Some question (as MAtrox did not answer ;(

[ccc]

Why 704/576 not 720/576 (MPEG2 std) , what will happen when converting to MPEG2 720/576 ? Black lines? Strechted screen ? ( ;( ) ?
Is it possible to make it possible to record at 720 in a future driver?

The new teletext program gives me a little window (at 1280*1024) , I 'd like to stretch it out... (is it possible? ) + Could it be possible to save the info to .txt file (+automatic recording...) ?

What's the difference between Ulead Video Studio and Media Studio, which is best?

[DrMordrid]

Any capure frame size is a comprimise in terms of the max supported data rate vs. cost vs. quality vs. the standard frame sizes.

720x480/576 is commonly used as it's the size defined in the specs as "full frame" video and therefore is the default sizes of DV and DVD/MPEG-2 video frames.

To lower the data rate for a consumer analog capture card, and thereby keep the costs down, the easiest things to do are to increase the compression ratio and reduce the size of each captured frame. Both of these lower the required data rate.

Lessening the width down to the next lowest "legal" value from 720 gives 704 (the width has to be a multiple of 16, 720-16=704). This while keeping the "legal" video heights of (240 and 480 for NTSC, 288 and 576 for PAL) minimizes the amount of data in each frame while maintaining vertical integrity, which is the more important of the two.

Taking the compression route the Matrox MJPeg codec supports 6.6:1 in 704x480. Comprimising in another direction some other devices use compressions as low as 3:1 or 4:1 with frame sizes of 640x480. For some purposes and eyes the lower compression seems better. To others the higher linear resolution of 704x480 is better. This is where things often roam into the subjective.

In any case it's all a series of comprimises. The "missing width" isn't really missed as it's well into the TV displays overscan area, that area of the image tube covered by the case. This area is not normally seen unless your TV is out of adjustment. In fact much of the width of a 704 wide frame will not be visible on your TV-Out for this same reason.

The quality difference due to compression also up for debate. A better TV-Output module often makes up for higher compressions.

As for what happens with the aspect ratio of your clips: no black lines when going from 704 to 720 width. As for major issues due to the stretching of a 704 frame to a 720 format...mathematically yes, practially no.


Comparing Uleads Media Studio Pro and Video Studio is easy, here is the breakdown:

Video Studio 3/4 (get 4 if you can):

An excellent entry level editor with DV and Ligos MPEG/MPEG-2 export support. Not too many fancy effects but it has all the basics and then some.

Media Studio Pro 5 (6 coming soon):

A semi-pro to pro level editor. Expandable with tons of effects, a rotoscoping program, titler, capture program, audio editor and it is VERY powerful. The choice of most folks here, especially those with Matrox cards.

and here's a freebie:

Premiere 5.1x:

Lots of features in a semi-pro to pro editor IF you can keep running. Premiere has a bad habit of being unstable with a lot of cards and on some systems. This is because of rather erratic device support outside of a very limited "approved" list, and often it's unstable with those cards ON the list.

The latest 5.1c update is much better but it's still a resource hog and somewhat unstable. Most of its problems are the result of a rather poor job of being ported from the Mac.

Dr. Mordrid


[This message has been edited by DrMordrid (edited 26 December 1999).]

[jeff b]

Doc: Isn't the aspect ratio for DV an overscan? It shows up on my TRV900 as such, since the extra part on both sides of the image is cropped out on the analog output to the TV. The 720x480 has been a bit of a mystery to me, since that's an aspect ratio of 3 to 2, as opposed to what I always thought NTSC was: 4 to 3.

I got used to the 704x480 overscan on the Rainbow Runner these past 2 years, and it made sense that they'd do that. But now with the mini-DV running at 720x480, I'm really just kinda confused as to why they did it that wide, only to crop it out on the analog output.

Maybe I'm just laboring under a long standing false assumption, though, in light of your post?

[JoakimC]

Whats wrong with 704x576 MPEG-2. The MPEG-2 MP@ML standard says that the maximum resolution is 720x576 not that it must be 720x576. The DVD specification allows for 704x576. You can encode the files as 704x576.

[DrMordrid]

JeffB

Almost anything over ~650 wide is well into the horizontal overscan area. Also the overscan isn't clipped by your TV, it's just hidden by the screen mask. What makes it necessary to have overscan at all is a multiple-choice answer.

