Last Modified:
 
email dave@allmon.com

First Things First
 

I had earlier said that the CCD´s I had were putting out 3 volts instead of a nominal 1 volt. I had a Kodak Moment and have subsequently seen the error in my ways. I had a gaping hole in my test fixture that was allowing light to get under the shutter and polute the CCD as it was being read.

With the fixed fixture, my new maximum output is 1.72 V. Still enough to cause a problem, but not enough to cause Mr. Eastman to roll over. It seems that these devices (in spite of what the data sheet says) are all tested to guarantee 100ke well depth. Hence they put out a voltage within +/- 10% of the rated voltage, and bloom at around 1.2 V.
 

Gain effects
 

The standard video amp gain for the Genesis is 10.1 - determined by the value of R20 (9.1k). This is down from the 15k specified in the Audine. There doesn´t seem to be a causal relationship between the Kodak datasheets and the 15k value chosen by the Audine designers. I suspect this value was chosen as a compromise.

The Genesis gain of 10.1 is not nearly so arbitrary. It is chosen to make the video signal of a typical KAF sensor just fit the input range of the ADC. Why wouldn´t that work? It diesn´t work for one reason - Kodak couldn´t make an accurate datasheet if they untied both hands. The datasheet lists the well depth as 85 ke~ minimum, 100 ke~ nominal, and 120 ke~ maximum. The product manager for full frame sensors says: "The sensor capacity is tested to be at least 100k on each part" That means the minimum value in the datasheet is nonsense, and the value given as nominal is actually minimum. So Kodak guarantees that the 10.1 gain will work only for units at one extreme of the range, and that any error in gain will be toward clipping the signal with the ADC.

There can be an imagined increase in signal by overamplifying it. A brighter image is seen, and brighter = more signal. True, but brighter also =  more noise in this case, since the noise is being amplified by the same amount.
 

Do you need it?
 

To even determine if the gain adjustment is required you must modify your camera. You need to solder a wire-wrap wire or other suitable jumper across R20 to reduce its resistance to zero. This has the effect of reducing the video amplifier´s gain to unity. The CCD output will have any offsets added, will be clamped, inverted, and passed along to the ADC, but it will not be amplified.

If you then shoot a series of flats or specular highlights (sunlight on a car windshield, etc.) you can find the maximum value the camera puts out at saturation. A star will work if you feel the need to be astronomical about it. The idea is to get the CCD to put out its maximum voltage. Don´t over do it though, or you´ll wind up with a gain too low (like I did). All you need is the highest ADU value in the saturated area. Do this and write the number down.

If the number is considerably higher than 3276 or so, then you could benefit from this mod. If the number is considerably less than 3276, then you could benefit from the mod.
 

Maybe you have a device like mine - maybe you don´t. You may have a perfectly matched video amp / CCD combination, or you may not. How do you know? The first clue is your output value at saturation. If it is 32767, you have the problem. How bad? It could be anywhere from 1 ADU to 100,000 ADU. You don´t know until you hack the camera. If your saturation value is less than 32767, you don´t have the problem. If it is much less, like 16384 or less, then you have the opposite problem. Either way, the solution is the same - match the video gain to the CCD output sensitivity.

Let´s assume that you have determined that your amplifier doesn´t match your CCD. The direction and magnitude of the error doesn´t matter. We want to match it as close as we can.
 

How to do it
 

The CCD output gets weird under extreme saturation conditions, so we want to do this under very subdued lighting.

First, we want to use a jumper to short across R20. This is safe, and sets the gain of the video amplifier to 1.

Button it up and bring the camera to temperature.

Shoot a frame either on the scope, or with a long tube attached to the front of the camera. The tube will help provide a colimated beam. Sneak the exposure time up until the image (or highlight) just starts to saturate. You can tell you have reached saturation when the ADU of the brightest point slows or stops increasing with increasing exposure time.

Having arrived at a maximum number (mine was 5222 ADU) you can then plug it into the following formula:

Gain = 32767/Value

In my case that is:

Gain = 32767/5222 = 6.27

Now that you have the desired gain you can determine the value of R20:

R20 = (Gain - 1) x 1000

or in my case:

R20 = (6.27 - 1) x 1000 = 5270 Ohms

I tested this value in my camera using two different CCDs:

Side effects
 

No stored dark frames or flat fields can be used after the modification. The offset and gain of the data from the camera has been modified, and no longer bears any relationship at all to previous data.

If this modification will lower your amplifier gain, it will also increase your dynamic range. The visual difference between a saturated pixel and a dark one will be increased. This means that even though more significant data, and less noise, is making it into the image, the image will not appear to be as bright. The brightness will need to be increased when in find mode to compensate.

The SNR will increase. Maybe bunches, maybe only a little, but either way, it will increase to the maximum the design will allow.
 

Things you should know
 

Binning changes the output characteristics of the CCD. Assuming no saturation in any transport entity, the output will be 4 times as high at 2x binning than at 1x. Under saturation conditions however, the situation is different. A saturation situation in 1x mode with a value of 16000 ADU may result in 31000 ADU in 2x bin mode because saturation may occur in either the horizontal register or the FD, or both.

Because there are multiple cases of saturation, the choice of which gain to set on the amplifier becomes less clear. I chose to set the gain based on saturation at 2x binning, and as a result, I worked with around 16k ADU in 1x binning. As a practical matter, it seemed that I use 2x only when I´m trying to go deep on a dim object. I prefer to shoot at 1x, so I changed the gain to a value appropriate to that use.