Do we need to switch to HD?

[Eric Olson]

Originally Posted by Geoffrey Cox

DVDs play interlaced material perfectly well if you downconvert correctly and I've never had any problems with this.

People who have trouble downconverting interlaced material often are mixing DSLR footage with HDV and DV all in the same project or using footage shot with a 35mm adapter that turned the image upside down.

[David Heath]

Originally Posted by Chris Soucy

Er, David, sorry, but I have to take exception to this.

The "squat or bust" technology upgrade scenario has been so discredited that I'm amazed anyone would suggest it in this day and age............

Integrating one different camera into an existing workflow may not be everyones cup of tea, but it's better than falling flat on your face by trying the "squat or bust" routine and having it go spectacularly down the toilet.

OK, but how would you suggest the original poster mixes footage from the two cameras? Ignore HD, codecs, tapeless etc for now, and tell us how you'd suggest they went about mixing 4:3 footage from the old cameras with 16:9 from the new?

It's one thing to produce material for client that's 4:3 throughout, but if it changes aspect ratio on shot changes.....? It may be acceptable when archive is cut into a new programme - but surely not just on camera cuts?

Maybe you could switch the PD150s to widescreen mode? No - they are poor in that mode, and you'll get a noticeable jump in quality between the old and the new.

Alternatively, keep the PD150s in 4:3 mode, and extract the centre 4:3 cutout from the new cameras? Technically, that should match - but the end product is then still SD 4:3. So then what's the point of having bought any new cameras in the first place!?!

Or is there a fourth possibility I haven't thought of?

And what is "squat or bust" about getting (say) 3 EX1s anyway? You'll need to (possibly) upgrade the editing setup if you want to take full advantage of HD, but if all were used widescreen SD (at least for now), that would give the widescreen (and tapeless) advantages without any neccessary upgrade to the NLE. It would just mean importing DV format files versus capturing off DV tape.

Yes, a little getting used to - but won't that be the case with a single new camera anyway? Surely the issues importing from 3 tapeless cameras are the same as from one?

[David Heath]

Originally Posted by Geoffrey Cox

But David, isn't DVD a natively interlaced medium, so there's no need to de-interlace - downconverted interlaced material plays perfectly on DVD in my experience?

and

All I can say is that I've shot interlaced material (50i), edited in FCP in Prores then sent it to Compressor to high quality downconversion to m2v files for use in DVDSP to make the DVD. No interlacing enacted ....

Ah, but the software DOES - must - deinterlace - it's just that you're unaware of it.

Eric is completely right when he says "one must first deinterlace the HD material to 1080p60, then rescale the progressive frames to 480p60 and finally weave the progressive frames back to interlaced 480i30"

Think about it. A given field consists (for 1080i/30) of 540 lines, the next field the other interlaced 540 lines. You want to produce two corresponding SD fields of 240 lines each. You can't simply do it field by field, taking each 540 lines to form the corresponding 240, because of the interlace. Hence it's as Eric says. Their needs to be an intermediate deinterlacing to 1080p/60, rescaling to 480p/60, then form a conventional 480i/30 signal.

It's all transparent to the user though - just hit "downscale 1080i/30 to 480i/30". Behind the scenes it's far more satisfactory to start with 720p/60, do a simple rescale to 480p/60 (simple frame by frame rescaling) then interlace to 480i/30.

[R Geoff Baker]

Originally Posted by David Heath

Think about it. A given field consists (for 1080i/30) of 540 lines, the next field the other interlaced 540 lines. You want to produce two corresponding SD fields of 240 lines each. You can't simply do it field by field, taking each 540 lines to form the corresponding 240, because of the interlace. Hence it's as Eric says. Their needs to be an intermediate deinterlacing to 1080p/60, rescaling to 480p/60, then form a conventional 480i/30 signal.

I'm a little lost in your flow -- why can't the software simply do it field by field? In fact I believe that is what it does -- I don't see the requirement to follow the path you suggest? You take Field A, downsize it, take Field B, downsize it ... you end up with a downsized 1080i>480i file ..?

Cheers,
GB

[David Heath]

Originally Posted by R Geoff Baker

I'm a little lost in your flow -- why can't the software simply do it field by field?

It depends how good you want it to be......!

It may be possible to achieve a result as you say - but initial de-interlace to 1080p/60 will produce a better result. I'd guess less likely to give line flickering, for one thing.

