Full details sought about capture of 24p at 60i

G

guyburns

Active member
I'm new to video -- never owned or operated a film camera, a VHS player or recorder, or a DV recorder. I've never owned a standalone DVD player either. I hope that's not going to be too much of a handicap getting into 1080p video.

I've been through every post in this forum, and haven't come across a detailed explanation of how the GH1 captures 24p into 60i. There has been quite a bit of comment about how to extract 24p once recorded, but no detailed description of the overall process.

QUES
Can anyone recommend a web-based article, PDF or book on how a video device captures 24p and incorporates it into a 60i data stream? The info should be applicable to the GH1 and should get rather technical.

I want to find out how much degradation is involved -- and how obvious it is -- when capturing 24p in 60i and then trying to extract it, because I am tossing up whether to buy a PAL version of the GH1 ($800 more expensive, but no time distortion [?] during the recording/extraction of 25p), and the NTSC model (see http://www.dvxuser.com/V6/showthread.php?t=198534).

I've been through the Final Cut manual, and that does a pretty good job over several pages of explaining what happens, but I can't be sure if it applies to the GH1. For example, it talks about recording with 2:3:2:3 pull down and also 2:3:3:2 pulldown, and says that one gives better results depending on the playback situation, but doesn't really go into much detail:

Several NTSC and 1080i60 HD camcorders can record using standard pull-down, though advanced pull-down is usually recommended when recording 24p video. However, for final playback on television or DVD, 3:2 pull-down is generally considered to have the most acceptable quality of motion.​

I'd like a document that goes on page after page about all that type of thing, and which has lots of timing diagrams. Any suggestions most appreciated.
 
guyburns said:
I want to find out how much degradation is involved -- and how obvious it is -- when capturing 24p in 60i and then trying to extract it, because I am tossing up whether to buy a PAL version of the GH1 ($800 more expensive, but no time distortion [?] during the recording/extraction of 25p), and the NTSC model (see http://www.dvxuser.com/V6/showthread.php?t=198534).
Click to expand...

If you live in the US, you most likely will not get a Panasonic factory warrenty for a PAL camera.

The PAL version does 25p, and as such does not need to be converted to 24 fps... unless you really want to... slow your film down by 4%...

For the NTSC version, the sensor produces 24p, which is then formated in the 60i stream with the 3:2 pulldown. This can be (shoulbe be) removed, and is the 'recommended' way to use the camera.

The 'advanced' pulldown indicated in the Apple Final Cut manual, was introduced with Panasonic's DVX100 and was to address problems putting 24p material onto DV tape. I believe the GH1 just does the the 'simple' 3:2 sequence.
 
FWIW, here's my understanding. For NTSC, the GH1 sensor provides 24 progressive 1920x1080 frames per second. Then each pair of progressive frames A,B is converted to four horizontally interlaced frames A1,A2,B1,B2 and the final interlaced frame is duplicated once to give you A1,A2,B1,B2,B2. There is no degradation or loss of information.

HOWEVER, many players and most editors do not try to undo this transformation, and for those that do try, few do it correctly. For example, Cineform NeoScene (see attachment) and Microsoft Expression Encoder 3 will mangle the deinterlacing process and give you progressive frames that appear to combine interlaced frames that do not belong together. This is most noticeable in any footage with lots of motion.

I have personally confirmed that Windows Live Movie Maker, Sony Vegas Movie Studio Platinum 9.0b, and Sony Vegas Pro 9.0c all mangle the GH1 footage; none are able to play the footage back correctly while editing or render it correctly (Vegas Pro being the worst, see below). AFAIK, the only editor that plays the Gh1 footage back correctly frame-by-frame and renders it correctly is Edius Neo 2 Booster.
 

Attachments

  • Still0124_neoscene-ghost.jpg
    Still0124_neoscene-ghost.jpg
    158 KB · Views: 0
  • vegaspro-ghost-tear.jpg
    vegaspro-ghost-tear.jpg
    356.2 KB · Views: 0
Thanks for the responses. It's a rather serious issue isn't it, all that software not correctly extracting 24p?

