On 12 Jan., 03:28, "Michael J. Mahon" <mjma...@aol.com> wrote:
Sync emulation is part of the emulation--it would work fine if
there were any DSP engineers today who understood how TV sets
used to work. ;-)
BTW, Grundig used analog sync and color decoders before a digital
double-scan is performed...
This yields a picture quality I've never seen on a double-scan TV.
I don't think an analog sync decoder is very expensive.
It's easy in one chip, but that's an *extra* chip for a digital TV.
The DSP code is pretty straightforward, but it's extra code. It's just
cascaded RC filters and thresholding.
I just found out: You're not done at this point.
You need to convert it to an image of fixed size. That means the DSP
SHOULD
- sample the signal at a rate to get the right width (perhaps use a
PLL to get the clock from the Hsync, this could be integrated into the
HPLL)
- store it line by line, start a new field after either Vertical sync
or after a certain number of lines has passed (350 or so).
- check the position of the Vsync to find out whether this is an even
or odd field.
- if there is interlacing, deinterlace and scale by the factor needed
to fill the screen
- if not, scale every field to fill the screen.
This is not done too.
The algorithms I saw did something like the following:
- sample at an arbitrarily chosen frequency.
- scale it to the minimum needed to achieve the desired video
bandwith
- store it in memory with the expected number of lines. Go crazy if
it is wrong.
- Always interlace the fields. If there is no interlacing, interlace
anyway.
- Blur - eh denoise the picture
- Apply nonlinear scaling to allow certain special display modes.
- Scale it to the size needed.
This is from the data sheet of a Micronas video processing chipset I
wanted to use to get high-quality video display on a VGA monitor. I've
given up...