Analog TV's don't hold their horizontal blanking adjustment forever. Components change their value over time. This can throw the HBI out of whack, causing the screen to shrink until it no longer completely covers the screen mask.

Also horizontal blanking variances can creep into VCR, camcorder, video game and cable signals that can cause the frame to appear narrow on your TV. Often the problem here is the manufacturers & content providers playing it fast & loose with the standards.

Without overscan these problems would cause a noticable border on the sides of your TV's display.

Some vertical overscan is also necessary to keep the vertical blanking interval (VBI) out of your TV's visible field. The VBI is that black band that shows if your picture rolls vertically. It's also where the closed captioning and telext information are located.

Macrovision can also cause problems in the form of bending or black level distortion at the top of the screen. Vertical overscan is used to hide this as well.

Call overscan a safety margin that will hopefully disappear once all TV's are digital and the need for it is gone.

Dr. Mordrid



[This message has been edited by DrMordrid (edited 27 December 1999).]

[ccc]

Thx for the answers
I'll try encoding mpeg2 704*576 then instead of 720

[Brian Ellis]

Hi!

I'd just like to a couple of Cyprus cents worth, here, if I may.

The number of frame lines is 625 or 525, with a 2:1 interlace, according to which system you use. A minimum of 8 lines (in practice, usually 15 to 20) is used for frame synch pulses which are about 20% blacker than black. These frame synch pulses are extended to contain a host of other things, such as the obligatory line synch pulses (the line time base must keep in synch even when no lines are visible, otherwise it will take an additional 30 or 40 lines before the picture becomes horizontally stable again), teletext and other information. These are repeated, of course, at the standard frame rate. There used to be a tolerance on the actual frame rate to allow it to be the same as the power supply frequency, to avoid hum lines creeping up or down the picture. This is now quartz fixed to very close tolerances and better smoothing techniques reduce hum lines. The total blanked-out vertical t

ime between the bottom of one field and the top of the next one is 0,07 V +/- 0,01 V (usually 37 - 40 lines but permissible from 32 to 42 with NTSC). With what we are doing, we lose about 5 lines of actual vertical picture signal, this being negligible. I believe the reason for this is to enable the compatibility between computer and TV requirements. These 5 lines are hidden behind the TV bezels, anyway.

Similarly, there is a blacker-than-black line synch pulse at a frequency of 15 625 Hz (PAL), 15 734,264 (NTSC colour) or 15 750 (NTSC mono) with a tolerance of +/-0,02% when operated synchronously or 0,0001% non-synchronously. These pulses set the black threshold level with a step (jargon used is the front and back porch) before and after the actual synch pulse itself and also contain the colour burst signal which tells the receiver how to reconstitute the colour for that line. How you fill those lines is totally arbitrary and whether you have 704, 720 or any other value will not be noticeable in the end result as the average TV set cannot resolve this kind of signal! Believe it or not, the time from the end of one line to the start of the next one may be as long as 0,147 H or as much as 1/7 of the line length to accommodate the porches, the pulse and the colour burst.

Now these are all integral parts of the analogue TV signal, whether this be generated from an analogue camera or converted from a digital signal and there is no way to get round this.

Another factor to be considered is that the bandwidth required to resolve the theoretical ideal of identical luminance resolution in both axes plus the chrominance plus the sound is much higher than any system can handle. We are tied in to broadcasting standards to ensure compatibility with TV sets. The total bandwidth is theoretically limited to a maximum of 6 MHz (usually 5 or 5,5 MHz), bearing in mind that the sound sub carrier is frequency modulated with a deviation of 25 kHz at the top end of the envelope. This limits the signal-to-noise ratio, which is possibly more important, visually, than the actual resolution. The theoretical maximum s/n ratio at 5 MHz is 52 dB and 56 dB at 4 MHz. In commercial and even pro TV monitors, the bandwidth is restricted to give a noise-free picture. This is why you will be lucky to get even 300 equivalent-lines horizontally on a commercial TV set, perhaps 400 on a professional monitor. So 704 or 720 pixels (352 or 360 equivalent-lines) is very largely theoretical, no matter what you do.

I agree that there must be wide tolerances within individual TV sets and other devices but these will not disappear with 100% digital TV (ask anyone who views digital satellite signals: the digital receiver is no better than anything else, often worse)

Hope this helps more than it confuses

------------------
Brian (the terrible)



[This message has been edited by Brian Ellis (edited 27 December 1999).]

