25p to 60i by 2:2:3:2:3 pulldown

When there is need to convert 25p footage to 59.94i, the usual advice is to first conform the 25p to 23.976p and then apply standard 2:3:2:3 pulldown. The result then runs about 4% slower than the original. The sound can be up-pitched by about 4% in a separate operation. This roundabout procedure is silly since 2:2:3:2:3 pulldown can directly transform 25p to 60i while preserving the original timing. How can you apply 2:2:3:2:3 pulldown? A simple way is to put the 25p into Compressor, make the frame rate 30 fps in Video Settings and set the interlace dominance in Frame Controls. Finally conform the 60i to 59.94i.

If the 59.94i is to have different resolution than the 25p, the rescaling should of course be done at the beginning (on the 25p).

A more common need is 50i to 59.94i conversion with PAL and NTSC resolutions respectively. One method is to deinterlace the 50i to 25p, then rescale, and then follow the method described above. The weakness in this simple method is that the deinterlace to 25p sacrifices temporal information in the 50i desired for the 59.94i.

Dennis Couzin
Berlin, Germany

When an edited 25p is transformed to 60i by 2:2:3:2:3 pulldown, and the player is 2-field frame-based, approximately 2/5 of all cuts will cause a double-image "frame". These can be quite ugly. So it is better to transform unedited 25p to 60i and then to do the editing in 60i, but this isn't always possible. The standard 24p to 60i transform using 2:3:2:3 is slightly meaner, making 1/2 of all cuts ugly. Neither transform offers a simple way of doctoring the editing to avoid these frames. For 2:2:3:2:3: pulldown the underlinings show where troublesome cuts lie. These cuts can be moved to the neighboring two cadence positions in the 25p. It is best to do this numerically. From a list of all the cut locations, instructions which cuts need doctoring, and how, can be easily calculated.

There is another trick available. The 25p can be converted to (fake) 50p by frame doubling. 1:1:1:1:2 pulldown can then be applied to the 50p to make 60i as described in another strand. Again 2/5 of the cuts in the original 25p will make double-image "frames", but the routine for doctoring the cut within the 50p (as described in the other strand) is easier than within the 25p.

Dennis Couzin
Berlin, Germany

Sorry, I missed this thread.

>approximately 2/5 of all cuts will cause a double-image "frame". These can be quite
>ugly.

If the player is configured properly to play an interlaced stream, you do not see frames and certainly not both fields mashed up in a frame, but rather fields, or de-interlaced fields (ie. double de-interlacing). So double imaging will not be visible on playback.

That said, IMO, the biggest sin in broadcast is not preserving the source frame rate of the master, but "masters" with pull down added. This leads to algorithms going all crazy later when you need to use shots from the master. Eg. De-interlacing footage with pull down for web.



www.strypesinpost.com

One more thing. I'm not sure if 2:2:3:2:3 results in visually smoother motion as opposed to 2:3:2:3. In theory, it may seem better as there is less repeated fields, but this would require testing, and my guess is that with a constant 2:3:2:3 pull down, the eye quickly gets used to the motion quality of the video. Eg. Duplicating a field every 6 frames looks a little strange, because the eye gets used to a certain motion, and then there is a slight delay.



www.strypesinpost.com

strypes: We'd have to make videos both ways, 2:3:2:3 and 2:2:3:2:3, and play them a variety of players to decide which is smoother.

My main reason for suggesting 2:2:3:2:3 is to avoid the 4% slowdown. I went to rent Fassbinder's "Berlin Alexanderplatz" in DVD. It was originally shot as 25 fps film for German TV. Criterion Collection, restoring from the original negatives, decided to make it NTSC for its USA audience. That's OK, but they decided to make it 4% slow using standard pulldown. Fassbinder is such an intense filmmaker that I'm concerned that 4% slowdown -- sound pitch must have been fixed -- could change the feel of the work. So I didn't rent the DVD.

There's no justification for slowing down 25p work intended for NTSC, since 2:2:3:2:3 pulldown works. If hardware converters are geared to 2:3:2:3 then software converters are superior.

Dennis Couzin
Berlin, Germany

>Criterion Collection, restoring from the original negatives, decided to make it NTSC
>for its USA audience.

If it was shot at 25p, they would have made it 23.976fps. The DVD player then adds a pull down if necessary, although I wish they didn't have to do that, as many TVs these days are able display a 25fps frame rate.



www.strypesinpost.com

Do you mean the DVD is 23.976p? (Is 23.976p regarded as NTSC?)
Under what circumstances must the DVD player then add a pulldown? When the display is 60Hz? Does the DVD player then add a 2:3:2:3 pulldown plus deinterlace?

I thought the most widely playable DVD in the USA would be strict NTSC 720x480 59.94i, and if Criterion Collection were looking for playability they could make that DVD. Then they should have gone straight from the 25p via 2:2:3:2:3 pulldown.

When a DVD player which does accept 25p DVDs outputs to a 60Hz display what exactly does it output? Does it use 2:2:3:2:3 pulldown and deinterlace?

For me, the display end of video is the mystery.

Thanks.

Dennis Couzin
Berlin, Germany

>Do you mean the DVD is 23.976p? (Is 23.976p regarded as NTSC?)

Yes.

>Under what circumstances must the DVD player then add a pulldown?

I believe it adds a pull down for composite and maybe all analog SD outputs (eg. S Video).

>When a DVD player which does accept 25p DVDs outputs to a 60Hz display what
>exactly does it output?

I'm talking about modern LCD displays today. Those support multiple standards, so displays are not strictly 50 or 60 Hz.



www.strypesinpost.com

strypes: is it safe to assume that an LCD monitor specified as 60Hz, e.g., my Samsung 305T, is able to refresh at all rates up to 60Hz?

Dennis Couzin
Berlin, Germany

I was talking about TV monitors. Not too sure about computer displays, but that should be handled by the graphics card, and it's usually 60Hz. However, since the 80s, PAL TVs were able to accept NTSC signals, it just didn't work the same across the pond for some reason or other. Today with LCDs, there is no reason why power line frequencies should affect frame rate, because the electronics are all different. All of the LCD broadcast monitors I've worked on are multi systems, and I don't need to add a pull down to edit 24p. I could be wrong about LCDs in North America, but I just don't see a reason for nt supporting all the ATSC standards now.



www.strypesinpost.com

I've learned from a monitor tester that almost all LCDs are natively 60Hz. They play 50p material with pulldown judder: 4 frames for 1/60 sec, 1 frame for 1/30 sec, etc. Very expensive European LCDs, including some but not all EIZOs, escape this defect.

Dennis Couzin
Berlin, Germany

Interesting. Which sort of LCDs? Broadcast monitors and TVs or computer monitors?



www.strypesinpost.com

Nothing inherent in LCDs makes them more suitable for 60Hz refresh than for 50Hz refresh. There's extra cost in making dual rate drive circuitry, and extra cost in making separate 60Hz and 50Hz models. (NTSC) Japanese manufacturers and (NTSC) USA consumers seem to have colluded to make 60Hz refresh standard. Ah, globalization, and tough luck for the far greater portion of the world on 50Hz video standards.

For example, NEC makes a deluxe (1000 Euros extra) version of the Multisync PA301W for just the European market, calling it the SpectraView Reference 301. It has the standard 60Hz drive circuitry and so the judder problem with PAL material.

Anything called "broadcast monitor" must have drive circuitry appropriate to its region or else have dual rate circuitry. There's a good online test magazine for monitors, etc. which routinely includes judder tests. It evaluates many kinds of monitors, but not broadcast monitors.

Dennis Couzin
Berlin, Germany