12/9/2002: I inserted the AO7 in the image chain. Another round of spacer arithmetic and I arrived within 0.25" of the optimal back focus distance. It was pretty gusty last night and I didn't expect much. With the open truss and the AO7 guiding at around 7 Hz or better, I was getting nice images with round stars. The wind eventually got so bad that it literally blew the scope away from the guide star between downloads so I had to shut down for the night. But there is no doubt that a closed tube would have been useless with those winds. Here is a single 10 minute exposure of NGC891 taken with the AO7 and wind blowing. FWHM's were as low as 2.5 arc-sec.
Fig. 1: NGC891
The only calibration was dark field subtraction, curves and levels (no sharpening) in Photoshop. As I didn't have any flats for this configuration, I used Ron Wodaski's new radial gradient removal plug-in and it did an admirable job of providing a quick and dirty flat field.
I did notice an issue that is probably AO7 related. In past nights when working with the PIR, it was kind of fun to expose an image while the PIR slews. Nice circular star trails resulted. However, with the AO7, the camera load is more off-axis and the target doesn't stay in the center. I suspect this is due to flexure of the CFW8 cover plate and its inability keep the off-axis load perpendicular to the optical axis. There is a product in development that may alleviate this issue, if it does indeed come from the CFW8 cover plate. A prototype should be available within a couple of weeks. If this doesn't resolve this issue, then there may be some decentration of the PIR due to the AO7-induced off-axis load.
The PIR is still awesome, though <g>
12/15/2002: Here are some images from rotation with the AO7.
Fig. 2: 0 - 270° CW Rotation Fig. 3: 270 - 0° CCW Rotation
These images were taken by starting the PIR rotating through a 270° cycle starting at 0 and then back. Once the PIR started rotation, the camera was exposed for 90 sec. The PIR took around 74 sec. to rotate through the 270° so there was around a 15 sec. dwell at the end to enable identification of the end.
It appears there is another factor, especially after looking at the right image. It appears the mirror in the AO7, which was not driven for this test, flopped significantly toward the 250° point of rotation. I happened to have 2 AO7 units and both exhibited the same behavior. Clearly this is not an issue while guiding but it is somewhat disconcerting while attempting to frame an object. See 12/23/2002 update, below.
12/20/2002: I repeated the above rotation exposure without the AO7.
Fig. 4: 0 - 270° CW Rotation Fig. 5: 270 - 0° CCW Rotation
The jiggles at the 250° point disappeared. Upon close examination of the image, the center of rotation is not on the center of the chip but about 40 pixels high (towards the guide chip. This seems to indicate the chip on this camera is mounted approximately .014" off-center, high. Not a big deal but interesting to note. For those putting optical stops in for the open truss, don't go too tight.
At Dan's request, I tried some repeatability tests for the TCC focuser. First I removed the AO7 to eliminate that as a potential variable. Using FocusMax v3.0.12 beta, I then ran 6 V-curves to get the best data. I ran 7 focus runs and discarded the highest and lowest count setting and averaged the remaining 5. I homed the focuser and repeated the 7 focus runs. I repeated this 3 times. Here is the data
|Run 1 (after V-curves)||12661 ± 4|
|Run 2 (after homing)||12656 ± 3|
|Run 3 (after homing)||12656 ± 3|
|Run 4 (after homing)||12663 ± 2|
After each FocusMax run I used my manual adjustment technique and in all cases came out 1 count higher than the FocusMax average. To put it in perspective, Dan told me that the width of the Critical Focus Zone is approximately 40 counts. I believe the that the beta version of FM gets one to pretty accurate focusing. Also, the repeatability of the TCC focuser is very high indeed, since the biggest difference after homing was between an average of 12656 and 12663.
I also recorded FWHM and HFD (Half Flux Density) during the tests. There was generally good agreement between the values for a given run, i.e., FWHM = HFD to a first order. The values themselves corresponded to 2.6 to 3.3 arc-sec., which I attributed to seeing variations.
Bottom line: TCC and FM work very well together and get quite close to what I would consider optimal focus. As I have reported before, I believe FM version 3 is definitely superior to older versions in achieving accurate focus with RC's. As I was doing manual focusing, I could clearly see some astigmatism when slightly out of focus. This astigmatic spot rotated by 90° from one side of focus to the other. In focus, there was no sign of astigmatism, at least with last night's seeing conditions.
Once I have the RoboFocus retrofit kit, I will run a similar test to see how the two focusers compare.
12/23/2002: Steve Mandel and I are developing a replacement CFW cover plate with the goal of a more robust and stable mounting with and without the AO7. We are calling this the Muscle Plate (MP). It replaces the existing cover plate and requires no modification to either the CFW8 or AO7. The prototype came in Friday and we had about an hour of clear sky last night. I repeated the PIR rotation with the AO7 attached to the MP. Compare these images to those taken with the standard plate.
Fig 6: -90° to +270° CW Rotation Fig. 7: +270° to -90° CCW Rotation
Note that there is absolutely no jumping, jiggling or other distortion in the image. It appears that the problem reported above with the AO7 may be related to the construction of the original cover plate. The MP design increases the strength considerably and provides a hard mount for the AO7. I will try to repeat the test again with the standard plate but I'll be reluctant to go back to it <g>
These exposures were 110 sec. at 1x1 binning. The PIR Takes around 100 sec. to cover 360°. There is some interesting low level movement throughout the rotation and it seems repeatable as a function of rotation. I believe the only way this movement can happen is if there is some very slight wobble of the PIR axis of rotation. Other ideas?
As a point of calibration, the PIR is mounted with its motor at the top, when the OTA is on the east side of the mount and pointed south. With that orientation, and the camera/AO7 combination is such that the camera is on top of the AO7 and that corresponds to the 180° position. In the above images, 180° as so defined, is up.
