Recently, I have had the opportunity to evaluate an Apogee Alta U16M. Apogee has been making large format cameras for a number of years and previouly shipped products based on the KAF16801. This model uses the newer KAF16803 large image sensor. The KAF16803 has 4098 x 4098, 9 micron pixels and is approximately 52 mm in diagonal.
Even though this was an engineering sample, the camera arrived on 4/25/2007 packaged as a finished product, with a complete installation manual, software, power supply and test data specific to this camera. All Apogee cameras have a standard two-year warranty and a lifetime guarantee against condensation in the camera. The sensor is sealed into an inner chamber filled with argon. The chamber has a lifetime guarantee against condensation. The latter is very significant since some other cameras having a sealed and inert-gas filled chamber have only a 90-day guarantee.
The low profile camera chassis is blue-anodized and uses a massive heat sink on the back to remove heat from the sensor. I had an adapter made by Precise Parts (highly recommended) to mate the U16M to another vendor's filter wheel and is shown in this image. (The evaluation sample I have was an older chassis design so that I had to have the plate that mounts to the camera face drilled out to accept a 2.5 x 20 tpi thread. From there, a Precise Parts adapter interfaced to the filter wheel. Current production U16M has t 2.5 x 20 tpi thread at the output of the camera chassis, eliminating the need for the transfer plate and reducing back focus by .18".)
The sensor is massive as well, as can be deduced from the Melles-Griot 63 mm shutter. The shutter has a minimum shutter open time of 0.03 s. For comparison, here is the U16M beside my FLI IMG6303E, which uses the KAF6303E with 3072 x 2048, 9 micron pixels and is approximately 33 mm in diagonal. Size does matter!
The U16M is considerably larger in cross-section but significantly less deep. Interestingly, the envelope for the two cameras are similar - 147 cu. in. for the U16M and 157 cu. in. for the IMG6303E. Here is a side view of both cameras for comparison.
Software installation went as smooth as could be imagined. Perhaps unique among imaging cameras and most amateur astronomy equipment, Apogee had their drivers Microsoft certified so there was no annoying "Continue anyway" button. The driver installed flawlessly and was recognized by both Maxim and CCDSoft (full version). The camera was then mounted to my scope for testing.
The cooling fans and heat sink can be seen to the left. The CFW-7 filter wheel is next, followed by my off-axis guider with an SBIG ST-402ME guide camera and finally the RCOS Instrument Rotator. This is all mounted to a 14.5" Ritchey-Chrétien telescope operating at F/9. It makes for a nice, compact assembly.
The U16M has a USB 2.0 interface for fast downloads. I have a Windows XP Home machine in the observatory with a modest Athlon 2600+ processor and 1 GB RAM. It is accessed by my office computer via Radmin 3.0. There is a USB 2.0 hub mounted on the OTA, along with an Edgeport/4 serial to USB converter. A single USB cable runs through the mount to the computer. The U16M is powered from my 13.8 volt power supply via a barrier strip also located on the OTA.
I tested the Apogee Alta operation with both CCDSoft 5.00.182 (Alta driver 1.00) and Maxim 4.58. Both performed satisfactorily. Apogee provides a lot of information on the camera in its low level driver. Maxim seems to report this information very completely and compactly while CCDSoft has almost the same data. See the following screen shots:
Maxim puts all the available info in a single screen; CCDSoft has most, but not all the info Maxim has. For example, CCDSoft does not show the input voltage. Maxim provides the complete set of FITS keywords. CCDSoft is lacking XPIXSZ and YPIXSZ.
I prefer to test download times at a combination of binnings with both CCDSoft and Maxim. I also tested 100 x 100 pixel download times, such as might be encountered in focusing. Testing consists of taking a number of bias frames and averaging the results. Taking bias frames only removes any shutter issues from the timing. Here are the results:
|Full frame, 1x1 bin||23.0||24.0|
|100 x 100 frame, 1x1 bin||0.9||0.8|
|Full frame, 2x2 bin||10.4||9.0|
|Full frame, 3x3 bin||6.0||6.0|
Thus there appears to be no significant difference between programs. It is worthwhile to note the U16M has a 32 MB frame buffer. This makes the data immune to any computer response issues, for example if the computer is doing something else during download. In my setup for example, frames are directly written from the observatory computer to my processing computer over the local area network, with the attendant Ethernet collision avoidance and latency issues. No artifacts were noted in any of the data.
