Professional Interlaced PAL to NTSC conversions

[kakomu]

I'm curious how professional authoring companies make the conversion from interlaced PAL video (50i) to interlaced NTSC video (60i). The best example I can think of off the top of my head is the IT Crowd which was recorded at 576i (considering its origin is the UK), but the NTSC DVD has a 480i transfer. The NTSC DVD doesn't appear to have any ghosting, or repeat frames from what I can tell, so somehow they made a fairly clean conversion from 50i to 60i.

I've seen a few threads discussing this conversion which seems to frequently use motion compensation filters, however, I'm curious if that's how professional authoring companies make the conversion or if there's another method that they use. If the method is different, is there any way of simulating this method in Avisynth? It seems that the method used for the IT crowd is a bit more sophisticated than avisynth's "changeFPS"

Moreover, what hardware do broadcast companies use to make on-the-fly conversions from PAL formats for broadcast in NTSC territories?

Thanks in advance.

[Ghitulescu]

Most professionals use the Snell&Wilcox converter.

[pbristow]

First, this isn't actaully an Avisynth Usage question. =)

Second, there's no single answer to this. Companies use various techniques depending on how much time and effort they want to (or are being paid to) spend.

A lot of TV shows these days, although both shot and broadcast as 576i (or 480i), are actually post-produced as 25p (or 24p) by de-interlacing the source, to give it "that film look" [SHUDDER]. I'm pretty sure the IT crowd is one of those. But if you have access to the 576i source tapes, EDLs etc, it would be easy to "re-post-produce" the show as either 24p or 30p, with no repeated frames or ghosting: It's just that sometimes the step from one frame to the next would be bigger or smaller than the "normal" size. Given that such shows have to made with the American market in mind as well as the UK/European one (to maximise sales), the original production company probably made the 30p version straight after, or at the same time as, the 25p one. Indeed, that may be one of the reasons for post-producing in progressive frames, to achieve a consistent look between versions but at the appropraite frame rate for the market in question.

So that's *one* answer to your question. There are plenty of others. :)

[kolak]

Most professionals use the Snell&Wilcox converter.

Which is 70K£ and also not that flawless as some may think.

[Ghitulescu]

First, this isn't actually an Avisynth Usage question. =)

Second, there's no single answer to this. Companies use various techniques depending on how much time and effort they want to (or are being paid to) spend.

A lot of TV shows these days, although both shot and broadcast as 576i (or 480i), are actually post-produced as 25p (or 24p) by de-interlacing the source, to give it "that film look" [SHUDDER]. I'm pretty sure the IT crowd is one of those. But if you have access to the 576i source tapes, EDLs etc, it would be easy to "re-post-produce" the show as either 24p or 30p, with no repeated frames or ghosting: It's just that sometimes the step from one frame to the next would be bigger or smaller than the "normal" size. Given that such shows have to made with the American market in mind as well as the UK/European one (to maximise sales), the original production company probably made the 30p version straight after, or at the same time as, the 25p one. Indeed, that may be one of the reasons for post-producing in progressive frames, to achieve a consistent look between versions but at the appropraite frame rate for the market in question.

Which is 70K£ and also not that flawless as some may think.

It doesn't matter at all. Most software solutions are generally slower than their hardware counterpart, and not necessarily better. What is also missing from most SW solution is the ability to "feed"/"stream" the audio-video in real-time.

This is generally a question of DVD-editing or DVB-broadcasting.

For DVB-broadcasting, they simply feed the programme into a S&W and that is it, they watch the result and this should be satisfactory. Nobody has the time to tinker various parameters and do several encodes in a row to pick up the best, no, they paid a load of money for one thing and they expect this thing to earn this money.

For DVD-editing they may use SW or SW/HW solutions. Depending on the money involved, the time is also constrained (time = money). Premium editions, anniversary ones, etc., may have been better treated, but this is not always the best one can get, because nobody in the field of commercial video processing has the time an amateur has (for free).

[2Bdecided]

I've seen a few threads discussing this conversion which seems to frequently use motion compensation filtersNo, for 24p/25p stuff, they just speed up/slow down, or repeat a strange number of fields. No blends. Wrapping 24p or 25p in 50i or 60i streams is trivial. Recovering 24p or 25p from 50i or 60i is trivial or easy or at least possible (depending on field repetition pattern).

You only get blending and ghosting when you use a dumb conversion that does a mediocre 50i<>60i conversion (i.e. intended for real interlaced sources, like news footage) and apply it to 25p-in-50i<>25p-in-60i or 24p-in-50i<>24p-in-60i. Then you get a field blended mess that can be tricky to unpick.

If it's shot at 24p or 25p, it's very rare for someone to create a 30p version. It's possible (especially from a 50i or 60i original before the filmic effect has been added), but rare.

Cheers,
David.

[kakomu]

A lot of TV shows these days, although both shot and broadcast as 576i (or 480i), are actually post-produced as 25p (or 24p) by de-interlacing the source, to give it "that film look" [SHUDDER]. I'm pretty sure the IT crowd is one of those.

The NTSC DVDs of the IT Crowd are 60i, as in 60 discrete fields (as opposed to 24p source interlaced into 29.97fps). When I rented the NTSC DVDs of the IT crowd, I remember bobbing the video and seeing that the fields didn't contain any conversion artifacts (ghosting or repeat frames). At the time, I just remember that it seemed pretty impressive that they (probably) recorded the series in a PAL standard and that the NTSC conversion looked very clean.

No, for 24p/25p stuff, they just speed up/slow down, or repeat a strange number of fields. No blends. Wrapping 24p or 25p in 50i or 60i streams is trivial. Recovering 24p or 25p from 50i or 60i is trivial or easy or at least possible (depending on field repetition pattern).

You only get blending and ghosting when you use a dumb conversion that does a mediocre 50i<>60i conversion (i.e. intended for real interlaced sources, like news footage) and apply it to 25p-in-50i<>25p-in-60i or 24p-in-50i<>24p-in-60i. Then you get a field blended mess that can be tricky to unpick.

If it's shot at 24p or 25p, it's very rare for someone to create a 30p version. It's possible (especially from a 50i or 60i original before the filmic effect has been added), but rare.

Cheers,
David.

I'm referring to interlaced sources, ie 50/60 fields per second. Not progressive video with 3:2 pulldown.

So, in other words, I'm curious how the conversion from 50fps to 60fps is made (and then packed into the appropriate interlaced container).

Most professionals use the Snell&Wilcox converter.
Thanks for the info.