Questions about HDTV MPEG2 encoding

Discussion in 'General TV Discussions Forum' started by Mandarin Man, Sep 22, 2005.

  1. Mandarin Man

    Mandarin Man
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    Hi,

    I have a couple of rather technical questions regarding MPEG2 encoding for HDTV. Here goes..

    1080i

    - Is 1080i from film sources encoded similarly to DVDs from film sources?

    - For example, does 1080i60 HDTV mpeg2 from film scources contain 2:3 pulldown flags (encoded at 23.976 fps), like NTSC DVDs?


    720p

    - Are there currently any 720p25 / 720p30 broadcasts or is it all 720p50 / 720p60?

    - Are most HD news and sports shows shot in 50p/60p?

    - How is 25 fps or 30 fps source video treated when broadcast as 720p50 / 720p60? Are frames repeated? Couldn't you broadcast 25p/30p mpeg2 containing flags that tell your TV to just double every frame?

    - Does 720p60 HDTV mpeg2 from film sources contain 2:3 pulldown (repeating full progressive frames)?


    Thanks in advance.

    Roy
     
  2. Stephen Neal

    Stephen Neal
    Well-known Member

    Joined:
    Mar 29, 2003
    Messages:
    6,532
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    133
    Ratings:
    +882
    I believe that just as MPEG2 SD allows for frames (and macro blocks?) to be flagged as progressive or interlaced, MPEG2 at HD does as well.

    I believe this varies broadcaster by broadcaster and encoder by encoder. Effectively it is a system/config thing?

    Some HD broadcasters correctly flag 3:2 pulldown, allowing them not to broadcast the redundant field, and the receiver converts the 48i/24p sequence to 60i. Others just broadcast the 3:2 sequence as complete fields, and treat it just like any other 60i sequence.

    I believe the latter is more common - as it copes better with broken 3:2 sequences? I suspect the former is more in use on dedicated film channels - though I have hear that both are used.

    The only country currently using 720p is the US - and they only broadcast the 720/60p variant as the final broadcast signal to viewers. Australia has 720/50p in their standard, but nobody currently uses it AIUI. Japan and Korea are 1080/60i only AIUI.

    Most 24p production in the US is done at 1080/24p, with 720/60p masters converted from this by scaling and 3:2 frame repetition. There are a few shows shot at 720/24p (I think Arrested Development is one) - these will just be frame replicated to 3:2 for TX.

    Either 720/60p or 1080/60i in the US - the key thing is the 50 or 60Hz capture rate for fast motion.

    Drama and Sit Com is usually shot 24p in the US.

    Sport and News is usually shot 60p or 60i in the US.

    Documentaries can be shot in either format.

    (60i sources are created from video cameras running internally at 60p, but with interlacing systems based on line offset line averaging in the main)

    In source video terms - yes the frames are replicated. The VTR or server source played into the MPEG2 encoder will contain 60 frames every second, and if the original source was 24p or 30p there will be repeated frames to create the 60p sequence.

    Whether the MPEG2 encoder is smart enough to spot the repeated frames I'm not sure.

    The source VTR will contain 3:2 repetition of the 24p source to create the 60p master for transmission.

    Whether the MPEG2 encoder can exploit this I don't know. I guess it could - as I believe the 3:2 detection is used on 1080/24p sources carried as 1080/60i 3:2 field pull-down

    NB you may see references to the US ATSC standard including 1080/24 & 30p and 720/24 & 30p transmission modes as well as 1080/60i and 720/60p. This is true - however these aren't used in the US. They don't change the broadcast video format on the fly in video terms - the input to the MPEG2 encoders is permanently 1080/60i or 720/60p.
     
  3. Mandarin Man

    Mandarin Man
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    Great info, Stephen. Thanks.

    Will we see 1080p broadcasts in the future and am I right in saying that 1080p will offer no advantages to 1080i as far as film source material is concerned?
     
  4. Stephen Neal

    Stephen Neal
    Well-known Member

    Joined:
    Mar 29, 2003
    Messages:
    6,532
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    133
    Ratings:
    +882
    You have to be careful when you use 1080p to describe a video standard - as there are multiple flavours.

    1080/50p and 1080/60p are not currently quite viable as production formats - they offer the same advantages of 720/50p and 60p - i.e. no interlace artefacts and full vertical resolution, and a high frame rate for sports etc. However the 1080 variant is sharper both horizontally and vertically than the 720p variant.

