Cinemacraft DV to DVD Template

[newbie33]

Where can I get sample templates for Cinemacraft to convert captured DV AVI to DVD compliant mpeg?

Thanks.

[DJRumpy]

There isn't much to do. Ensure your using the Bottom Field First. Use the MPEG Standard matrix.

Make sure under the VIDEO section that you select 'Alternate' for the scanning order. This assumes of course that your video is interlaced. The rest of the options are noise filters and such.

I would definately suggest mutipass VBR for the encoding method, rather than CBR. DV from a cam tends to have a lot of motion if it's a hand held.

[BJ_M]

yep i agree with DJ , there are also specific matrixs you can use with DV -- why this is see here: http://ricksz.swsites.net/DVinternals.html

this is the resulting matrix (to be used with interlaced DV material ONLY) , setting 6000or more) avg VBR / 9000 max / 1500 min ..

 8,13,13,17,17,21,21,28,
13,13,17,17,19,21,23,30,
13,17,19,19,21,23,28,34,
13,17,19,19,21,23,28,48,
17,19,19,19,23,28,34,48,
19,19,23,25,28,32,34,48,
19,21,23,25,28,32,34,48,
21,21,25,25,28,32,34,48

 8,11,11,15,15,17,17,24,
11,11,15,15,17,17,21,24,
13,15,15,17,21,21,26,34,
13,17,15,17,21,21,26,48,
17,21,21,23,21,30,34,48,
17,21,21,23,28,30,34,48,
19,19,25,26,28,30,48,48,
19,19,25,26,28,30,48,48

[tphaggerty]

All of that is fine EXCEPT if you are using CCE Basic. Then, you can't set Top/Bottom field encoding and as far as I know you can't change the quant. matrixes. (At least I can't find the options to change this stuff).

Using CCE Basic, I have found DV AVI encodes to be very soft and "flickery". All of the posts you will find indicate this is a problem with Top Field encoding, which CCE does by default, again, as far as I can find there is no way to change this in CCE Basic.

[BJ_M]

cce basic has a setting for field order .. 0 is top and 1 is bottom , or you can use a even or odd number likewise ...

you have to enter matrix tables (i think) in your registry - not sure ,i dont use basic ..

[tphaggerty]

There is a field that has 0 or 1 called First Line or something like that. Doesn't seem to make a difference, but I'll have to try it again. It says that 1 should be the default for DV AVI (which is Bottom Field First), which is what I left it at, but it still seems to encode Top Field First (at least according to BitWise and MF2).

This wouldn't be a problem if it just threw away the first field, but according to posts on doom9, it apparently rearranges the fields and then sets the TFF flag in the MPEG file.

I really like the speed of CCE and actually the interface is quite good once you get used to it, but all of my DV AVI encodes are very soft so far, to the point of being fuzzy (compared to TMPGenc, ProCoder and MainConcept, all of which I have as well).

[newbie33]

Thank you for your replies. I am using CCE SP. I do not see any place
to set the Field Order. Is there any registry key that I can set directly?

I am using 2 Pass VBR. It is giving results better than CBR, although CBR is three times faster. Even if I use 2 Pass VBR, I still cannot get good video quality at 5Mbps. At 8Mbps 2 Pass VBR, I am getting reasonable quality, but I can still see the deterioration. I moved the Simple Video Setting from Easy to Difficult which modifies the filters, but I did not see any difference.

I left the luminance level set at 16-235? Should I change it to 0-255?

IDCT precision is at 10, I think this should be better than 8, although this may depend to the decoder.

I also left the M=3, N/M=5. Alternate scan order is selected.

I have experimented with both Standard and MPEG Standard quantization matrices, but I did not see any difference. I will try the DV matrix.

To capture the DV, I am using the IEEE firewire card. Some people suggest using different DV codecs. I think I only have the DirectShow Microsoft DV decoder. Does it really make any difference? How can I force CCE to use another DV decoder? It was giving me error to update my qdv.dll file. I think this is the DV decoder.

What I am trying to achieve is one of the the following which I am not sure possible:
1) Encode the DV at 8Mbps 2 Pass VBR with no visible video quality loss 2) Encode the DV at 8Mbps CBR with very minimal distortion
3) Encode the DV at 5Mbps 2 Pass VBR with very minimal distortion

Thanks.

