I'm watching an episode of The Simpsons on a local channel right now, and it has a different look than usual, which is particularly noticeable when the characters or camera moves. I can't quite describe it. Is it possible that it was converted from PAL to NTSC? ________________________________________________________________
If love of money is the root of all evil, why do churches want it so badly?
Remove "bination" to reply.
Drunk and full of pie, Sydney Assbasket wrote:
If so, you'd notice a lot of "frame-jumping". NTSC is 29.97 frames per second, while PAL is 25 frames per second. You almost have to re-sample the PAL version... meaning show it, and record it from the screen.
(I've gone through lots of frustration trying to convert between the two. It's nearly impossible to do with software, anyway, without resampling the file as if it were a source. Some dual-deck VCRs will do the job, and some other specialized equipment. Too much hassle).
"The Rack" wrote...
it
the
that it
How does this work? Do they have both Pal and NTSC masters?
Here in Argentina we use PAL (well, PAL-N). Fox broadcasts in NTSC, and the cable company converts the signal to PAL. It looks pretty crappy. But the open channel which has The Simpsons also broadcasts in PAL, and the picture quality is much, much better.
Cheers,
Drunk and full of pie, Nicol�s Di Candia wrote:
A good question that we'll have to ask... FOX. I've never seen a good, quality tape that's been converted from PAL to NTSC; but, about a year ago I looked into some equipment that does the conversion. Expensive as hell (like 10 grand). There were certain websites advertising NTSC to PAL/PAL to NTSC conversion for a fee.
I believe the actual process of conversion is called "telecine" / "inverse telecine". (Well... that actually involves going from TV to movies and vice-versa. (24 fps to around 30 fps)). But the process is similar.
If you're in America, definitely not. It's possible your station is using one of those loathsome "dynamic time compression" boxes from Prime Image Corporation. I've read that if they're cranked up too high, they'll mix a lot of frames and cause a visual effect similar to what video format conversion looks like. More info at: http://abcnews.go.com/sections/scitech/TechTV/techtv_trimtv_020124.html
That may have been done in the dark ages of television. :-) But format conversion has been an electronic process for many decades. And there's no single way about it: the more the hardware costs, the more complex its conversion process will be, yielding better results. With the real $$ equipment, no frame jumping should ever be visible -- all sorts of stuff like "temporal interpolation" and motion smearing is done along with the usual PAL color to NTSC color conversion and scan rate conversion. Just watch BBC America, or if your PBS station shows them, BBC comedies like "Keeping Up Appearances" or the Deutsche-Welle news service from Germany. That's all PAL to NTSC. All you really notice is that motion is blurred (like on vertical credits scrolled), and that the video doesn't have the same "raw, live" appearance analog NTSC video tape has.
Believe it or not, but television is still broadcast in black and white, and PAL and NTSC are just the names of two different methods for sending the color information separately from the main picture. (Think of it like TV sound, where you have normal stereo, plus the SAP channel that can only be detected by SAP-enabled sets.) Color TV was designed that way so old B&W sets could continue operating after color broadcasts began. (So B&W sets see the same old black and white signals as always, while color sets know where find and extract the hidden color from the signal. Then they mix it into the B&W picture and display the colorized result. I guess it should call it colorized TV instead of color TV, then.) If you wanna see what this looks like out of some excessively weird curiosity, take a look at <http://www.snpp.com/staff/brian/ntsc.html> for an example.
So really, the least that needs to be done to convert video is re-encode the color information and then change the scan rate of the B&W video. I guess systems that leave out all the extra processing are the ones which produce the frame jumping you refer to.
You're thinking of kinescope. Telecine is from film to video; kinescope is from video to film.
Fox? Only NTSC. They send NTSC dubs overseas and let the receiving end deal with converting the episodes to PAL or SECAM on their own equipment.
now _That's_ ... Interesting!
btw i'm actually Not being sarcastic.
In article , "Brian Petersen" wrote:
That's not quite true. In this instance, PAL and NTSC are more than just mechanisms for encoding color data- they are two distinct formats. PAL is a 50 Hz progressive signal with 625 scanlines, a 4.43 MHz subcarrier frequency, and uses the CCIR TV standard, while NTSC is a 59.94 Hz interlaced signal with 525 scanlines, a 3.58 MHz subcarrier frequency, and uses the EIA TV standard. The encoding of other data, such as stereo channels and subtitling are also handled differently. Only a handful of countries use 60 Hz PAL.
