Will downconverting a hdtv stream to dvd be any better than a real dvd?

[yoda313]

Ok here is a trivial question that is sure to be a simple task of what your personal taste are.

Suppose you take a movie broadcast in high def and downconvert to dvd 720x480. Is this going to be identical in quality to a store bought pressed dvd? I haven't done a high def downconvert in a long time and I'm wondering if it is really worth the hassle or not.

I have a cable high def dvr and I do have a firewire connection for it that works. However I also have a pinnacle usb capture device that I can use to capture in real time to dvd spec.

I guess what I'm wondering is if it will be worth it to do the digital transfer via firewire and then the headache of converting to 720x480 mpeg for authoring. The only thing I'll get out of this way is if it has 5.1 audio.

Is that correct? Is the 5.1 the only reason to do a digital transfer rather than an analog dub to dvd spec mpeg2?

Opinions are welcome. As I've said I have done it a few times in the past but not for a long time. I'm just wondering if the extra downconversion time is really necessary. Albeit aside from the 5.1 issue I don't know (again if its a stereo broadcast than there seems to be even less reason to go through the hassle). THanks.

[edDV]

Nothing from an OTA tuner or cable/sat box will match a DVD made from a digital master.

For local tuners it depends on the format and quality.

- Depends first on the tuner/set top box. If it converts HD to wide format analog and you have a way to cap AC3, then you have a quality analog (S-Video or analog component) capture. This would be good enough quality for most. The problem is most cable/sat boxes output letterbox over S-Video so force a more extreme YPbPr analog component capture solution (e.g. ADS Pyro for wide analog component 480i). If you adjust the set top box for wide 480i out, you can't be watching high def at the same time.

- Many high def cable boxes will do an internal A/D to the IEEE-1394 port from analog channels. My Motorola DCT-6xxx boxes A/D to a variable bit rate transport stream (m2t ~3-7Mb/s). Quality varies locally. Here it has become quite good since TV stations now digitally feed to the cable company. The cable company converts that to analog or feeds it to the set top box as digital and the box creates analog. But again you are dealing with letterbox on analog channels or full bit rate 16:9 HD from the digital channel.

Analog channel converted to m2t in cable box

[vhelp]

yoda313, I'll be brief, because I was going to say a few things and its turning out to a
lot. Oh well.. that's me. Anyway. Maybe I'll post it later or not. I'll see how it turns out. In
the mean time..

I wanted to give you a heads up for tonights HD broadcast on ABC. AS you know, (or don't know)
ABC broadcasts movies every Saturaday nights, mostly. And usually starting around 7pm, 8pm or
9pm, depending on the length of the movie and weather or not they apply TEC to it.

--> TEC -- Time Expansion/Compresion.

The movie that is planned for tonight (at 8PM) is:
** "Miracle"
** Coach Herb Brooks (Kurt Russell) leads the U.S. Olympic hockey team to victory
** over the Soviet team in 1980. (AS, L)
** Drama, 2004, PG, *** stars,
** 02h:15m, Color, English, United States

You can be sure that it will most likely be in -/+ AR 1:78.1 and in fairly good quality, and it will
contain the resolution of 1280x720p and 59.94 fps. The Telecine Cadense patterm for this type
of HD format is:

--> PDPDD PDPDD PDPDD ...

And, to restore it back to 24p, (provided the above pattern is consistant and uncorrupted) can
be applied with the following AVIsynth script:

--> selectevery( 4,0,3 )
--> selectevery( 5, 0,1,2,3 ) or selectevery( 5, 1,2,3,4 )

Now, assuming that you DGIndex it, the rest is up to you to incorporate into the avisynth
script (completely) and run it through its paces and finally, re-encode it to your MPEG-2 for
Standard DVD. If you don't want to bother with the above setup, you can always experiment
with the other IVTC methods for avisynth that do it more or less, automatically as best they
are capable of.

But, the main problem with that broacast will be in its chosen format for that particular time slot,
tonight. The cincirn here (IMO) will be weather or not they leave the standard 3:2 cadense telecine
intact or apply the TEC (a new re-telecine cadense pattern system) to the broadcast so that the
movie can fit inside the time slot. The movie is 2 hours and 15 minutes long, so i'm not sure yet
what they will chose. I'm not that knowledable in the new avenue of video to feel my that
answer for mylself just yet. So, we'll see in the first few minutes of the broadcast what they
chose to for this program content.

I'm going to give this movie a go and see (for myself) what they did choose in the end. If you
don't find out by then, I'll let you all know when I complete a quick analysis.

Until then..

-vhelp 4538

[edDV]

Please explain "--> PDPDD PDPDD PDPDD ... "

Are you saying motion reverses at the second P?

My experience with 720p shows frame repeats 2 then 3 in the nominal case.

23.976P starts ABCD

59.94P repeats AABBBCCDDD

But I've noticed on ABC series that frame repeats vary either for seeming editorial reasons or as you say for time compression. FOX seems to follow normal 2-3 frame repeats (e.g. "House").

I just cap ABC as is for now at 59.94. This results in about 7.5GB/hr before commercial removal or about 6GB/44min after commercial removal. Ideal decimation back to an ABCD sequence would reduce that to ~2.5GB/44min.

[vhelp]

Hi edDV,

--> PDPDD PDPDD PDPDD ...

is what PPPii PPPii PPP ii is for telecined 29.970 frame rate video, and to restore it back to 24p
you would proceed with your usual inverse telecine (ivtc) method and apply it to the PPPii PPPii
source, converting it back to 24p.

thus, in (snipped) avisynth script, you would write the following to restore 29.970 -> 24p:

--> selectevery( 5, 0,1,2,3 )
or
--> selectevery( 5, 1,2,3,4 )

(depending on your field order or where you start from inside your cadense pattern)

So, in practice, PDPDD.60p-HD is the equivalent of PPPii.30i-SD. They are in essense, the
3:2 pattern for Film that was telecined for television.

Hi edDV, I would ask you to capture some of ABC's movie broadcast tonight, it starts
in a coupld of minutes, and I'll let you decide what you think is done to the video's telecine
cadense and maybe you can bring it back to 24p using your prefered method ?

Well, for now, have fun.. I'm eating dinner now..

-vhelp 4539

[jagabo]

I just cap ABC as is for now at 59.94. This results in about 7.5GB/hr before commercial removal or about 6GB/44min after commercial removal. Ideal decimation back to an ABCD sequence would reduce that to ~2.5GB/44min.

Not really. Exact duplicate frames take almost no bits to encode in an MPEG stream.

[edDV]

I just cap ABC as is for now at 59.94. This results in about 7.5GB/hr before commercial removal or about 6GB/44min after commercial removal. Ideal decimation back to an ABCD sequence would reduce that to ~2.5GB/44min.

Not really. Exact duplicate frames take almost no bits to encode in an MPEG stream.

True if you recode. But raw cuts with programs like HDTV2MPeg2 preserve the original GOPs for something better later.

[vhelp]

But I've noticed on ABC series that frame repeats vary either for seeming editorial reasons or as you say for time compression. FOX seems to follow normal 2-3 frame repeats (e.g. "House").

Yes. But I don't watch 'house' so I'm not familiar with their pattern layout, but if you say that their
telecine cadense pattern is not the usual 3:2 (PPPii for 30i, or PDPDD for 60p) then it is safe to assume
that you have a source which has been re-telecined with a newer cadense for time manipulation.

-- I noticed that commercials are now applying this to their sources as well. I discovered this
a couple of days ago while working with the tv series, LOST. During one of their commercials the
pattern was not consistant, and it was not edited for scene-changes in the usual sense. If you
don't believe me, then go check out the Pepsi's latest commercial

For any broadcast video source that is 30i (29.970 fps) ...

The term PPPii is more or less the standard clue for a 3:2 cadense leading to source video's origin to be
from Film, and also serves as the primary signature for restoring back to 24p through the avenue of an
inverse telecine procedure. And anything other than that pattern that you see above that is
showing any signs of inconsistant or irratic cadense pattern will most likely not restore back to 24p
flawlessly unless it is a consistant TEC cadense. And then you have to determine if this TEC source
pattern is consistant with its unique cadense pattern and a procedure to decode that
patterns keys for a proper restore back to 24p. At this time, I am not skillful enough to come up
with a regimin to unlock (all) those (varying) types of unique patterns for inverse telecining work.
And the best alternative for those would be through the useage of avisynth scripts and those auto-
mated ivtc plugins available. Many to choose from, there.

And, for any broadcast video source that is 60p (59.940 fps) ...

The more newer term (for HD at 60p fps) PDPDD is more or less the standard clue for a 3:2 cadense
leading to the source video's origin to be from Film, and also follows the same path as that above in
processing for an inverse telecine procedure back to 24p film, including but not limited to TEC sources
as well, again, noting the last passages above.

The IVTC procedure is available in many fronts. AVIsynth scripts and even virtualdub, and other
utility type applications dealing with video, etc. The easiest and obvious path is avisynth scripts.
I like to use the built-in selectevery() function because it is farely straight forward in the ivtc
processing. Speed is not my concirn. I don't process whole movie sources daily like perhaps
some people do. So this built-in function serves me well.

When I process some of my videos through that avenue of avisynth, I go through the longer and
laborous steps to get to where I want to go. So I do the following:

To make a .d2v project (can be opend directly inside avisynth) from your HD .mpg source:

A - run DGIndex
B - open the hd source
C - select the option: \Video\Field Operations\[x] Ignore Pulldown flags <-- (cause there are none)
D - select the option: \YUV->RGB\[x] TV Scale
E - select the option: \File\save project\[my_hd_source.mpg]

Now, to create the acutal VFAPI psuedo avi file from the .d2v source:

A - run VFAPIConv.exe tool
B - open the .d2v project file you created in the above steps
C - select the [Convert] button, then exit. You're finished.

From here, you can open this pseudo avi video in any video editing application. Be sure you have
the vfapi driver of course, installed for this to work properly.

So why do I go through the above laborous process ? Simiple. Because I can. Actually, its a long
story and I just prefer this method for various reasons.

Once I have my fake avi file, I can now open it in virtualdub or avisynth for further processing.

To begin an aspect of one of my methods for ivtc analysis or other nonsense, I write a short avs
script. I already have plenty of templates for various types of projects, including HD.

But, since I have a newer method of analizing telecine cadense patterns without avisynth, I use
that avenue instead. That is a tool I developed in pascal. Anyway. Assuming I have found the
proper pattern to decode, if I find that I can utilize an avisynth script to process it, I will.

I have a generalize avs script that I use for HD sources. It goes something like this:

01 x = directory path + filename
02 AVISource( x )
03 # separatefields().selectevery( 5, 1,2,3,4 ).weave() # for 30i video
04 # separatefields().selectevery( 5, 0,1,2,3 ).weave() # for 30i video
05 # separatefields().selectevery( 4,0,3 ).selectevery( 5, 0,1,2,3 ).weave() # for 60p video
06 # separatefields().selectevery( 4,0,3 ).selectevery( 5, 1,2,3,4 ).weave() # for 60p video

Once this script is saved as a text file, you can open it in your video editor. I like to use the
familiar virtualdub, and I mostly just drag n drop the script file onto its window, and go from
there.

If the source is HD @ 60p, then I use the line (by uncommenting out the '#' char) with the
selectevery (4,0,3) in it. This one removes one of the duplicates frames, (bring 60p down to 30i)
and then follows it with an ivtc using one of the methods, selectevery (5, 0,1,2,3) to bring it down
to 24p. Otherwise, if the source is SD @ 30i, then use the single line for the selectevery(5, 0,1,2,3)
or other selection.

That's it. HD.60p->24p or SD.30i->24p restoration.

The above is for NTSC type videos. I don't know much about the PAL region, so I don't know
the exact processes for those.

Oh, and by the way, that ABC movie was broacasted with a clean 3:2 telecine. It was easily
restorable back to 24p and observing line 05/06 in the above avisynth script, though I did not go
through the whole movie in its intirety to verify for 100% accuracy -- it wasn't my intention.

-vhelp 4540

[edDV]

But I've noticed on ABC series that frame repeats vary either for seeming editorial reasons or as you say for time compression. FOX seems to follow normal 2-3 frame repeats (e.g. "House").

Yes. But I don't watch 'house' so I'm not familiar with their pattern layout, but if you say that their
telecine cadense pattern is not the usual 3:2 (PPPii for 30i, or PDPDD for 60p) then it is safe to assume
that you have a source which has been re-telecined with a newer cadense for time manipulation.

You are thinking about 480i or 1080i. 720p is progressive just like 480p. 23.976p comes off the film scanner as progressive frames (PPPP). Then full frames are repeated 2 then 3 to get to 59.94 fps exactly the same way 480p is output from a progressive DVD player. There are no split fields for 720p and no interlace at all.

[edDV]

-- I noticed that commercials are now applying this to their sources as well. I discovered this
a couple of days ago while working with the tv series, LOST. During one of their commercials the
pattern was not consistant, and it was not edited for scene-changes in the usual sense. If you
don't believe me, then go check out the Pepsi's latest commercial

-vhelp 4540

Almost all commercials are distributed on 1080i tape or data. For tape, a HDCAM or DVCPro deck can output 1080i or 480i with telecine or 720p or 480p at 59.94 full frames per second.