[ccc]

Brian Ellis:
So for TV it is no use capturing at 704 ?
360 would do the job ?
Or am I fault... ?

PS: Digital 8 has a greater resolution (DV std 720...) but i am capturing via the Analog input of the marvel not digital thus...
Is the bandwith than high enough to let the 720 lines through and is it then better to use 704... ?

[jeff b]

Thanks Doc & Brian for the ascent into this technical info, as it's not always easy to get anyone to run with it in lay terms, at least enough to gain an understanding of what's going on with this stuff. Believe it or not, I supervise (among others) a video test area where I work, and I know how to repair our fiberoptic link cards for video. But for me it's more of a rote procedure, as opposed to a deep theoretical understanding of what's going on with it. I certainly couldn't design things like that!

As far as CCC's concerned, though, I think that there's one aspect of this whole thing that should be pointed out: the higher the digital resolution that you capture, edit, and output, then the better the final analog product will probably look.

I've always used 704x480 for projects on my Mystique220/Rainbow Runner Studio. Anything less has always shown up at the final stage of output back to tape looking not as good, even if I used a lower compression ratio up front. The smaller frame sizes always looked 'enlarged' and not as good.

As has been pointed out here, a little sliver of the edges will be cropped, once it gets back onto a TV, though. It's rarely going to be of any importance, unless you put a still picture into your project that fills the frame.

[jeff b]

Now for the MPEG-2 aspect issue...

If you capture at 704x576 PAL, and want to create a 'standard' MPEG-2 out of these clips at 720x576, it seems to me that you'll probably end up with a very slightly stretched image. It will be so slight, though, that you'll hardly notice it. I've created MPEG-1 projects at 320x240 from clips captured at 704x480 (NTSC), and the aspect change is so slight that it's not an issue.

[DrMordrid]

Agreed. That stretch or squash amounts to no more than a few percent.

I think that once PURE digital TV is a reality, which satellite and digital cable are currently not, then the overscan will be an anachronism.

For this to occur there will have to be a 1:1 relationship between pixels on the CCD in the cam and a solid state display with no analog signals in between to get out of alignment. Only then will overscan be unnecessary.

Dr. Mordrid




[This message has been edited by DrMordrid (edited 28 December 1999).]

[Brian Ellis]

ccc

No! 360 pixels will give you 180 equivalent-lines (alternate pixels black and white) which is well under the resolution of even the very cheapest TV.

If your TV is able to resolve, say, 330 horizontal equivalent-lines, you would need a MINIMUM of 660 pixels to profit from that. Anything over that would marginally improve the picture up to about 3 times that value, because there would be less dithering, but whether this would be a real enhancement of visual quality is not certain, in most cases. Certainly, I very much doubt whether anyone could say, looking at an image, that it used 704 or 720 pixels at a digital stage, so little would be the difference, even if you had the two images side-by-side on identical TVs.

Doc

Sorry, but even pure digital TV must break down an analogue signal into a digital format and then reconstitute it back to an analogue form. As an analogy, look at digital sound: the microphone is an analogue device. Its signal is analogically amplified and the goes into an analogue/digital converter before being master-taped. The resultant CD then produces a digital signal which goes through a digital/analogue converter and the resultant signal is passed through an analogue power amplifier to your loudspeakers. You may very well ask why change the format. There are four reasons: tape has a very finite signal-to-noise ratio of the order of 65 - 70 dB (this may be enhanced another few dB by special pre-emphases, such as Dolby); the frequency response of a slow tape is inadequate; tape has a total harmonic distortion in the region of 1 - 3% due to the non-linearity of the magnetic hysteresis curve and digital media can be made more robust against careless handling. So it is with video. Theoretically, there are two ways of reconstituting the analogue signal: 1) converting the whole analogue signal, as it is today, with synch pulses, black level porches and so on, into digital and 2) recording just the video data and allowing the rest to be reconstituted at the receiver end. Both are very feasible. 1) is the most reliable and cheapest method but will require a much greater bandwidth whereas 2) will require quite complex electronics to ensure that the locally generated parts of the signal are slotted into the transmitted signal correctly and, IMHO, this will produce just as much lost space as 1). I therefore cannot follow your reasoning why you state digital TV will eliminate overscan which is a function of analogue defects which will persist. Can you please elucidate?

