I was wondering if my experience is shared by others who are using 6303 based CCD cameras. I have an SBIG STXL6303E camera. While doing some tests to try to quantify the accuracy of my photometry, I was looking more closely at my images. I noticed some faint horizontal lines in some of the calibrated images and followed these back to the flats. I posted on the SBIG support forum here: http://forum.diffractionlimited.com/threads/stxl6303-horizontal-and-vertical-lines.3804/#post-22206
I sent my camera to SBIG who have been looking into this. They are finding these faint image lines in all of their 6303 cameras. On the cameras with newer chips, the lines are always 36 pixels apart as are mine.
I am interested to know if anyone else using the 6303 can take a close (highly stretched) look at their flats and tell me if they see similar results? Anyone else see them in their calibrated images? When I get home Sunday, I can post a screenshot of what I see. If you happen to be in Nashville, I’d love to compare notes.
best regards,
Cliff Kotnik
Seems like something that SBIG could find with an oscilloscope.
Maybe the chip specs hold a clue, but it could be the amp sagging.
Cliff--I just e-mailed you a recent master flat FITS file from my 6303E. It has no regular artifacts that I can't flat-field away.
Although with SBIG's STXL camera series, all problems get worse with low voltage, so you'll want to verify that your power system is actually delivering 12 V to the camera, under 100% cooling load. That rules out using Bisque's cable set which, however convenient, inflict big voltage drops for these newer cameras.
Thanks Eric. I have not seen the email, but then I only have my iPad while at the conference and would have not way to see a FITS file. I’ll check when I get home.
I do run the SBIG power all the way to the camera and returned the power supply to them with the camera. I’m sure they would have tested that. Additionally, they have reproduced similar patterns in older and newer cameras of their design and others. They report the exact nature of the pattern varies from chip to chip. This must be new information for them or they would not have had me send the camera in.
The real question I need to dig into is if this can be calibrated away for my specific camera.
Best regards,
Cliff
Reading through the emails here, it looks like Arne has a handle on it as a chip defect. But it seems inconistant from user to user. That might suggest aging components in the power supply. And, I'll bet these power supplies perform differently at temperature extremes. Also the suggestion has been made here that the cables provided are not large enough gauge. If things are that close, then the little voltage drops on connectors could come into play. The connectors supplied are not exactly military grade and maybe the subcontractor that crimps connectors occasionally spits out a higher loss connection.
The other thing we don't know is what the camera does with those voltages. Is the 5 volts regulated to 3.3 volts? 12 volts regulated to 10 volts? Manufacturers could provide schematics. The electronics is something any good technician could cook up. The tough part of the thing is the cooling and software integration.
I never had the courage to access the camera electronics to apply probes. Maybe if you had a friend with a few decades of experience.
It is possible to check out the power supplied at the end of the cable. You can make a short extension with taps to access the voltages and grounds. Need to find the plugs and sockets online and have the correct crimping tool or avoid cold solder joints. You will want to look with a meter and an oscilloscope with the camera at 100% cooling. If the meter looks noisey, the scope will tell you what is going on.
If you can get the camera specificatios for current draws for all voltages, you can provide an equal draw with power resistors. Don't use a 1/8 watt resistor for 5 amps. Make use of Ohm's Law. ( Remember, once all the smoke comes out of the resistor, it won't work any more. )
Another thing is to just try a new power supply that is made for the camera or a bench supply that can supply all the voltages at rated current. Is is +5 and +/-12 Volts? 5V for the logic and +/- 12V for the OP amps. Global used to have just what I needed for half the cost of one made for the camera.
If the power supply is a switcher, you have 60 to 140 KHz (approximately) switching noise. It gets worse if output filter capacitors go bad. Seems like frequency that could make lines during the CCD scan. Check the suggested scan rates on the chip's spec sheets. See if they rhyme with typical switcher speeds.
As a general rule, I like to see less than 2 to 10 millivolts of noise on the scope while plugged into a supply. Better than 1% regulation. On a scope, you may discover that a switcher does not run at a constant rate. The industry actually makes dither chips that sweep the switching speed a little bit so that all the noise doesn't appear at a single frequency. That is how manufacturers comply with the FCC's E&M radiation limits.
Electronics has a way of failing rather than healing. So you could just watch your images to see if the problem gets worse as the power supply and camera get old.
I would use a break-out box or cheater cord to study power at the camera end of the cable with an oscilloscope capable of at least 10 MHz. You could look to see what's coming out of the cable with no load on it. Might be something obvious. But that is not the whole circuit. Test with a load.
But then it looks like you can safely ignore it as long as it doesn't bug you.
RayTRE
Hi,
in the past there were banding issues with SBIG cameras. I am pretty sure a search of the Yahoo group will bring up some discussion about it. I knew about astrophotographers who had this problem. I am using since several years now FLI related cameras and have no problem with this. My chip is the 16803.
Josch
Thanks for the suggestions here and in Nashville. I'd like to add a bit more detail.
First, let me say I don't think this is reflection on the SBIG camera design. According to their service folks, it is not a problem with my particular camera either. It seems more like a question about the 6303 chip itself.
If you zoom into the attached Flat.PNG file,you will see the series of regular horizontal lines. Now I have compared notes with Eric and his flats look very similar. Eric reports flat calibration removes these for him. I too see these artifacts removed by calibration - but not always.
I came at this initially from another angle. I was participating in a little experiment from Tom Calderwood to compare photometric measurements with Landolt standard stars. It seems I caught a case of "Arlo Envy" and was keen to look for ways to improve my measurements. I had taken a long time series of V1326 Cyg and was reviewing the calibrated images in AIJ's image sequence animation when I saw horizontal lines flash on and off as the animation cycled through the images. It was no where near a majority of the images that exhibited these lines and they are not very large in depth. You have to have the right stretch to see them. You can see what I mean if you zoom into the other image I have attached.