To get to 480p/720p, a full inverse telecine to 23.976 fps is done follwed by 1280x720 resize and then 2:3 frame repeats to 59.94.

That isn't to say some equipment doesn't get it wrong but network promos and commercials should be correct.

For non-film action, 1080i/29.97 is deinerlaced to 1080p/59.94, then resized to 1280x720p/59.94

[vhelp]

But I've noticed on ABC series that frame repeats vary either for seeming editorial reasons or as you say for time compression. FOX seems to follow normal 2-3 frame repeats (e.g. "House").

Yes. But I don't watch 'house' so I'm not familiar with their pattern layout, but if you say that their
telecine cadense pattern is not the usual 3:2 (PPPii for 30i, or PDPDD for 60p) then it is safe to assume
that you have a source which has been re-telecined with a newer cadense for time manipulation.

You are thinking about 480i or 1080i. 720p is progressive just like 480p. 23.976p comes off the film scanner as progressive frames (PPPP). Then full frames are repeated 2 then 3 to get to 59.94 fps exactly the same way 480p is output from a progressive DVD player. There are no split fields for 720p and no interlace at all.

Ok. Don't confuse me

I assume when you refer to FOX that it is the same as ABC 's 1280x720p @ 60p. Because I do
not receive all the various types of channels out there, I don't know all the specs for them.

In any case, that is why I separated the two.. (PPPii for 30i video frame rate, and PDPDD for 60p
video frame rate, though the 60p is unique for HD source contents, hence the not so familiar
(cryptic) 3:2 pattern, pdpdd) In order to restore the 60p->24p, you have to first follow it with the
function, selectevery(4,0,3) -- this removes one of the duplicate frames. Then, you follow
it up witht the actual ivtc implementation method using selectevery(5, 0,1,2,3) -- this will give
you 24p.

For example. I know that NBC (in connecticut) is using 1920x1080i @ 30i. So, the 3:2 patterns is
PPPii and would apply to that source type, again, using the avenue of avisynth script function,
selectevery(5, 0,1,2,3)

Now ABC airs 1280x720p @ 60p, thus the 3:2 pattern (when content is film origin) the usage for
that spec is PDPDD and would apply to that source type. So, we use the avenue of avisynth
scripts again, here: selectevery(4,0,3).selectevery(5, 0,1,2,3).

Of course, if the source program is all Interlace or Progressive in origin then the above
applications for ivtc will not work, as these are not film based, but video based.

Note, we round 59.940 to 60p for less typing

-vhelp 4541

[edDV]

ABC-DT, FOX-DT and ESPN-HD (cable) use 1280x720P at 59.94 fps 24x7 for everything.

PBS distributes to stations at 1920x1080i/29.97 but many locals convert to 1280x720p/59.94 plus two 704x480i/29.97 subchannels (second local +Vme).

All the rest use 1920x1080i/29.97.

Conversion of film source 720p/59.94 to 720p/23.976 is a frame decimation process.

[edDV]

Frame decimation strategies vary but often follow this method

Film Frames ABCD @ 23.976 fps

720p Frames AABBBCCDDD @59.94fps

drop every other frame (SelectEven/Odd in AVISynth script)

drop even #frames
A_B_B_C_D_

Drop second duplicate frame

ABCD

or
drop odd #frames

_A_B_C_D_D

Drop second duplicate frame

ABCD

[vhelp]

Conversion of film source 720p/59.94 to 720p/23.976 is a frame decimation process.

Ahh, yes, it is..

--- on another note..

First of all, this movie was actually enjoyable, and was one reason why I didn't show up until
later on when it was finished. Plus, I'm a fan of Kirt Russell, so there. Anyway

I was hoping that the quality aspects of the picture detail would be better than the usual
HD programs (being that is was movie based) but to my surprise, that was not the case.

So, I put a small section through a few paces to see what I could come up with for a down-
convert. Actually, pretty far down -- youtube'ish to be exact. Anyway. After all said and
done, I was quite surprised at the results I obtained. Trial n erorr is the usual approach in
things of this sort. So, tomorrow, I will try and upload a short clip on youtube for whatever
that is worth. Maybe i'll find a respectable section to work with and post a larger size when
I have access to fast than dial-up. In the mean time, that what I have been doing since I
last stopped by.

As I said earlier, this ABC movie special was in a clean HD 3:2 telecine cadense (PDPDD pattern)
and restoring it back to 24p was no problem. The mepg errors were not too noticable and a
few filtering applications worked out ok.

