STXL-6303e Incorrect Max ADU

Something is wrong with your data or your math. There's no way the median could double with a small change in average.

 

Tom, I'm another customer, and was trying to follow your logic and your math, and it`s not right.

Start by doing your tests again with 1x1 binning, because that is what matters.
Cool the camera and let it stabilize for at least 30 minutes or so. So, if you are at room temperature, maybe set the setpoint to -35C to be below a 20C room temp. (I think there is 60C delta-T cooling)
Remember to mask out any hot pixels from your calculation.
Take a look at what the EGAIN value is in the FITS file, that will give you the amplifier factor. It is measured in e- / ADU.
This will come in handy to help you understand the max dynamic range of the camera, details below.
If you're trying to test this in the house, you might cut up some 40+ squares of white paper and stack them over the nosepiece of the camera, to get even lighting and better dimming.

With a 16bit Analog-to-Digital converter, the max readout value the ADC provides is 65535.
The KAF-6303 CCD sensor has a vertical pixel saturation of typically 100,000 e- (electrons), although it could be from 85,000 to 120,000 according to the data sheet.
So, if you massively overexpose the chip, it will saturate, and you`ll get garbage out. eg the electron well is full, and it spills into neighbouring pixels.
It varies by horizontal and verticals. The vertical number is lower on the KAF6303.

There's a calibration in the camera that is used to go transform the CCD pixel Vout signal range to the input range of the ADC.
e.g. an amplifier factor to map 100,000 electrons to 65535 (0x0000 to 0xFFFF) counts, however, it also has an offset to get rid of the noise level/dark current etc.
e.g. ideally, you want the ADC 0 value to be as close to the minimum Vout that you get from a properly biased, dark, cooled chip.

That minimum has to be measured and calculated.
From the data sheet, the CCD noise floor is specified at 15 to 20 e- .
There's readout noise 11e- too.
So, you've got a baseline of maybe 21e- that have to be allowed for/

Then the real monster is the dark current.
There is a dark current of 15 to 50 e- per pixel per second.
So best case for a 30 second exposure, 15x30=450 e-. For 30 minutes 15 x 1800 = 27,000.
Worst case, 50x30=1,500 or 50 x 1800 = 90,000.

I think the "art" of designing the camera is taking in all of these factors.

So, I went looking for a real-world example. A friend's 6303 cooled to -30C with an EGAIN of about 0.37.
The 65536 count on his camera works out to a total system range of about 177,000 e-.
If his CCD has a full well depth at the upper limit of the data sheet value of 120,000, he would get a maximum pixel readout of 44,431, before the pixel was saturated.
Add in the noise, dark current, etc., and you can see that 120,000 goes way up.

(I havent allowed for offset from noise or dark current etc), which could easily bump that beyond 65535, but the ADC will just top out.
So, if you're trying to figure out what ADU is the saturation point for your camera, I'd try adjusting light levels for an ADC count below 44,000.

Binning is done different ways on different cameras. Some chips have multiple Vout pins (eg RGB cameras used in Mono mode), and they get summed in the camera by hardwware.
I think your camera does the binning in firmware/software, and so mapped 2x2, your software pixel value could be 4x higher, although the individual pixels still max out the ADC.

And if you are still getting bad data, share a sample FITS file, and then we'll have the details.

Data sheet etc here:
http://www.onsemi.com/PowerSolutions/product.do?id=KAF-6303


Cheers
Colin

 

Well depth varies a bit sensor to sensor. We adjust the A/D converter gain on these cameras so that the maximum ADU level of 65535 is around the point of CCD saturation. Unfortunately customers think if the camera doesn't go all the way to 65,535 that something is wrong. In fact the opposite is true - if you want to get the maximum dynamic range from your camera, the CCD sensor should saturate first. That means the converter won't show 65,535 unless you are binning.

Unfortunately the average consumer complains if the converter doesn't go all the way to 65535, and this leads to a lot of time wasted explaining how to correctly set up the converter. So a little while ago we adopted a policy of adjusting the cameras so they always fully saturate the converter, and lose a bit of the dynamic range of the sensor. We actually have to over-saturate a bit to make sure most of the pixels go all the way to 65,535, which makes things even worse. But it makes customers happy.

There is more than sufficient oversampling of the noise with a 16-bit converter that it doesn't matter much what exactly the full scale is set to. So you're not losing anything at the bottom end no matter how we adjust it.

From my point of view, your camera is actually set up correctly.

Now, if your camera has a linearity problem, that's a different story. We do run a linearity test on every camera manufactured. If you can provide your serial number then I can look up your test results. If there's a major discrepancy then we may want to investigate further.

 

That is by design. The sensor has more well depth in the readout than in the pixels. That gives it more dynamic range when binned - i.e. the output voltage goes higher - so we automatically adjust the gain to make use of that added dynamic range. As the FITS header will indicate.

Well of course.