50p vs. 50i

Ron Evans · December 13th, 2019, 08:55 PM

Quote:

Originally Posted by Steve Game

..
1080 25p means: 1080 lines of video at 25 frames per second, progressive scan
1080 25i means: 1080 lines of video at 25 frames per second, interlaced scan sent as 2 fields (i.e 50Hz)
1080 50i means: 1080 lines of video at 50 frames per second, interlaced scan sent as 2 fields (i.e 100Hz)*

Useful reference https://en.wikipedia.org/wiki/Interlaced_video

Cary Knoop · December 13th, 2019, 11:09 PM

Quote:

Originally Posted by Steve Game

Not really. With 50p the sampled video has double the number of samples per second as 25i (or 25p), i.e. 1920x1080x50 vs. 1920x1080x25.
However, with interframe compression, only the 'I frames' have twice the amount of data. The 'b frames' and 'p frames' have data content proportional to the difference between samples, so 25 fps will have differences in the scene represented by samples taken 40ms apart encoded whereas with 50 fps, if all other parameters are equal, the differences will be less owing to half the time between the samples.
1080 25p means: 1080 lines of video at 25 frames per second, progressive scan
1080 25i means: 1080 lines of video at 25 frames per second, interlaced scan sent as 2 fields (i.e 50Hz)
1080 50i means: 1080 lines of video at 50 frames per second, interlaced scan sent as 2 fields (i.e 100Hz)*

*as far as I know not used in any mainstream broadcasting system and presumably, apart from generating slow motion video, not used anywhere else. The letter 'i' stands for 'interlaced' just as the 'p' in 25p stands for 'progressive', and relates only to the type of scan. It doesn't change to represent anything else just because the frame figure is above a particular value.

It looks like you got it completely wrong, the temporal resolution of 50p and 25i are exactly the same, the only difference is the vertical resolution, 25i is half the vertical resolution of 50p.

Ron Evans · December 14th, 2019, 06:31 AM

Many years ago it was simple when almost everything was interlaced. Then 50i and 60i refereed to interlaced fields. However in both cases the timecode incremented every two fields so 25 fps for PAL and 29.97fps for NTSC. Conventions were changed to refer to timecode. Very confusing since the temporal motion is the same as the field sample rate not the timecode. It is just that each sample ( field not frame ) is half the vertical resolution. So if now you want to shoot video with the same temporal motion as old TV most people were used to then in PAL it would 50P and in NTSC 60P ( 59.97 fps )

Steve Game · December 14th, 2019, 05:52 PM

Quote:

Originally Posted by Cary Knoop

It looks like you got it completely wrong, the temporal resolution of 50p and 25i are exactly the same, the only difference is the vertical resolution, 25i is half the vertical resolution of 50p.

Slightly wrong as this line is in error:
1080 25i means: 1080 lines of video at 25 frames per second, interlaced scan sent as 2 fields (i.e 50Hz)
That is what 25psf is, i.e. 25p interlaced to satisfy transmission channel capabilities or standard HD Blu-ray specifications.*
With 25i, the odd and even fields are sampled at 25Hz each so for any given line or individual sample, the repetition rate is 25Hz. Although not always practiced, shooting with a 180 degree shutter for both 25i and 25p (20ms) gives the same amount of motion flow because the motion is only updated at the full 1080 resolution every 40ms. Using a narrower angle (10ms) because every 20ms half of picture is updated would give a less smooth motion and grittier image detail..

*Note, in 50Hz regions, HD Blu-ray video is generally distributed as 25i or 25psf as there is no 25p in the standard. More recent TVs can display 23.976/24Hz framerates, but distributors generally only produce one version for 50Hz markets.

Rainer Listing · December 14th, 2019, 07:15 PM

Also 50i nowadays typically (sadly) is taken to refer to 50 (interlaced) fields per second, i.e. still 25fps.
Getting back to Roland, at the same bitrate each single frame captured at 25p is going to be higher quality than the single frame captured at 50p. If you are delivering 25i, it shouldn't matter whether you shoot 25i or 50p (your observation that 25i and 50p files are the same size). OTOH If you are delivering 25p and shooting 50p, your NLE drops every second frame so you need to shoot higher bitrate than you would at 25p to get the same quality (but I suspect you already know that).

Ron Evans · December 14th, 2019, 07:39 PM

As I mentioned above the confusion is that with interlace the timecode is updated every two fields a full interlace sequence. It is the same for PAL as NTSC. sample rate for PAL is 50hz and for NTSC 60hz conveniently the mains power frequency for each source country. Each field is sampled at these frequencies. There is no difference in temporal motion between 50i and 50P or 60i and 60P just the picture vertical resolution. You expose both the same way as the camera is effectively shooting progressive frame and only recording a field. However convention says timecode for 50i is 25 and for NTSC 60i is 29.97fps. I believe now quoted as 25i for interlace and 29.97 fps interlace. Unfortunately interlace NTSC has the same timecode as 29.97P but they definitely do not look the same. On a good TV 60i looks just like 60P. On my upsampling Sony TV I cannot tell the difference as the TV interpolates the missing scan lines of the interlace signal to display a progressive image anyway.

