ASTRONOMY TECHNOLOGY TODAY. Your Complete Guide to Astronomical Equipment

ASTRONOMY TECHNOLOGY TODAY Your Complete Guide to Astronomical Equipment THE ORIGINAL TEETER’S TELESCOPE • GLATTER’S LASER COLLIMATOR TWO LOOKS AT T...
Author: Rudolph Manning
1 downloads 2 Views 8MB Size
ASTRONOMY

TECHNOLOGY TODAY Your Complete Guide to Astronomical Equipment

THE ORIGINAL TEETER’S TELESCOPE • GLATTER’S LASER COLLIMATOR TWO LOOKS AT THE STELLARVUE SV102ED

Testing QSI’s 532ws Scientific CCD Camera

Volume 2 • Issue 2 February 2008 $5.00 US

TESTING

QSI’S 532WS SCIENTIFIC CCD CAMERA

By Richard Berry

I wrote this story for savvy amateur noise and Fourier Transforms reveal every CCD cameras for scientific research, so I astronomers who are looking for a topproblem—and he was showing results figured their products had to be reasonnotch CCD camera that is suitable for from an internal engineering test. Cusably good. And because I realized that both imaging and science purposes, espetomers usually don’t get to see things like Kevin actually wanted me to do a serious cially if you want a high-quality camera that. I was impressed. and demanding evaluation of the new that seems to help you make great camera, I agreed to do so. images. Since I’ve been involved with CCD cameras for quite some The QSI 532ws Arrives time, I’m amazed at the number of The test camera arrived in fine imaging cameras available to mid-July. My initial impression amateurs on the market today. But on receiving the camera was exthis story is about just one: Quantremely positive. It arrived in a tum Scientific Imaging’s 532ws sturdy double box, inside which CCD camera. was an airtight Pelican case conOur tale begins at the Society taining the test camera for Astronomical Sciences meet(S/N#502137), power supply, ing, held each year before RTMC USB 2.0 and guider cables, a Tin Big Bear, CA, when I met Kevin mount to 2-inch adaptor, a set of Nelson. Kevin is a founder and the Allen keys, plus an optional liqmarketing director for Quantum FIGURE 1: The QSI 532ws CCD camera arrived neatly packaged uid heat exchanger for summerScientific Imaging. In the vendor in an airtight case with foam cutouts. Even after several months time cooling. Everything was of use, all of the components still fit into the case. One touch I area, Kevin was showing his comsnugly bedded in foam cutouts really liked is that the body cap on the camera body has an Opany’s new line of scientific imag- ring that keeps dust and moisture out of the camera. (see Figure 1). ing cameras. He pulled out a QSI’s 532ws camera is based notebook and proudly showed me the Kevin picked up on my interest, and on Kodak’s 3.2-megapixel KAF-3200ME Fourier Transform of a bias frame—and asked whether I would be willing to test CCD. This CCD has 6.8-micron square that really got my attention. Why? Beand evaluate the new 532ws camera. As pixels in a 2184 by 1472 array, and boasts cause bias frames are sensitive to system the company’s name implies, QSI makes a peak quantum efficiency of 80% or Astronomy TECHNOLOGY TODAY

29

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA higher. The camera, was configured to make images with an overscan area, common in science data cameras but rare in amateur imaging cameras. By now you are probably wondering why, as an amateur astronomer, I wanted to test a “science” camera? The answer is that, because scientific applications demand more from a CCD than amateur imaging, I expected better overall performance. Amateur cameras can be noisy and display rather severe faults before they seriously interfere with making pictures of

the Orion Nebula—but a scientific imaging camera has to meet a higher standard. I wanted a camera that would do a good job on really faint objects, that offered professional-quality firmware and electronics, a camera that would never stand in the way of taking great data or making great images. The test camera was a model 532ws. QSI offers the 532 in two types, the 532s (with mechanical shutter), and the 532ws (with mechanical shutter and five-position internal color filter wheel). The filter

wheel in the test camera came populated with Astronomik type II LRGB color filters. In size, the 532ws is 4.45 by 4.45 inches and 2.5 inches deep; the 532s is somewhat thinner. On one side are the power, USB 2.0, and guider control sockets. You can see specifications for all of QSI’s cameras at http://www.qsimaging.com/. Setup went smoothly. Speaking as a professional writer/editor, I would award the manual that came with the 532ws the highest marks for clarity. It is clear, well written, nicely illustrated, and complete. Using it as a guide, someone who had never set up and used a CCD camera before would probably encounter no problems. You can download the manual from QSI’s website; it’s a 2.0 MB PDF document. The arrival of the camera brought with it a lengthy spell of cloudy weather— perfect weather for testing! I set up the camera indoors and shot hundreds of bias frames, dark frames, and light-box flatfield frames. These would allow me to measure the camera’s basic performance characteristics (see the sidebar: Performance Testing a CCD Camera on page 45). Gain, Readout Noise, and Dark Current Three numbers tell you a lot about how a CCD camera performs. These are the gain, readout noise, and dark current. They were the first things I measured. The camera manufacturer sets the gain to maximize camera performance; you need to know the gain to convert pixel values in your images to the number of detected photons. The CCD adds readout noise; the lower it is, the better. And dark current is a parasitic signal generated by the CCD itself; lower is better. In The Handbook for Astronomical Image Processing (by me and Jim Burnell, published by Willmann-Bell, Inc. at http://www.willbell.com) we devoted a chapter (“Measuring CCD Performance”) to this subject, so I won’t repeat the tech-

30 Astronomy TECHNOLOGY TODAY

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA nical details here. Suffice it to say that by taking two bias frames, two flat frames, and one long-exposure dark frame, anyone can determine the gain, the readout noise, and the dark current of a CCD camera. Furthermore, The Handbook for Astronomical Image Processing comes with full-featured processing software, Astronomical Image Processing for Windows, called AIP4Win for short. AIP4Win includes a CCD Calibration Tool so you don’t need to struggle through the math. Simply select the bias, flats, and dark frame, then click to get your results. The gain of the test camera was 1.33 electrons per ADU, its readout noise was 10.5 electrons r.m.s., and the average dark current of 0.08 electrons per pixel per second at operating temperature of 0 Celsius. These are excellent figures. Kodak specifies a readout noise between 7 and 12 electrons r.m.s. for the KAF-3200ME, so the QSI camera and Kodak CCD met Kodak’s specifications. I’ll say more about

FIGURE 2: A single bias frame from the evaluation camera displays only the random readout noise across the whole CCD. This is a textbook illustration of how bias frames from a CCD camera should look.

dark current later in this story; suffice it to say here that dark frames turned out to be much more interesting—and very much better—than I anticipated.

Once you have the gain and readout noise, you can calculate two more basic parameters: the saturation signal and the dynamic range.

Astronomy TECHNOLOGY TODAY

31

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA Saturation occurs when the count in ADU reaches 65,535. Kodak’s specified minimum for the saturation signal in electrons is 50,000 electrons; the test camera reached saturation at 49,300 electrons. I queried QSI about this: to accommodate the capacity of the horizontal shift register for 2x2 binning, QSI sets the gain somewhat high to provide the optimum tradeoff between saturation and dynamic range when the CCD is binned. This engineering decision benefits amateur astronomers shooting color images, and it does not significantly compromise Kodak’s saturation specification. Dynamic range is the ratio between the maximum signal and the readout noise; in the test camera the ratio was 4930:1. Engineers sometimes think in decibels; in the test camera, the dynamic range is 73.8 dB, that is, better than Kodak’s specification of 72 dB. Well, that’s a lot of numbers—gain, readout noise, dark current, saturation signal, and dynamic range! By measuring

32 Astronomy TECHNOLOGY TODAY

them the first day I had the camera, I did what any professional astronomer would have done: I had characterized the camera. By the end of the first day, I knew that the QSI camera and its KAF-3200ME CCD met specifications, and were likely to perform very well on the stars. After working with the evaluation QSI 532ws for another ten weeks, I ordered one for myself. When mine arrived (S/N#502314), the first thing I did was measure the basic parameters: they came out almost identical to those of the test camera. Shortly after that, a friend received his QSI 532ws, and it tested out with the same basic parameters. I must say that I am impressed with consistency of QSI’s product, and personally very pleased that my own camera tests every bit as good as the evaluation camera. Analysis of QSI 532ws Bias Frames With the basic parameters established, I began testing in earnest. While

the bad weather continued, I shot hundreds of bias frames, dark frames, and flat frames. I’ll talk about the bias frames first. Bias frames reveal subtle aspects of a camera’s “personality” that make a big difference when you try to take images at the limit. Bias frames are exquisitely sensitive to noise from faulty camera electronics, poor shielding from radiofrequency interference, faulty grounding, and a host of other problems. And if noise appears in the bias frame, you can be sure that every image you shoot—darks, flats, and images—will be noisy, too. What impressed me in looking over the mass of data I collected is that the QSI 532ws’ images are exceptionally clean and free of readout artifacts. Because I write image-processing software and collect sample images whenever I can, I’ve seen bias frames from many different CCD cameras. Bias frames should be perfectly flat, showing no features at all except random pixel-to-pixel variation due to readout

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA

FIGURE 3: Hot Pixel Image and Histogram. Hot pixels comprise only about 0.2% of the pixels in this image; the other 99.8% of pixels have extremely low dark currents. In the log-scale histogram, the peak on the left stands three orders of magnitude above the shoulder where the hot pixel population begin.

noise. In reality, many cameras produce bias frames with quasi-random horizontal stripes, vertical patterns, plus a few (or sometimes many) dark or light columns. Taken together, these artifacts are called pattern noise. In some cameras, the pattern is the same in every bias frame; in others, the pattern is different in every bias frame. Repeating pattern noise is better than changing pattern noise, but all pattern noise is bad. Pattern noise points to shortcuts or unsolved problems in the design or construction of the CCD camera. However, the QSI 532ws’ bias frames

looked so close to ideal that I was amazed. Figure 2 shows a typical bias frame. Histogram of bias frames show pure Gaussian noise with a standard deviation of 10.5 electrons r.m.s.; that is, they showed pure random readout noise. However, I was not satisfied to say the bias frames from the QSI 532ws were perfect. After all, every dog has some fleas. To beat down readout noise and reveal whatever tiny residual pattern noise might be present, I combined hundreds of bias frames. Ultimately, I found semi-random patterns with an amplitude of 0.2 electrons, or about 2% of the readout noise.