Likewise don't forget the chroma information. In the range of cameras we are talking about here that's likely to be 4:2:0, so 270 lines of chroma information to downscale - 4:2:2 may not make that much improvement with progressive material, interlace is where 4:2:2 most shows benefit.

The real point is that starting with 60p makes the job easy. You've got an individual 1280x720 true progressive frame to correspond to every NTSC field.

[R Geoff Baker]

I take your point -- if you start with a progressive image, it is 'easier' to create an interlaced one than vice versa. But if your source is interlaced, you still have the problem of deinterlacing, and all the bob and weaving choices that have to be made -- good software will do the job without creating an interim 'faux' progressive frame, in my experience. And that's all I intended -- there is no reason why the faux progressive frame is necessary.

But if you've tested, and your software does better work with this interim step, have at it!

Cheers,
GB

[David Heath]

Originally Posted by R Geoff Baker

..........if your source is interlaced, you still have the problem of deinterlacing, and all the bob and weaving choices that have to be made -- good software will do the job without creating an interim 'faux' progressive frame, in my experience. And that's all I intended -- there is no reason why the faux progressive frame is necessary.

Ah sorry! No, I wasn't inferring that there ever should be a 1080p/60 de-interlaced file as such, deliberately made. Rather that it's a step (maybe a virtual one) that good downconversion software should go through on the way to the final result. That's what I really meant by "Ah, but the software DOES - must - deinterlace - it's just that you're unaware of it."

To the user you just specify "downconvert", and a NTSC file pops out the end. But under the bonnet the engine is doing a deinterlace, rescale, and reinterlace, even if the maths and algorithms are all rolled into one and the user's not aware of it.

[Eric Olson]

Originally Posted by R Geoff Baker

I'm a little lost in your flow -- why can't the software simply do it field by field? In fact I believe that is what it does -- I don't see the requirement to follow the path you suggest? You take Field A, downsize it, take Field B, downsize it ... you end up with a downsized 1080i>480i file ..?

It is not quite this simple. The top 540 line field in a 1080i video stream is shifted in space by 1 scan line from the bottom 540 line field. One scan line in a 1080i video stream is about 0.09 percent of the field of view. If you rescale each 540 field to 240 lines, the two fields will still be offset from each other by 0.09 percent of the field of view. However, the top and bottom fields of a 480i video stream should be offset from each other 0.20 percent of the field of view. This error could result in noticeable aliasing and loss of resolution.

[R Geoff Baker]

Well if your software doesn't know what a line off set is, you've got troubles. For the record, neither FCP nor Premiere have given me problems, but YMMV

[Eric Olson]

Originally Posted by R Geoff Baker

Well if your software doesn't know what a line off set is, you've got troubles. For the record, neither FCP nor Premiere have given me problems, but YMMV

I expect you are right on that one. It would be interesting to compare the results of different software and settings when converting the same 1080i30 source to 480i30.

Getting back to the original post, producing a widescreen DVD doesn't immediately require an investment in new computers and software. If you shoot HD, the footage can be down sampled to widescreen SD prior to editing.

[David Heath]

Originally Posted by R Geoff Baker

For the record, neither FCP nor Premiere have given me problems, but YMMV

No, but do they deinterlace before downconversion, or just do a simple field by field translation? The user has no way of knowing (and neither do I) but I suspect they deinterlace, even if the user is unaware.

It's worth thinking of NTSC-PAL standards convertors as a comparison. I remember seeing converted US TV shows in the 70s and technically there was no doubt that strange things had gone on, both relating to jagged edges on straight lines, loss of definition and jerkiness of motion when converted to PAL. Come the digital world and it became a different and much better story - the new techniques allowed much more complicated and cleverer algorithms to be used.

In this case, downconverting 540 lines to 240 COULD be done as simply as dropping 13 lines in every 27. It wouldn't look very good! Amongst other problems you'd get a step on diagonal lines every 12 th line. At the very least, prediction needs to be done such that every output line is formed from a number of input lines to smooth things out. If that was (say) 5 lines, that could lead to a noticeable softening, but if the 5 was 5 out of a deinterlaced 1080 there should be a marked improvement.

It's not that you can't convert interlace to interlace on a field by field basis, it's that there are better (and far more complicated) ways to do it. And starting from a 60p original takes a lot of the pain away.