Capturing legacy 60i SD footage with correct cadence

[Charles Papert]

Hi everyone! Been ages since I've posted here. Hope all are surviving the Bad Times one way or another.

I have been keeping myself busy with archiving, digitizing and organizing all types of my old media...audio cassettes, photos and videos. One issue I'm having with the last is how to best capture standard def 60i footage for archiving. If one captures as a straight 60i file, the interlacing artifacts look pretty rough in a progressive display environment such as a computer. If one de-interlaces, the result is a 29.97p file but it changes the cadence of the footage from 60 images per second to 30, which gives a faux film-look and I'd rather retain the classic 60i look.

Now, I know the arguments: 60i IS 30p, for all intents and purposes (so tempted to write "for all intensive purposes" just to be annoying, but can't bring myself to do it). Yes, the two field collapse into a single frame just fine. But it doesn't have the correct cadence.

The way I've been dealing with it is to run it through JES Deinterlacer, that hoary old piece of shareware that somehow is still working out after all these years, which frame doubles to 60p and interpolates the missing info to turn each field into a frame. It works pretty damn well, all things considered. Remember this is largely home video footage I'm working with, so it's not super precise material But, it is a software process so it takes time. What I'm trying to figure it is whether there is a realtime hardware solution that can do the same thing?

I tried using my Decimator MD-HX to do a realtime conversion from 60i to 60p, and the result appears to be that it combines the fields into frames, then repeats each frame to fill out the 60p. Which means it is effectively the same visually as a 30p file, which doesn't help.

Anyone know of a solution that won't break the bank?

[Charles Papert]

And here is the second part of my quest: I'd like to capture real time into H264. I'm using an Atomos Samurai Blade for my initial captures, which was their last generation that did standard def, but H264 isn't an option. ProResLT is fine for the stuff I really care about, but I have many tapes that I'd like to archive but not take up so much storage space. I can do the conversion to H264 in Resolve, but just like the deinterlacing, it's another process that takes time and resources and I'd love to do realtime. I've been trying to do it using a Teradek Cube I already owned that has an H264 record feature, but I have yet to find a reliable result with it, it is quirky.

I know there are a lot of pretty cheap H264 recorders designed for the surveillance market however I feel like those will introduce more artifacts and be less reliable than I'd hope for. Need something that can produce a high quality result similar to processing through Resolve but in real time.

One product I was looking at was the Datavideo HDR1. If anyone has notes on that, would be great.

[Christopher Young]

Do you have a 60i clip that you know well that you can judge a result from? I guess you do. Can you upload a clip where you definitely see a difference in cadence and let some of us try converting it and re-uploading it for your evaluation? The reason I'm suggesting this is that for the last twenty years or so I have been working on a lot of documentary archive material that has come in every shape, size, frame rate and format you can imagine. All of which had to be brought into a progressive 21st century. Always prepared for a challenge to find suitable workflows. I guess 15-20 seconds should give us a reasonable crack at it.

Chris Young
Sydney

[Bruce Watson]

Originally Posted by Charles Papert

If one captures as a straight 60i file, the interlacing artifacts look pretty rough in a progressive display environment such as a computer. If one de-interlaces, the result is a 29.97p file but it changes the cadence of the footage from 60 images per second to 30, which gives a faux film-look and I'd rather retain the classic 60i look.

Ah, the archivists' dilemma. This is, perhaps, the essence of trying to hold onto old source material without holding onto all the equipment that was used to create it and view it. Because once you've "rescued" the source, it never looks the same on modern equipment. Sigh...

But surely this has already been "best practiced" by some group or other -- there's a heck of a lot of old programming on TV, especially on OTA subchannels. This is the same thing you're asking about -- old interlaced source being seen by people on progressive HDTVs. For example, there's one channel in my area called "Grit TV" which is all about old westerns. About half their programming is old B&W TV stuff. It's possible it's a scan of the original film, but I doubt it all is. Some of it is bound to be video tape. Anyway, that's where I'd start looking for a solution, because it would seem that someone already has a reasonable one.

[Charles Papert]

Christopher Young, good suggestion and I will put up a test clip when I get the chance.

Bruce Watson, good point. The reason I haven't looked that route is because a few years back I directed a presentation pilot that took place in a newsroom, and I wanted the simulated on-air footage to have the live 60i look and the behind-the-scenes footage at 24p to differentiate it. We've seen this for years...my reference was the Larry Sanders Show. However, this was pretty easy when the delivery format was 60i, you'd just do the 3:2 pulldown with the film footage to convert to 60. I learned during Key & Peele that this gets to be a problem if you have a 24p or 29.97p deliverable, because you can't push more than 30 frames into that cadence.

My thought with the piece I was making was that since it would live on Vimeo or Youtube, both sites accommodated 60p clips so I would post everything to that timeline. The trick was converting the 60i footage (from broadcast cameras) to 60p and keeping the original cadence (dropping it on the timeline converted it to 30p with duplicated frames).

I went to a post house I have worked with that has some very sophisticated processing gear that was designed in-house because I thought they could just run it through for me. Turned out, they couldn't...but the engineer fought me tooth and nail on what I was trying to achieve (again, the old "60i is 30p, there's no difference" argument). Ultimately I ran a test using JES Deinterlacer and realized it would work, so I went that route. Hilariously, when I sent the test to that engineer to point out to him that it was indeed possible, his response was predictably defensive as he tried to pretend he didn't get that was what I was looking for. Sigh...

Here's that clip (there are several pieces of 60i originated footage in there, most obviously at 3:07):

Anchors Away on Vimeo

[Andrew Smith]

Yeah, I'm sure there must have been some 60i in there.

