View Single Post
  #322  
Old 12-01-2011, 08:32 PM
Penthode's Avatar
Penthode Penthode is offline
VideoKarma Member
 
Join Date: Jun 2009
Location: Kitchener/Waterloo Ontario Canada
Posts: 1,462
Quote:
Originally Posted by tubesrule View Post
Modern video encoders use purely digital techniques to generate the composite signal. Writing the algorithms for my converters to do both quadrature modulation for NTSC/PAL and FM modulation for SECAM was relatively trivial. You can also add FIR/IIR filters for limiting the subcarrier bandwidth to the correct spec.

Implementing a discreet analog approach is certainly possible, but unless the source for the component video is coming from a very old piece of equipment, it's being sourced as digital data anyway, so it's just easier and cleaner to do all the work in the digital domain.

The great thing with fpga's is they can be thought of as equivalent to a software based solution but for hardware. You are still physically wiring low level hardware blocks together, but using firmware to do so, so you can change and add to the design ad infinitum without having to actually change physical parts or wiring. They are also extremely cheap for applications like this, so not only are they very flexible, it would be difficult to do a discrete design for a lower size or cost.

Darryl
Hi Darryl,

Thanks for the reply. I won't deny that the FPGA is ideal for spatial/ temporal - rescaling/ interpolation. And certainly generating a composite signal may be straight forward. But what is your signal source?

The source of the signal I was considering would ideally come from a modern DVD player. The quality loss through analog to digital and digital to analog conversion is minimal if the video remains as component Y' Cb', Cr'. Analog component, serial digital (SMPTE259C) or HDMI, is okay. If the signal source is composite NTSC, the results will be compromised. Cascading composite codecs is lossy.

Why I suggested that the FPGA may be over the top is that to actually generate a CPA signal from analog component video should only take a dozen or so discrete transistors. If rescaling and interpolation was necessary, I would not hesitate to say FPGA. But with 525 line 59.94Hz video in and out, I do not think an FPGA is necessary.

My curiosity is now raised and I will ponder a hardware based solution.

Terry
Reply With Quote