Couple of quick Telecine questions

[russta]

Hi,
I'm going to get a few of my 64t reels telecined soon, onto miniDV i guess (would prefer it as an image sequence tho).

One question I had was is it possible to not have the 3:2 pulldown speed correction done, or is it possible to just get frame by frame onto DV? (i don't mind if it runs the 18fps footage at 25fps, i can adjust the speed myself afterwards)

Also will the footage be interlaced on the DV tape?

Finally, i know of this website that has some Super 8 demos:
http://www.super8widescreen.com/
But they are really small resolution, does anyone know of a good, preferably full resolution example of some telecined super 8 footage?

many thanks

[audadvnc]

Depending on the telecine setup you may be able to get the film transferred at 15 fps, 25fps or any other speed you want. Give the post house a call, explain what you want and ask them if they can do it.

[wado1942]

Also will the footage be interlaced on the DV tape?

Yes, DV MUST be interlaced but if you do a progressive transfer to DV then you can treat it as such. Now if you plan to shoot 18fps, they wouldn't do a 3:2 pulldown, it would be a 4:3:3 pulldown (in conjunction with a 1fps increase in speed) which would be dang near impossible to remove in post (actually it's even more complicated if you go to 25fps). On the other hand, if you have 18fps footage transferred at 25fps, you have to slow down the footage in post which will do one of several things.

#1, you set the programs internal settings to 18fps (assumming it's supported) and slow down the footage, you work with the footage in that realm and export at 25fps. The software will either duplicate frames to fill in the holes causing jerky motion or it will progressively blend frames together. Since the greatest common factor of 18 and 25 is 1, that means only 1 frame per second will not be blended with other frames. That makes for some very blurry images.

#2, you keep the workflow at 25fps and simply use a velocity envelop to slow down the footage. This will have the same effect as #1.

#3, if your software supports it, you can export at 18fps and leave it that way but PAL output will be impossible.

Honestly I think your options are a lot better if you shoot at 24fps and had it transferred at 25fps. Or if you must save money on film, you could shoot 18fps and let them do the 4:3:3 pulldown with the 1fps speed increase but you will never be free of the interlacing artefacts.

Or if you were in North America, you could buy a 5-blade projector and do the transfer yourself. But alas in Europe, there's no mechanical way around the 25fps requirement for video unless you mod the 5-blade projector to run slightly faster, but then you wouldn't have AC sync anymore which means you'd get flicker.

[MoonstruckProductions]

... is it possible to not have the 3:2 pulldown speed correction done, or is it possible to just get frame by frame onto DV? (i don't mind if it runs the 18fps footage at 25fps, i can adjust the speed myself afterwards)

Also will the footage be interlaced on the DV tape?

As Wado mentioned, DV tape must be interlaced. In fact, any company that uses a video signal to transfer film will end up with an interlaced frame, but as long as it is captured 1 frame at a time, you will not get interlace artifacts.

I would suggest using a transfer house that will transfer your footage at 50i (25fps) and saving the captured files to hard drive. The PAL standard (50i) seems to work a lot better when applying speed/rate changes. Just make sure you are getting a frame-by-frame transfer.

It's also as simple as setting your project settings in your NLE to "Progressive" to de-interlace your transfer. Other NLE's make you add a filter for de-interlacing, but it does the same thing.

You can view examples of DV film transfers at our web page: http://www.moonpro.us ->Digital Conversions ->Super8mm -> Examples
Keep in mind that those are examples of 29.97fps transfers at 20fps using a cinemate. They are not frame-by-frame. I should have some examples of full resolution frame-by-frame up soon.

I'm sure there is someone on this forum that has some good full res examples of frame-by-frame right now. Hopefull they will respond with an address for you.

-Scott

[Chris-B]

Are you based in the UK?
How much footage are you needing transfered?

I have a frame-by-frame (workprinter type) setup. I maybe able to do a cheap transfer for you.

Let me know if your interested,
Chris.

[Chris-B]

I don't use frame blending when slowing footage tranfered frame-by-frame (25fps) to 18fps.