The U16M is rated to cool at least 40°C below ambient using programmable, intelligent cooling. In a nominal 30.7°C ambient, I set the camera set point to -30.0°C to see where it would limit. During cool down, the U16M firmware revised the set point to -9.6°C, apparently based on getting the cooler duty cycle to approximately 80%. Maxim had reported the lowest cooler temperature to be -12.4°C. The cooler drive, which had been at 100%, slowly backed of to 82% at the revised set point of -9.6°C. At this point, the heat sink temperature was at 40°C with the fans on at medium speed. The heat sink is cooled by 4 Sunon magnetic levitation fans for low vibration.
The U16M takes a considerable amount of time to cool down since the thermal transient is carefully managed. This is different than most users are accustomed to but shouldn't present any problems. Simply turn on the cooler at the same time you start your OTA cool down. It is not unreasonable to cool a telescope for 45 minutes to an hour. The U16M should reach its operating temperature in that period. I recently modified CCDAutoPilot to accommodate the camera manufacturer's thermal transient management approach.
Gain and Read Noise determine optimal sub-exposure duration for color imaging and the noise floor for narrow band imaging. These key parameters were measured using AIP4WIN v2.0. Three measurements were taken and averaged. Gain measured 1.4 and Read Noise measured 9.6e when taken at a sensor temperature of -15°C. The U16M data sheet had values of 1.3 and 8.9e respectively when taken at -18°C. This is substantial agreement.
I checked the average bias level at ambient temperatures of 20°C and 30°C. I found the average bias stability to be within 1% over the 10°C range. I believe this is excellent and helps insure reusability of bias (and dark) libraries.
Quantum efficiency (QE) is a good metric of overall camera sensitivity. Heretofore, the KAI11002 represented the largest format, anti-blooming image sensor, albeit with low QE. The more sensitive KAF6303E has a smaller sensor area and is not anti-blooming but higher QE. Both have 9 micron pixels, as does the KAF16803. So how to the sensors stack up?
|x pixels||y pixels||megapixels|
Here is a combined QE curve for the three sensors. The data represents overlays from the respective Kodak data sheets.
The 16803 improves on both sensors, except for the red end of the 6303E. It has anti-blooming capability and uses a second generation micro-lens design. Thus far, I have not seen the vertical artifacts that appeared common with the older KAF3200ME, first generation micro-lens design.
I have a set of AstroDon E-line filters that are used with the IMG 6303E. I measured the G2V response (2 measurements) using CCDStack and arrived at an RGB ratio of 1.04/1.0/1.07. This is essentially balanced and compares to 0.8/1/1.2 for the IMG6363E. The MgF2 coated fused silica output window, coupled with the KAF16803's QE seems to be a good match for the E-line filters.
With such a large chip, I had no delusions of being able to fully illuminate the chip since as a minimum, its diagonal exceeded the diameter of my 2" filters! I also knew the pick-off prism for the off-axis guider would intrude into the optical path. However, I was interested in seeing how bad it would be.
My first test was with my OAG set up as it was for the IMG6303E, knowing there would be some significant optical intrusions. I used M101 as my test subject and these images represents 12 x 10 minute sub-exposures at an image scale of 0.55 arc-sec./pixel.
This image shows the resultant image with my OAG unmodified and set up for the IMG6303E. You can see the expected vignetting at the corners of the image as well as the impact of the pick-off prism from the OAG immediately below the galaxy. In looking at the vignetting, there appears to be two stops involved. Additionally, the pick-off prism for the OAG is limiting the quality area.
This image shows the result of opening the apertures in the CFW-7 cover plate and the OAG. Additionally, the pick-off prism was reworked so that it could be moved further out of the optical path. The usable area is further increased in this image. This represents the usable imaging area with 48 mm filters. It is clear that to take advantage of the full area of the sensor, the larger, square filters are necessary. (The data was taken through high clouds and the camera had to be rotated to get a suitable guide star location.)
I decided to add a feature to CCDAutoPilot that I called Smart Sub-framing. Using the camera control program, a cropping is defined and maintained in CCDAutoPilot for all subsequent image acquisition as well as for calibration frames - dark, bias and flat field frames. For my case, I used a 3402 x 3402 crop area centered on the chip. More could probably be squeezed out but this left a comfortable, square image that gave a FOV of over 0.5° on a side at an image scale of 0.55 arc-sec./pixel.