As well as researching the GH1 in depth before I buy, I also have to decide on software. So, does Final Cut extract 24P okay? And Premier? Or does Premier use Expression Encoder 3?
 
I haven't tried Adobe Premiere or Apple Final Cut.

Another option that works is to use VoltaicHD to convert your AVCHD footage into an editor-friendly 24p progressive frame format, eg. AVI/MOV/WMV. AFAIK, the only mistake VoltaicHD makes is to double the first frame of the clip; everything else looks correct. Once you convert with VoltaicHD, you can use any editor.

Note that this problem only exists with FHD NTSC footage (24p in 60i wrapper). If you shoot in 720p60, you won't experience this problem, and you can use any modern editor. And you won't experience this problem in the PAL version of the GH1.
 
It looks like Premier does not natively do the deinterlace/reverse telecine but perhaps can be configured to do so:

http://www.dvxuser.com/V6/showthread.php?t=197892

Personally I think, "Why bother with these torturous workarounds when it works correctly in Edius Neo 2 Booster?" It's not like any one of these editors is that great anyway, its only a question of time before you run into some horrible bug or missing crucial feature. And if you're really committed to a specific editor, then VoltaicHD solves your problem for only $35.
 
Personally I think why bother with a NTSC camera when living in a PAL country as the OP does. Both 24 fps and 60 fps will constantly put you at odds with the 50 Hz AC current when shooting under fluorescents and other ballasted lights. Like this:
IMPORTANT NOTICE: No media files are hosted on these forums. By clicking the link below you agree to view content from an external website. We can not be held responsible for the suitability or legality of this material. If the video does not play, wait a minute or try again later.
I AGREE

http://www.youtube.com/watch?v=njHVkG8nekA

If the GH1 is still high priced in Australian check out Hong Kong retailers, they sell the same PAL version without the EU tax time limit on recordings.
 
The flicker occurs when the shutter speed is different from the mains frequency, and I suppose it has a similar explanation to the "beating" you hear when you listen to two tones of similar frequency. If you shoot at 1/50 in PAL countries and 1/60 in NTSC countries, beating won't occur.

Correction: Yes it might! As I now understand it, whether flicker is present does not depend on the shutter speed; it depends on the frame rate. In PAL countries, to avoid flicker the frame rate must be an integer multiple of 25 Hz; in NTSC countries a multiple of 30 Hz.

According to one source, shooting at 59.94 fps under certain types of 60.00 Hz lighting, even though the frequencies are very close, the frequency mismatch may cause a variation in brightness with a period of about 8 seconds (see http://www.pixim.com/news-and-event...-fluorescent-flicker-in-security-camera-video).

Then there is the problem of variation in the mains power itself. Where I live (Tasmania), the power authority tries to keep the mains frequency in the band 49.85 to 50.15 for 99% of the time (see http://www.gpoc.tas.gov.au/domino/otter.nsf/LookupFiles/Position_Paper_Frequency_Operating_Standards_2005.pdf/$file/Position_Paper_Frequency_Operating_Standards_2005.pdf). So, even in PAL countries, shooting at a frame rate of 25 Hz may cause problems under artificial light whose intensity fluctuates with mains frequency.
 
Last edited:
Ummm... I'm a near fifty year old filmmaker who's been dealing with this my entire work-life and trust me it's not that easy in practice. Especially with GH1 - if you use the 720 60p mode and shoot at 1/50 you'll end up with a 30 fps file as as soon as the shutter speed goes below the frame rate the frames get doubled. And you loose the slo-mo option in editing.
 
I've done what I think is a fairly extensive web search and have come up with very little that is definitive about the GH1 and how it records 24 fps into a 60i data stream. I'm hoping that with other people's help, I can come to a better understanding. I like to know how things work, even if I don't end up with the NTSC version.

First thing: does the manual's comment "sensor output is 24fps" mean 24.00 fps or 23.98?