[BJ_M]

Thank you for your replies. I am using CCE SP. I do not see any place
to set the Field Order. Is there any registry key that I can set directly?

what version of CCE SP are you using ? upgrade to the latest. and you can set field order correctly .

I am using 2 Pass VBR. It is giving results better than CBR, although CBR is three times faster. Even if I use 2 Pass VBR, I still cannot get good video quality at 5Mbps. At 8Mbps 2 Pass VBR, I am getting reasonable quality, but I can still see the deterioration. I moved the Simple Video Setting from Easy to Difficult which modifies the filters, but I did not see any difference.

well there is a difference --

I left the luminance level set at 16-235? Should I change it to 0-255?

if you are going to watch it on a tv but the source was DV .. DV is already 16-235 so set it 0-255 so it doesnt clamp it further .. if it looks to bright or to dark on your tv -- set it to 16-235 (im not sure what else you are doing)

IDCT precision is at 10, I think this should be better than 8, although this may depend to the decoder.

though many think otherwise -- 9 is the correct setting for DVD based on the bit rate used ..

I also left the M=3, N/M=5. Alternate scan order is selected.

it can be set to 3 and 4 , its DV so it is interlaced, why are you selecting progressive ? the source is 29.97i so setting P here will make the picture look jerky on a tv.

I have experimented with both Standard and MPEG Standard quantization matrices, but I did not see any difference. I will try the DV matrix.

To capture the DV, I am using the IEEE firewire card. Some people suggest using different DV codecs. I think I only have the DirectShow Microsoft DV decoder. Does it really make any difference? How can I force CCE to use another DV decoder? It was giving me error to update my qdv.dll file. I think this is the DV decoder.

when you capture it doesnt mater WHAT codec you have on your system as firewire capture is NOT capture but a digital transfer ... (how many times this has been repeated - it should be a sticky)..

it doesnt mater which DV codec CCE uses for its encoding, the DV codec only mater if you are re-rendering DV on your system ..


DV is a lossy format and so is Mpeg2 .. some loss will always occur transcoding from one format to another .. ..

[DJRumpy]

BJ_M, I hadn't heard that about the DC Component Precision setting. I've always assumed more precise = better. Of course you know what assumptions can get you

Can you point me to your refernce? I'm interested in the info.

[newbie33]

I used the DV quantization matrix, but I have not seen any improvement.

I downloaded the latest demo CCE SP 2.67, but it does not have the Upper Field First option either.

Could you look at the registry:
[HKEY_CURRENT_USER\Software\Custom Technology\CCE SP\template\0]

and let me know which key corresponds to Upper Field First option?

Thanks.

[BJ_M]

I used the DV quantization matrix, but I have not seen any improvement.

I downloaded the latest demo CCE SP 2.67, but it does not have the Upper Field First option either.

Could you look at the registry:
[HKEY_CURRENT_USER\Software\Custom Technology\CCE SP\template\0]

and let me know which key corresponds to Upper Field First option?

Thanks.

its right in the video settings , look for a 0 near the bottom and change it to a 1 ..

[DJRumpy]

It's under the VIDEO button. The option is labled "Upper Field First".

[BJ_M]

BJ_M, I hadn't heard that about the DC Component Precision setting. I've always assumed more precise = better. Of course you know what assumptions can get you

Can you point me to your refernce? I'm interested in the info.

Thanks.

see http://forum.mainconcept.com/viewtopic.php?t=79 regarding low bit rate dvd's and as the bit rate goes up to the dvd max (say over 6000 avg) , 9 is more suitable.

10 could in theory be used for a CBR 9800 encode , but that situation is pretty rare .

the people at main concept are extremly knowing about mpeg standards.

there are other ref. to this around also on the web but i first got the info from a lecture at a compresionist meeting ..

[BJ_M]

It's under the VIDEO button. The option is labled "Upper Field First".

its not labled that way anymore .. you just enter a odd or even number representing first field .. a 0 for top and 1 for bottom

[DJRumpy]

Hmm. I'm using 2.66 SP. I thought I was on 2.67. I can't believe they change the interface on a point release. Any idea if they fixed that old bug regarding top field first when your source was actually bottom field?

[BJ_M]

they call it offset line and yes its fixed - well its still a little odd on how they do it but it works so what can i say .

at least it doesnt do oppisite of what you expect

[newbie33]

I set the Offset Line to 1, but I still get Top Field First in the mpeg stream.
Are they cutting out the first line with this option to simulate top field first?
That does not seem right to me.