This is pretty much the usable square area with 2" filters in my situation, around 70% of the available imaging area. A slightly larger area is possible with a rectangular format.
So how does this camera work with real imaging? With the low dark signal at -15°C and the quiet background, I did some imaging. I was interested in taking advantage of the large FOV and quiet background. All of these images were sub-framed to 3402 x 3402 pixels, resulting in a FOV of approximately 0.5° on a side. Here are some images taken thus far:
Click on each of the above images for a larger image, exposure details, etc. Having such a sensitive, anti-blooming, large sensor is a dream to work with. Dealing with the file sizes is another matter altogether. See the Data Reduction topic, below.
The various connections are shown below.
On the left is a custom interface connector for special applications. The next two jacks are RS-232 connections that can be used to bring two serial port data over the USB connection. Unfortunately Windows COM port drivers don't appear to be available for these ports, limiting their usefulness. The two LED's can be programmed to light for various status conditions and can be disabled if desired during exposure. Finally, there is the USB 2.0 connector and the power connector.
I found the camera very responsive in operation. Due to the continuous flushing of the sensor, image capture started instantly. The advantage of this over "flush at the start of capture" (Manual flushing), was most apparent during a FocusMax run. Manual flushing adds around 5 sec. to each exposure, regardless of focus exposure binning and sub-frame size. Here, there is no such limitation. Since FocusMax uses small sub-frames of 100 x 100 pixels by default, a six frame focus run takes around 40 sec. with Manual flushing and 8 sec. with continuous flushing.
Since the camera is USB 2.0, I replaced my USB 1.1 Icron Ranger USB extender with a PC in the observatory and put it on my LAN. Using CCDAutoPilot, all image acquisition was directly downloaded to the RAID 5 drive in my office PC over the network. Despite the usual network latency, the images appeared on my office PC clean and without any download artifacts. I suspect this is due in large measure to the camera's onboard buffer.
The data files for such a large camera are also huge. A 1x1-binned image is 32 MB. Processing a stack of such data is challenging for any 32-bit computer. I recently upgraded the motherboard in my office computer to an Athlon X2 5600+ dual core processor with 3 GB RAM, which supports 64-bit operating systems. I then installed Windows XP-64 bit Professional on a separate partition. When I want to process images, I boot into the XP-64 partition and take advantage of CCDStack's 64-bit support to easily and quickly process the data. I have easily processed a stack of 12 sub-exposures. Once the masters are available, I switch back to XP 32-bt and continue my processing.
In the short time I have used the U16M, I have been very impressed. From first opening the box through installing the software, the entire user experience was very pleasant. The fit and finish is outstanding. The highly flexible mounting arrangement almost guarantees a successful mount to any OTA or filter combination. Ashley at Precise Parts can probably build anything you need if Apogee doesn't have it. Documentation and software installation was the easiest I have seen for any camera.
The camera operation has been smooth and responsive. The continuous sensor flushing makes downloads for focusing and other short duration exposures very efficient. The bias point stability insures the ability to use a master bias for accurate flats a reality. Anti-blooming is a treat when it comes to image processing for aesthetic imaging. The large FOV opens up considerable imaging possibilities, even with the 2" round filters I have. Of course the large square filters are to be preferred if your system and wallet(!) allow. In my particular arrangement, I find the camera form factor much more suitable, due to limited observatory space. Since it runs from a nominal +12 volt power, I can use the system voltage I have on my OTA and eliminate one power supply "brick". With the built-in serial ports, I could eliminate the Edgeport/4 serial-USB converter I currently have and send the focuser and rotator control down the U16M's USB cable.
At first blush, cooling capacity seems to be a downside, especially in warm southern Arizona temperatures. It should be noted that Apogee makes a liquid cooling back that can increase the cooling differential to 60°C but I am not sure this is needed. Read noise is measured with typically short exposures and was measured at a respectable 9.6e. However, when taking a long exposure, dark current noise adds to this number. Fortunately, the 16803 has a dark signal that is only 20% of the KAF6303E and 0.6%(!) of that of the KAI11002. Recall the dark signal doubling temperature is 6.3°C. Thus, a KAF16803 can operated at least 13°C warmer than the 6303E and deliver approximately equivalent dark signal as the 6303 and similar sensors. I explore some of the issues in this paper.