Second thing: since the way the GH1 records 24p does not appear to be publicly documented by Panasonic, I'll have to go on what others have stated in this thread and in other threads in this forum, which is that:

1. Images, called frames, are captured 24.00 times per second at 1920 x 1080.

2. Call the first four frames A, B, C, D. The odd-numbered lines of A and the even-numbered lines of A are extracted and stored as A1 and A2. The same is done for the other frames. These are called "fields".

3. The fields are then put together in the sequence A1, A2, B1, B2, B2, C1, C2, D1, D2, D2, and stored. The extra fields B2 and D2 are required to flesh out the sequence so that the traditional "inverse telecine" method can be used when extracting the frames in the next step.

4. When it comes times to extract the frames: A1 and A2 are combined to form a full A-frame, and so on for B, C and D, with the proviso that the extra B2 and D2 must be accurately located and dropped, otherwise the proper sequence will be destroyed. This process of extracting the frames is called "inverse telecine".

Is that how the GH1 handles 24p/60i?
 
guyburns said:
3. The fields are then put together in the sequence A1, A2, B1, B2, B2, C1, C2, D1, D2, D2, and stored. The extra fields B2 and D2 are required to flesh out the sequence so that the traditional "inverse telecine" method can be used when extracting the frames in the next step.
Click to expand...
That's correct except the reason for the extra frames is that they are trying to create 59.97 interlaced frames per second, which is a standard for video.
 
Okay, now that it seems like I have a reasonable grasp of the process, I'd like to know why Panasonic uses this method. I want to understand as much as possible about this process.

QUES 1
What situation are they catering for by using 24p/60i. Is the final 60i data stream suitable for presentation on an NTSC TV without doing anything to it? If that is the case, are they catering for the American novice user who wants to take HD footage and then display it on his TV, but doesn't want to have to edit it?

QUES 2
If they are trying to simplify things for novices for displaying on a TV, why choose 24 fps? Why not 25 fps or 30? Were the engineers hard up against the data rate limit so couldn't use 25? What I am getting at is: capturing at 24 fps sort of implies to me they are also catering for someone other than a novice; someone who actually wants the 24 fps images and who is prepared to go to the bother of reverse telecine to get them. If that is the case, why not just give users a 24p option, like they give PAL users the 25p option?

I'm unclear about why Panasonic would do what looks like to me -- a complete PAL video novice -- something that doesn't look very sensible. But maybe it is a sensible approach and I've missed something.
 
guyburns said:
What situation are they catering for by using 24p/60i. Is the final 60i data stream suitable for presentation on an NTSC TV without doing anything to it?
Click to expand...
Yes.
guyburns said:
If they are trying to simplify things for novices for displaying on a TV, why choose 24 fps? Why not 25 fps or 30?
Click to expand...
I agree, 24p-over-60i is lame. Canon, Sony and Panasonic offered 25P/30P in their DV cameras for years, as a reasonably simple and clean way of recording progressive video inside interlaced stream, because DV is interlaced only.

AVCHD supports native progressive, so I would expect either 30p-over-60i for the mass market, or native 24p for wannabe filmmakers. Even Canon understood this, and now offers native 24p. Why Panasonic cannot get it is beyond me.
 
24p/60i is slowly becoming clearer to me, but I'm still not completely happy. Theres's the timing problem, what I will call "jitter". I brought this up in another post (http://www.dvxuser.com/V6/showthread.php?t=198534), but I pulled the plug on that one and began again here.

One of the responses to my other post, re 24p/60i, was: "If you play it back at 60i, yes, it looks a little weird." How weird? Really obvious to anyone, or only to a discerning eye? Given that Panasonic has probably implemented this scheme to allow novice users the ability to play back their videos straight from their GH1 on an NTSC television (see below), surely they wouldn't want the video to look "a Little weird".

Back to the timing problem. I'm going into this in great detail because I really want to know how it works. I will use two decimals to indicate the speed, and I will use a hypothetical 60.00i (instead of 59.94i) for simplicity.

