[opendtv] Re: Sampling Frequency of Image Signals

  • From: Mark Schubin <tvmark@xxxxxxxxxxxxx>
  • To: opendtv@xxxxxxxxxxxxx
  • Date: Tue, 29 May 2007 21:08:20 -0400

dan.grimes@xxxxxxxx wrote:

I have a couple of questions regarding the digitization of the signals coming from the CCD.

I realize that many cameras out there today process the signal coming from a CCD in analog and only after completely processing it in analog is the signal sampled at some frequency to put it on an SDI interface or record it to some form of digital media. There are some cameras that go to the digital domain earlier and do much more digital processing. I have several different cameras that do some processing in the digital domain but at various stages of the path.

Are there any cameras that are sampling each pixel as they come out of the CCD, and then processing in a true native matrix in the digital domain?

All CCDs emit analog signals.

Or is this not recommended, possible, or perhaps practical?

The nature of a CCD precludes quantization on an individual pixel-sensor basis.

I would think that if the digitization happened before the compression circuit, it would require a large number of bits per sample (or the use of a floating point number) to keep the dynamic range of the CCD, so perhaps it is not practical at this time.

There is a trade-off between A-to-D bit depth and analog-processor quality.

I think what I am asking is when is earliest practical point in the processing system of a camera that the analog signal can be digitized in today's high quality HD video cameras?

After the analog preamp.

I know the designation such as 4:4:4 means different things depending on the video system. For instance, in regards to digitizing an analog signal it would mean sampling at four times the subcarrier frequency.

Not quite. That's what was called 4fsc. But you are correct that that's where the "4" came from. In (then) CCIR Rec. 601, the sampling rate was (and is) 13.5 MHz, not far from 4 fsc's 14.3 MHz for NTSC.

In the digital domain I assumed it referenced the spatial frequency, meaning that it was full resolution and 1920 pixels per line were sampled from 1920 imager or 720 samples were taken from a 720 pixel imager).

Very close. In SDTV, 4:4:4 means sampling each color at 13.5 MHz (4:2:2 means sampling luma at 13.5 MHz and color-difference signals at 6.75 MHz). In HDTV, it means sampling each color at 74.25 MHz.

But considering that most cameras process an analog signal, perhaps that isn't true. And the 4:2:2 has been used with so many different formats, I'm a little confused to its true meaning in today's digital formats.

See above.

FYI, 4:1:1 means full-bandwidth luma sampling and quarter-bandwidth color difference. 4:2:0 means full-bandwidth luma, half-bandwidth color difference, and half the vertical resolution in color difference, too.

TTFN,
Mark



----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:

- Using the UNSUBSCRIBE command in your user configuration settings at FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word 
unsubscribe in the subject line.

Other related posts: