NEW! PaintShop Pro 2019

Bitrates for DVDs

Discuss anything about video editing, HD, codecs, etc......

Moderator: Ron P.

Bitrates for DVDs

Postby Devil » Sat Oct 20, 2007 1:36 pm

A number of years ago, I did some tests regarding the artefacts introduced by encoding DV footage to DVD-compliant MPEG-2. I've just found that I have this here on my website.

As this is a perennial question, I'm putting it here as a "sticky"

My conclusion was that, using a DV-compressed original, there was little to be gained by using bitrates higher than 6000 kbit/s CBR for the video. This applies to practical viewing of the DVD on a 32" LCD or a 28" CRT TV from 3 DVD players, as well. This point of view has been disputed by a member of this forum, who claims that the quality visibly improves at higher bitrates. I would never dispute this assertion with input from an uncompressed source; in that case, higher bitrates should improve the picture quality, at least up to the limit of the DVD player's ability to read DVD¡ÓR discs without glitching (in some cases, this may be as low as a combined video+audio bitrate of about 7500 kbit/s). However, there is a possible explanation why my colleague's assertions are different from mine. His DV original was in NTSC, 720 x 480 pxl, 29.97 f/s and, most importantly, with a colour space of 4:1:1, which is converted to 4:2:0 during encoding. My DV original was in PAL, 720 x 576 pxl, 25 f/s and with a colour space of 4:2:0. It may well be that the massive conversion of the colour space with NTSC works differently from the minor conversion with PAL. However, this is speculation.

While on the subject, I have also analysed the effectiveness of VBR (1- and 2-pass) compared with CBR. I have no documentation to substantiate what I am about to say. What I found was that, if I encoded at an average bitrate of 5000 kbit/s with CBR and VBR, the latter with a max of 7000 kbit/s and a min of 3500 kbit/s, the results were rarely better with either of the VBRs. The test vehicle was an edited DV project lasting 55 minutes. It included both fast moving and slow moving scenes with a number of sudden butted scene changes, some with very contrasting changes of scene. It also included some transitions of varying types (including some 3-D types) over times varying from 1 second to 15 seconds. I used a bitrate viewer to examine what happened and I ran the three DVDs through a player to see what the visual difference was. At the time, my TV was a good 28" CRT type. I honestly could critically see little difference between the three, nor could my wife with a more casual viewing. On the bitrate viewer, I could certainly see a difference, but the high bitrates were visible mostly during the transitions, especially the fast ones where there was also a quantisation change. Even street scenes with fast moving traffic rarely showed a bitrate much higher than the average. I concluded that VBR, under these conditions, did not really help. Incidentally, it was this test that allowed me to discover that CBR is not really constant, either, although the bitrate changes were relatively small and apparently random (no discernible reason why one scene would be different to another).

This inspired me to look at a couple of "Hollywood blockbusters" (original pressed disks) under the bitrate viewer and what I saw there contrasted with my homegrown DVDs. The average bitrate was only about 4500 kbit/s, but the fast scenes went up to 9000 and the slow ones down to 2500 and the bitrates changed largely even within the scenes. The transitions were typically about average or slightly above. Knowing that the very high bitrates were verboten with DVD¡ÓR and some players, I copied one short (single layer) commercial disc and, sure enough, it would not play in my Grundig player, where the original did. The copy did play in the Philips player, though, although there were a few minor glitches and one chapter went slightly out of synch.

Hope this info is of help.
Devil

P4 Core 2 Duo 2.6 GHz/Elite NVidia NF650iSLIT-A/2 Gb dual channel FSB 1333 MHz/Gainward NVidia 7300/2 x 80 Gb, 1 x 300 Gb, 1 x 200 Gb/DVCAM DRV-1000P drive/ Pan NV-DX1&-DX100/MSP8/WS2/PI11/C3D etc.
User avatar
Devil
 
Posts: 3522
Joined: Fri Mar 18, 2005 8:06 am
Location: Cyprus

Re: Bitrates for DVDs

Postby GeorgeW » Mon Mar 03, 2008 12:54 pm

Devil wrote:My conclusion was that, using a DV-compressed original, there was little to be gained by using bitrates higher than 6000 kbit/s CBR for the video. This applies to practical viewing of the DVD on a 32" LCD or a 28" CRT TV from 3 DVD players, as well. This point of view has been disputed by a member of this forum, who claims that the quality visibly improves at higher bitrates.


Devil, I believe I am aforementioned colleague :)

Just to be clear, I have never disputed your findings. I have always asked you to qualify your findings. For instance, I always asked you to mention the following (and I note you have started to include some of these items) :
-dv camcorder being used (not all DV camcorders are created equal)
-screen size you are viewing
-viewing distance
-PAL vs. NTSC (I mentioned colorspace differences between PAL and NTSC, and that might contribute to my findings being different that yours)
-encoder used (i.e. not all encoders are created equal)
-settings in encoder used -- speed vs. quality, frame seek, motion search, etc...


Now, with all the TV's that folks will be buying in the states to meet the eventual cutover in 2009 -- 40", 46", 52", 60"+ HDTV's -- some will start to see the differences between 6,000kbps and 7,000kbps encoded mpeg2. I can clearly see the differences on my 60" HDTV upscaled by a Toshiba HD-DVD Player (which does a great job upscaling SD-DVD's).

Regards,
George
User avatar
GeorgeW
 
Posts: 3262
Joined: Sat Dec 11, 2004 5:25 am

Postby Ron P. » Mon Mar 03, 2008 1:36 pm

George,

Unless I misunderstand this "cutover in 2009", the only people that may be required to purchase new TVs are the ones using a antennae to receive their signal. TV stations will be dropping the analog signal transmission, thus those antenna will no longer be able to pick it up.

However other than some folks living in rural areas, most now use cable TV, or Satellite TV. Cable and Satellite transmit digital signals, not analog. So they don't have to do anything to continue receiving their programming. The ones that still want to use the old TV antenna, can get a converter box. But there's no mandatory call to go out and purchase a new TV.
Ron Petersen, Web Board Administrator
User avatar
Ron P.
Site Admin
 
Posts: 12197
Joined: Tue May 10, 2005 12:45 am
Location: Kansas, USA
operating_system: Windows 10
System_Drive: C
32bit or 64bit: 64 Bit
motherboard: Hewlett-Packard 2AF3 1.0
processor: 3.40 gigahertz Intel Core i7-4770
ram: 16GB
Video Card: NVIDIA GeForce GTX 645
sound_card: NVIDIA High Definition Audio
Hard_Drive_Capacity: 4TB
Monitor/Display Make & Model: 1-HP 27" IPS, 1-Sanyo 21" TV/Monitor
Corel programs: VS5,8.9,10-X5,PSP9-X8,CDGS-9,X4,Painter

Postby GeorgeW » Mon Mar 03, 2008 2:07 pm

Ron,

I agree that not all folks will have to buy new TV's. My point is that many more households will have larger screen TV's come 2009...

Regards,
George
User avatar
GeorgeW
 
Posts: 3262
Joined: Sat Dec 11, 2004 5:25 am

Postby Devil » Mon Mar 03, 2008 2:30 pm

George

Your belief was quite right :D I didn't wish to mention you by name, to avoid introducing any polemic into the thread.

I appreciate your questions.

The camera used was arguably the best mini-DV prosumer 3-CCD Panasonic ever made, their first one, the NV-DX1EN. This did not have DV in or out, so I used the Sony DRV-1000P DVCAM drive to read the tape. This cam was made before Pana started using head-shrinkers to make their cams ever-smaller. It has 3 x 1/3" CCDs. I also have its successor, the NV-DX100, which is MUCH smaller, has 3 x 1/4" CCDs, but is not as good as the DX1EN and can hardly be classed as prosumer but high-end consumer.

Of course, the encoder used was that shipped with MSP7, the MainConcept, which is what I used then for the tests. The "quality setting" was always 95% (this is my standard setting).

The viewing distance for my tests were both close to the screen for detailed scrutiny and at normal viewing distance of about 2 m. Of course, only the latter is of real interest because if an artefact is invisible at normal viewing distance during ordinary viewing, it is of no consequence. I did the scrutiny first to be able to identify when they happened, to be as critical as possible at the normal viewing distance.

Since changing the TV, I have not done such critical viewing of test items, although I have looked closely at the results of some parts of the same scenes. I have a problem with the new TV, a JVC 32" LCD, in that it upscales all SD input via the Scart plugs to 720p. This not only improves the pics, it eliminates a lot of artefacts. For example, the artefacts round the bill of the flying Yellow-billed Stork in the samples referred to above are smoothed out in the upscaling process and the bill is sharper. I therefore conclude that upscaling is a no-no for the critical detection of artefacts. I cannot switch off upscaling with this model. The only time I've seen artefacts with this TV during ordinary viewing is when there is a thin (less than 3 vertical pixels) line within about 15¢X from the horizontal (e.g., a violin string): this "staircases" horribly (probably the upscaling does not attempt to change it, so that the "steps" are probably 1/576th the height of the active screen). I have no idea how the upscaling works, but it is certainly impressive.

I think that all your other points are dealt with in mmy earlier post.
Devil

P4 Core 2 Duo 2.6 GHz/Elite NVidia NF650iSLIT-A/2 Gb dual channel FSB 1333 MHz/Gainward NVidia 7300/2 x 80 Gb, 1 x 300 Gb, 1 x 200 Gb/DVCAM DRV-1000P drive/ Pan NV-DX1&-DX100/MSP8/WS2/PI11/C3D etc.
User avatar
Devil
 
Posts: 3522
Joined: Fri Mar 18, 2005 8:06 am
Location: Cyprus

Postby GeorgeW » Mon Mar 03, 2008 4:50 pm

Devil wrote:The camera used was arguably the best mini-DV prosumer 3-CCD Panasonic ever made, their first one, the NV-DX1EN. This did not have DV in or out, so I used the Sony DRV-1000P DVCAM drive to read the tape. This cam was made before Pana started using head-shrinkers to make their cams ever-smaller. It has 3 x 1/3" CCDs. I also have its successor, the NV-DX100, which is MUCH smaller, has 3 x 1/4" CCDs, but is not as good as the DX1EN and can hardly be classed as prosumer but high-end consumer.


And that is my point exactly -- your results are based on high-quality dv source. Shoot the same video with a low-end miniDV/digital8 camcorder, and encode that at 6,000kbps vs. 7,000kbps...

My point is there are too many variables to make a blanket statement using a DV-compressed original, there was little to be gained by using bitrates higher than 6000 kbit/s CBR for the video.

Regards,
George
User avatar
GeorgeW
 
Posts: 3262
Joined: Sat Dec 11, 2004 5:25 am

Postby Devil » Mon Mar 03, 2008 5:34 pm

George

I'm not sure I follow you here. If I don't get significant losses when encoding MPEG from a good DV source, why should I get more from a less good DV source? DV is DV is DV is DV, as Gertrude Stein might have said. A low quality cam may have a poor lens, small CCDs (some have 1/6" now!!!), i.e. a poor analogue signal, but the DV compression is the same as with the best one; in fact, it is invariable for mini-DV, with no choice of parameters. I understand that there are only a couple of makes of encoder chips for cams of all marques and price ranges. I agree with you that software encoding, especially MPEG, can be very variable in quality, but not the DV hardware encoding.
Devil

P4 Core 2 Duo 2.6 GHz/Elite NVidia NF650iSLIT-A/2 Gb dual channel FSB 1333 MHz/Gainward NVidia 7300/2 x 80 Gb, 1 x 300 Gb, 1 x 200 Gb/DVCAM DRV-1000P drive/ Pan NV-DX1&-DX100/MSP8/WS2/PI11/C3D etc.
User avatar
Devil
 
Posts: 3522
Joined: Fri Mar 18, 2005 8:06 am
Location: Cyprus

Postby GeorgeW » Mon Mar 03, 2008 6:13 pm

Brian,

The point is that different folks will have different starting-quality dv source. Was it shot in low light (where the 3ccd would win over the 1/6" CCD), was it an analog-to-dv conversion, what type of optics were used, any extra lenses, any compositing in the edit stage, any titles or transitions, etc... So, in that sense, dv is not dv is not dv. The resulting attributes might be the same (i.e. the attributes of a dv video file), but the material compressed to those attributes can vary greatly. In other words, 6000kbps CBR could be good for YOUR footage, done in your lighting, in PAL shot on 3 CCD cam encoded with Ulead mpeg encoder in MSP7 at 95% Quality viewed on 32" LCD from "x" feet away, without compositing and/or transitions or added titles, etc...

I'm glad you have decided to at least include some of the "variables" when posting your findings...

Regards,
George
User avatar
GeorgeW
 
Posts: 3262
Joined: Sat Dec 11, 2004 5:25 am


Return to VIDEO & DVD

Who is online

Users browsing this forum: No registered users and 5 guests