Now, as far as getting this source to a respectable SD DVD format comparitable to standard
commercialized dvd is another story. I could prob do a test scenario of this but I do not have
the actual dvd to test as guage against the progress for this uisng this ABC's broadcast source
as the example.

If anyone has this movie on dvd we could use that as a gauge for yoda313's to observe
as an example.

Also, I'm wondering if there is a 'pattern' to detect when a source might be a candidate
for TEC prior to it being broadcasted. I'm starting to work out some clues already. One of
those clues is the movie's length against the time slot it is fitted for. This movie was aired
at 8pm today and ended aprox 11pm. The guide said 2h:55m. In addition, they may have
cut some unimportant parts out. I'm sure there are other aspects to all this but darn if I
have any more energy left for this endeavor tonight.

With that I close down, good night

-vhelp 4543

[guns1inger]

I have an even simpler issue that is a lot less technical and a lot more aesthetic. HDTV broadcasts over here crop movies to fill a 16:9 screen. If a movie was made with a wider aspect ratio, I don't want a butchered version. I would rather have the DVD original, including whatever black bars may be appropriate, than the dumbed down, fill the screen version.

[edDV]

From tonight's ABC-DT movie "Miracle". This shows normal 2-3-2-3 frame repeats.

click on frame to enlarge

[vhelp]

You are right, guns1inger.. they did (and do) crop these movies. However, as much
as I *do* agree with you, its far better than was SD television has been doing for years,
cropping out to 4:3

According to IMDB, the movie is 2.35:1, though ABC broadcasted it as 1.78:1 (rounded of course)
And, the play length was 115 minutes. I think that's about 2 hours.. (115/60)= 1.92 hrs. The
guide said it was 2h:15m. So, add in the commercials, and a few carefully planned scene cuts,
and you are in business, i guess.

I don't always like the cropping that goes on by these guys because they always leave in a few
pixels on the horizontal lines, and they do *NOT* help the conversion process (in terms of the
MPEG re-encoding) with this extra black lines. MPEG makes perfect encoding video detail when
the source is evenly devisable by (16x16) not 8x8 or 8x16, (that is internal to mpeg) but the
16x16 applies to all the video, letterbox/boarders and video portions.

When ABC (or other broadcasters) air a 16:9 aspect ratio source video there is always these
extra few lines. I would guess that it *IS* actually better to leave them in, and later during
the pre-processing (on our end) would could better deal with them, i guess.

Well, see you guys tomorrow or so..

-vhelp 4544

[edDV]

I have an even simpler issue that is a lot less technical and a lot more aesthetic. HDTV broadcasts over here crop movies to fill a 16:9 screen. If a movie was made with a wider aspect ratio, I don't want a butchered version. I would rather have the DVD original, including whatever black bars may be appropriate, than the dumbed down, fill the screen version.

It looks like our ABC did exactly that. According to IMDb the original aspect ratio was 2.35:1 "Super35"
http://www.imdb.com/title/tt0349825/technical
http://en.wikipedia.org/wiki/Super_35



This was a direct cap of the TS stream.

[jagabo]

Frame decimation strategies vary but often follow this method

Film Frames ABCD @ 23.976 fps

720p Frames AABBBCCDDD @59.94fps

drop every other frame (SelectEven/Odd in AVISynth script)

drop even #frames
A_B_B_C_D_

Drop second duplicate frame

ABCD

or
drop odd #frames

_A_B_C_D_D

Drop second duplicate frame

ABCD

I just use SelectEvery(5, 0, 2) or SelectEvery(5, 0 3). One of those two will work regardless of where in the dup pattern the capture starts.

AABBB CCDDD - either
ABBBC CDDDE - either
BBBCC DDDEE - SelectEvery(5, 0, 3) only
BBCCD DDEEF - either
BCCDD DEEFF - SelectEvery(5, 0, 2) only

[yoda313]

Wow I diddn't expect this to turn out this detailed.

I've decided on simply analog dubbing stereo broadcasts and using firewire for 5.1 broadcasts.

THanks for everyones input.

A side note I am actually using a hybrid video output for the analog dubbing from my comcast motorola dvr. I have a svideo cable going out but I need so much cable lenght that I ended up using a svideo to rca coupler to convert to a rca cable to get the extra cable length I needed.

With this configuraion it indeed seems to be a letterbox 4:3 signal from the box on hd channels. That's ok for me on stereo broadcasts. My tv has a nice fill/stretch mode that will compensate and "fill" the screeen properly without distorting.

HOpefully this will be useful to other people as well.