    HOWEVER there are also 1080/24p, 1080/30p and 1080/25p formats used for production - they are the common HD formats used to shoot drama on tape, or to post-produce 24p or 25p. They are usually variants of 1080/48i, 1080/60i and 1080/50i (and are called segmented frame when treated this way) - and have the reduced frame rate issues of film when it comes to motion rendition.

    A 1080/25p or 30p signal can be carried as a 1080/50i and 60i signal - but will have a higher vertical resolution if not vertically filtered for interlace compatibility.

    You are right that 24/25/30p material doesn't benefit from being carried via a 50p or 60p video signal, and only loses information vertically if carried in a 50i or 60i signal.
     
  5. Mandarin Man

    Mandarin Man
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    Actually I was also thinking that the 24 or 25 full frames of film can perfectly be broadcast as 60i or 50i, provided you have quality frame-based encoding and correct 2:3 pulldown detection in the case of 60i. The decoder/TV would simply perform a weave to restore the full resolution progressive film frames, which would effectively be exactly the same as 24p or 25p from film sources. So therefore 1080p will offer no advantages to 1080i as far as film source material is concerned, right?

    1080p60 on the other hand will always be better than 1080i60 because of the higher possible framerate..

    Stephen?
     
  6. Stephen Neal

    Stephen Neal
    Well-known Member

    Joined:
    Mar 29, 2003
    Messages:
    6,532
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    133
    Ratings:
    +882
    You are ignoring the vertical filtering that is introduced when interlaced video signals are produced. Therefore a 1080/24p signal converted to 1080/60i or a 1080/25p signal converted to 1080/50i will be lower resolution vertically - this is to reduce the visibility of interlace twitter on high-frequency vetical detail. (It IS not a function of the 50i or 60i video signal per se - as 25p can be carried unfiltered in 50i - as is the case with segmented field progressive recordings. However this is NOT the case for material intended to be broadcast in 50i or 60i and displayed on an interlaced display - which IS filtered)

    Therefore a 50i or 60i broadcast signal will not contain as much vertical information as the 25p or 24p source - and will not be a perfect source.

    No - it would be a good de-interlaced version, but it couldn't replace the vertical detail filtered out as part of the pre-processing to optimise video for interlaced displays.

    No - 1080/24p, 25p and 50p and 60p sources will all be sharper vertically than a 1080/50i or 1080/60i signal carrying a 1080/24p or 25p source de-interlaced back to 1080/25p or 50p, 24p or 60p.

    Frame rate of 1080/60p is the same as the field rate of 1080/60i, so the benefits are full resolution on moving information (1080/60i drops to 540/60p equivalent on faster movements), and vertical resolution. The motion rendition will be roughly the same (60i video signals are sourced from 60 frames a second not 30 frames a second capture - often people think that 60i is sourced from 30 frames a second capture and thus is less fluid. Not the case)
     
  7. Mandarin Man

    Mandarin Man
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    I didn't know about the vertical filtering! So it's introduced by the broadcast MPEG2 encoder? So only 720p HDTV is truly progressive?

    So do frame-based encoded DVDs also carry 24p or 25p filtered??

    How much verical detail is actually filtered out?
     
  8. Stephen Neal

    Stephen Neal
    Well-known Member

    Joined:
    Mar 29, 2003
    Messages:
    6,532
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    133
    Ratings:
    +882
    No - it is introduced in the video origination chain.

    1. With video cameras, the filtering is inherent in the process that converts a 1080p or 4320p CCD image to a 1080i video signal using line averaging and offset techniques.

    2. In telecines, when transferring film to 1080i video there will be a level of vertical filtering incorporated into the transfer AIUI.

    3. In conversion from 1080/24 or 25p to 1080/60i or 50i there will be a degree of vertical filtering in the output stages of a VTR AIUI.

    Effectively any device that outputs a 1080i signal but is sourcing this from a progressive source (video camera, film telecine, VTR playing 24/25p material but outputting in 60i or 50i) should perform this interlace filtering to reduce interline twitter. Some electronics graphics devices don't - and you can see the results as flicker on fine-detail - SD or HD.