[BJ_M]

I set the Offset Line to 1, but I still get Top Field First in the mpeg stream.
Are they cutting out the first line with this option to simulate top field first?
That does not seem right to me.

yes its weird as i said, but it works correctly and if you step through the fields (only real way to tell correctly) you will see it is rendering correctly ..

i had to check this myself as it is different than before but no mater what cce puts top field first in stream .. ..


it has no bearing on playback on a dvd player (what the stream says for field order) .. it just decodes in chunks anyway .

[newbie33]

It looks like the quality is as good as I can get from CCE.

I am curious, do you know what tools do holywood studios
use to compress the DVDs?

[BJ_M]

we have spruce encoders , sonic 2000 encoders , CCE SP, optibase encoders and some others -
we also have HD encoders ...

[DJRumpy]

newbie33, can you give us the settings you are using? List each under Video, Audio, Etc (We don't need audio though..I list it only as an example).

Also, what software are you using to verify the field order of your output? Output that is wrong will be very obvious. Does your output video appear to jerk or shudder on fast motion? Wrong field order will not cause a 'softening' effect.

You should be avoiding the 1-click options under the Quality button. They will override the other options selected. Turn off all of the clickboxes. Set your Quantizer characteristics to somewhere between 16-32. Set your scanning order to Alternate. Ensure the 'Progressive' checkbox is unchecked.

BJ_M, do you have the actual info from the FAQ regarding the DC Precision setting? This suggested it might affect quality, but doesn't state why.

[BJ_M]

ill see if i can dig it up , in the meantime some good reading on the subject is here:

http://viswiz.gmd.de/DVP/Public/deliv/deliv.211/mpeg/mpeghome.htm

http://bmrc.berkeley.edu/frame/research/mpeg/mpeg2faq.html

http://www.siggraph.org/education/materials/HyperGraph/video/mpeg/mpegfaq/mpeggeneral-1.html

[DJRumpy]

Please do. Searches on Google didn't yield much, but I did find an interesting link: http://www.digvid.info/digitalvideo/mpeg.php
Good reading for mpeg buffs (I should get another life...).

Some points of interest:

DC values in MPEG-1 are limited to 8-bit precision ( a constant step size of 8 ), whereas JPEG DC precision can occupy all possible 11-bits. MPEG-2, however, re-introduced extra DC precision.

This reads as if MPEG-1 was originally spec'd to produce only 8-bit precision anyway, which would support that a setting of 8 would be appropriate for low bitrate sources. It does not indicate that high bitrate could benefit from a higher setting. It also does not indicate that a low bitrate could not benefit from a higher setting either {*sigh*}. The impression I get from this, is that a higher DC component setting increases the data in the stream, due to lower compression, but with an increase in quality (assuming there is enough bitrate available). If this assumption is true, then using a higher DC component setting should be fine, as long as your not 'bit-starved' in your mpeg. A low bitrate MPEG-1 (say VCD), which is very bit-starved would actually benefit from a lower DC Component precision (better compression).

MPEG video frames are broken into blocks of 8x8 pixels which are DCT processed and quantized as outlined above. Blocks are combined into macroblocks of 16x16 or 16x8 (MPEG-2 only) pixels.

This was interesting. I was under the impression that even MPEG-2 used 16x16 macroblocks, not 16x8. This would kill the old reasoning that all resolutions should be divided by 16, as the vertical could be divided by 8 without issue.

It also has a very interesting section on Audio compression. I wasn't aware that MPEG audio used temporal compression, and or frequency masking. I though it only filtered out freqencies above hearing, and when the bitrate was set low enough, progressivly into the hearing range. According to this, it also removes freqencies cancelled out by other (louder) frequencies. It also removes audio which the human ear cannot hear due to recovery time (A loud pop, followed by a soft noise, which can't be heard because the eardrum has not recovered from the louder of the two noises).

I'm going to keep digging, but please do post the FAQ if you can find it. I would think that DVD, having a broad range of bitrate available, would benefit from a higher DC precision. Again, I'm guessing here, as I haven't got all the facts yet. I'll post more if I can find it.

[newbie33]

Thank you for your time on this issue. I used ReStream to find the field order. When I change the order with TMPGEnc, it is showing it correclty.

Here are the settings I use for CCE
-----------------------------------------------------
Video Setting:

Mode: MPEG-2 ES, Multipass VBR
Bitrate: Avg: 5000 Min: 2000 Max: 9000
[Avg: 8000 is better but still has distortion]
Passes: 2 [tried 3 pass no difference]
3:2 pulldown detection: (None)
Letter box hint: (None)
Panscan: (None)
Aspect Ratio: 4:3

MPEG Video setting:
GOP sequence M=3 N/M=5 [tried N/M=4, I think it was worse]
GOP Header every 1xN frames
SEQ Header every 1 GOP
Add sequence end code: Checked
DVD video compliant: Checked [tried (None), did not see any difference]
DVD Multiangle: (None)
Close all GOPS: (None)
Restrict auto I-frame insertion: (None)
Equalize Each GOPs bit lenght: (None)
Half Horizontal Resolution: (None)
Half vertical resolution: (None)
Make Room for USCC: (None)
Luminance Level: 16 to 235 [tried 0-255, did not notice any difference]
Offset line: 1
Frame rate: 29.97
Quantization Matrices: MPEG standard [tried Standard, DV matrices, no difference]
Packet size: 2048

Quality Setting:
Natural picture: (None)
Computer Graphics: (None)
Animation: (None)
Low Pass Filter: (None)
Effect restricted vertical filter: (None)
Dithered quantization: (None)
Quantizer characteristics: 16
Intra block DC precision: 10 [tried 9 no difference]
Block scanning order: Alternate
Progressive frame flag: (None)

[DJRumpy]

Your settings all look fine. Other than lowering your MIN, and raising your MAX, I wouldn't change anything. You should utilize all of the bitrate available to you. Try a MIN of 0, and a MAX of 9800. Chances are it will not drop that low, but the option is there if the encoder needs it.

I would stick with the DV matrix they've suggested. The DVD Compliant checkbox will do nothing if your input is already DVD compliant. I tend to leave it unchecked so any inconsistancies will show up in my output, letting me know I have a bad input somewhere.

The luminance level setting is still something I'm not happy with yet. Adam will tell you to leave it at 0-255. I still have encodes where the luminance scale looks compressed using this setting. I've never noticed it using 16-235. I understand the logic behind the 0-255 setting, but I can't understand why my output doesn't exhibit the compression artificact on the 16-235 setting. I'd suggesty you try both to decide which you like. Look at dark areas of a scene to see if they have a 'fog' or 'glitter' of gray squares appearing in them. If they do, try the other setting.

For DVD Video, Cinemacraft suggests you use a DC Component precision of 10.

What resolution was your input DV file?

[BJ_M]

not to sound argumentitive - but the reason we never set min values to 0 on most encoding sessions is that very very dark (over 75-80% black content scenes) can "trick" the encoder into going to low and causing these scenes to look bad .. a moving starfield is a good case in point -- lots of pinpoint lights moving quickly on a black background (why also filtering should be off for that type of scene) ..

M=3 N/M=5 [tried N/M=4, I think it was worse]

im suprised you feel that 3/4 looked worse .. that is odd indeed

[newbie33]

DV resolution is 720x480 30 frames per second.
The original DV looks good, I have problem with
the compressed MPEG clip. I did not edit the DV
files, they are exactly the same as the raw firewire
capture. I used the Windows Movie Maker to capture the
DV.

[hrpinman1]

Hello, hoping to get some confirmation from all you expert CCE and TMPG folks......
I've been capturing using my Sony DV camcorder into my computer's Firewire port. Once I have the AVI files, I have experimented with both TMPG and CCE 2.66 to encode. My question has to with the FIELD ORDER. TMPG seems to be fine when I specify bottom field first. CCE seems to work well (i.e. not jerky) only when I specify TOP field first. From what I've read here in this forum, DV is supposed to be BOTTOM field first. So can someone confirm that basically CCE has the field order thing REVERSED, or am I missing something in the CCE settings? Thanks much for your help.

[DJRumpy]

BJ_M. No worries. To make a point though, even a black frame requires bitrate. On top of the video bitrate, there is also the audio bitrate, which is CBR for MPEG-1 audio. Figure at 1 second clip will have at least one key frame, possibly two, plus the audio bitrate. It should never drop to 0 in theory.

hrpinman1, I think someone else mentioned that no matter what, CCE will output top field first. I'm beginning to wonder if it doesn't auto detect it anyway. If your output looks good, then I wouldn't worry about it. It's obviously handling it without issue. I wasn't aware that the top field bug was still in versions as new as 2.66 though.