Zeke’s Seat’s WHERE BUSINESS IS ALWAYS LOOKING UP

www.AdvancedTelescope.com

Advanced Telescope Systems Manufacturing Quality Portable and Permanent Telescope Piers

The Sky Looks So Sweet, If You

GET A ZEKE SEAT!! We take PayPal, Cash, Check, M/C, Visa & Trade?

Our astro seats are all TIG welded to last a lifetime. Great seat - Great price. Four full inches foam padding 18-24 ins for all night comfort. One handed adjustment. About 10Lbs, 0-33 ins. Special Purchase (239) 995-9353 • www.zekesseats.com • [email protected]

Toll Free

(877) 96-SCOPE Astronomy TECHNOLOGY TODAY

33

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA This is remarkably small compared to anything that I’ve seen in bias frames from other cameras; in some, the pattern noise is larger than readout noise. Having found and measured this tiny residual, I can say that for all practical intents and purposes, the test QSI 532ws bias frames are textbook perfect. And when my own 532ws arrived, its bias frames were just as good. I mentioned earlier that the QSI 532ws test camera was configured with an overscan area. In science cameras, an overscan area is included because it provides valuable insight into the operation of the camera and CCD. After clocking out the image area of the CCD, the electronics stop clocking the vertical shift registers but continue to clock the horizontal shift register. In the test camera, there were an additional 400 columns on the right-hand side of the image. These columns have lower pixel values and less noise than the area of the bias frame read from the image

34 Astronomy TECHNOLOGY TODAY

area of CCD. The random variation in the overscan is actually the “true” readout noise; it measured 8.1 electrons r.m.s. in the test camera (versus 10.5 electrons r.m.s. in the image area). This puts the “true” readout noise in the test camera at the low end of Kodak’s specified range of 7 to 12 electrons r.m.s. While measuring large numbers of bias frames, I noticed that the mean value of the entire image sometimes changed by plus or minus one ADU. This is quantification noise. It comes about because the analog signal from the CCD’s on-chip amplifier is converted to an integer value, and then auto-zeroed to a fixed value. Quantization noise adds approximately 0.5 electrons r.m.s. of random noise to the signal, and auto-zeroing quantization noise is present only when you combine a large number of images. Quantization noise is present in all digital imaging systems. In the QSI 532ws, quantization noise (0.5 electrons

r.m.s.) is negligibly small compared to readout noise (10.5 electrons r.m.s.) and shot noise (often a few hundred electrons r.m.s.). I detected quantization noise only because I set out to measure every source of noise I could find. Analysis of QSI 532ws Dark Frames and Dark Current To investigate the QSI 532ws’ dark current, on cloudy nights I shot bias and dark frames at temperatures ranging from 0 Celsius to -25 Celsius with exposure times ranging from 10 seconds to 1000 seconds. This produced many gigabytes of images for analysis. The first thing I checked was the stability of the 532ws’ temperature. Dark current depends strongly on temperature, typically doubling every 6 degrees Celsius. If the CCD temperature oscillates around the set point temperature, dark frames have either more or less dark current than those taken at the nominal set point tem-

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA perature. If you combine multiple dark frames at slightly different temperatures, in effect you’re hoping that the highs and lows will average out. Even if this is true, temperature oscillation contributes unwanted noise to your images. The QSI 532ws has a thermoelectric cooler. When you turn on the cooler, the camera cools rapidly, reaches the set point temperature, and then stays right at that temperature without oscillation above and below the set point. In checking over 500 dark frames, I found only two in which the CCD temperature did not equal the set point temperature, and then by only 0.2 degrees. I was quite satisfied with the stability of the 532ws’ temperature control. Dark frames I took in July worked perfectly with images that I shot in September. Dark current in the KAF-3200ME CCD proved more interesting than I expected. In the olden days, every pixel was a hot pixel. In the two QSI cameras that I tested, the vast bulk of pixels had such low dark current that it was difficult to measure accurately. To get an accurate measurement of the dark current at a working temperature of -20 Celsius, I had to resort to extreme measures. I shot a sequence of 256 bias frames and 256 dark frames with an exposure of 60 seconds. Using AIP4Win’s Sigma-Clip Files function, I combined these into a master bias and a master dark frame, then subtracted the bias from the dark to create a master “60-second thermal frame.” A histogram (shown in Figure 3) produced by AIP4Win’s Pixel Tool shows that 99.8% of the pixels follow a Gaussian distribution, with a long-tailed distribution toward higher values. In this image of 3,124,848 pixels, one pixel had a pixel value 5995 ADU, 18 pixels had values ≥1000 ADU, 119 pixels had values ≥100 ADU, 2931 pixels had values ≥1000 ADU, and 5982 pixels (0.2%) had values ≥3.5 ADU. The remaining pixels lie in a Gaussian peak centered at 0.265 ADU, and of those pixels,

FIGURE 4: Non-Hot Pixels versus Temperature. Because hot pixels are rare on the KAF3200ME, large areas of it contain no hot pixels at all. When I examined the mean pixel value of hot-pixel-free areas as a function of temperature, I found that cooling below -15 Celsius does little to reduce the dark current. For the QSI 532ws, the optimum operating temperature is somewhere in the range of -20 to -25 Celsius.