    Well 1080/24p and 1080/25p source material is also progressive (though at half the frame rate or less than 720/50p and 720/60p), and will be viewable as such in the edit suite etc. However when it is converted to 1080/60i or 1080/50i for delivery via MPEG2 or MPEG4 broadcast then there should be vertical pre-filtering to allow it to be displayed correctly on an interlaced display. (Which in the US are pretty common - as CRT HDTVs were really the only HD display option until DLP/LCD/Plasma caught up with the resolution issue in the last two years or so)

    I would expect DVDs in particular to be optimised for interlaced displays - after all they are the bulk of the TVs around the world still (CRTs!).

    The level of filtering may not be the same - but AIUI it is around the 75-80% mark.

    Video cameras that convert progressively captured 50/60Hz progressive CCD captured sequences to 50/60 field interlaced format typically average two lines from the source frame to create the source field, but they then offset the averaging process by a single frame line to generate the next field, so it isn't a 50% resolution reduction vertically over time, nearer a 75% one I believe.

    It is a filtering process when it comes to 25p film transfers to 50i AIUI - and I don't think it is as simple as a brick wall filter.

    If you want to consider why this filtering is needed - think of the worse case scenario as a source image. Imagine a 1080p source that consisted of 540 black and white line pairs. When this is interlaced all of the black lines appear in one field of the interlaced signal, and all of the white lines appear in the other field. This will give you an entirely black field followed by an entirely white field - and rather than looking like a fine grid, it will look like a flashing black and white signal, cycling at frame rate (25Hz in the case of 50i).

    If you now consider what happens if this signal is suitably vertically pre-filtered. The alternate black and white lines are filtered and in an ideal world generate a grey average across both lines - so instead of a flashing black and white signal, you get a uniform grey signal, with no detail but no difference between fields...

    Obviously this is a worse case scenario, but you can imagine what would happen with detail less fine than this, but still finer than the field resolution - you'd still want a degree of filtering. (Which would also avoid aliasing issues as well...)

    These issues are all quite different from PC video resolution issues. In PC video you want to see each pixel individually quite often - whereas with video you don't want to know that the video signal is pixel based, and often a slight filtering improves the subjective picture quality close to the sampling rate.

    In recent years broadcasters have begun to use PC/Mac graphics solutions rather than graphics solutions produced by broadcast video equipment manufacturers. One of the side effects of this has been a drop in the perceived quality of the graphics, as aliasing and interlace flicker have returned... I recently watched a repeat of the 1987 General Election on BBC Parliament, which used some very nice graphics generation kit for the time. Comparing it to this years election - the graphics were far less sophisticated - but the rendering quality was IMHO higher. Text on captions was cleaner and less flickery, and some of the animated graphics were far less jagged-edged than the more recent stuff. The 1987 stuff may have been slightly softer - but it also looked nicer in rendering terms, if not in sophistication.
     
  9. richard plumb

    richard plumb
    Well-known Member

    Joined:
    Jan 27, 2002
    Messages:
    14,478
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    133
    Location:
    Windsor
    Ratings:
    +1,046
    What about HD-DVD (whatever flavour)?


    Will it take advantage of the medium? I.e will it be mastered for 1080p/24, then packaged into the necessary formats by the player or the authoring?

    i.e, will the player output, say, 1080i/50, but the receiving display/player be able to accurately recreate a full resolution 1080p/24 image from that?

    I would hope that the bluray/HDDVD specs would provide for a more robust flagging system compared to DVD, and I would hope that mastering would not be filtered for interlaced displays, as this could surely be done relatively easily on the display device anyway, as most will be progressive panels.
     
  10. Mandarin Man

    Mandarin Man
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    Wow.

    So if you were to export a DVCPRO HD/1080p24 NLE project (true progressive) to MPEG2/1080i60 for HDTV broadcast, the MPEG2 encoder would perform this interlace filtering?

    Also curious about HD-DVD. 1080p24 would seem the way to go..?

    Also, will we see 1080p HDTV broadcasting in the near future?
     
  11. Stephen Neal

    Stephen Neal
    Well-known Member

    Joined:
    Mar 29, 2003
    Messages:
    6,532
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    133
    Ratings:
    +882
    No - it isn't a function of the MPEG2 encoder, it should happen earlier than this.

    Either the conversion from 1080/24p to 1080/60i for export should include a vertical filter process, or the output from the NLE will be nasty for 1080/60i viewers and have lots of twinkly, flickery detail (assuming the detail is present and the camera source hasn't actually pre-filtered!), and theoretically non-broadcast...

    I suspect that a decent broadcast NLE manufacturer - Quantel for example - might include decent filtering. A PC/Mac variant might not and would, theoretically, output non-broadcast quality video (i.e. overly sharp). These days there may not be anyone in QC in a broadcast environment who would spot this...

    Intellectually it strikes me that unfiltered 1080/24p and 1080/25p are the way to go for pre-recorded formats, as long as the vertical filtering can be delivered to a decent quality on replay. (Remember that many existing HD owners in the US and Japan are watching on 1080/60i native CRT displays)

    1080/50p and 1080/60p are currently impractical for broadcast production (the bandwith is too high for the standard 1.2Gbps HD-SDI interconnect used for uncompressed connections in the broadcast environment) - and they aren't currently part of any proposed broadcast standard AIUI.

    1080/24-30p are included in some broadcast standards (ATSC for example) - but the difficulty of running a mixed frame rate standard channel and the poor quality of 24-30p for sports etc. means that they are unlikely to be used.
     
  12. Mandarin Man

    Mandarin Man
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    One last question regarding interlace filtering and then I'm wrapping it up..

    Take the Panasonic HVX200 for example. It records 24p/30p in both 720p and 1080i recording. So when using 720p you would get the full 720 resolution, but when using 1080i you would lose some of the 1080 resolution due to interlace filtering..
     
  13. Stephen Neal

    Stephen Neal
    Well-known Member

    Joined:
    Mar 29, 2003
    Messages:
    6,532
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    133
    Ratings:
    +882
    Without knowing exactly how the camera is implemented on the HX200 it is difficult to say for sure how the 720p and 1080i compare.

    No - you don't "lose some of the 1080i resolution" - you get the full 1080i resolution, which is equivalent to that of a progressive system running at around 800 lines. You have to stop thinking that a 1080i system will deliver 1080 lines of vertical resolution - it won't if it is a proper interlaced signal. If the interlaced signal contained the full resolution of a 1080p source signal it would potentially look quite nasty on a 1080i display when it comes to HF information vertically, and you would get quite nasty vertical aliasing as well.

    A 1080i video camera, by design will take a 1080p (or 4320p in the case of Thomson/Philips/GrassValley LDK 6000s) and convert it to a 1080i signal by line averaging (which delivers the vertical filtering)

    The line averaging is often implemented (along with an optical low-pass filter in the camera to reduce sampling aliasing) as follows :

    50 or 60 1080 (or 4320) line frames are sampled each second.

    For field 1, lines 1&2 from frame 1 are averaged to create a field line. Lines 3&4 of frame 1 are averaged to create a field line etc.

    For field 2, lines 2&3 from frame 2 are averaged to create a field line. Lines 4&5 for frame 2 are averaged to create a field line etc.

    (With 4320p sensors then 8 lines are averaged rather than 2. 4320p sensors are used by Philips/Thomson/GrassValley to create a 1080p - 4 lines averaged, 1080i - 8 lines averaged, 720p - 6 lines averaged - signal. A very neat solution.)

    In isolation this looks to be converting a 1080p signal to a 540p signal, and it would be if there wasn't the frame-line offset between fields.

    I would expect the HVX 200 to convert the output of its 1080line CCD to a 1080i video signal using this method - and thus deliver around 800 lines of vertical resolution.

    When the camera is running at 1080/24p or 1080/30p (assuming it does) - I would expect that it scans the CCD at a lower vertical refresh rate, but clocks out every line in the progressive frame. This will deliver the same number of lines a second (30p compared to 60i, 24p would deliver fewer). However I wouldn't expect there to be any line-averaging in the p modes, unless they are cheating and combining fields and shuttering somehow.

    In 720p mode I would expect that either the 1080 line sensor is clocked out at frame rate and scaled, or possibly that the 1080i signal is treated as 540p and upscaled (the latter would be quite nasty...)

    However it is also worth remembering that this camera looks to be based on DVCProHD/100 which significantly subsamples the image horizontally. AIUI it uses 960x720 sampling rather than 1280x720 sampling at 720p, and 1280x1080 sampling rather than 1440 or 1920x1080 sampling at 1080i. Ironically this is actually a lower resolution recording format than the MPEG2 TS/PES based HDV system which runs at 1280x720 and 1440x1080, though it uses far higher levels of compression to squeeze this into a quarter-or-less of the data rate. (DVCProHD/100 is frame based, HDV is GOP based, and so less suited to lossless editing)

    I'm not an expert on domestic and professional cameras - my experience is more at the broadcast end I'm afraid (and a little bit out of date)
     
  14. Mandarin Man

    Mandarin Man
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    Thanks for your input, Stephen.
     
  15. Stephen Neal

    Stephen Neal
    Well-known Member

    Joined:
    Mar 29, 2003
    Messages:
    6,532
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    133
    Ratings:
    +882
    I'm just re-reading your question and I may not have properly addressed one aspect.

    1080/50i and 60i video signals CAN carry 1080/24p/25p and 30p video signals without a vertical resolution loss - there is nothing to stop you splitting the frame into two fields without applying the filtering. This is the so-called "segmented frame" progressive standard - and is used by many broadcast VTRs from Sony to record progressive video. The key thing to remember is that although the 1080p signal is being carried in an interlaced video stream, it isn't actually an interlaced signal.

    Thus it is possible that a 1080/30p captured signal recorded as a 1080/60i signal may be higher resolution vertically than the same image captured 1080/60i and recorded 1080/60i. (The former will not have had interlace filtering applied - though any optical filtering will remain.)

    (Home and Away is edited to take advantage of this. H&A is shot 1080/25p - and this is down-converted to 576/25p for editing. The editing is actually carried out using 576/50i standard def video gear - but with great care not to pass this segmented frame video via any processing that will filter vertically as if the signal were interlaced - so no DVEs etc. For transmission the 576/50i video is converted back to 576/25p and then upconverted to 576/50p by frame repetition. I have assumed that the 25p video is filtered to create the 50i masters for UK transmission - though it might not be - and in fact the 25p may be slightly softer than theoretically is possible?)
     
  16. dsb

    dsb
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    i'm sure i've read that blu ray and hi def dvd will store images in a 1080p 24frame format for movies - even if most players will not output the image in this format.
     
  17. Stephen Neal

    Stephen Neal
    Well-known Member

    Joined:
    Mar 29, 2003
    Messages:
    6,532
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    133
    Ratings:
    +882
    Yes - I've read this as well. It will be interesting to see if the on-disc recording is full resolution 24p, with the player adding filtering for interlaced outputs, or if it is pre-filtered to allow the interlacing process to be simpler.

    (There is of course the option that there is no interlace filtering - and an assumption is made that most 1080i fed displays will actually be de-interlacing to progressive...)
     
  18. Dutch

    Dutch
    Well-known Member

    Joined:
    Dec 5, 2002
    Messages:
    2,601
    Products Owned:
    2
    Products Wanted:
    5
    Trophy Points:
    106
    Location:
    North Yorkshire
    Ratings:
    +284
    This EBU technical paper may be of interest to you, Stephen.

    http://www.ebu.ch/CMSimages/en/tec_text_r115-2005_tcm6-37869.pdf

    Steve
     
  19. Stephen Neal

    Stephen Neal
    Well-known Member

    Joined:
    Mar 29, 2003
    Messages:
    6,532
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    133
    Ratings:
    +882
    Yep - had stumbled across that. Saw a demo of a BBC R&D solution to the 1080/50p via an HD-SDI connection problem at a recent open day. They have developed a low latency (8 lines of delay) mezzanine compression scheme that compresses a 1080/50p signal into a 1080/50i HD-SDI video signal. Even more clever is that the compression is devised in such a way that you get a low quality 1080/50i signal as part of the compression - so can monitor the feed for identification purposes with a regular 1080/50i monitor.

    It uses a bit more than 2:1 compression - but does actually deliver a solution that would be compatible with existing routing infrastructure - though peripheral devices like vision mixers and VTRs would require compression and decompression units.

    (NB Am I the only one who finds the EBU nomenclature of 1080i25 and 1080p25 misleading? It implies that 1080i25 only provides 25 images a second - whereas, although it runs at 25 frames per second, there are 50 separate fields, all potentially taken independently at 50 different points in the second. I much prefer 1080/50i personally...)
     

Share This Page

Loading...
  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice