BobDiaz
01-01-2006, 08:04 PM
With any video camera, Tube, CCD or CMOS, as the light is reduced, the gain in the video amplifiers can be increased to increase the signal level. Unfortunately, not only does this amplify the signal, BUT it also amplifies the noise in the image sensor. I have a prediction about the amount of noise for 1080@60i compared to DVCPRO50 @60i. IF all the camera settings are the same between the two modes of recording, the DVCPRO50 should show less noise in the image compared to 1080.
For those who hate reading long posts, stop here and wait for someone who has (or will have) the camera to try it out and see if I was right. For those who want to know why I believe this, read on...
WHY THE DIFFERENCE:
Let me start with something simple, you roll one dice, the odds of the number 6 showing up is 1 chance out of 6. If you roll and record the results, the average should come out to 3.5; you'll have to roll the dice at least 10,000 times, but the average should be close to that number.
Now if we have two dice and roll, take the number rolled, and divide it by 2, the results could be anywhere from 1 to 6, BUT only 3 combinations are close to 6: (6+6)/2, (5+6)/2, (6+5)/2 The number of combinations is 36 possible. Of the combinations, only 3 out of 36 (or 1 out of 12 times) is 6 or close to 6 (5.5).
As we add more dice, then roll, add then up, and divide by the number of dice, the chances of reaching the extreme number, like 6 decreases, BUT the chances of reaching the mid point (3.5) increases. You may ask, "What has this to do with CCD Noise?"
Simple, the random noise in a CCD is like the roll of a dice for each element in the CCD. (It's a little more complex than that, but let me keep this simple.) When shooting in the 1080 resolution at +18dB Gain, it's like taking ONE dice and adding it to the video signal. When shooting in the DVCPRO50 mode with +18dB Gain, it's like taking FOUR dice (each with the video signal) adding them together and dividing by four. The video signal level should be about the same, but the noise is reduced by 4. There are other factors that get in the way, but overall, the noise level is reduced.
For those of you who get the camera soon, could you please try the test and post the results of the test. As a side note, when you lower the detail, you allow the camera to take some information from neighboring pixels. This softens the image a little, but it also allows you to reduce noise. The idea is to find the right balance between best detail (sharpness) and least noise. Thankfully the camera allows us to set it to our liking.
Bob Diaz
For those who hate reading long posts, stop here and wait for someone who has (or will have) the camera to try it out and see if I was right. For those who want to know why I believe this, read on...
WHY THE DIFFERENCE:
Let me start with something simple, you roll one dice, the odds of the number 6 showing up is 1 chance out of 6. If you roll and record the results, the average should come out to 3.5; you'll have to roll the dice at least 10,000 times, but the average should be close to that number.
Now if we have two dice and roll, take the number rolled, and divide it by 2, the results could be anywhere from 1 to 6, BUT only 3 combinations are close to 6: (6+6)/2, (5+6)/2, (6+5)/2 The number of combinations is 36 possible. Of the combinations, only 3 out of 36 (or 1 out of 12 times) is 6 or close to 6 (5.5).
As we add more dice, then roll, add then up, and divide by the number of dice, the chances of reaching the extreme number, like 6 decreases, BUT the chances of reaching the mid point (3.5) increases. You may ask, "What has this to do with CCD Noise?"
Simple, the random noise in a CCD is like the roll of a dice for each element in the CCD. (It's a little more complex than that, but let me keep this simple.) When shooting in the 1080 resolution at +18dB Gain, it's like taking ONE dice and adding it to the video signal. When shooting in the DVCPRO50 mode with +18dB Gain, it's like taking FOUR dice (each with the video signal) adding them together and dividing by four. The video signal level should be about the same, but the noise is reduced by 4. There are other factors that get in the way, but overall, the noise level is reduced.
For those of you who get the camera soon, could you please try the test and post the results of the test. As a side note, when you lower the detail, you allow the camera to take some information from neighboring pixels. This softens the image a little, but it also allows you to reduce noise. The idea is to find the right balance between best detail (sharpness) and least noise. Thankfully the camera allows us to set it to our liking.
Bob Diaz