There are two possibilities when considering how 24.00p/60.00i could be implemented:

POSSIBILITY 1
Full frames of 1920 x 1080 are captured at 24.00p and encoded in the form: A1, A2, B1, B2, B2. If this data stream is played back at 60.00i, there is no overall time distortion because each group of the original two 24.00p frames (2/24 = 1/12 second) fits exactly into five of the played-back 60.00i fields (5/60 = 1/12 second).

HOWEVER, there is time distortion ("jitter") within each group of two 24.00p frames because they are being displayed at 60.00i. For example, starting the clock at zero milliseconds, at the end of A2 the time should be 1/24 seconds, but will actually be 1/30 second; at the end of the first B2 the time should be 2/24 second, but will actually be 2/30 seconds. The frames are being temporarily speeded up. It is only at the end of the second B2, that the times are identical and equal to 5/60 seconds.


On the other hand, if instead of being played back at 60.00i this stream is properly decoded using inverse telecine and played back at 24.00 fps there is no jitter.

POSSIBILITY 2
Half frames of 1920 x 540 are captured at 60.00i. If this data stream is played back at 60.00i, there is no overall time distortion.

HOWEVER, if frames are extracted in trying to obtain 24p, there is time distortion ("jitter") within each group of two 24.00p frames, but of a different kind to that described above. For example, starting the clock at zero milliseconds, at the end of A2 the time should be 1/30 seconds, but on playback will actually be 1/24 second; at the end of the first B2 the time should be 2/30 second, but will actually be 2/24 seconds. The frames are being temporarily slowed down. The times come back into alignment at the end of 4 frames (4/24 = 10/60).

TWO QUESTIONS
Q1: Even though the GH1 manual states that the images from the sensor are 24p, is it beyond the bounds of possibility that Panasonic has implemented Method 2 to ensure that unedited 24p/60i video from the GH1 shown on a TV looks good? Is it possible to tell which method is being used? Does the data stream contain timing signals?

Q2: It comes down to this: which user is having their video compromised: the novice, for whom the 24/60i system has been implemented (I assume), who just wants to watch videos straight from the camera on their NTSC television; or the user who extracts 24p?
 
Last edited:
guyburns said:
Is the final 60i data stream suitable for presentation on an NTSC TV without doing anything to it? If that is the case, are they catering for the American novice user who wants to take HD footage and then display it on his TV, but doesn't want to have to edit it?
Click to expand...

Their initial concern is not as much american novices but the rich domestic buyers who just pull the card out of the camera and into the Viera™ HD television from Panasonic and view the eight pretty clips from the last trip abroad. GH1 is intended as a high-end consumer camera (almost luxury class), not a professional tool. The manual even says that AVCHD is for television viewing and for 'computer use' one should opt for the MJPEG mode. It's just us crazies here who realised that with a bit of TLC one can pull very good looking footage out of it.
 
guyburns said:
One of the responses to my other post, re 24p/60i, was: "If you play it back at 60i, yes, it looks a little weird." How weird? Really obvious to anyone, or only to a discerning eye? Given that Panasonic has probably implemented this scheme to allow novice users the ability to play back their videos straight from their GH1 on an NTSC television (see below), surely they wouldn't want the video to look "a Little weird".
Click to expand...
Every Hollywood film ever shown on NTSC or ATSC television in the last 60 years has used 2:3 pulldown to embed 24p within 60i. If you've ever seen any 24fps footage on broadcast television, whether a movie or a high-budget TV show, you've watched 24p within 60i.

As for your proposed methods, method 1 is correct. The sensor runs at 23.976 fps and it scans progressive frames at that interval. They are embedded within 60i using 2:3 pulldown.

Q2: It comes down to this: which user is having their video compromised: the novice, for whom the 24/60i system has been implemented (I assume), who just wants to watch videos straight from the camera on their NTSC television; or the user who extracts 24p?
Click to expand...
The compromise is the television. The camera images the footage properly. If you author a 24p DVD and play it on a progressive-capable display, it will be displayed properly.

It's been done this way for 60 years. It's not a problem.
 
You must log in or register to reply here.
Back
Top