12/26/2002: I repeated the above experiment by removing the MP and reinstalling the standard CFW8 cover plate.
Fig 8: 0° to 360° CW Rotation Fig. 9: 360° to 0° CCW Rotation
The resultant images looked more like Figure 6 and 7 than Figure 2 and 3. I assume this is due to differences in tightening the three screws used for the standard AO7 mounting. It is worth noting that the center of rotation is closer to the center of the chip, but that could be due to mounting centration. I do believe that the more rigid plate gives an axis of rotation that is more consistent. The waves in the rotated image in Figure 6 and 7 appear to be a residual of the undriven mirror. The next image shows the result of exercising the Y axis during rotation.
Fig. 10: 360° to 0° CCW Rotation while exercising Y at 100%
The standard plate requires careful attention to mounting, especially the screws that mate against the ring to mount the camera. These point contacts must be secured and not attach on a previous point that has raised aluminum or depressions. I believe the MP will provide a more reliable and repeatable mounting of the AO7 to the camera.
Brad provided me with a RoboFocus (RF) kit, which included a motor mounting bracket customized for the RF stepper motor. Note that the TCC focuser uses a secondary focuser with absolute position feedback after it is homed. The RF has no position feedback; it depends on counting the pulses sent to the motor and assumes none are lost.
The RF was mounted to the secondary positioner as shown below.
Fig 11: RoboFocus on 10RC secondary positioner
The RF motor is 2.3" shorter than the TCC servo motor. Also, it may extend up to 0.1" into the optical path due to the motor being off-center and larger in diameter than the servo motor. For the 10" RC and a 4" secondary, the RF measures approximately 2.1" from the shaft axis to the outside of the housing. This won't be an issue for larger scopes and is probably only a slight issue for the 10". I selected an RF setup that would approximate the TCC for comparison. I trained the RF over a travel of 0.8" from the homed position of the TCC. Here are the parameters that I used:
These are the parameters I used for training and operation. Based on the RF having a resolution of 3600 counts per turn, a lead screw of 20 tpi and a step size of 2, the resultant resolution is 36,000 counts per inch, which is close to the TCC resolution of 40,000 counts per inch. This resolution results in a critical focus zone width similar to the TCC/servo - around 30 counts at F/9.
Next, FocusMax was used to obtain focus. 6 V-curve runs were taken and averaged on a night of reasonably good seeing with focused star FWHM's of 2.2 - 3.0 arc-secs. FM was set up to use the convergence feature with a convergence within 2 steps for 5 samples. Two separate sets of 12 focus runs each resulted in a an average focus of 10644 and 10645 ±4 counts. I saw no problems with the resultant focus point obtained with FM and it agreed very closely, within a count or two, of my best manual focus effort.
My conclusion is that the RoboFocus secondary focuser has acceptable performance with the added advantage of being 2.3" shorter than the servo and the slight disadvantage of intruding slightly into the optical path. The latter issue is limited to the scopes with smaller secondary mirrors, such as the 10".
1/13/2003: I was interested in taking some unguided exposures, especially based on the way the Paramount was performing. I had read and understood Stan's posts about having the noise contribution of the background count exceed the readout noise and that lead me to use 10 minute sub-exposures for my ST-8E/10RC combination. Nevertheless, I wanted to do some experiments at lower exposure times. I took 5 sub-exposures at 2 min. each unguided, followed by a 10-minute exposure AO7 guided at 30 Hz. I then measured an average count and standard deviation in a star-poor area of the image. The data was reduced with 10 darks, 20 bias frames in Mira. For this exercise, signal-to-noise ratio is defined as mean/standard deviation.
|5 2-minute frames
|5 2-minute frames
Min/Max Clip combined
|1 10-minute frame|
The slight reduction in SNR between average and min/max clip is an expected result. What is unusual is the SNR similarity between the combined short exposures and the single long exposure. This is a somewhat surprising result to me and I would be interested in any comments. It may be the sky glow (the moon was up), was sufficiently high to partially mask out the readout noise. I need to repeat this experiment without the moon in the sky.
During the course of these tests, I came across another issue relating to the TCC secondary focuser. I was getting some apparent "trailing" even at very short exposures. Suspecting astigmatism, I refocused the secondary and arrived at the same focus point. However, subsequent exposures showed no trailing. I had not moved the scope. I only went from my original focus count of 10210, move in by 30 counts to 10181 or so and then back to the original 10211. Here are cropped images taken before and after the 30 count in-and-out movement.
Before 30-count move in-and-out After 30-count move in-and-out
The images above were taken at an altitude of 48° and a declination of -8°. Exposures were 2 minutes. I am not sure whether this is a "settling" of the secondary transport after a slew or what but I've now programmed FocusMax to come at focus from the out direction. It is not at all unusual to have a preferred direction to get to optimum focus and with the secondary focuser, it seems to be "out" for me. I would be interested in hearing if others see something similar. I repeated this process at a different part of the sky with similar results, so I suspect it may be real.
Finally, just for fun, I took an image of the Eskimo, just to see if the color was as I remembered it looking through an eyepiece on the WIYN Telescope (3.5 meter) on Kitt Peak. That was an unbelievable experience, by the way!
NGC2392, The Eskimo Nebula
Unguided exposures of 5x3m red, 5x3m green and 5x5m blue. Processed in Mira, combined in Maxim and curves, levels and USM in Photoshop. The combined green channel had average FWHM's of around 2.5 arc-sec. Maybe a little too green but pretty close to what I remembered.
Go back to the Impressions page
Hidden Loft Observatory Home page
More to come...