405 to 625 conversion

[peter scott]

That is remarkably good! When you reduce the brightness to get the camera to use a slow enough shutter speed you need to ensure that the contrast setting does reproduce blacks at black level.

Peter (the critic)

[ppppenguin]

It's not just the line number conversion; that only affects vertical resolution which isn't very obvious on most pictures. It's also passing through two ADCs and two DACs, none of which are really high grade. Though they are not bad (to be fair, the Aurora and Hedghog use very similar ADCs and DACs)

Most test signals aren't going to show very much if vertical resolution is degraded. This is probably because in traditional TV systems, there was very little apart from the camera and display that affected vertical resolution. A zone plate is probably the most sensitive test signal because it represents all vertical and horizontal frequencies. A vertical frequency sweep would also work. Neither would show phase or time domain effects very well. I'm trying to envisage the vertical equivalent of pulse and bar. An ordinary crosshatch signal might well have visible effects. Each horizontal line is just a single line on each field.

That reminds me of trying a 625>405 converter with a crosshatch signal where the horizontal lines were out of order on the 2 fields. I can't remember which field is meant to be above the other but this signal was the wrong way round. The result was truly horrible after conversion.

[`FRANK.C]

I have noticed that when a cross hatch is converted from 625 to 405 some of the horizontal lines can flicker.  This will depend on where the 405 line is positioned between the 625 lines.

The test card in the Hedghog is a field that is repeated twice to make up the frame. Each horizontal line is made from one horizontal in each field.
The test card in the Hedghog II is similar except the center horizontal line is made up from lines taken from opposite fields than all the other horizontal lines in the test card. It is difficult to explain the illustration below might help.
I believe that the center line is done that way in the PM5544 test card.
I read somewhere that it is supposed to help show up interlace problems.

Frank

[ppppenguin]

This is part of the usual problem with interlace. Spatially adjacent lines are a whole field apart in time. It's very difficult to disentagle this when manipulating pictures. Converters such as the Aurora and Hedghog don't even try. They just treat each field separately. The Pineapple converter had switchable modes. There was one mode that interpolated spatially adjacent lines. Lovely on still images but horrible as soon as anything moved. You've probably seen something similar on some flat screen TVs where movement looks horrible.

A standard crosshatch pattern with a line of white on each field is actually vertically aliased, witha sudden transition from back to white. This can also be true of computer originated graphics. There is no perfect way to change the number of lines in the presence of aliasing. Either you have to blur the picture vertically or accept overshoot and ringing.

[Mike Watterson]

If you can detect it was a progressive source such as film, then de-interlacing is relatively easy. As Jeffrey knows most professional gear and better TVs will do some sort of de-interlacing if frame rate or resolution changes are needed.
NTSC to PAL used to be ghastly.
24 fps film to NTSC used to be horrible hence 480p as a USA digital mode and USA component video DVD players with progressive output as DVDs from actual cinema are stored either 24 fps 480 line or "24 fps for playing at 25 fps" 576 line (with usually sound pitch changed).

Any DVD or BD player with HDMI will do progressive if the disc is. I bought a progressive component out DVD player years ago for NTSC and it outputs 1440 x 480p or 1440 x 576p for suitable DVDs. The newer 4K TV has no Component progressive input, but the older 1920 x 1080 TV has VGA and RCA Component.
I got a cheap component progressive to HDMI box, but it fails if the DVD player disc source is interlaced. The TV Component input works on progressive or interlace sources from at least 480i to 1080p

[ppppenguin]

There are two things that have made standards conversion harder, especially where interlace is involved. When th original framestore converters were designed in the 1960s and 70s (BBC analogue, DICE and ACE) the main sources of images were cameras and telecine. Both had limited vertical resolution, you didn't get really sharp vertical detail of the sort you get from digitally originated graphics. Cameras (though not TK) also had limited temporal response. In other words they smeared motion somewhat. In film camera terms that's equivalent to a shutter angle approaching 360 degrees.

As cameras improved, notably with solid state sensors replacing tubes, you could work at the equivalent of narrow shutter angles. This sharpens up movement nicely but makes life harder for a standards converter. It's harder to get away with assumptions that both vertical and temporal frequency response will be smoothly rolled off. Motion is nearly always aliased anyway, wagon wheels going backwards is a common artefact. Make that sharp rather than a bit blurred (narrow shutter angle) and the conversion compromises don't really work properly.

This is all going well beyond what we need for simple line rate conversion.

[ppppenguin]

Update after a long break. Jon Collis from the Kaleidoscope project came over today with a Sony CV2100 VTR and some 405 tapes. The waveforms were grim and replay wasn't entirely reliable on a good 405 monitor. The latest iteration of my 405 to 625 converter worked as well as the monitor. Better in fact, since the output syncs are 100% guaranteed steady even if the picture breaks up. The changes I made to make the field sync separator more rugged really work in practice.

Should work with 405 on any VTR format with the possible exception of CV2000. People recorded 405 on all sorts of things from Shibaden 700 to VHS.

The CV2000 was a 405-only VTR that used a crude skip-field recording system. I have no idea how that would be tolerated but only one way to find out.

PS: A DICE will be on display at the BECG open days on 9/10 September. I don't think it will be working but it's an impressive beast. The framestore used over 4000 chips on nearly 100 cards.

[ppppenguin]

I should add that the code I wrote to ruggedise field sync separation for my converter is equally applicable to Frank's Hedghog. Frank, just ask if you want the VHDL.

[`FRANK.C]

Well done Jeffrey!
That's really great news.
To work as good as a good monitor is about the best that can be ever hoped for.
From what little I've seen of them, even at the best of times off tape syncs are pretty rubbish.

Thanks for your kind offer of your code. Yes please I would love to have it.
I haven't been in the workshop in ages and probably wont be in the near future.
But it would be great to have it to hand when I eventually get back to it.

Frank

[Mike Watterson]

Before Digital Timebase Correction and similar was a thing, I remember people in BBC Belfast feeding a VTR through a Sound-in-syncs coder and then decoder back to back. maybe 1976-1977