After seeing these in a calibrated image I looked more closely at the flats and saw the regular pattern there. I cannot really figure out how a problem in my calibration process could remove these lines from most images of a sequence while generating them in a fraction. It is not the end of the world. I do examine my images and can reject any where I see these lines near a target or comp star.
Hence, my question is really for owners of 6303 based cameras regardless of the brand. Do you ever see these horizontal lines in calibrated images?
thanks
Cliff K
Hi Cliff,
Thanks for attaching the png's, as I was not sure from your description as to what the problem looked like.
My experience on this kind of uniform pattern is that it is inherent to the sensor itself. Usually it is a square pattern, representing the photomask used to create the pixel structure, and is then stepped across in a raster scan to make a full array. The stepping is not quite uniform, and so the border pixels tend to be slightly different - more sensitive, more dark current, etc. That said, this may or may not be the same effect as I don't immediately see any vertical pattern (though maybe things are perfect in that direction!).
My first question is what filter were you usuign for these flats? I'll look at the SRO flats, where we are using an STL-6303.
Remember you are dealing with something that might be obvious to the eye, but is of extremely low light level and most likely won't impact your photometry at all.
Arne
Arne,
In the flats, I see the pattern of horizontal lines about 36 pixels apart. There are verical lines, but they are perhaps 7-800 pixels apart. I see these in all my flats - B V R I L and Ha. The particular flat screen shot I posted was V. The pattern in Eric's flat is very similar, although Diffractin Ltd reports they see the pattern not as regular in an older generation chip they happen to have.
I see the failure of my calibration to remove the artifacts appear only for the horizontal lines, but I have found this in a fraction of both V and R images which I oftern use for long sequences. Again, it is far fewer than half the calibrated images that show the horizontal lines and not nearly every file that appears in my flat. The lines are a single pixel wide. I agree it is a small imperfection. I need to stretch and zoom just right to see it.
thanks,
Cliff
Cliff, from what I see here, my 6303's pattern is the same in shape but perhaps a little less in intensity--hard to judge accurately.
But as far as I can tell, my calibration gives full removal of this faint pattern. For the record, my process every 2 months is to take 16 flats in each filter, median-averaging them to filter master flats in MaxIm's Set Calibration, applying very many dark and bias frames taken at the same time, all at max cooling (-35C for a SBIG STXL).
Hi Cliff,
I just looked at the master flats from SRO, which uses the STL-6303. No horizontal lines are present.
If you zip your fits image, I think you can upload it here to the forum. I'd like to see the raw flat and the dark/bias frame you used to calibrate it.
Eric, flats every 2 months is out of the norm. I highly recommend flats with at least a 1-week cadence, and if you can take them every night, even better. They cost you nothing, as you use twilight time for flats (either sky flats or any projection system). With AAVSOnet, we see new dust, dust moving, moisture, etc. on a regular basis, which is caught by the more frequent flats. Just a suggestion.
Arne
Arne,
Here are the calibration files.
Cliff
Arne, thanks. When I had my own roll-off roof, I did take flats every night. Now, in a New Mexico hosted facility where the roof opens before sundown and opens after sunrise, panel flats must wait for cloudy nights. But your advice is sound, of course--I can and will take flats during our 1-2 cloudy nights each week.
Hi Cliff,
I do see your pattern in your master flat that has been dark subtracted.
There is nothing obvious in the dark or the bias; good!
I notice that the STX's 6303 is being run in a 3072x2047 mode, rather than the STL's 3072x2048 mode. Interesting. You are running at -35C, while SRO runs at -20C. Our STL camera dates from 2009; your sensor is newer. I'm wondering if the foundry made some changes. Other than that, I have no suggestion. You might try running at -20C and see if it is a temperature effect. The pattern is very low level and should calibrate out.
Arne
Arne,
Thanks for the time to look into this. I will try the different temperature. The 2048 vs 2047 is a mystery to me. I do not know of any control I have over it.
best regards,
Cliff
A higher temperature will reduce the current draw which reduces voltage drops on cables and connections.
So far the emphasis has been on the CCD. It could be the ADC chip responding to power supply noise. Skipping a bit. What is the number of ADU's between lines and adjacent pixels? If it is one ADU the ADC could be skipping a bit. ADC's are a whole nuther science. There are specs for stepsize, min/max voltages, and much more. The amplifier also has a spec for sensitivity to power supply noise.
But then you could do the analysis to see how many bits it takes to become significant. This forum suggests that it is not important to measurement accuracy.
RayTRE
Ray,
Thanks for the suggestions.
To respond a bit, I will confine my comments to those lines that have found their way through my calibration process. Just to makes sure we have the same focus, I find a fraction of my images in a long sequence have a fraction of the horizontal lines that show up in the flats appear in the calibrated image.
So for a typical image with a typical aperture/annulus setting, I find the total ADU count in the aperture minus total ADU in the annulus when measuring a patch without stars is overstated by 1,000 if there is one of the worst horizontal lines passing through it. To put this in context, a 12.3 mag star with SNR around 700 in the same image has a aperture-annulus of 175,000 and a 13.6 mag star has a count of 55,000. So it seems to me that I need to reject images where the lines pass through the target or comp star. Of course eliminating the lines altogether would be ideal.
I do like the idea of cooling to -20C going forward. If I find the intermittent lines in calibrated images go away I can dig into the power supply and associated cabling more.
best regards,
Cliff