(Think of it
The biggest problem with converting NTSC to PAL is converting the 59.94 Hz signal to a 50 Hz signal.
Nope. <g> PAL/NTSC really are just the methods by which color encoding works, and are entirely unrelated to the black and white video and its number of scan lines, fields per second, etc. (NTSC defines color as a phase- and amplitude-modulated sine wave, where the phase determines hue and amplitude dictates saturation. PAL defines color almost identically, except that phase is alternated 180 degrees between lines.)
Just look at Brazil's television system for example. They use PAL color encoding but with the field rate and scanline count typical of NTSC areas (525/60). On the other hand, it's perfectly possible to create a 625/50 system which uses NTSC color encoding. It's just that nobody has done so because there is no reason to. NTSC color is inferior to PAL color.
Regarding stereo, the encoding method is also unrelated to NTSC/PAL. In broadcast signals, audio is placed on its very own carrier (not in the main video signal), where the stereo encoding standard can be decided on case-by-case bases. For example, Japan and the United States are both NTSC/525/59.94 Hz countries, but MTS stereo encoding is used in the USA, and Matrix encoding in Japan. Sweden, Switzerland, and Uruguay are all PAL/625/50 Hz countries, but Uruguay uses MTS, Sweden uses Nicam stereo encoding, and Switzerland uses no encoding -- I believe they just have two mono FM signals per TV channel, one for Left, the other for Right.
You aren't wrong about there being standards which specifically link one form of color encoding to specific black and white parameters -- such as "NTSC-M," which means AM-modulated negative black and white video with 525 scanlines at 59.93904761905 fields per second plus NTSC color added to the B&W video carrier at 3.579545 MHz. :-) All I said originally was that NTSC and PAL themselves are strictly color schemes which have no functional dependence on the other parameters of what today are still black and white television broadcasts.
(Is everyone getting this? <g>)
On the other hand, maybe you can help as far as EIA and CCIR. Being an American, I know little of CCIR, but to my knowledge, there are actually several EIA standards, each defining different aspects of American video. E.g. one defining the 525/59.94 standard, another defining which scanline closed-captioning should be placed on (and in what format), etc. Do you have any links to white papers or other techy/nerd papers on them? I've tried the usual Google searches, but only seem to come up with results that mention "EIA," not copies of the EIA standards documents themselves.
In article , Brian Petersen wrote:
As long as you have no follow up questions, yes. Yes we are.
Can I ask this question of someone who might know: Between HDTV and the internet, will NTSC, PAL, SECAM etc. all go away in favor of a single worldwide TV standard, or will incompatibilities still exist?
David Z Houston TX
How were these images made? Can this be done with Adobe Photoshop? ________________________________________________________________
If love of money is the root of all evil, why do churches want it so badly?
Remove "bination" to reply.
Yes. I just used PhotoShop to re-create what NTSC color looks like all by itself. It's very simple.
(1) Open an image. (2) Duplicate it into another window (Image > Calculate > Duplicate). (3) Invert the duplicate image (Image > Map > Invert). (4) Un-invert the duplicate's color (Image > Adjust > Hue Saturation, and then slide the Hue control -180 or +180 -- it doesn't matter). (5) Paste the duplicate over the original at exactly 50% opacity.
What this does is create a second copy with polar opposite brightness, but unmodified color, as compared with the original. By blending the two 50/50, you cancel out the luminance and leave only the color. The result is exactly what NTSC color looks like minus the black and white signal it's ultimately added to for display.
Oh, yeah. And if you want the black and white portion as well, you must re-open the file in a new window and change it's mode to Greyscale and then back to RGB to get it. (Don't just use the Hue/Saturation box to reduce saturation to 0, since it doesn't produce the same result.) Then, to combine the color-only image and the B&W image back into a full-color picture, paste the color-only image over the B&W image at 100% opacity but with PhotoShop's brushes box set to "Color" mode.
"Brian Petersen" wrote...
Very interesting. One thing that I don't get: how does this work with all the different formats? I'm currently under the impression that black & white TV doesn't have NTSC or PAL. If they send a black & white signal with the 625 or 525 lines of the local format which is then combined into the final frame, shouldn't that affect the reception on B&W sets?