FIGURE 5: Individual Hot Pixel Values versus Temperature. The hottest hot pixels are individualists, each one of them different. Plotted as a function of temperature, the hottest hot pixels show differing behaviors. The less-hot hot pixels (at the bottom of the graph) behavior in a predictable fashion.

46% lay within ±0.5 ADU of the peak value. Converted to electron units, the hottest hot pixel had a dark current of 130

electrons/pixel/second, while the mean dark current for all pixels was 0.006 electrons/pixel/second. In practical terms, the dark current at -20C for 99.8% of the pix-

Astronomy TECHNOLOGY TODAY

35

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA els on the CCD proved so small as to be negligible. Out of a total of 3.2 million pixels on this CCD, only ~6000 hot pixels displayed significant dark current. With this in mind, I examined the behavior of “good pixels.” I made sets of bias frames and accompanying dark frames with exposures up to 1000 seconds. Using the Pixel Tool in AIP4Win, I measured mean pixel values of the bias frames and the 1000-second dark frames in a region that was entirely free of hot pixels. The results are shown in Figure 4. Best-fit fourth-order curves represent these data well. At temperatures of -15C and lower, dark frames nearly match the values found in bias frames. Converting the difference between the 1000-second dark frames and the bias frames to dark current yields the following mean dark current for a region free of hot pixels: Temperature Mean Dark Current [Celsius] [e/pix/sec] -25 0.0010 -20 0.0013 -15 0.0016 -10 0.003 -5 0.01 0 0.08 These figures are remarkably low, a dramatic tribute to Kodak’s mastery of CCD technology. From a practical view-

36 Astronomy TECHNOLOGY TODAY

point, they suggest that there’s no reason to run the QSI 532ws at temperatures lower than -20C to -25C because for 99.8% of the pixels on the CCD, the dark current nearly matches the bias, and is effectively zero. We used to see performance like this with the Sony HAD CCDs; now it appears that Kodak has mastered lowdark-current technology too. But what about the remaining 0.2% hot pixels? I wanted to know: How do they behave? To sample the temperature behavior of hot pixels, I located six pixels that were either saturated or more than halfway to saturation at 0 Celsius, and three pixels with values below 5000 ADU. Using AIP4Win’s Pixel Tool, I measured the pixel value from a set of median-combined 1000-second exposures. The results are plotted in Figure 5. All hot pixels become “hotter” with higher temperature, but the rate of increase among the group of six appears to differ. Several curves cross—implying differing temperature coefficients—and, of course, some of the hot pixels reached saturation at 65535 ADU. Although too low to show well in the graph, the low value pixels appeared to behave with greater consistency. However, the unpredictable behavior of the hottest hot pixels suggests that dark-frame subtraction may not work

for a handful of hot pixels on each CCD. To sample the exposure time behavior of hot pixels,. I used AIP4Win’s Pixel Tool to measure a selection of hot pixels from a sequence of dark frames with exposures ranging from 10 seconds to 1000 seconds. The resulting plot, for a temperature of -20 Celsius, is shown in Figure 6. Although these curves appear quite well behaved, they are curves rather than straight lines; that is, the dark current of high-value hot pixels is non-linear with respect to exposure time. Among other things, these data suggest that with the KAF-3200ME (and quite possibly, other CCDs in the same device family), scaling dark frames linearly with time during dark-frame subtraction doesn’t correct a handful of hot pixels. This behavior can be countered by dithering the telescope pointing when taking images and using an algorithm such as sigma-clip to eliminate outlier values. The QSI 532ws camera provides a stable thermal environment for the KAF3200ME. The CCD itself displays remarkably low dark current in ~99.8% of its pixels. Of the remaining 0.2% (~6000) hot pixels, it is likely that fewer than 100 of them respond erratically with temperature or non-linearly with exposure time. Used with care, standard calibration procedures can deal with this small number

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA of high-value hot pixels. In any case, kudos to Kodak for making CCDs with such low dark current. Analysis of QSI 532ws Flat Frames Flat frames provide a window into a camera’s response to light, specifically, they allow us to measure the linearity and uniformity of the CCD. For the KAF3200ME, Kodak specifies nominal values for non-linearity and non-uniformity of 1%, with maximum allowable values of 2% non-linearity and 3% for non-uniformity. Although my methods did not exactly match the test methods that Kodak employs, my results suggest that QSI has succeeded in squeezing every bit of performance possible from Kodak’s CCD. Linearity Linear response means that the digital output from the CCD camera is strictly proportional to the incident light. CCDs with anti-blooming gates are intentionally made non-linear: they spill excess charge from pixels as they approach saturation. For astronomical science imaging, linear response is necessary for accurate photometry. In imaging, non-linear response in the CCD can lead to gradients in the image. However, the QSI 532ws test camera passed the linearity test with flying colors. In The Handbook of Astronomical Image Processing, we described a technique that can be used to determine linearity. To test the QSI 532ws, I wanted a technique that would more closely mimic taking real astronomical images. So I took a series of flat-field frames of increasing exposure with matching dark frames. After darksubtraction, I measured the mean value of a uniform region in the center of a flatfield image. This allowed me to test the camera’s linearity as if I were taking celestial images, which is what both aesthetic imagers and scientists really care about. For consistent results, I first had to make an extremely constant source of light. I tried an LED used with the constant-current circuit described in the

FIGURE 6: Individual Hot Pixel Values versus Exposure Time. Conventional wisdom holds that the pixel value of hot pixels increases linearly with exposure time. However, my investigation revealed that rate of increase of the hottest hot pixels slows as the pixel value rises.

FIGURE 7: Linearity. The QSI 532ws CCD camera responds very nearly linearly. In the central portion of the curve, the greatest departures from perfect linearity are +1% and 0.6%, well within Kodak’s specification for the KAF-3200ME CCD.

Handbook, running from a 12-volt 5 amphour battery, but it declined by several percent over a two-hour period. Using a constant voltage circuit to run the constant-current circuit solved this problem. To guard against systematic brightness

shifts, I made the exposures in rotating sets so that zero-exposure and maximumexposure images occurred at regular intervals. Analysis of the images showed that no systematic shift in LED output occurred while the camera was taking data.

Astronomy TECHNOLOGY TODAY

37

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA In the setup that I used, the CCD saturated at roughly 26.5 seconds exposure. I set the minimum exposure to 0 seconds, the maximum exposure to 26 seconds, and between the minimum and maximum, selected 20 exposure times from 0.1 seconds to 25 seconds. I used the camera’s shutter to time the exposures. I calibrated each image with its dark frame, then plotted the median value of the entire image against the exposure times. The response curve is shown in Figure 7. To the eye, this plot looks perfectly linear. But a linear least-squares fit reveals a non-linearity of +1% at 8 seconds exposure, and a minimum of -0.5% at 26 seconds exposure. Kodak specifies that the worst-case deviation from a straight-line fit between 2% and 90% of saturation is nominally 1%, so the system performance lies well within the Kodak’s specs. The QSI 532ws camera should be capable of 10-millimag photometry or better over its entire dynamic range. When I

38 Astronomy TECHNOLOGY TODAY

did a photometry run with the test camera, I achieved 5-millimag accuracy quite easily. Observing exoplanet transits should pose no problem for the QSI 532ws camera. Uniformity Uniform response is crucial in making pictorial images. Can you imagine what would happen if one side of your CCD were more sensitive than the other side? Or if it were more sensitive to one color over another? You could get gradients and strange color effects in your images. I tested the color uniformity of the QSI 532ws, and once again, the camera and its CCD proved their merit. Kodak’s specification for uniformity requires a nominal 1% standard deviation from the mean value of a 128x128 sample when the CCD is illuminated uniformly. I wanted to verify that the KAF-3200ME was uniform in all standard LRGB colors. I made multiple flat-field frames using the

Astronomik Type II R, G, B filters supplied with the test camera, and averaged them to make high-quality master flats. When I measured sample blocks of the specified size, I found that one standard deviation was 0.25% for luminance flats, 0.32% for red-filtered flats, 0.35% for green-filtered flats, and 0.55% for blue-filtered flats. In other words, the CCD in the QSI 532ws camera bettered Kodak’s specification by two to four times. Departures from perfect uniformity took the form of short horizontal streaks and periodic vertical banding in the red and green flats, while the blue flat showed only the vertical banding. Since the pattern remains the same from frame to frame, it should be removed during routine flat-fielding. Although Kodak does not specify the uniformity of CCD with different color filters, it is an important consideration for making good-looking color images. Using AIP4Win’s Join colors Tool, I made a

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA color image from flat-field frames. The resulting color image was almost perfectly gray. When I measured RGB values in a 24-bit color TIF image, they were usually identical, and seldom varied more than 1 ADU over the image. Averaged over the whole image, the response of the CCD in blue light is approximately 0.1% higher in its center than the mean response, and around 0.2% lower around the edges in green and red. It was quite difficult to detect and measure such tiny non-uniformities; you would never detect them in normal astronomical imaging. Summary Evaluation Based on Test Images Examination of bias frames, dark frames, and flat frames revealed that two QSI 532ws cameras and the KAF3200ME CCDs in them performed to specification. I was unable to find any problem or noise source in the test images that could be attributed to any failing or fault in the QSI camera. Properties of the Kodak KAF-3200ME Image Sensor This part of my review focuses on the KAF-3200ME image sensor itself. The 3200ME is described in a technical paper, “A 3.2 Million Pixel FullFrame True 2f CCD Image Sensor Incorporating Transparent Gate Technology” by E.J. Meisenzahl, et al., available from (www.kodak.com/go /imagers), and

