Before I get heavy handed and potentially blast a company and their manufacturing QC, let me ask you what amount of error would you consider okay to accept (not that I have any recourse at this date) and live with? Consider that the camera is being used with a rotator.
The test I perform is to put a star dead center on the CCD and then take a time exposure for a full rotation. If the camera and rotator are "perfect" the star will remain dead center. In my case I have a rotator and adapters that are as close to perfect in X-Y shift as can be made. At the camera the rotator+ adapters move laterally ~ 7 microns as measured by a dial indicator! With my camera that translates to just over 1 pix!
So, the question is how much error would you accept if the CCD is off center?
I'd expect you'll see 0.1 of a mm or so variation, as the sensors themselves are simply inserted into a IC socket. All of those pins will likely seat differently each time.
For any manufacturer to then add another level of complexity by having an x-y stage that is subsequently calibrated down to the resolution of each sensor...well....as they said in The Castle: "tell 'em their dreamin'"
Thanks for your reply! With my camera 100 microns would be about 16-17 pix. Take a look at the attached photo! This was not with my best adapter and it shows ~ 232 pix error at 180 degrees rotation. My much better adapter reduces this to 190 pix which is over 1 mm in error! (the chip is a KAF16200)
According to the manufacturer this is considered "normal." I was completely mystified by this answer. They said the CCD has to "float" on top of the cooler to avoid thermal paths from the case, that there are a few pins to keep it reasonably centered. I cannot believe this is "normal" by any standard for other astro-cameras.
I suppose I'm being obsessive, but this is a pain to live with. Pointing sucks unless I'm at the rotator PA that the model was generated at.
It would be interesting to see how say a DSLR camera behaves if placed in the same rotator. If memory serves, they have remarkably tight sensor placement tolerances......
As an experiment I placed a smaller chip camera by the same manufacturer on the filter wheel (i.e., all same adapters save the adapter from FW to camera) and got essentially the same 1 mm error on rotation. I actually did this experiment before doing the dial indicator checks and I was certain that I had revealed an error in the FW, but no, the dial indicator confirmed only ~ 7 microns of error on full rotation after the FW. The only thing remaining was/is the CCD (on both cameras by this manufacturer).
If the reasons for the error are due to the difficulties surrounding cooling and thermal paths, a DSLR test wouldn't have these issues given no cooling.
I'm doubting, but without any proof, that an FLI or SBIG camera would have such a large error.
I believe QHY works to a 20 micron tolerance for orthogonality.
My FLI 16803 and Microline 16200 seem to have the same slight tilt so I use the same packer for both and I get round stars in the corners.
I think that would be the adapters or the filter wheel, MMOAG itself. I have had it that the same Proline camera gave round stars to the corners of an AP140 which you would expect to have superior machining and squareness.
No big deal, just work out the tilt and put a packer between the camera and the filter wheel on the offending corner or side. In my case its about a 0.55mm error on the top of the camera. I use a spark feeler gauge to pack it out with. That works well. Or if you want a more sophisiticated solution get the Gerd Neuman CTU tilt correcting unit. That seems to be the tilt adapter of choice.
With increasing MP counts on these CMOS cameras and faster optics on average these days, tilt is becoming increasingly obvious. Its quite common to see images with one or more corners showing tilt and distorted and bloated stars.
Peter,
The 1mm error corresponds to 0.5mm displacement of the chip. Considering how some of these cameras are made, I would not expect the centering to be better than 0.25mm.
I pulled apart my old QHY8 at one stage to correct the tilt of the chip, not worrying about the centering, and I was disappointed by the methods used for mounting it. The housing was made to high accuracy, which should have allowed the mounting of the sensor in an accurate manner. Instead the position of the sensor was determined by the PCB and the PCB was quite distorted by all the insulating materials that were jammed underneath it.
Another issue with these sensors, that I don't really understand, it that they seem to have more pixels than what is actually being used - at least some of the sensors that I have looked at. If that is the case then maybe the physical center of the chip is not the same as the optical center?
It does seem like camera manufacturers would care more about tilt rather than orthogonality. In my older square QHY8 the sensor is attached to a metal plate that is then screwed to the heat sink, allowing for some correction of tilt. I was able to greatly improve it using just a dial test indicator. If your camera uses a similar setup it seems plausible that it could shift in the x and y as well. Maybe there is some room for adjustment?
...
Another issue with these sensors, that I don't really understand, it that they seem to have more pixels than what is actually being used - at least some of the sensors that I have looked at. If that is the case then maybe the physical center of the chip is not the same as the optical center?
Yes interesting point, I've often wondered the same thing, but maybe within your comment we part of an answer. Perhaps during manufacturing those unused pixels are used for some form of individual camera calibration to move the sensor's geometric center in X-Y (electronically/via firmware/software) in order for the sensor center to be coincident with the optical center or at least for it to be at the geometric center of the camera's flange mount. That way when manufacturing lenses they could do similar by aligning the optical and flange center and thereby produce, in the end, a system which was as optically aligned as best as possible, manufacturing tolerances notwithstanding, with whatever lens is fitted.
Of course that's in high volume camera manufacturing, manufacturer's may do otherwise in different sectors of the market.
BTW ... just spit ballin' here
Best
JA
*Another possibility: Of course it's also possible/necessary for camera sensors with a pixel shift or electronic vibration control function to have some "spare" pixels to use for such functions in order to "maintain" the sensor size whilst actually "moving" the sensor, by using different pixels. Such usage (electronic vibration control) may not be used in astrocameras, but astrocam manufacturers will often source their sensors from suppliers to the higher volume photographic camera market in which they are used...... BTW and food for thought: I think I would be a great idea to try to use such as a form of pixelshift in a real time-ish system to provide some form of adapative optical control, be it only in X-Y, (certainly nothing like a flexible optic.).
I didn’t check my thread for a while so was surprised today to see a flurry of responses. Thanks guys!
Greg, I fear you may have misread my post. This is not about tilt at all. It is about X-Y shift that is revealed during rotation.
I cannot add anything to the current comments other than to once again state the bleeding obvious...so much shift (.5mm) really defeats pointing with a high end mount. Sure, a closed loop slew fixes it for sure. It ought to be unnecessary!
I’ve been trying to imagine how a simple, reliable, rigid X=Y adapter might be made?
I'd have to observe with sensor placement errors notwithstanding, my QHY600M fits into the CFW flange with about 1mm or so of play. Torquing up the grub screws without being mindful of this guarantees the sensor will not be perfectly central to the optical axis.
That said, building a pointing model with a mount, while doing so at random camera angles would not be a good idea....but the resulting pointing model with a fixed orientation should still be quite good.
I'd have to observe with sensor placement errors notwithstanding, my QHY600M fits into the CFW flange with about 1mm or so of play. Torquing up the grub screws without being mindful of this guarantees the sensor will not be perfectly central to the optical axis.
That said, building a pointing model with a mount, while doing so at random camera angles would not be a good idea....but the resulting pointing model with a fixed orientation should still be quite good.
Absolutely! But, one can have a great model that is out by 100s of pixels if you slew at a different rotation angle, no? In my case if I make a model at North up all is perfect as long as I slew at North up. If I rotate 180 degrees my pointing is off by 1 mm!