Cheers,
Correct.
Yes, it would affect reception, but not destroy it. The main B&W picture uses a huge amount of bandwidth (up to 4.2 MHz IIRC) while the NTSC and PAL color signals are mere sine waves (no bandwidth) added at 3.58 or 4.43 MHz (respectively) to the B&W picture's signal.
NTSC/PAL color subcarriers damage the B&W signal by creating a "beating" effect -- a floating "grid" or "screen" of dots covering the whole picture. Color TVs include circuitry to minimize or eliminate the interference, but old B&W sets -- which were never manufactured with the anticipation that a color signal would one day be added to broadcasts -- did not have any such corrective circuitry. So you saw the beating and simply put up with it. (I believe that B&W sets manufactured after NTSC came out had band-pass filters added to them, preventing them from seeing the B&W waveforms above 3.58 MHz. Which would have prevented the sets from displaying finer detail in pictures, but which also would have eliminated the beating.)
On the other hand, the B&W signal itself also causes damage to the NTSC/PAL color subcarriers. Depending on the B&W picture's resolution, more or less bandwidth is used. (The waveforms for a low-res picture might go as high as 2.5 MHz, while those for a sharper one might extend up to 4.2 MHz.) Any parts of the picture sharp enough to reach 3.58 MHz destroy their own color (the waveforms at 3.58 MHz literally disrupt the color sine wave also found at 3.58 MHz). The visual result of this is a rainbow effect that shimmers over the high-resolution portions of the picture. For example, here is a screen capture of a "multiburst" test pattern, which is used to test the resolution of analog signals. As you can see, once the lines reach a point where they're fine enough (high enough to reach 3.58 MHz), an oily rainbow pattern appears. <http://www.snpp.com/staff/brian/multiburst.jpg> (In real life, the rainbow colors would be much more vibrant -- the frame grab tended to kill them off. But you get the idea.)
Drunk and full of pie, Brian Petersen wrote:
Maybe I shouldn't have been so glib. The point is, you're dealing with a set of discrete samples (the frames), not a continuous wave. You most certainly have to resample to do format conversion, whether it's done by hardware or not. Any continuous signal representing a wave had to be constructed from samples. It's this reconstructed wave that is sampled (resampled), not a pristene, continuous video signal. Such a thing does not exist in electronics. IOW, there's no way to have an infinite number of frames. These signals may be present in nature (photons bouncing off everything), but they cannot be exactly duplicated in electronics. Even The Simpsons could be considered a "sample", since even if we saw Gracie's masters, we wouldn't see a continuous video image--we'd see 24fps :)
"Motion smearing" involves alpha-blending frames between two boundaries by an ever increasing percentage.
Consider that to convert from 30 fps video to 25 fps video, you'll need to decimate one out of every six frames (25/30 -> 5/6). That means you skip every sixth frame in the original 30fps file, and play back the remaining frames at 25fps...
If that were all there was to it, it would be simple. This is where we have the frame jumps, however. What if, for instance, you skip the only frame where a character's mouth is open? It will look very odd when played back. This is where the alpha-blending comes in.
Conceptually, we would multiply the 30 fps video by 5/6 (we want 5/6 of the frames). Thus, we'd raise the number of frames by a factor of 5. Then, between each of the two regular frames, we would put four "tween" frames. The first would consist of 80% of the original image and 20% of the second, the next would be 60%/40%, 40%/60%, and 20%/80%. Once this is done for the whole set of samples (the whole movie/show), then we would take one out of every six frames to represent the new 25fps version.
All we're really doing: (for a six frame sequence)
Frame 1 (100% Original Frame 1) Frame 2 (80% Original Frame 1, 20% Original Frame 2), Frame 3 (60% Original Frame 1, 40% Original Frame 2), Frame 4 (40% Original Frame 1, 60% Original Frame 2), Frame 5 (20% Original Frame 1, 80% Original Frame 2), (Skip Frame 2) Frame 6 (100% Original Frame 3) <-- notice skipped 6th frame
Therefore, we're decreasing/increasing alpha-blending by 1/5th for each new frame (20%). So, all we need to do is change the alpha-blending by this factor for each frame in the sequence, and drop the last frmae. This is much easier than actually creating 5 new frames per frame!