Kodak’s technical specification, “KAF3200E / KAF-3200ME Performance Specification.” Before launching into technical details, I want to point out that in my tests, the KAF-3200ME was a shining star. It is roughly twice as sensitive as many front-illuminated CCDs, and it is sensitive across the entire spectrum from deep violet right into the infrared. Furthermore, its dark current is extremely low, its readout noise is very low, and its bias frames are exceptionally uniform and smooth. Having established its qualities, we’ll pick a few nits. Indium Tin Oxide Gate Technology By using microlenses in conjunction with standard polysilicon for one clock phase and indium tin oxide (a blue-transparent gate material) for the other clock phase, the KAF-3200ME attains a quantum efficiency of ~60% at 400 nm (blueviolet), peaking at ~80% at 600 nm (orange) wavelength, and falling to ~60% at 760 nm wavelength. This means that each pixel consists of two areas, one that is roughly ten times more sensitive to blue/violet light than the other. Cylindrical microlenses over the array direct all light falling on the device to the indium tin oxide side of the pixel. In 2005, several members of the AASVO Photometry group raised questions whether the dual gate structures might produce artifacts such as those reported by Christian Buil (see http:// astrosurf.com/buil/us/fiber/fiber.htm). No

further discussion or testing appears to have followed. Buil’s investigation suggests that users should avoid making undersampled images, especially in green, blue, and violet. Micro-Lens Artifacts Micro lenses are tiny cylinder lenses designed to concentrate light on the indium tin oxide half of each pixel in the KAF 3200ME. Although I can imagine scattering from one micro lens to adjacent micro lenses could account for some artifacts, they may also be the astronomical equivalent of the “urban myth.” Because I aligned my camera precisely north-south and east-west, I was not able to attribute possible low-intensity extensions along rows and columns in my images with any certainty to spider diffraction or blooming. If the effect is real, it is very small. Residual Images When you take an image that saturates the CCD, some photoelectrons diffuse into the silicon substrate. When you shoot

Astronomy TECHNOLOGY TODAY

39

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA frames. Using AIP4Win’s Multi-Image Photometry tool, I measured the mean pixel value of the residual image. The decay curve is shown in Figure 9. In 5 minutes, the residual image falls to the level of the readout noise and becomes hard to see. Using precision photometry, I was able to follow the decay for about 15 minutes. The moral of the story is that after making images with bright stars, you should avoid making dark frames or images of faint objects for 10 to 15 minutes.

FIGURE 8: Retained Charge. On the left is a 60-second exposure of a zero-magnitude star. On the right is a 60-second dark frame taken immediately after the first image. The grossly over-saturated image of the star has left an obvious ghost image. Over the course of 15 minutes, the ghost image fades to nothing.

FIGURE 9: Decay of Retained Charge. This graph shows how the ghost image in Figure 8 fades with time. After roughly 5 minutes, the ghost falls to the level of readout noise. After fifteen minutes, it becomes undetectable.

the next image, some of those electrons diffuse back into the charge well, appearing as a “ghost” of the first image. Figure 8 shows effect; the left image is a 60 second exposure of a bright star; the right

40 Astronomy TECHNOLOGY TODAY

image is a 60-second dark frame. In time the residual image fades. I investigated residual images by making a single 60-second exposure of a bright star followed by sixty 60-second dark

Cosmic Rays and Particle Tracks While acquiring dark frames and deep-sky images, I observed a large number of “cosmic rays” in the images. In dark frames, they appear as bright spots, tracks, and worms that make corresponding dark spots, tracks, and worms when subtracted from images. Calibrated images thus end up with a scattering of irregular bright and dark blemishes. In a test run using dark frames totaling 7680 seconds of exposure, I calculated a mean rate of 650 x 10-9 events/pixel/second, which means that KAF-3200ME images accumulate roughly 120 “cosmic ray events” per minute of exposure in images and dark frames. Cosmic rays were discussed on the AAVSO Photometry list serve; Ben Davies found that the radioactive isotope potassium-40 (K40) in the cover glass of the KAF-3200ME is the likely source of these interlopers. Kodak uses Schott D263 glass containing 6.9% potassium rather than common optical glass BK7, which contains 11% potassium. The most relevant study of these “cosmic rays” is found in a paper, “Radiation Events in astronomical CCD images,” by A. R. Smith et al., available on the web at http://en.scientificcommons.org/677140 6. This paper claims that the rare straight tracks are real cosmic-ray muons, but the curvy worms and star-like spots are caused by radioactive decays that produce alpha-

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA particles and low-energy electrons. Because they occur randomly, cosmic rays can be removed when multi-image exposures are stacked by using an algorithm such as a normalized sigma-clip. This removes statistical outliers while retaining statistically normal pixel values. Stellar Photometry with the QSI 532ws Measuring stellar magnitudes is probably the most common type of scientific observation undertaken by amateur astronomers. It seemed natural, therefore, to evaluate the performance of the QSI 532ws by making the light curve of an interesting star. I selected the star BL Camelopardalis because it’s reasonably easy at 13th magnitude, it changes rapidly, and its northern location means it’s possible to follow it for many hours. And it’s easy to find: BL Cam is two-thirds of the way from Polaris to the bright stars in Perseus, in the middle of a prominent line of four stars. If you get the middle two stars in your field of view, you’ve also got BL Cam. I used the telescope and mount I normally use for imaging, and on two nights shot some ten hours worth of clear-filtered 60-second exposures. I auto-calibrated the images and performed aperture photometry using AIP4Win’s Multi-Image Photometry Tool. The longer of the two light curves is shown in Figure 10. The measured amplitude of variation is 0.35 magnitude and varies somewhat, and the period is under an hour. The curve looks as smooth as any of the light curves I found in the literature. The internal consistency of the data was excellent: the standard deviation of the differential magnitude between the comp and check star was 0.0055 magnitudes, in agreement with the expected error due to shot-noise and readout noise. While two nights of clear-filter photometry are hardly definitive, the QSI 532ws certainly did produce excellent light curves and magnitude uncertainty

FIGURE 10: Light Curve for BL Camelopardalis. This 13th magnitude variable star goes through its cycle of variation in less than one hour. The upper curve follows the star for over six hours; the lower curve shows standard deviation in the differential magnitude between the comparison and check stars is only 0.0055 magnitudes.

