Downconvert ProRes 1080 60i to ProRes SD

A question regarding software-based field dominance swappage:
I've been slightly dissapointed with the "shift fields" filter. What is your best recipe for field dominance swappage when downcoverting 1080 60i ProRes (not HQ) to 720x480 anamorphic Standard Def ProRes?

The situation:

I've cut a sequence using footage captured as 1080 60i ProRes. Field dominance for 1080 60i is "upper".

Delivery will be a standard def DVD as well as a standard def tape. Field dominance for SD ProRes is "lower".

For the broadcast SD tape delivery I'll have a sequence set to SD ProRes and the frame size conversion will use letterboxing.

For the SD DVD delivery I have a separate SD ProRes sequence, different from the sequence for tape layback in that it will have the "anamorphic" check box activated in the sequence settings. This is because SD DVDs burned through DVD Studio Pro always have better resolution when output to mpeg-2 as "anamorphic" rather than being forcibly letterboxed into a 3:2 sequence within FCP; SD DVD players are intelligently able to "unsqueeze" as they understand the anamorphic format.

Again, the "shift fields" filter does a decent job at reversing the field dominance. However, is there another filter or workflow that produces higher quality field dominance swappage?

To be specific:
When I'm watching Standard Def television programming produced by top-dog suppliers I never notice interlacing "jaggies". This is true even on large television sets. However, I have noticed aliasing / jaggies / interlacing bad guys on projects output by FCP. I'm NOT blaming FCP in any way. I understand that the DV codec is one cause of this problem. This is why using ProRes or uncompressed is a better method, obviously. I'm just trying to make sure I'm doing everything I can within the software to produce a second-to-none downconversion, without having to use a Teranex.

Thank you.

I am fully aware that my beef is possibly just a case of staring at computer screens and NTSC monitors for too many hours per day. This could be a case of obsessive perfectionism. Many people would say "there is no problem, what are you talking about? You're one of those 'frame f%@#*ers'"

Maybe the interlacing is reduced through 120Hz processing. Maybe it is something called "dynamic rounding".

Some people would say "Yeah you should use an Avid". I don't need no stinking Avid. I know that someone understands the question.

Thank you.

Quote

I don't need no stinking Avid. I know that someone understands the question.

Maybe. But I can't say that person is me. I'm hearing three things.

1. You're concerned about doing a good downconversion. You're already aware that good downconversion is done in hardware, though, so that's good.

2. You refer to "interlacing jaggies" which, of course, aren't artifacts at all, but a natural part of the interlaced format.

3. You've asked a question about field order.

Going from bottom to top: No, you can't do a better field order reversal than Final Cut's, because Final Cut reverses field order exactly the same way everybody else does. The picture is shifted either down or up by one scan line, effectively putting odd fields on even lines and vice versa. This has the net result of reversing the field order. It's not something that can be done well or poorly, because it's not an algorithm. There's no math involved. It's just a linear shift, sample-for-sample.

As pointed out, interlacing is natural when working with interlaced material. A CRT television (with an interlaced picture tube) won't give you visible interlacing, since that's how televisions work. LCD televisions do a better or worse job of handling interlacing, depending on the manufacturer. This stinks, but it's life, and there's no way around it.

Final Cut won't do the best downconversion from 1080 to SD, but it's just fine when you're working in 24p. When working in 60i, you get better results from using Compressor's scaler.

I'd be happy to take another swing at it if you could explain the question a bit more clearly.

I'll get back to the post later, as I'm taking a short break from cutting right now.

>Field dominance for SD ProRes is "lower"

Is that true? Coz ProRes in PAL is upper.



www.strypesinpost.com

For example:

To my eye the result of the Nattress filter called "G Reinterlacer" is superior to the result produced by the "shift fields" filter in that the interlacing jaggies are less noticeable. They should not be noticeable when the video is playing. I know that the fields are visible when parked on a single frame. I am using a high-quality professional Sony broadcast monitor for viewing the SD signal, through an AJA IO via Component YPbPr.

It is true, yeah, Strypes. NTSC is lower-field-first, PAL is upper-field-first.

Quote

To my eye the result of the Nattress filter called "G Reinterlacer" is superior to the result produced by the "shift fields" filter

Respectfully, that makes no sense. G Reinterlacer converts frames into fields. You would never use it on interlaced material to make it ? erm ? differently interlaced.

Quote

I am using a high-quality professional Sony broadcast monitor

Unfortunately, that doesn't really mean anything. The days of just being able to say "Sony" to mean "BVM20F1U" are over, and Sony shipped some deeply crappy LCD broadcast monitors.

Hell, even a Cinetal does a remarkably bad job of displaying interlaced material compared to a good old BVM.

It sounds to me like you're seeing fields because there are fields there. This is the nature of the beast when you work with interlaced material. Odds are you weren't asked to vote on the shooting format, but if your client is complaining, rent a BVM for a few days so they can see what NTSC looks like on an NTSC monitor.

Thank you all.

The "G-Reinterlacer" looks better than the "shift fields", no matter what it was designed for.

There are no client complaints.

I am just asking questions because one can never know everything about everything.

The question is:

What is the best method for downconverting 1080 60i to SD from within Final Cut where the interlacing is least noticeable? Of course standard def is interlaced. What do you know about how to make a better downconversion that I do not know?

The answer is.....................?

Thank you Jeff. I'll give Compressor a shot.

Peace.

See, that's the part where I'm getting tripped up: "where the interlacing is least noticeable." Noticeable in what context? If you're monitoring properly, interlacing is never noticeable, ever. If you're not monitoring correctly, then of course the interlacing will be noticeable, because you aren't monitoring correctly. And if you're trying to deliver into some non-interlaced format, then the "best" answer will vary wildly depending on what you're trying to do.

Context:

The problem is most noticeable along people's shoulders.

And:

The problem is also noticeable on a horizontally-striped shirt being worn by someone who is in the background, far from the camera position.

Conversely, there is no problem in any region of solid color.

Thank you sir.

Okay, I don't understand this part about making it look differently interlaced. On a monitor that displays interlaced fields (CRTs), the motion would look different.

You shouldn't down convert in Final Cut. Go out through a Kona card if you're going out to tape. For DVDs, go through Compressor, set "Resize" to best" in the frame controls tab and do not deinterlace.



www.strypesinpost.com

One day I'll purchase a mini-Teranex.

This thread is kind of going off the rails.

Just tell me how you're monitoring this material, please. That's the piece of information I need to make a guess as to whether you're talking about something that's really a problem, or something that's intrinsic to interlaced footage.

I've used Compressor for this procedure in the past and it takes an astronomical amount of time to render when I really tweak the settings of the frame controls and set everything to "best"; of course this is partly because I'm still on a Dual 2GHz PPC G5.

I'm monitoring the standard def signal via YPbPr cables going from an AJA IO to a Sony CRT.

Thank you for the good advice about using the Kona 2 for layback and the Compressor workflow for DVD.

Thanks Strypes.

The NTSC monitor is a Sony PVM-20M4U.

>I've used Compressor for this procedure in the past and it takes an astronomical amount of time
>to render when I really tweak the settings of the frame controls and set everything to "best"

You can try un-checking "adaptive details" to speed up the conversion process.

But I'm suspecting you're looking at aliasing. Aliasing is always a problem with down/up conversion on interlaced footage, since there's more interpolation involved. You need to crank up the anti-alias slider to fix that, so the adaptive details section will have to be turned back on. The anti-aliasing slider will soften the image a bit, but that's how anti-aliasing filters work, then you counter balance that by cranking up the details slider slightly until you get a result that is acceptable.



www.strypesinpost.com

i wonder if what you are seeing are jaggies from the low-quality down-SCALING?

you should try setting your timeline to Motion Scaling = BEST (in the sequence settings, Video Processing Tab)

of course, if the Nattress filter gives better results,
why don't you just use that???


nick

Upon taking the advice of Strypes to use Compressor with the adaptive details / anti alias / details level option for reducing aliasing.....the Batch Monitor gave me an ETA of 246 hours, 50 minutes to downconvert a clip with a TRT of 30 minutes.

The Nattress filter, just like the Compressor adaptive details / anti alias / details level option, slightly softens the image.

Why should any video professional's best option be to soften an image? I know from experience that the Teranex produces impeccable results in real time, with no softening. Unfortunately I cannot afford a Teranex.

I am greatly appreciative to everybody for their ideas.

Apple has delivered a great product for a very reasonable price. If hardware conversion is necessary from time to time then who can complain?

It is nobody's fault. It is what it is.

Perhaps I'll shoot 30P next time.

I'm going to try and downconvert using After Effects.

if were going the hardware route,
then sure, you'd convert your whole 30mins and have a new set of rushes to work with.

as for a software conversion,
but why spend all that time converting it all when you can easily just convert your final shots?
(also you should do a small test first to see if you will like the results)

are you mixing HD and SD footage?

if NO then just cut in a sequence that matches your clips.
when done, export and down-convert with compressor.

if YES, then there are ways to isolate the used HD shots.
make sure you label the HD clips by clip NAME.
this can be searched for in the timeline.

do a "find all" in the timeline, then once selected,
use the MEDIA MANAGER to make copies of just the used sections.

run those through compressor.


nick

>Why should any video professional's best option be to soften an image?

Because the trade off after interpolation is between having jagged edges (and sharper image) or smoother edges (softer image). Anti aliasing invariably softens the image. It will never be perfect. Same as with frame rate conversion. Motion estimation technology makes it a tad better.



www.strypesinpost.com

I'll also try and convert 10i to 60p to eliminate the aliasing using this:

[www.nattress.com]

[www.nattress.com]

For a deinterlace, you can actually toggle off the adaptive detail checkbox in Compressor. That reduces compressing time greatly and deinterlacing does not require the adaptive detail filters, which is used mainly for scaling operations.



www.strypesinpost.com

After Effects produced a flawless downconvert. No aliasing. No MacGuyver missions.

After Effects was the answer.

Again, although standard def is obviously interlaced, there should be no aliasing.

If one is downconverting from HD to SD then the image should look great, period.

Thank you all for your comments.

I suggest using After Effects for downconverting via software.

good to know, swami.

how was the speed of AE compared w compressor?
my guess it was pretty good.


nick