Why not Firewire ?

[This_Idiot]

Hi Grigory,

It's ages since we've heard from you - I'm glad that it's because you are getting good results after all the struggles, and not because you're fed up with us

Nice to see you back

Chris

[Grigory]

Hi,

Anatoly:
"DV captured clips are dark and bloodless on computer monitor."

Yes, it is. Don't you think that it is because in DV you have the same digits as on the tape, and therefore no way to adjust the brightness during the capture?
In analog captures, you always get from DV another digits. So, your better looking digital video, produced by capturing from analog output of digital camcorder is just another way to apply the filter itself, and nothing more?

All PC monitors are different from TV. They have different gamma coefficient, well suited for computer screens. The dark parts of picture are always too dark on monitor compared to TV. It is easy to see by making simple grey bars and cpmparing how they look on TV and on PC screen. Different monitors are also working different. Compare say Sony and Viewsonic, and Siemens, add here comparison between trinitron and other tubes. What type of display do you prefer for mpeg1? Sony trinitron monitor, or non-trinitron models?

Now about analog captures. Yes, you can adjust levels on input and and output to get simultaneously:
a) better look of captured video on PC monitor (incorrect levels, but looking better)
b) correct output for TV (by correcting incorrect levels on analog output adjustment)

Well, using analog signals capture and output you can feel yourself comfortable comparing PC monitor and TV screen.
For this you are loosing signal to noise ratio, introduce errors in ADC and DAC, and loose at least 20% of dynamic range for those level and contrast adjustments.

Believe me, if you capture in DV, make mpeg1 or mpeg2 from that DV, and then look on PC monitor and on TV screen while playing mpeg on hardware decoder, you will see the same level difference. If you adjust the levels to get better picture on PC screen, you'll get too bright image on TV. It is not a DV fault, it is a difference between PC monitor and TV.

With DV only, you are sometimes limited to TV screen while editing. But you can always be sure that you capture without distortions, you do not need to adjust input levels (a nightmare for some scenes and with RR_G) to escape oversaturation or even green-flashing frames.

About formats. DV has either 4:2:0 or 4:1:1.
For PAL, 4:2:0 is matching 4:2:0 of MPEG 1 or 2. There is minimal possible loss: most good encoders take YUV video of 4:2:0 format from DV decoder and use it directly, without conversion. For NTSC, there is conversion, but only one time 4:1:1 to 4:2:0.
Do not speak about HDTV 4:2:2 mpeg2: it is not practical as final format.
I am absolutely sure that nobody can expect better quality from 4:2:0 or 4:1:1 converted to 4:2:2 by any filter. It is impossible. You cannot ADD additional samples to chroma.

As for analog "4:2:2" format, typical analog output uses something like 4:1:1 (6 to 1.5 MHz bandwidth for luma and chroma in PAL) format, so you inevitably loose 1/2 of chroma resoluion. Yes, you get doubled chroma samples, but the original analog signal does not provide you independent samples.
More, 4:2:2 MJPEG format with equal data rate has higher compression ratio because of higher uncompressed data rate. DV format compresses much better, so you cannot expect lower artefact level from mjpeg. Of course, if, in the case of PAL, you actually have 4:1:0 format after DV-analog-yuv conversion (taking into account that you get one color value per four in line, and two lines on afield with the same chroma values), you actually have less "informative" data to compress.

Software filtering. I can agree that converting from DV to mpeg1 may require filtering and level adjustment. Both actions can be easily performed by VirtualDub or AVIsynth. The compression time overhead can be up to 100% in the case of very fast Cinemacraft encoder, and 1-5% for Tmpeg. Is it a real problem?
I do not adjust the levels. Filtering for SIF size movie is well done by VirtualDub deinterlace-blend fields, and later precize bilinear resize to SIF size. The same is possible with AVIsynth script.
For mpeg2 I use the frame width from 352 to 720, typically 352, 384, 480 or 720, depending on the movie content and required datarate. Resizing and filtering are done by
avisource("d:\movie.avi")
crop (16,16,688,544)
separatefields
blur(0.15)
bilinearresize(368,272)
weave
addborders(8,16,8,16)

This eliminates unnecessary content outside the visible on TV area, smoothes a bit horizontal edges and provides very good quality at reasonable data rate. For talking head video it is possible to put one hour on 700 MB CD. The quality is better than VHS.

DV is better. The explanation why there are less topics concerning DV is simple - we in DV have less problems. This explains why, for example, I stopped writing here frequently.
Do I have problems? Yes, I have hard drive space limitations. I can capture entire 14 GB of DV on my drive, and next hour, and next... You know, the drive space is always used , regardless how many you have.
Another is that precise DV tape automatic indexing can take the same time as capturing the tape. By the way, can you recapture in analog exactly the same piece of video, frame in frame. In DV, you can.

You can like the look of analog video on screen more than DV, but if you count how many times you had to re-capture with analog, and how many useful features are impossible for analog captures and count your time, you will see that DV is better.

Grigory

[mynx]

dchip: MPEG2 and DVD use the same lossy compression as DV. Discrete Cosine Transform (DCT) is used in all current digital video.

Professional DV camera's do record 4:2:2. They decided that consumers don't need this amount of color. All camcorder CCD's read 4:2:2. It gets down converted to 4:1:1 just after the image is captured.

[Grigory]

Anatoly,

Your approach is useful, but I doubt that it helps to get the best possible quality of mpeg. The most questionable for me is the compression quality, because original DV must be definitely better as mpeg source. I agree that it might require different filtering, but the level of compression artefacts is the lowest possible, because DCT is used only once instead of two times for DV-analog-mjpeg conversions. If you use yuy captures, you probably can get good quality due to filtering of sharp edges on analog instance of video.

mynx:
The analog video bandwidth is even lower than equivalent of 4:1:1. Even S-video connection is limited by current standards. Most TV sets also have chroma bandwidth limitations. You can easily check this by making color bars on grey background so that the luma is not changing. Look at these bars on TV and see that you don't have more than ~100 bars per screen.
For PAL 4.43 and for D/K band it is possible to keep approximately 4:1 luma to chroma sampling ratio along the line, but remember the delay line that also reduces the resolution by a factor of 1.5 in vertical direction, and that reduction takes place on per field basis.
With DV, you can get as high chroma resolution as the camcorder can record.
CCD sensors produce analog signals. These signals are processed to get one luma and two chroma difference analog signals. Then all three are digitized. The sampling rate for luma is x4 higher than for chroma components. To reduce the noise, analog signals are filtered to match the digitizer sampling rates.
2:1 sampling ratio is used in very expensive professional models, but the reason is to keep as much as possible color resolution for chroma keying and similar effects on production stage. On transmission stage that resolution is reduced to what we really have in consumer DV devices. Remember: satellite digital video transmission uses mpeg2, and it has 4:2:0 color format.



SECAM:
SECAM is color transmission format for video broadcasting. It is supported by the tuner.
The sound to video frequency shift is 6.5 MHz in formal Soviet Union. It is called D for meter band and K for decimeter band, or D/K. Western Europe uses 5.5 MHz called B/G.

Matrox uses tuner receiver that cannot tune to sound carrier in "SU domain". This choice helped to save 5 cent for production cost, and made the RR_G tuner useless for 250,000,000 potential market. Not too big loss, however. 5 cent per card is better.
Most TV tuner cards sold in SU do have similar Philips tuner box, but it supports D/K. You can buy one of these cards and ... capture in yuy2 in exactly the same way as with RR_G
You can dig at http://ixbt.stack.net and find an article how to replace the filter for sound. This operation requires soldering and you will lose the warranty.

Grigory

[This message has been edited by Grigory (edited 19 December 2000).]

[mynx]

Grigory: I know that the digital signal has more bandwidth than even S-video, I was responding to the mention of the lossy compression.I know the CCD is analog, but the associated IC converts it to Digital 4:2:2. It is then down converted to 4:1:1 by a second IC. It has been this way for years. Analog camcorders use the same process as digital ones. The DV format can record the full 4:2:2. The format is a standard. The extra information simply isn't needed for consumers. Thank you for your version, being a D8, Mini-DV and DVD technician my training was a lot more in depth. I however left out most of the technical info to make it more friendly to read.