[mythtv-users] 25p vs 50p compression was (Re: OT: LED or Plasma (was Advice on choosing a TV))

Andre mythtv-list at dinkum.org.uk
Thu Jun 16 11:26:44 UTC 2011


On 16 Jun 2011, at 11:54, Paul Gardiner wrote:

> On 16/06/2011 11:10, Andre wrote:
>> 
>> On 16 Jun 2011, at 08:57, Paul Gardiner wrote:
>> 
>>> On 25/05/2010 14:37, Andre wrote:
>>>> It turns out it's actually really easy to send high frame rate video so no halfway  houses are really required, 1080p50 needs no more bandwidth than
>>> 1080i50 less sometimes, 1080p300 needs little more than 1080p50, less
>>> sometimes, when the frames are clear of motion blur and little change to
>>> the next one they compress really well! As a handy little side benefit
>>> if there is a lot of sensor noise (which there will be at 300 or 600fps)
>>> that changes every frame the real picture doesn't so much, so the first
>>> thing that gets removed in compression is the sensor noise, proper noise
>>> reduction for free!
>>> 
>>> I've now confirmed that myself. I've been experimenting with JYA's new
>>> mythtranscode with --fifodir --passthrough piping the result into ffmpeg
>>> using libx264. That for the first time allows me to get good transcodes
>>> of some 1080i content without losing the interlacing.
>> 
>> I've been following that thread, looks like a useful addition, does this mean you can properly transcode full motion 1080i50 to 1080p50?
> 
> Sure. Although, so as not to overstress my FE, I've been keeping with
> 1080i50, unless downscaling.

Nothing more stressful for a FE than de-interlacing, 1080p50 should be much less work than 1080i50, even IONs should be fine with it, final quality all depends on how good your transcode de-interlacing is of course.


> 
>>> But also I've
>>> been trying scaling down to 720p25 and 720p50. The interesting thing
>>> was how little difference there was between the file sizes for the two:
>>> 
>>>   720p25   140MB
>>>   720p50   150MB
>>> 
>>> Tiny difference in file size, but huge difference in motion
>>> quality of course.
>> 
>> You won't see a similar progression to 720p100, 720p300 etc with standard encode settings, the gop needs to stay at the same time duration regardless of the frames included, ie if gop=12 at 720p50 it should be 24 at 720p100 or 72 at 720p300, this also requires a great deal of ram in the decoder, definitely a 1GB vdpau card!
>> 
>> Good luck finding any p100 or p300 to experiment with ;-) I tried to convince BBC R&D to release their high frame rate test material but they weren't interested.
>> 
>> A customer recently trialled some equipment that was capable but although they are a sports broadcaster the production guys were more interested in shooting at 24p than 50,60 or 100,120 or 300!
> 
> Interesting, although p50 seems really smooth to me. It's 25p and 24p I
> don't really like, although loads of people prefer it for being more
> film-like.

Supposedly 24p creates a "dream like state in the viewer's mind" or so the film school people maintain, it creates an "annoyed state in the viewer's mind" when we are talking about me ;-) They seem to have all forgotten that early film was hand cranked at different rates depending on the action portrayed, the projectionist got a crib sheet of crank speeds shot by shot. 24 was eventually chosen as a compromise to de-skill film projection without incurring too much film stock cost or flickering too badly, purely economics no inherent dream like magic.

There is still a feeling in the industry that anything more than 25p or 30p is a "waste" for drama and that 1080i is "great" for sports, I often feel I will have retired before all this BS is unlearned and we get TV and movies that are pleasant to watch.

Maybe it's not so surprising when many people on this list understand much more about how TV really works than most professionals I work with!

Andre


More information about the mythtv-users mailing list