My goodness that was well written and very funny! Tell me they didn't all crack up whilst filming it. That piece was so entertaining to watch.

Andrew

[Charles Papert]

Originally Posted by Andrew Smith

My goodness that was well written and very funny! Tell me they didn't all crack up whilst filming it. That piece was so entertaining to watch.

Oh nice, thank you Andrew! It was a challenging day, lots to get through, can't remember too much cracking up between takes from my perspective at least! Continuity from studio to control room, plus coordinating all of the screens...lots of moving parts. Plus trying to get our "celeb" cameos in and out quickly.

[Ron Evans]

I have used TMPGenc Mastering Works 7 that will deinterlace to double frame rate, 60i to 60P maintaining cadence/temporal motion. One can choose method and I have used "interpolate high precision" that seems to work reasonably. I am not a fan of slow frame rates so always shoot 60P now.

[Charles Papert]

Interesting, Ron. Looks like Windows only though, yes?

[Ron Evans]

Yes I think I am Windows based.

[Christopher Young]

Originally Posted by Charles Papert

One issue I'm having with the last is how to best capture standard def 60i footage for archiving. If one captures as a straight 60i file, the interlacing artifacts look pretty rough in a progressive display environment such as a computer. If one de-interlaces, the result is a 29.97p file but it changes the cadence of the footage from 60 images per second to 30, which gives a faux film-look and I'd rather retain the classic 60i look.

Download these two files and compare them. One is NTSC DV 29.97 and the other is the same file De-interlaced and converted to 59.94 and it maintains the cadence of the original 29.97 DV file. Well, I believe it does. Have a look and see what you think. This particular DV file shows very high-speed movement which when transcoded correctly can still deliver a 60i cadence feel. The resulting MP4 is also less than 10% of the size of the original DV AVI file. I've made these files downloadable so as to avoid web transcoding issues such as can happen with YT and Vimeo.

Original DV interlace file:
https://www.sendspace.com/pro/dl/q63m0k

De-interlaced DV file with a tiny amount of NR as it is such a high-speed video image.
https://www.sendspace.com/pro/dl/jivmxv

Anyone any feedback?

Chris Young

[Bryan Worsley]

That NTSC DV 29.97 clip doesn't look interlaced to me Chris. Theres no movement between the fields. Looks like it was recorded in 'progressive segmented frame' (29.97 PsF) format with 2-2 pull-down - what Panasonic called 'Frame Mode' and Sony called 'Progressive Scan' i.e 29.97 progressive frames signalled (‘wrapped’) as 59.94i, so on playback each frame gets duplicated.

All you are doing by 'double-rate' deinterlacing is creating imperfect (interpolated) duplicate frames. It doesn't need deinterlacing. If you want to keep the original 2-2 pull-down cadence you would treat the clip as progressive and apply 'hard 2-2 pull-down' (i.e. duplicate each 'progressive' frame), which is what most, if not all, editors do (by default) to convert 29.97p footage on a 59.94p project timeline; I can’t see any other reason for wanting to do so.

Originally Posted by Ron Evans

I have used TMPGenc Mastering Works 7 that will deinterlace to double frame rate, 60i to 60P maintaining cadence/temporal motion. One can choose method and I have used "interpolate high precision" that seems to work reasonably.

The deinterlacer in Mastering Works 7 is pretty good, but it is Windows only. I've always used AVISynth/VaporSynth deinterlace filters (QTGMC for highest quality). VapourSynth can be set-up on MacOS but it's more convoluted. I recall the RE:Vision Effects 'Field Kit' plugin (FCP included) was highly regarded back in the day, but I've no experience with it.

As for hardware deinterlacing - well there's BMD's Teranex AV. Again, no experience with it myself, but I've read where some have had issues with deinterlace results:

https://forum.blackmagicdesign.com/v...hp?f=3&t=84285

[Christopher Young]

It was indeed interlaced, Bryan.

Yes agreed QTGMC is the best basically but even on Windows it's not the most friendly way of doing the transcoding. Basically the same quality I can get from QTGMC can be attained by using a different workflow in a much quicker simpler workflow.

Here are grabs of Media Info's details of two files. The first being the info from the original XDCam 4.2.2 1080i footage and the second being the info from the subsequent DV interlaced file that that was transcoded from the original XDCam. That being the DV file I uploaded. If you want to be convinced and see more upload a DV interlaced, or any interlaced file for that matter of about 10-30 seconds and I will process it for you. Old saying. Seeing is believing :)

Chris Young

[Bryan Worsley]

I'm not convinced Chris. Yes, it is signaled as 'interlaced' scan type, but I see no motion between the separated fields.

Care to upload a short sample of the original XDCAM footage ?

[Christopher Young]

Hi

Another example for you Bryan. This time fairly quick moving sport. Aussie Rugby League.

This is AVCHD 1080i 50 at its very low bit rate of 24Mbps. Converted to 1080p 50. Though at a higher bit rate than the original AVCHD as for this kind of sport to maintain the best compromise between size and quality I find I need to bump up the bit rate. You can see the encoding parameters used on the MP4 file, red box. The 50 frame look-ahead it what really helps on the motion estimation.

These files get delivered to the referees, coaches, managers, and the National Judiciary Panel for judgment of game code violations.

The 50i file can be downloaded here:
https://www.sendspace.com/pro/dl/nzi8sf

and the 50p file here:
https://www.sendspace.com/pro/dl/wg9mey

So far everybody feels it looks like the 50i from a motion cadence point of view so I'm are happy with the result.

Chris Young