... that given the present bandwidth limits on a US TV channel, 441 line
images would yield higher resolution than the standard 512 lines.
Assumptions:
������������
4.0 MHz video (baseband) bandwidth.
Monochrome (like RS-170) signal.
30 Hz frame rate, interlaced.
To simplify interlace, an odd number of lines.
To simplify the synch dividers ("step counters"), a line count that is
the product of small integers.
Aspect ratio 1s 4:3
RS-170 Horizontal and vertical blanking intervals are set at 10% of the
time of a scan time. At 441 lines, the same time is a smaller fraction,
but I keep 10% for simplicity. I will round the numbers freely.
Line counts.
������������
525 is 3 x 5 x 5 x 7. 441 is 3 x 3 x 7 x 7. 525/441 = 25/21 ~= 5/4
Resolution with 525 lines.
��������������������������
The horizontal scan frequency is 30 Hz x 525 = 15,750 Hz. The time of a
scan is 1/15750 Hz = 63.5 microseconds, of which 90%, 57, is visible.
4 MHz x 57 microseconds = 228 cycles/line, horizontal resolution.
To sample this frequency, we need 456 samples horizontally. The vertical
signal is already sampled by the raster. Of the 525 lines, 472 are
visible, corresponding to a vertical spatial frequency of 236 cycles in
the vertical direction. Because of the 4:3 aspect ratio, only 171 cycles
are needed in the vertical to match the horizontal. The vertical
dimension is oversampled.
Resolution with 441 lines.
��������������������������
The horizontal scan frequency becomes 30 Hz x 441 = 13230 Hz. The time
of a scan is 1/3230 Hz = 75.6 microseconds, of which 90%, 68, is
visible. 4 MHz x 68 microseconds = 272 cycles/line, horizontal
resolution.
The vertical resolution is .9*441/2 = 198 cycles. To match the vertical
resolution, we need 3/4 x272 = 204. The vertical is undersampled, but
not by much. (It is still better than what a 525 line system needs.)
The horizontal resolution went from 228 cycles to 272, better than 19%
more.
Why?
����
RCA proposed a 441-line standard before 1940, and was sure enough of
its merit to go ahead with commercial production of transmitters and
receivers. Their argument was that on the small CRTs then available,
visible line structure was acceptable, and that techniques were
available to render the scan lines in larger tubes invisible.
<sidebar>
If the vertical intensity profile of the beam is a raised cosine of
width zero to zero is two lines, the raster disappears. A raised cos(f)
is also {cos(f/2)}^2 shifted and scaled. Relative to it, the adjacent
scan profile is {sin(f/2)}^2 similarly shifted and scaled. Their sum is
everywhere constant in the region of overlap. The necessary spot profile
is easily simulated by a small high-frequency exciter winding on the
vertical part of the yoke, a procedure known as "spot wobble".
<\sidebar>
So why did the FCC overrule the engineers and mandate an inferior
system? Ignorance of DSP fundamentals accounts for much of the reason,
and politics for a little. An important part was basic misunderstanding
of sampling in general and rasters in particular.
Jerry
Disclaimer: If I made a fundamental error in my thinking, I hope to be
quickly corrected. Corrections from all are welcome. I will consider
Alan Herriman's judgment definitive, but I may argue with others. :-)
--
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������
Be it resolved ...
Started by ●March 17, 2004
Reply by ●March 17, 20042004-03-17
"Jerry Avins" <jya@ieee.org> wrote in message news:4058698A.4010303@ieee.org...> ... that given the present bandwidth limits on a US TV channel, 441 line > images would yield higher resolution than the standard 512 lines.Jerry, There is no argument with your arithmetic. But your claim might be better stated: "... in the same bandwidth it would have been possible to increase horizontal resolution at the expense of vertical resolution for a better match of those two resolutions." That you said: "would yield higher resolution" just isn't substantiated because you reduced the vertical resolution. Dithering won't bring it back. That it's "matched" doesn't make it "higher". What you didn't talk about were the "psychooptics" (I guess ... I take it freely from psychoacoustics). Is human perception more sensitive to vertical structure in a horizontal scan system? Is human perception more sensitive to the discrete nature of the lines than the lack of resolution in the horizontal? Well, if you've ever pondered video for quality time after time then you'd know that "softness" (lack of horizontal bandwidth) is much harder to see than the line structure. Thus the commercial efforts to create line doublers, triplers and quadruplers with no hope of improving *resolution* all that much. In signal processing terms they're interpolators. In viewing terms they make a huge difference! For that matter, why bother to interpolate anything? One reason is because the data is a lot easier to look at. Video interpolation has a benefit for the same reason. So, if you subject either scheme to a zone plate then you might be able to extract technical measures. But I just don't believe that reducing vertical resolution in order to improve horizontal resolution in a horizontal scan viewing system is all that great an idea. I have no idea what the historical steps and missteps might have been. Fred
Reply by ●March 17, 20042004-03-17
Jerry Avins <jya@ieee.org> writes:> ... that given the present bandwidth limits on a US TV channel, 441 line > images would yield higher resolution than the standard 512 lines. > [...] > > Disclaimer: If I made a fundamental error in my thinking, I hope to be > quickly corrected. Corrections from all are welcome. I will consider > Alan Herriman's judgment definitive, but I may argue with others. :-)You need to research "The Kell Factor." The Kell Factor reduces the vertical detail in any raster system by some amount, typically 70% for interlaced systems. There are many factors included in the Kell Factor and I recommend the "Television Engineering Handbook" edited by K. Blair Benson (now deceased) and Jerry Whitaker. In NTSC there are usually 480 active scan lines (some are not visible at the top and bottom, but let's ignore that for now). If we apply the typical Kell Factor of 0.7 to 480 lines, we get 336 lines or 168 line pairs. Your picture bandwidth is also low at 4 MHz. Typically the figure 4.2 MHz is used, but the real limit is the FM audio at 4.5 MHz. I seem to recall that the broadcast limit is that video must be 30 dB down by 4.5 MHz, but my references are not at hand.
