Video Sampling Rate question

[0x34]

I'm designing a circuit which samples (ADC) a composite video source, turns it into a bit stream, then returns it (DAC) to composite video. The Video source is 420 lines NTSC color. Can I successfully sample this at 10.8 Msps? Or, is that too slow to correctly sample the color? Is there an industry standard minimum sample rate to correctly sample video?

At 10.8MHz, I'll only get about 3 samples per cycle at 3.58MHz. Is that enough to correctly acquire color phase? Or, is my math wrong?

Note,
I am not using any JPEG type scheme for this circuit. I'm simply slicing the signal into many small slivers, digitizing each slice into 10 bits of data, transmitting over a twisted pair, the reconverting to composite using a 10-bit DAC.

 

[crutschow]

Here's an article on NTSC signal digitizing where they use a sample rate of 12-15MSps.

 

[Ubergeek63]

I'm designing a circuit which samples (ADC) a composite video source, turns it into a bit stream, then returns it (DAC) to composite video. The Video source is 420 lines NTSC color. Can I successfully sample this at 10.8 Msps? Or, is that too slow to correctly sample the color? Is there an industry standard minimum sample rate to correctly sample video?

At 10.8MHz, I'll only get about 3 samples per cycle at 3.58MHz. Is that enough to correctly acquire color phase? Or, is my math wrong?

Note,
I am not using any JPEG type scheme for this circuit. I'm simply slicing the signal into many small slivers, digitizing each slice into 10 bits of data, transmitting over a twisted pair, the reconverting to composite using a 10-bit DAC.

much to slow.

get a chip meant to do it.

over twisted pair? are you kidding? Video and Imaging - Video Encoders / Decoders - TVP5151 - TI.com 13.5MHz pixel rate 2 9 bit words per pixel for a data rate of 243MBPs

 

[PCBWING]

Color TV Digital parameters

 

[Ubergeek63]

Color TV Digital parameters

Seems to me that borders on CGA that is 320x240 and that is only 360...

Rec. 601 - Wikipedia, the free encyclopedia

 

[0x34]

Yes over a twisted pair. Again, I'm not using any type of format. For the sake of argument, lets just say that I need to sample a 4MHz signal (dynamic 1Vpp level), send it via a twisted pair, then reassemble the source signal at the other end. The bit stream of this project will operate at 750Mbps.

Now, the basic rule of thumb is that you need to sample at twice the source frequency. Since the highest frequency in composite video is 3.58 MHz, I figure 10.8 MSPS is adequate. However I'm concerned that the reconstructed signal might appear frequency shifted after the DAC due to the small sampling rate (3.017 samples per each 3.58 MHz cycle). I know this will work fine for the luminance and sync portions of the signal. I'm just concerned about the chroma information.

I am familiar with the TVP5151, which formats the sampled signal into 8-bit 4:2:2 YCbCr. However I'm not doing that. I just wanted to know if anyone's had any experience in sampling a signal at this frequency and get a consensus as to phase distortion due to low sampling rates.

So, again, my question is, at 10.8MHz, I'll only get about 3 samples per cycle at 3.58MHz. Is that enough to correctly acquire color phase? Or will it construct even enough information for the keying circuit to lock?

 

[crutschow]

This reference says a sample rate of 3 to 4 times the color subcarrier frequency is adequate. The complete article requires a purchase, but that may be worth the cost to understand the theory..

 

[mneary]

Sampling of NTSC must be exactly locked to the subcarrier, and the lower the sampling rate the more critical this becomes. Any drift will make rainbows. Sampling at 3.017x the chroma frequency will only give you psychedelics.

I think you'll find it cheaper to demodulate the chroma and encode it YCbCr.