Building a Standards Converter

[ppppenguin]

Don't forget the 405/567 switch in this part of the design. It's possible to include it in the main 405/567 switch by doing something like this.

if (V_COUNT = VXX1 or V_COUNT = VXX2) and ODD_FIELD then RD_EN <= '0'; else RD_EN <= AVID_LINE;

and in the if/then/else add the alternative values for VXX1 and VXX2

if SW_8A = ???? then
VXX1 <= 16;
VXX2 <= 17;
else
VXX1 <= 18;
VXX2 <= 19;
end if;

[`FRANK.C]

Jac
563,564, 565, 566..... are half lines. So the difference between syncing to 565 and 563 is that the picture will be moved up the screen by one line if syncing to 563. Which is probably difficult to notice.

Frank

[ppppenguin]

As Frank implies, when you're working with interlaced standards, the vertical counters are normally counting half lines. So on 405 there are 405 half lines per field. You can get some very strange effects if you inadvertantly count an even number of half lines.

This doesn't apply to Hedghog which only has line stores but if you're working with a framestore it all gets more complicated. You have various ways of organising the memory. You can store the lines in vertical order. That means that if you read the memory linearly the result won't be interlaced. So you have to increment the vertical address by 2 lines per line to get interlace. Or you can store lines in temporal order, efffectively 2 blocks of memeory, one for each field. Both methods have pros and cons.

[Jac Janssen]

Thanks Frank,

That explains it clearly. So I'll best go for 565.
Is it possible to compensate for the vertical shift when using the eq pulses?

Thanks Jeffrey for your addition.

Jac

[ppppenguin]

EQ pulses on systems designed without them are a bit tricky. You can tweak the vertical timings but the EQ pulses have to fit somewhere. This usually means moving the picture slightly vertically.

[`FRANK.C]

Systems either had EQ pulses or not and the television would be adjusted (height/vertical shift) accordingly.
It is only when having EQ pulses switchable on a standard that the vertical shift is noticeable, which would never have happened in reality.

Here in Ireland at the time when dual standards were in use the televisions were usually adjusted so they would over scan both horizontally and vertically. This help reduce the effects of fluctuating mains voltage. So any change in vertical position due to changing standards would not be noticed( System I having EQ pulses and System A not having them).

I can remember on frequent occasions even with over scanning the picture getting smaller than the screen. Here, back then, what would be considered a "good" picture is very different to today.

Frank

[ppppenguin]

EQ pulses were a controversial addition. Blumlein maintained that they were not needed. He said well designed TV will give good interlace without them. So 405 never had them. Others disagreed. EQ pulses certainly make sync separator design less critical.

[`FRANK.C]

I thought that it is interesting that there was a later version of the modulator chip MC44BS373CA the MC44CC373CA. The data sheet is here.
They are not pin comparable. The  MC44CC373CA appears to be even more obscure that the MC44BS373CA.
I have never heard of it before.

Frank

[ppppenguin]

The 2 chips are claimed to be functionally equivalent but not pin compatible. If the MC44CC373CA is not available then it's of no interest anyway.

[`FRANK.C]

I suspect the reason for the horizontal shift in the picture is due to the need to stop reading from the time redistribution memory for one line during the frame blanking period.
I cant think of the reason right now but the line chosen is critical or horizontal shift or other problems will occur.  

The reason is:

The interpolater delays the video which pushes the picture down the screen.
To counteract that, on the 405 side FRAME_RESET resets to 403 rather than 0. Which has the effect of pulling the picture back up the screen.

The TR (time redistribution) memory needs to be cleared at the start of each frame.
Clearing the TR memory with FRAME_RESET  would mean losing the last line or two of the field. So another point needs to be found during field blanking period to clear the TR memory.

At the time the TR memory is cleared it must not be either written to or read from or there will be an incomplete line stored in the TR memory. Which will have the effect of shifting the picture horizontally. For that reason the start of 625 line 12 was chosen as the TR memory clear point.

On the 405 side after the TR memory is cleared the memory is not read for one line to allow the memory to fill a little.
As the 567 lines are shorter in duration than the 405 lines more of them are required to get passed the point the TR memory is cleared.
resetting to 563 with FRAME_RESET as opposed to 565 is one way of achieving this.

Frank