Interlace is not a progressive frame split in two fields then recorded or transmitted one after the other. Interlace is the same sample rate as progressive but only fields of the progressive image are recorded or transmitted. That is why it is not possible to add them together since in PAL area the fields are 1/50 second apart and in NTSC 1/60 second apart. Initially done this way since all equipment was mains powered and the sync to mains frequency was easier than expensive oscillators that we have now.

Cary Knoop · December 14th, 2019, 08:39 PM

Quote:

Originally Posted by Steve Game

Slightly wrong as this line is in error:
1080 25i means: 1080 lines of video at 25 frames per second, interlaced scan sent as 2 fields (i.e 50Hz)
That is what 25psf is, i.e. 25p interlaced to satisfy transmission channel capabilities or standard HD Blu-ray specifications.*
With 25i, the odd and even fields are sampled at 25Hz each so for any given line or individual sample, the repetition rate is 25Hz. Although not always practiced, shooting with a 180 degree shutter for both 25i and 25p (20ms) gives the same amount of motion flow because the motion is only updated at the full 1080 resolution every 40ms. Using a narrower angle (10ms) because every 20ms half of picture is updated would give a less smooth motion and grittier image detail..

*Note, in 50Hz regions, HD Blu-ray video is generally distributed as 25i or 25psf as there is no 25p in the standard. More recent TVs can display 23.976/24Hz framerates, but distributors generally only produce one version for 50Hz markets.

You are completely wrong and you now cause, presumably unintentionally, more confusion by adding PsF to the discussion.

It is very simple:

NTSC has a temporal resolution of 59.94, PAL has a temporal resolution of 50. The only difference between progressive and interlaced is that interlaced has only half the vertical resolution and that both fields are packed into a single frame.

In other words if you deinterlace an NTSC video and the result is not 59.94p but 29.97p you are doing something wrong.

Quote:

Originally Posted by Ron Evans

That is why it is not possible to add them together since in PAL area the fields are 1/50 second apart and in NTSC 1/60 second apart. Initially done this way since all equipment was mains powered and the sync to mains frequency was easier than expensive oscillators that we have now.

Exactly!

Rainer Listing · December 14th, 2019, 10:29 PM

Quote:

Originally Posted by Cary Knoop

In other words if you deinterlace an NTSC video and the result is not 59.94p but 29.97p you are doing something wrong.
!

Uh uh. Other way round.

Cary Knoop · December 14th, 2019, 10:34 PM

Quote:

Originally Posted by Rainer Listing

Uh uh. Other way round.

Nope!

...

Rainer Listing · December 15th, 2019, 02:21 PM

Gary, sorry, missed the p on the 59.94, the way I read your post it seemed to say that the purpose of deinterlacing 59.94 fields, NTSC being 29.97i, is to result in 59.94 progressive frames, which of course is incorrect. .

Cary Knoop · December 15th, 2019, 02:28 PM

What I mean is exactly what I wrote:

In other words if you deinterlace an NTSC video and the result is not 59.94p but 29.97p you are doing something wrong.

If the result is 29.97p you just threw away half of the temporal resolution and merged fields from two different time events.

Here is an example of a properly deinterlaced PAL source, it's 50p and if you download the video and inspect each frame you will see that each frame is unique.

And here is an example of a properly deinterlaced NTSC source becoming 59.94p:

Rainer Listing · December 15th, 2019, 03:13 PM

OK, figured it out, probably showing my age, NLE deinterlacers, Resolve, Vegas, PP, blend or merge fields, 50 interlaced fields per second results in 25 progressive fields. In VirtualDub, QTGMC, rerender, you can uprez the fields to frames and seems nowadays that's also called deinterlacing. I guess either way is "proper", just need to be clear about terminology.

Ron Evans · December 15th, 2019, 08:39 PM

The merge fields etc in the NLE is their way of creating the full frames. Do they double the nearest etc. Interpolation is the correct way to do it and needs a more elaborate process. It means taking several fields and creating the missing fields to complete the progressive frames . The interpolating TV's do this. I believe my Sony uses 7 fields and of course from this information can annoy the purists using 24P it can interpolate extra frames to smooth out the motion. Easily making 30P look like 60P. But the same process is used to replace the missing fields from an interlaced source to create the 60P that the TV needs for display.

Cary Knoop · December 15th, 2019, 09:06 PM

Quote:

Originally Posted by Ron Evans

Interpolation is the correct way to do it and needs a more elaborate process.

Even better is to use motion adaptive techniques.

QTGMC with Vapoursynth (or AviSynth) is your friend!

December 14th, 2019, 06:31 AM	#18
Ron Evans Inner Circle Join Date: Mar 2003 Location: Ottawa, Ontario, Canada Posts: 4,222	Re: 50p vs. 50i Many years ago it was simple when almost everything was interlaced. Then 50i and 60i refereed to interlaced fields. However in both cases the timecode incremented every two fields so 25 fps for PAL and 29.97fps for NTSC. Conventions were changed to refer to timecode. Very confusing since the temporal motion is the same as the field sample rate not the timecode. It is just that each sample ( field not frame ) is half the vertical resolution. So if now you want to shoot video with the same temporal motion as old TV most people were used to then in PAL it would 50P and in NTSC 60P ( 59.97 fps )

December 14th, 2019, 07:15 PM	#20
Rainer Listing Major Player Join Date: Jun 2008 Location: Nowra, Australia Posts: 440	Re: 50p vs. 50i Also 50i nowadays typically (sadly) is taken to refer to 50 (interlaced) fields per second, i.e. still 25fps. Getting back to Roland, at the same bitrate each single frame captured at 25p is going to be higher quality than the single frame captured at 50p. If you are delivering 25i, it shouldn't matter whether you shoot 25i or 50p (your observation that 25i and 50p files are the same size). OTOH If you are delivering 25p and shooting 50p, your NLE drops every second frame so you need to shoot higher bitrate than you would at 25p to get the same quality (but I suspect you already know that).

December 14th, 2019, 07:39 PM	#21
Ron Evans Inner Circle Join Date: Mar 2003 Location: Ottawa, Ontario, Canada Posts: 4,222	Re: 50p vs. 50i As I mentioned above the confusion is that with interlace the timecode is updated every two fields a full interlace sequence. It is the same for PAL as NTSC. sample rate for PAL is 50hz and for NTSC 60hz conveniently the mains power frequency for each source country. Each field is sampled at these frequencies. There is no difference in temporal motion between 50i and 50P or 60i and 60P just the picture vertical resolution. You expose both the same way as the camera is effectively shooting progressive frame and only recording a field. However convention says timecode for 50i is 25 and for NTSC 60i is 29.97fps. I believe now quoted as 25i for interlace and 29.97 fps interlace. Unfortunately interlace NTSC has the same timecode as 29.97P but they definitely do not look the same. On a good TV 60i looks just like 60P. On my upsampling Sony TV I cannot tell the difference as the TV interpolates the missing scan lines of the interlace signal to display a progressive image anyway. Interlace is not a progressive frame split in two fields then recorded or transmitted one after the other. Interlace is the same sample rate as progressive but only fields of the progressive image are recorded or transmitted. That is why it is not possible to add them together since in PAL area the fields are 1/50 second apart and in NTSC 1/60 second apart. Initially done this way since all equipment was mains powered and the sync to mains frequency was easier than expensive oscillators that we have now.

December 15th, 2019, 02:21 PM	#25
Rainer Listing Major Player Join Date: Jun 2008 Location: Nowra, Australia Posts: 440	Re: 50p vs. 50i Gary, sorry, missed the p on the 59.94, the way I read your post it seemed to say that the purpose of deinterlacing 59.94 fields, NTSC being 29.97i, is to result in 59.94 progressive frames, which of course is incorrect. .

December 15th, 2019, 02:28 PM	#26
Cary Knoop Major Player Join Date: Jan 2017 Location: Newark, CA Posts: 324	Re: 50p vs. 50i What I mean is exactly what I wrote: In other words if you deinterlace an NTSC video and the result is not 59.94p but 29.97p you are doing something wrong. If the result is 29.97p you just threw away half of the temporal resolution and merged fields from two different time events. Here is an example of a properly deinterlaced PAL source, it's 50p and if you download the video and inspect each frame you will see that each frame is unique. Carole King - (You Make Me Feel Like) A Natural Woman (1971) [Restored] - YouTube And here is an example of a properly deinterlaced NTSC source becoming 59.94p: Gary Burton / Chick Corea - Crystal Silence - YouTube

December 15th, 2019, 03:13 PM	#27
Rainer Listing Major Player Join Date: Jun 2008 Location: Nowra, Australia Posts: 440	Re: 50p vs. 50i OK, figured it out, probably showing my age, NLE deinterlacers, Resolve, Vegas, PP, blend or merge fields, 50 interlaced fields per second results in 25 progressive fields. In VirtualDub, QTGMC, rerender, you can uprez the fields to frames and seems nowadays that's also called deinterlacing. I guess either way is "proper", just need to be clear about terminology. Last edited by Rainer Listing; December 15th, 2019 at 04:16 PM.

December 15th, 2019, 08:39 PM	#28
Ron Evans Inner Circle Join Date: Mar 2003 Location: Ottawa, Ontario, Canada Posts: 4,222	Re: 50p vs. 50i The merge fields etc in the NLE is their way of creating the full frames. Do they double the nearest etc. Interpolation is the correct way to do it and needs a more elaborate process. It means taking several fields and creating the missing fields to complete the progressive frames . The interpolating TV's do this. I believe my Sony uses 7 fields and of course from this information can annoy the purists using 24P it can interpolate extra frames to smooth out the motion. Easily making 30P look like 60P. But the same process is used to replace the missing fields from an interlaced source to create the 60P that the TV needs for display.