Monday, January 17, 2011

Why is NTSC color carrier frequency 3.57954545 MHz and not some other number that can be remembered easily?

Hello guys. I am back after a long break. Now I will continue to tell a little more about NTSC before I start writing blogs on the actual Game Console implementation on LPC1768. I would suggest you to read the previous post before reading this one for better understanding.

I mentioned that the NTSC color carrier frequency is 3.57954545 MHz. If I want to create a game console for NTSC TV then I need to generate this frequency and use it as the color carrier. But why is 3.57954545 MHz chosen as the carrier and not some other frequency which can be easily remembered or generated?


Before telling the answer, I will discuss briefly about the spectrum of the video signal. Video signal has a total of 6 MHz bandwidth. 1.25 MHz is used by the vestigial side band (It is a small portion of the other side band). Of the remaining 4.75 MHz, around 0.6 MHz is allocated for the audio in the form of FM. So the reamaining 0 to 4.2 MHz band is left for video signals. This spectrum was used from the days of black and white transmission. When color transmission started, the NTSC thought of some amazing methods to incorporate color in the same bandwidth so that the black and white TVs can reject the color signal. The chroma information required around 0.6 MHz of bandwidth. So the color carrier frequency should be around (4.2 MHz - 0.6 MHz) = 3.6 MHz.




We just cannot use 3.6 MHz blindly. There are some other things that we need to consider. Since crominance signal, luminance signal and audio signal are all very near to each other there will be interference between them. The main concern here is the interference between the chrominance and luminance. Due to this interference the luma signal is sometimes considered as chroma signal and chroma signal as luma by the TV. This is very much visible to the naked eye. But by doing some artifacts it can made invisible!!

If the interference in one horizontal line is complimentary to that of that in the next line, the total effect will make the interference invisible. Also if the interference in the current frame is complimentary to that in the next frame, the total effect will not be visible. To make this possible, the carrier phase must change by 180 degree for each line and also for each frame. Each scanline length in NTSC is 63.5 uS. So for carrier phase to change by 180 degree for the next line, this line must have "odd multiple of one half the scan rate"s of cycles of the carrier. That is if this line contains one cycle of carrier, then it will have completed 360 degrees and starts from 0 degree in the next line. But if this line contains one and a half cycle of carrier, then it will have completed (360+180) = 540 degrees, which is equivalent to 180 degrees and start from an angle 180 degree in the next line.

Scan Frequency = 1 / 63.5e-6 = 15748.0315 Hz

Half of scan frequency = 7874.01575 Hz

Number of carrier cycles in 7874.01575 Hz = 3.6e6 / 7874.01575
                                          = 457.19999


We can choose 457, then we get the carrier frequency as 3.598425 MHz. But some experiments concluded that we get better results if we choose 455. But still for 455 we do not get 3.57954545 MHz as carrier frequency. There is one more thing to be considered. While implementing color NTSC frame rate was dropped to 29.97 frames per second instead of 30. So now there are (29.97 x 525) = 15734.25 lines per second. This is nothing but the line frequency. Now (15734 x 455 / 2) = 3.57954545 MHz.

Since there are 525 lines per frame, there will be (525 x 455 / 2) = 119437.5 cycles of the color carrier. The 0.5 in the number of cycles indicates that the carrier at the start of next frame will be differing in phase from that of the one in the current frame by 180 degrees. So now we have made both interline and interframe interference invisible!!

I mentioned in the previous post that in the Game Console I will be using fake progressive scanning. That is I will not be using interlaced scanning. So I get only half of the 525 scanlines = 262.5 scanlines. To utilize the advantage of the 3.57954545 MHz carrier (i.e use the interframe artifact of NTSC i.e to get a phase difference of 180 degree for each frame) each frame should have odd number of lines. So I will be using 263 lines per frame and will be leaving 261 interlaced lines.

So now we know the reason for choosing 3.57954545 MHz as color carrier is to 
  1. Adjust the color information in the NTSC black and white spectrum to maintain compatibility.
  2. Minimize the visibility of the interference effects.
For more information about this and also about PAL and SECAM visit http://www.videointerchange.com/pal_secam_conversions.htm. 

Special thanks to http://www.videointerchange.com/ for providing very good information about the television video standards.
 

1 comment:

Kent Lopez said...

Why is the 'I' portion of the chrominance signal sent with non-symmetric side-bands of 0.5 MHz upper and 1.5 MHz lower, whereas the 'Q' portion is sent with only 0.5 MHz side-bands? Someone from the NSC bureau told me that this was a result of time-dispersion which can lead to color bleeding from red and compensates for phase-seepage between the delta and gamma functions of the inter-symbol (analog) beta function. Perhaps you can clarify.