FIGURE 11: Top view of the QSI 532ws with the top cover removed. Filter slots 1 through 4 have Astronomik II RGBL filters; the fifth slot has no filter. The CCD itself is behind the shutter wheel visible through rectangular opening. The metallic objects at the right are miniature motors that rotate the shutter and filter wheels.

FIGURE 12: Top view of the QSI 532ws with the top cover removed. The intricately nested circuit boards, motors, shutter, and filter wheel inside the QSI 532ws are visible in this side-on view. Note how the circuit board is recessed into the solid aluminum block of the camera body.

was consistent with the expected performance.

and flat frames that I had no doubt the QSI camera would perform extremely well with astronomical subjects. I used a Vixen R200SS 200mm f/4 Newtonian reflector with a Paracorr coma corrector on my faithful old Byers 812 mounting. The small pixels (6.8 microns square) turned out to be a good match for a fast Newtonian with a coma-corrector, and I must say that I was impressed with

Imaging with the QSI 532ws Although the month the test camera arrived was cloudy and rainy, when the skies finally cleared, I set out to make test images of astronomical objects. By then I had done enough work with bias, dark,

Astronomy TECHNOLOGY TODAY

41

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA

Introducing the New Multicart! Transforms in seconds into 8 different configurations Tough, corrosion-resistant textured powder coat finish Non-skid frame bed helps prevent cargo from shifting Cable quick release allows folding down for compact storage and transport Spring-loaded wing bolts that won't vibrate out Stair climber rails make going up and down stairs and curbs easy Strong yet light G-Force castors with smooth, quiet thermoplastic tread Shock-absorbing pneumatic wheels with polymer hub and roller bearings We also custom build the best, strongest, most well constructed scope transport cases available!

www.scopeguard.com 42 Astronomy TECHNOLOGY TODAY

the performance of the Paracorr—nice round images right to the corner of the frame. I made sequences of 60-second exposures with the Astronomik II LRGB filters that come with the standard QSI 532ws camera kit, and stacked them with AIP4Win. In addition to images, I usually made 16 60-second dark frames, 16 light-box flats, and 16 flat-darks each night. Since the KAF-3200ME blooms, the bright stars in my images do exhibit blooming trails. Since my interests lean in the direction of astronomical science, I regard blooming trails as a small price to pay for excellent linearity. I have included several of these images with this story (Figures A through F). At the telescope, the QSI 532ws was a pleasure to work with. I never needed to fight the camera to get good results; instead, it made my job easy. Its integral filter wheel and shutter are completely transparent to the observer—you set them

as parameters in the acquisition software and the images get taken. My own QSI 532ws camera arrived about a month after I had sent the test camera back to QSI—just in time to catch Comet 17/P Holmes! The image in Figure G was made on November 2. I was able to perform many different analyses on the images, including cross-sectional brightness profiles, color-ratio images that revealed the color neutrality of the inner coma, the strong blue coloration of the outer coma, and the kinky and distorted tail. Design, Build Quality, and Overall Assessment QSI asked me to do a serious and demanding evaluation of their camera and report any and all problems to them. When I got the camera, I found it to be a real winner. Bias frames were noise-free, cooling was rock stable, and all-around performance was excellent. It has a com-

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA pact footprint and the workmanship is impressive. I ordered one for myself, and when it arrived, I ran more tests on it, and found mine every bit as good as the test camera. The QSI 532ws met expectations when I tried CCD photometry of a variable star. When I made images of standard Messier and NGC objects, I got excellent LRGB results with total exposures sometimes under an hour. And when Comet 17/P Holmes appeared, my own QSI 532ws performed like a champ. I did not compare the QSI camera to other cameras in the amateur market. What I did was look for camera-generated noise sources, and check that the KAF3200ME chip met or exceeded Kodak's specified performance levels. The camera tested out very clean, and the images met all of Kodak's specs. Although it's not relevant to science uses, the design of the 532ws is aesthetically pleasing. The camera is compact, rounded, and feels solid and good when you handle it. The machining is accurate, and parts fit very well. Changing the internal filter wheel is easy and takes just a few minutes. Figures 11 and 12 show the tight integration of the internal filter wheel and shutter into the camera housing. The compact body design and short internal light path mean you can use the 532ws with optical systems as fast as ƒ/3.1 without vignetting using standard LRGB or narrow-band filters. The standard Tthread mount makes it equally easy to use on a telescope or with 35-mm camera lenses. I was impressed with the high sensitivity, the clean bias frames, the low dark current, the uniformity of the CCD’s response to light, and the smooth, reliable operation of the shutter and filter wheel. My overall assessment is that QSI 532ws is a laboratory-quality camera that’s an outstanding value for amateur astronomers who demand top-notch imaging performance.

44 Astronomy TECHNOLOGY TODAY

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA

Performance Testing QSI’s 532ws Scientific CCD Camera Capturing images from the sky is a complex process. Starlight passes through the atmosphere, through your telescope, through the filters in your camera, until it finally strikes your camera’s image sensor. In the controlled environment of the camera body, photoelectrons accumulate on a CCD image sensor, your camera’s electronics read out the CCD, digitize the signal, and pass data to the driver software in your computer, until it finally reaches the image acquisition software. Since my goal was to test the QSI’s 532ws performance, I needed to simplify the system. Rather than taking sky images and introducing a host of unknowns, I tested by taking standard calibration frames—bias frames, darks, and flats. This is the way professional astronomers test their cameras and CCDs. The performance of the camera and the performance of the CCD image sensor are all but inseparable. In testing QSI’s 532ws camera, I had to test Kodak’s KAF-3200ME CCD image sensor as well. The camera and the CCD inside it enjoy an intimate relationship. The camera not only provides a benign physical environment for the CCD, but it also controls the electronic environment. For top-notch performance, the camera and the CCD must work well together. Put a good CCD into a bad camera, or put a bad CCD into a good camera, and performance suffers. Only when the camera and CCD are both topnotch can you expect truly great imaging performance. What does the camera do? Here is a list of key functions: • Provides physical connection to external optics. • Provides a clean, dry, airtight container for the CCD. • On command, moves optical filters into the light path. • On command, shutters light entering the CCD. • Cools the CCD to desired operating temperature. • Provides all voltages necessary for correct operation. • Controls CCD functions by executing internal firmware. • On command, reads, digitizes, and stores image data. • Communicates data to external computer software.

a camera does its job properly, the CCD—not the camera— becomes the performance-limiting factor. My approach was to test the QSI 532ws as a “black box,” that is, I explored the characteristics of the camera through the images it produced. I compared these images to the performance of a hypothetical “ideal” image sensor and against Kodak’s published specifications for the KAF-3200ME image sensor. If the images were degraded from the “ideal” CCD, or worse than the specifications for the Kodak image sensor, I held the camera responsible. Key characteristics for the image sensor in a CCD camera are: • Readout noise • Dark current • Saturation signal • Non-Linearity • Non-Uniformity • Dynamic range • Quantum efficiency Kodak specifies these in a performance specification document; specs for all of Kodak’s CCDs are published on the web (http://www.kodak.com/go/imagers) as a PDF documents that anyone can access. The only specification that I was unable to test was quantum efficiency, which requires calibrated laboratory equipment. I am happy to report that the QSI 532ws met or exceeded Kodak’s performance specifications in all categories I tested.

If any of these functions fail, the performance of the camera suffer. When you buy a CCD camera, a good part of what you’re paying for is the know-how that the camera manufacturer brings to the design, assembly, and configuration of the camera. A well designed camera should be nearly transparent to the user. It should extract every bit of performance that the CCD can deliver. If you become aware of the camera’s presence in a negative way—if the camera adds noise to the signal—the camera is not functioning as well as it should. In other words, when

Astronomy TECHNOLOGY TODAY

45

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA

A

C

B

D

FIGURE A: Andromeda Galaxy. This picture of the Andromeda Galaxy is one of the first I took with the evaluation camera, and still my favorite from the testing program. I'd like to think that this is how this lovely galaxy would look if our eyes could see color at low light levels. L=12x60sec, R=10x60sec, G=12x60sec, B=13x60sec.

46 Astronomy TECHNOLOGY TODAY

FIGURE B: The Pleiades. On this damp night, the telescope fogged up so the total exposure time was quite short. In this image, the rows and columns on the CCD, the diffraction spikes, and the blooming trails were aligned with north is at top. L=5x60sec, R=5x60sec, G=5x60sec, B=5x60sec.

FIGURE C: NGC 6992 of the Veil Nebula. What a rich and starry region we have in Cygnus! This image was acquired in less than one hour of data collection. L=10x60sec, R=11x60sec, G=12x60sec, B=15x60sec.

TESTING QSI’S 532WS SCIENTIFIC CCD CAMERA

E

F

G

FIGURE D: Abell 2151. This negative image took just one hour to image with an unguided 8-inch f/4 Newtonian and mounting. It shows a rich cluster of galaxies with many interacting pairs and tidally distorted objects. L= 52x60 sec.

FIGURE E AND F: Wide-Angle Views of central Cygnus in H-alpha. The field around gamma Cygni is rich in well-known "objects" connected by an underlying continuous matrix of hydrogen gas glowing in H-alpha with foreground and background strands of dark material. Taken with my 60mm f/2.8 Nikon macro lens through an H-alpha filter. H=4x600sec.

FIGURE G: Comet Holmes on November 2, 2007. You can see the milky tone of the inner coma and the distinct greenish-blue of the outer coma. Wisps of the tail are faintly visible between the 4 o’clock and 6 o’clock positions. The inner coma has been detail-enhanced to show the subtle streamers from the nucleus. L=16x60sec, R=16X60sec, G=16X60sec, B=16X60sec.

Astronomy TECHNOLOGY TODAY

47

Suggest Documents