Of course, the real MAJOR advantage of digital audio and TV is the ease of editing. An anecdote, as an illustration: I recorded a concert in a church in Lausanne about 30 years ago, using a Nagra. During the actual live concert, there was an adaptation of a Mozart clarinet concerto with organ and the clarinettist overblew a fortissimo note on the adagio, fortunately during a solo passage. After the concert was finished and the audience had left, I asked him to repeat the faulty phrase, comprising three bars and, with razor blade and Scotch tape, I was able to splice in the correct note where the false note was. This took me about four hour's work as it had to be done on the master tape, running at 15"/sec, and the spliced-in section had to be accurate in length and timing to better than 1 mm to ensure its imperceptibility, remembering that the background noise was different with and without audience. When the musician heard the result, he was amazed at what I had done. In reality, I started the splice about 1/10 of the way through the note to about 2 milliseconds before the overblowing started and ended it about 2 ms after, still within the note. There was a level difference of about 1,5 dB, but I let this slide as being imperceptible. If I had had modern digital techniques, this repair of a faulty note could have been done even better in less than two minutes.

------------------
Brian (the terrible)



[This message has been edited by Brian Ellis (edited 29 December 1999).]

[Grigory]

The resolution in lines is determined as a (number of black) + (number of white) lines at 30% contrast level drop due to resolution (bandwidth) limitation. This is typical -3db gain drop.

The number of lines is counted on a line width, equal to screen height, 3/4 of total.

Typical TV can show ~330 such lines, due to luma-chroma separation filter working as simple 3.5-4 MHz low-pass filter in most TV's.

720 pixels per full line give theoretically 540 lines of resolution, not 360. You cannot see then on normal TV. Another limit for TV is a number of color stripes on a screen. I could count 600 tri-lines on 29" TV. So, it simply cannot show 720 pixels at all!

Actual resolution of D8 for PAL, with better CDD design than for NTSC, is about 420 lines.

Half width video has ~240 lines of resolution, which is not enough for TV's. So, half width captures are enough only for VHS.

Another problem is in the block size on the screen. Decoder makes analog signal and this signal takes one length in mm per 8 pixel block for full resolution video, and x2 more in mm for half resolution. Any compression artefact become better visible for half width, just because its size is bigger and it is not "filtered" by TV itself. So, full resolution video has better quality at the same compression ratio.
704 vs 720 is not a difference - it is a task of decoder to produce correct aspect ratio on TV screen.

On my TV, at least 16 pixels from each border are not visible. So, I alsways substitute them by black border in bbmpeg encoder. For LSX, it is possible to make only vertical borders of 16 pixels each. This can reduce datarate by (576-32)/576 =0.944 factor.

Grigory

[Nick C]

Brain Ellis,
As far as the overscan bit goes--well, as far as I understand it at least:

It's not so much there because of analog-digital conversion and digital-analog conversion headaches, as a simple matter having to do with the fact that video information comes to present-day analog TVs with analog cables routed to all sorts of devices manufactured by countless companies that all have very much their own ideas on exactly how strictly video standards should be followed.
If you were to crack the TV's case and see the full frames it's getting, they'd be all over the place. Your RCA vcr might fill the frame, but your playstation has a black area. There might be odd bands above and below because there's so much variation in the analog standard signals that come to the set. the overscan is just like matting on a photograph--hide the ugly borders.
Digital TV, on the other hand wouldn't really need this, assuming it's accepting digital video. If a digital signal comes into the set, regardless of how it was converted to digital format in the first place, it would be taken as data, and converted to something the set can display, the difference is, the TV set is in control of the frame size. Since a TV manufacturer now has complete control of the exact amount of the screen will be filled, there's no need for an overscan.

[Brian Ellis]

Nick

Please, Brian, not Brain : flattery will get you nowhere

You are assuming that the digital signal produced by your various devices is identical. This will never be the case. If you have a digital satellite receiver, just look at the difference in the video signals from different stations, presumably all operating using the same MPEG2 standard. Nearer home, look at Sony never guaranteeing that a D8 tape produced on one camera will be playable on another one (although I suspect that this is probably wide mechanical tolerances). I'm not in the least convinced that overscan will not be necessary. Sorry!

------------------
Brian (the terrible)