I just slow down the clip speed to 72.00% in premiere and make sure frame blending is unchecked.
I don't think using frame blending looks too good in PAL from 25 to 18fps, but it maybe different for NTSC.

Chris.

[wado1942]

Frame blending is good if you're going for like 50% playback but even on NTSC, it looks pretty bad for non-equally divisable frame rates. If you turn of the frame blending, Premiere will simply repeat frames like 2,1,2 to get 24fps. That's another good option but it's a bit jerkier than using a 4:3:3 pulldown.

[Chris-B]

I prefer the slight jerkyness of when it's not blended, which I don't think looks too much (if any) more jerky than when blended at 18fps in PAL. I know what you mean about slowing as far as 50.00%, blended can look at lot smoother then. Non blended at 50.00% can look like some bad jerky effect from an old sci-fi film, whereas at 50.00% blended it has a nice smooth dreamy quality.

I think often blending seems to make the picture too blurry for anything moving, either objects in the frame moving or the frame itself moving (ie panning, tracking). Blending can make the picture look really blurry with shaky hand-held stuff (which will have a motion blur anyway).

I think it works best non blended for 25fps to 18fps in PAL. As for 24fps in PAL, it is normally just scanned at 25fps and just left that way so that every thing is speeded up very slightly.

I guess it's up too the individual what they prefer, but with a frame by frame scan you can go either way and see what works best for you.

Chris.

[Chris-B]

I see in another post 'russta' mentions living in England, so I guess PAL will be used.

Chris.

[mattias]

Yes, DV MUST be interlaced

not true. dv is a frame based codec, with adaptive tweaks to handle interlaced source material with a lot of motion correctly. the progressive data is converted to interlaced again at output, since all sd video is interlaced by definition, but the dv format is actually progressive.

your post is informed and to the point, and probably much more helpful than this techno rant of mine, but i just wanted to correct the slight error that was there. :-)

/matt

[unit22]

Yes, DV MUST be interlaced

not true. dv is a frame based codec, with adaptive tweaks to handle interlaced source material with a lot of motion correctly. the progressive data is converted to interlaced again at output, since all sd video is interlaced by definition, but the dv format is actually progressive.

your post is informed and to the point, and probably much more helpful than this techno rant of mine, but i just wanted to correct the slight error that was there.

/matt

Your posts, and your icon just make me think Giant Floating Head of Knowledge. I like your footage I've seen too. Thanks for clearing that up.

[mattias]

thanks and you're welcome. the scheme is called dvc and here's a further explanation how it works:

both fields of video are stored in a 720x480/576 frame buffer and passed to the codec. the first step is determining the difference between the two fields is greater than some threshold and if so the image is again split into two fields that are compressed separately. this really only occurs when there's a misaligned edit or if there's *a lot* of motion. in most cases the frame is compressed as is, *even interlaced frames*.

the image is divided into blocks of 8x8 pixels that are dct encoded. each 2x2 "macroblock" is then compressed using its own quantization table, which is essentially a way of throwing away or reducing the dynamic range of the less important frequencies. this is the magic used by most compression schemes. older scemes like mjpeg use one q-table for the entire image, and newer schemes like mpeg-4 use variable sized macroblocks for better quality. (some also use wavelets and such rather than or in addition to dct's but that's beyond the scope of this post)

color is done at a lower res but using the same scheme. in pal the color resolution is 360x288 and in ntsc 180x480. the pal resolution might seem impressive but it's subsampled from the fields before sent to the compressor, not the frame, so the first line actualy belongs to the first and third line of the luma channel and the second to the second and fourth, thus it's not optimal for progressive images.

/matt

[wado1942]

Interesting.
BTW, when I first heard about 4:2:0 color-space, I thought it was a great idea. Then I actually saw how it worked. For a TRUE progressive image, it would be much better than the 4:1:1 of NTSC but in interlaced video, I think 4:1:1 is better. I just wish I could afford something that did 4:2:2 :roll: