the bottom half of the front of the box. The placement of the hole and the mounting point of the camera were carefully measured to ensure that both of my lenses would fit correctly. On both lenses, the focusing ring is on the objective end of the lens so the foam I will use to seal the gap around the lens will not interfere with focusing.

Shown below is my largest lens, an EF 70-200mm/f4L. When used with the cooling box, I removed the tripod foot and just let the camera support it. For larger lenses, I should be able to use a longer Losmandy plate extending out in front of the box, mount the lens on a foot directly to the plate, and leave the camera unfastened within the box.

I left enough room behind the camera for the display to remain flipped out, reasoning that this would leave less plastic between the cooling element and the CMOS chip.

The bottom assembly was painted in black acrylic paint which should harden after some number of weeks ... I hope.

First Few Heatsink Designs - I followed the basic strategy of using a Peltier TEC element with heat sinks and circulating fans on both the hot and cold side. In my first attempt, I mounted a pair of the Cooler Master CPU heatsinks which are shown below, back to back, sandwiching a 40mm 15V/6A Peltier between them. The results were rather poor, I only cooled the test enclosure down by 10 deg F. The hot-side heat sink stayed around 100F with ambient room temperature at 70F.

I my second attempt, I tried to boost the cooling by using two of these Peltier units in parallel. For each, I used a pair of brick heat sinks as shown below. The cold-side heat sink was 61mm x 58mm and the hot-side heat sink was twice that size.

On the inside of the enclosure, I used a small muffin fan to circulate air in the enclosure and, on the outside, I used a 140mm case fan to push air over the hot-side heat sink.

This did not work so well either. The internal temperature reached only 65F with the external temperature soaring to 140F. I tried forcing cold air from a portable AC unit over the hot-side heat sink. This helped, getting the temperature down to 49F and cooled the heat sink to 100F. Clearly, I I could dissipate heat faster, the Peltier could do the work of cooling the enclosure.

Liquid Cooled Design - In reconsidering how to dissipate heat, I first considered a "heat pipe" style CPU heat sink. They are reputed to be much more efficient than simple conductive heat sinks because they leverage evaporative effects. However, they don't work so well upside-down. Putting one on the back of my box would not place it in a favorable orientation.

I found a great deal on a refurbished Corsair H60 CPU cooler. A copper water-block with internal fins transfers heat to a liquid and a small pump inside the water-block circulates the hot liquid out to an efficient radiator with a 120mm fan circulating air over it. The whole thing is packaged as a pre-sealed unit.

As shown below, I had to rebuild the back of the box to accommodate the different shape and had to recess the mounting points of the water-block into the back panel so that the water-block surface would be flush with the inside of the case.

I resurrected one of my pair of CPU heat sinks from my first test for use inside the enclosure. Here it is with the Peltier attached to it with thermal compound. In this case, I used a different Peltier originally scavenged from a automobile beverage cooler which I had used in previous cooled webcam experiments. I do not know the maximum temperature drop or maximum voltage for this unit. It is another 40mm unit.

I then built an extension to the top part of my enclosure box in which to mount the radiator and fan. I felt like I was building an awning on the back of a little house. Oh well, whatever works.

I oriented the radiator such that the tubes entered on the back side. In this configuration, no matter where I am pointing the camera in the sky, this part of the radiator is pointing down or to the side, never up. This is the specified orientation for mounting this radiator in a computer case.

The initial room-temperature test of the box showed that it could bring the internal temperature to 41F while maintaining the surface of the radiator at 79F! Very impressive little gadget, that H60.

The finished product is shown below. The radiator housing is open at the top and the fan moves air up through this opening. On the top of the enclosure I attached two finder feet, one to attach a red-dot finder and one to attach my mini Orion guide-scope.

In order to control the cooling level of the Peltier, I used a MaxxTronic MXA033 motor speed controller which provides a 100Hz pulse-width modulation (PWM) signal for controlling the speed of a motor. It can be used in either 12V or 24V systems. I previously used the bi-directional version of this circuit to control the speed of the motor which rotates my observatory dome. The technique switches power to the load on and off very quickly; this is the recommended approach for controlling a Peltier device.

The output stage of the controller is shown in the schematic below. Note that the MOSFET provides switching of the ground line, not the 12V line. The connector shown above on the side of the box is a 3-pin XLR plug to which I run lines for Ground, +12V/+M, and -M (the switched-ground line).

To control the internal humidity, I put a reusable indicating desiccant pack made by DriBox tucked into a corner of the enclosure. I can recharge the media in the little plastic box by microwaving it couple of times for 45sec.

Here is the final mounting on my Losmandy GM-8 mount. As an added bonus, the extension I built for the radiator serves as a counter-weight for the camera and lens. The balancing point as configured was almost exactly at the center of the main box; no counter-weights needed ... sometimes you get lucky!

Field Testing - Having completed the first round of building, this evening I took it out to the Austin Astronomical Society's dark site at Canyon of Eagles. Before doing any real imaging, I performed some tests to get an idea of whether my efforts have payed off!

Power Consumption - The supply voltage applied to the fans, pump, and Peltier is 13.6V from a benchtop power supply. I am not sure if this is over-driving the H60 pump, or even the Peltier, so I will be reducing this to 12V. At 13.6V and maximum PWM duty cycle, the total current draw is 3.8A and at minimum duty cycle, it is 0.65A. An AC adaptor rated at 5A should be adequate to power the box.

Temperature Profile - As the box was cooling, I measured the internal temperature and humidity with a TEMPerHUM USB sensor mounted within the enclosure. Conveniently, reading of this sensor is integrated into the Astrophotography Tool software that I am currently using to capture DSLR images. The chart below shows the profile over time. The blue temperature curve is measured in the enclosure. The orange dots are the camera temperature reported in the EXIF metadata for three selected frames described below. The red line is the relative humidity in the box.

Dark Frames - At the beginning and end of the cool-down, I took dark frames to evaluate the effect of cooling. To see the noise visually, I imported the RAW files into Photoshop, converting to 16bit. I reduced the white point from 255 to 34 to highlight the dark current. Applying the same stretch to each test image allowed visual comparison. I then imported the original RAW frames into MaximDL and generated a histogram of each, noting the mean and standard deviation of the pixel levels. I also used IrfanView to extract the internal camera temperature from the EXIF data. I am not sure where these are measured but they are generally higher than the ambient temperature.

Warm, Long Exposure - This was the initial dark test consisting of a 10min exposure at ISO-1600. The temperature reading in the box was 83.1F, relative humidity 62%. The EXIF temperature was 80.6F. The pixel values had a mean of 2065 ADU with standard deviation of 132 ADU and variance of 17428.
Warm, Shorter Exposure - In this second test, I reduced the exposure length to 5 minutes at ISO-1600. The temperature reading in the box was still 83.9F with relative humidity 61%. The EXIF temperature had increased to 87.8F. The pixel values had a mean of 2056 ADU with standard deviation of 96 ADU and variance of 9216.
Cooled, Shorter Exposure - This last test was performed at the end of the imaging session. This is one of the dark frames I used for calibration as well. As with all of the "light" frames, this was an exposure of 5 minutes at ISO-1600. The temperature reading in the box has stabilized to 52.5F with a relative humidity of 66%. The cooling system managed a 30 deg F drop within the enclosure. The EXIF temperature had reduced to 64.4F, a 23.4 deg F difference from the beginning of the imaging session. The pixel values had a mean of 2049 ADU with standard deviation of 38 ADU and variance of 1444.

Full Frame Image - the following comparison shows the full frames

Dark Frame - 600 sec exposure, before cooling

Dark Frame - 300 sec exposure, before cooling

Dark Frame - 300 sec exposure, after cooling

Frame Center Images - these images show a crop of the central portion of the dark frames

Dark Frame Center Crop - 600 sec exposure, before cooling

Dark Frame Center Crop - 300 sec exposure, before cooling

Dark Frame Center Crop - 300 sec exposure, after cooling

Dark Frame Histogram - these plots show histograms of the dark frames grouped into 50 ADU bins.

Dark Frame Histogram - 600 sec exposure, before cooling

Dark Frame Histogram - 300 sec exposure, before cooling

Dark Frame Histogram - 300 sec exposure, after cooling

Summary - I am thrilled that this new design is able to drop the enclosure temperature by 30 deg F as this was as much as I was hoping for. It does take a good two hours to approach steady state. I was also surprised at how much more of a difference in the images cooling the box had over my usual strategy of halving the exposure time.

When the exposure length was halved from 10 minutes to 5 minutes, the variance in the dark-current signal variance reduced by a factor of 0.53, or nearly half. The change in internal camera temperature between the two 5 minute tests, as reported in the EXIF data, was 23.4 deg F (13 deg C). I am assuming that this is more correlated to the noise than the 30 deg F change in the enclosure temperature. This change reduced the dark-current signal variance by a factor of 0.157 corresponding to a halving for every -5 deg C.

In his article on SNR, Craig Stark comments on dark current stating "The intensity [of dark current ] should double as you double the exposure duration and it should also double for every 6 degrees Centigrade or so." If I interpret "noise" to mean variance in the dark signal, then my results are consistent with these rules of thumb. However, my reading of the article would indicate that it is the standard deviation of the dark current signal that I should be comparing.

I would like to repeat these tests with the bias signal subtracted from the dark frames to see what the actual thermal contribution is.

Results of the evening's imaging are shown in the next post.

Update 2013-10-27 - I have posted the build details for a second version of the camera box.

On the evening of August 12, I watched the Perseid Meteor shower. Visually, I only saw a handful of meteors. I saw more last week at the Canyon of Eagles dark site.

To catch some meteors on camera, I set up the DSLR on a fixed tripod pointing to the northeast. I exposed a continuous series of 700 frames, each a 15sec exposure at ISO-1600. This was the maximum exposure length I could get on an 18mm lens without the stars trailing. I used Astrophotography Tool to program the capture sequence and to help focus the camera lens on a bright star. The download time between images was about 10sec, so the camera had a 60% chance of catching any given meteor, assuming that the meteor was in the camera's field of view. Just after setting up, I visually saw one large bolide that unfortunately was not in the part of the sky at which the camera was looking ... worse, it was where I had just moved the camera away from in order to get less shots of planes landing at the Georgetown airport.

The next morning, after carefully combing through the images, I found 10 frames with meteor streaks and a lot more with airplanes and satellites. It's simple to tell the difference, as the real meteors never appear on consecutive frames, which would require them to be in the sky for more than 10 seconds.

On the computer, I attempted, and failed, to make a large mosaic stack using all of the images. This was because none of the software I tried were capable of correctly distorting the images to compensate for the projection of the spherical sky onto the camera focal plane. I gave up that approach and, instead followed this workflow:

I selected three sets of 16 frames from the beginning, middle, and end of the imaging session. Selecting more frames turned out to be a disaster as the distortion-induced misalignment of stars caused the sigma-clipping stacking algorithm to remove all stars from the periphery of the resulting frame.
Since the star-field was slowly drifting up, I cropped the bottom part of each frame to remove the trees.
Using Deep-Sky Stacker, I stacked each of the sets using the "standard" alignment mode.
I saved the three stacked images as as 32bit TIFF and then imported each as layers into Photoshop.
Using the GradientXTerminator plug-in, I removed the sky-glow gradient from each layer. Trying to piece together a set of frames with mismatched sky-glow would have made it impossible to get seamless transitions.
I converted to 16bit mode using the gamma/exposure mode of the HDR tool.
I rotated and scaled each frame to best align the stars. Due to the relative distortions of the frames, this could not be done perfectly.
I imported a single frame containing the trees and horizon from near the end of the imaging session, aligning with the previous three.
I used the Auto-Blend Layers tool to create a smooth mosaic across all four layers.
Next, I found each of the individual frames containing meteor trails and imported these as additional layers, aligning each with the overall mosaic.
I applied a layer mask to each meteor layer in order to select just the meteor trails.
Using adjustment layers, I applied levels, curves, color-balance, and contrast changes to enhance the image.
Finally, I added an adjustment layer to re-introduce the sky-glow gradient across the full mosaic which I had removed earlier on each of the individual stacks.

The final image clearly shows the radiant point in the western portion of the constellation Perseus. I was surprised how precisely the meteors seem to emanate from a single point. This is never quite the impression I get when catching them visually.