If we wanted to apply this to actual 29.97 NTSC video to go from NTSC to PAL, we'd have to get a new ratio and start over. BTW, 29.97 is actually 1000000/33367 (29.96973)
Turns out the PAL to NTSC frame ratio rate is 33367/40000. Conceptually, we'd increase the number of frames by a factor of 33367(!), alpha blend, then sample 1 out of every 40000. WAAAY too much work. Since we know we're changing the alpha-blending by 1/33367 for each new frame in the sequence, this is much easier. Decrease the original/increase the next alpha-blend percentage by .000029969730572 for each frame. After 33367 frames, we skip one.
(Roughly, that looks like this)
Frame 1 (100%), Frame 2(99.99% Frame 2, 0.001% Frame 3), Frame 3(99.98% Frame 3, 0.002% Frame 4)
Obviously, this still takes time... but it's much more manageable. Also note that NTSC hardware may not play back at EXACTLY 100000/33367 frames per second; many software AVI samplers don't exactly sample at that rate either. In that case, Microsoft compatible AVI files have a "scale" field that (when divided into 1000000) gives the exact frame rate.
Alpha blending.
NTSC uses 256 colors, while PAL is 16-bit color.
Not exactly. The color information is separate, but the frame information is still a continuous wave RECONSTRUCTED from samples. You still must interpolate.
Well, the extra processing deals with the frame blending.
Realizing, of course, that cartoons are not created in NTSC ;-) The standard for animation in America is 24 fps. In Japan, for anim�, it is generally 10 fps, though some studios use 15 fps. All film (including cartoons) must be converted to NTSC for them to be shown on TV. So truthfully, every episode has already been resampled at least once... if not more... before we see it on TV. If stations use digital masters, they may be getting a first generation sample. If not, they video taped it, and resampled the resampled version.
HTH
Rack The Engineer
Drunk and full of pie, The Rack wrote:
In the U.S., that is :)
Drunk and full of pie, The Rack wrote:
Whoops. "1000000/33367"
You *really* misinterpreted me. :-) When you said "record it from the screen," the concept of somebody pointing an NTSC camera at a PAL set literally popped into my head. (Well, you *did* say "record it *from the screen,*" after all. <g>) That's why I said "dark ages" and then went on to explain that:
Which in turn was meant to refer to the process you described above, a la frames as "samples" dealt with as individual things. I never for a moment analogized video to a continuous wave. (Well, actually, it is -- it's one continuous scan line rendered into frames by display/processing devices according to appropriately-spaced sync pulses along it. But now I'm just splitting hairs.)
Thank you. =) That was pretty much how I conceptualized the process -- though the term "alpha blending" is new to me, anyway.
Where are you getting the "10000000/33367" figure from? I've never heard of that one before. The scan rate for NTSC 525 video is exactly 15734 Hz. There are 262.5 scanlines per field (525 scanlines per frame), and exactly 29.969523809525 frames per second (59.93904761905 fields per second). So:
29.969523809525 x 525 = 15734 Hz 59.93904761905 x 262.5 = 15734 Hz
I did a little Google surfing and only found 33367 mentioned as the frame rate in this paper about digital video capturing: http://egor.spb.ru/doc/BookCD/DrDobbs/articles/1992/9207/9207a/9207a.htm
That would be unknown territory for me, so perhaps the 33367/29.96973 figure applies to some applications of digital video to standard 525/29.96952 NTSC analog? (Certainly, given the negligeable difference between 29.96952 and 29.95973 frames per second, going between the two wouldn't even make any difference.) But nevertheless, I'm looking at this entire thread from the point of view of the normal NTSC analog video everyone has lived with for the last 40 odd years. And the figures above (15734 Hz and its products) all apply to that.
Are you trolling me? <g> That's plain ol' not true. NTSC's and PAL's hue and saturation attributes are infinite in their variability. They are, by definition of analog, mere sine waves, where amplitude (at least between 0 and whatever is defined as the maximum amplitude) may vary in a perfectly linear way to define saturation, and phase may vary on its axis in a linear fashion to any point between 0 and 359.99999999999999999999[...] degrees.
The only situation where the normally linear variability would be quantized into a specific "palette" (e.g. 256 colors) would be if analog NTSC or PAL color signals were processed digitally. So if for example a video camera employed digital signal processing (e.g. photons --> lens --> 8-bit DSP --> D/A --> NTSC), then the RESULT would be a 256 color analog NTSC signal, as the source would have been a 256 color digital palette. But that doesn't mean the capabilities of NTSC and PAL themselves are limited 8- and 16-bit color; only that any DSP through which they are sent would create a sort of "bottleneck" reducing infinity to a numerically defined colorspace.
In fact, as far as 8-bit color, the only mass digital video application I'm aware of which limits color to a 256 palette is DirecTV. (Or maybe that was when they were on the MPEG 1.5 standard -- I can't honestly recall.) Either way, they do (or did) use 256 colors with some dithering algorythm to hide the appearance of gradients, so the video wouldn't look overly "posterized."
I'm starting to wonder -- is most of your knowledge on this subject based around digital platforms, or analog? Mine is in the latter. So maybe, for all I know, it's the standard to employ 256 color palettes in many digital "NTSC" video applications, explaining why you've asserted 256/16-bit for NTSC/PAL. 'Cause I assure you that that isn't true at all of the NTSC and PAL systems themselves.
That was implied by "change the scan rate of the B&W." <g> There wouldn't be any point to altering the scan rate of a signal (which really just means different sync pulses, given the speed of light/voltage is constant whether you're in America or the UK <g>) if you weren't going to also read and re- generate the visible picture content. :-)
Yes. Which is why I said previously that the better systems employ extra techniques like motion smear, et al.
No arguements. All I was trying to do was introduce "kinescope" into the vocabulary section of this thread, since "inverse telecine" sounded kind of funny. :-)
Tchyeah, apparently so! :-) Who do you toil for? (Uh, what kinda job?)
Drunk and full of pie (not really), Brian Petersen wrote:
My fault, actually. The "record from the screen" thing was an oversimplification, done as kind of a joke. Won't happen again :)
True, but you're just compositing frames... the *signal* is continuous, but it was reconstructed from samples. They're not called "samples" in the analog world, of course :) ("Moving pictures?")
Alpha-blending is a generic term meant to discribe blending transparency channels in digital images.
It's just a convenient ratio for NTSC digital video. Microsoft's AVI format specifies a "rate" and a "scale" field, which are set to 1000000 and 33367 respectively for NTSC AVI files. Note that dividing 29.969523809525 into 1 million gives you 33367.230202108754638434198679824. It's just rounding error.
Yes, the figures correspond to digital video. You're certainly right about the analog stuff. In this digital age, NTSC->PAL/PAL->NTSC in the digital realm makes sense. My post was just to show a practical way to convert NTSC video to PAL video. Given that much of TV is digital nowadays, it seemed appropriate.
No :)
Well, you answered the question there :) Digital video will never be capable of infinite color variation. It's become clear that we're talking about the differences between analog/digital video.
Well.. digital :) I've never studied analog video signals, truthfully.
I probably shouldn't have said "not exactly", since you said "the LEAST that needs to be done"...
Sorry; it sounded to me like you were talking about the frame rate. Of course, simply cutting frames and playing everything back slower doesn't work too well... as I pointed out.
That'd be nutty, huh?
That term's actually used by the popular shareware program "VirtualDub" (www.virtualdub.net)
It's all a lie :D A facetious comment, given the content of the post. I studied digital video signals, sampling, resampling, interpolation, decimation (etc) for a digital video editor application I'm writing (BEGAN to write). Most of the information comes from a book called "Understanding Digital Signal Processing" by Richard Lyons. It began as an application that could stretch/compress time in digital sound files, and ballooned into a digital video project.
Won't all of this be moot when HDTV obseletes our TV sets in 4 years? ;-)
The Post Office, actually--but thanks for thinking I could have a job as an engineer. (How about those two misinterpreted jokes? Hey; I've got an engineer's sense of humor!)
To end it all... let's point out a mistake in my previous post: where it says to drop each 33367th frame when resampling, it should say the 40000th.
On Sat, 09 Mar 2002 01:29:10 GMT, [email] (The Rack) wrote:
The only thing scarier is an accountant's sense of humour... where "Good Will Hunting" isn't the name of a movie, but a joke in itself...
[AcK!'s head continues to spin]
TTYL