Reply by ●March 17, 20042004-03-17
Jerry Avins wrote:> To sample this frequency, we need 456 samples horizontally. The vertical > signal is already sampled by the raster. Of the 525 lines, 472 are > visible, corresponding to a vertical spatial frequency of 236 cycles in > the vertical direction. Because of the 4:3 aspect ratio, only 171 cycles > are needed in the vertical to match the horizontal. The vertical > dimension is oversampled.Who said TV has to have square pixels? Normally is not, 720x576 (601 for 50Hz) is not square pixels, neither 720x480 (601 for 60Hz). So you can have mismatches between resolution and aspect ratio. bye, -- Piergiorgio Sartor
Reply by ●March 17, 20042004-03-17
Fred Marshall wrote:> "Jerry Avins" <jya@ieee.org> wrote in message > news:4058698A.4010303@ieee.org... > >>... that given the present bandwidth limits on a US TV channel, 441 line >>images would yield higher resolution than the standard 512 lines. > > > Jerry, > > There is no argument with your arithmetic. But your claim might be better > stated: > "... in the same bandwidth it would have been possible to increase > horizontal resolution at the expense of vertical resolution for a better > match of those two resolutions."I'll buy that.> That you said: "would yield higher resolution" just isn't substantiated > because you reduced the vertical resolution. Dithering won't bring it back. > That it's "matched" doesn't make it "higher".I implicitly took "resolution" to be lowest resolution, discounting the benefit of resolution along one axis unmatched along the other.> What you didn't talk about were the "psychooptics" (I guess ... I take it > freely from psychoacoustics). Is human perception more sensitive to > vertical structure in a horizontal scan system? Is human perception more > sensitive to the discrete nature of the lines than the lack of resolution in > the horizontal?I don't know> Well, if you've ever pondered video for quality time after time then you'd > know that "softness" (lack of horizontal bandwidth) is much harder to see > than the line structure. Thus the commercial efforts to create line > doublers, triplers and quadruplers with no hope of improving *resolution* > all that much. In signal processing terms they're interpolators. In > viewing terms they make a huge difference!Line structure is an artifact made visible by poor design. I have seen displays where it is completely absent. Surprisingly, the resolution of NTSC video exceeds that of black-and-white 45 mm movie images. I can show you pictures, but scanning halftones of large-format photos of CRTs may leave much to be desired. I lent Otto Schade's book to jj when he graced us with his presence. I can lend it to you too. Interpolation, if you want to do it, is best done in the image tube by collecting for each line an integrated line's width of information. (Some solid-state cameras do this by making each pixel square. Others reduce pixel height to make room for shifting circuitry.) Then with Gaussian or raised-cosine beam-intensity profiles, the raster disappears and the averaging is as good as can be had.> For that matter, why bother to interpolate anything? One reason is because > the data is a lot easier to look at. Video interpolation has a benefit for > the same reason. > > So, if you subject either scheme to a zone plate then you might be able to > extract technical measures. But I just don't believe that reducing vertical > resolution in order to improve horizontal resolution in a horizontal scan > viewing system is all that great an idea. I have no idea what the > historical steps and missteps might have been.Do you always favor vertical resolution? You could trade away more horizontal for vertical by using 735 lines. It seems to me that trading to equalize horizontal and vertical resolution would give best performance. The fly in the ointment is visible line structure, but that's an artifact of inadequate design.> FredJerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●March 17, 20042004-03-17
Piergiorgio Sartor wrote:> Jerry Avins wrote: > >> To sample this frequency, we need 456 samples horizontally. The vertical >> signal is already sampled by the raster. Of the 525 lines, 472 are >> visible, corresponding to a vertical spatial frequency of 236 cycles in >> the vertical direction. Because of the 4:3 aspect ratio, only 171 cycles >> are needed in the vertical to match the horizontal. The vertical >> dimension is oversampled. > > > Who said TV has to have square pixels? > > Normally is not, 720x576 (601 for 50Hz) is not > square pixels, neither 720x480 (601 for 60Hz). > > So you can have mismatches between resolution > and aspect ratio. > > bye,There are often sound engineering reasons for such mismatched resolutions. I imagine that those reasons usually arise from a need to compromise; to trade one good for another. Do you see a widespread use for anisotropic photographic film, that is, film with a grain size that varied with direction? When my eyes exhibit anisotropic resolution, I know that I need a new prescription for astigmatism. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●March 17, 20042004-03-17
bko-no-spam-please@ieee.org wrote:> Jerry Avins <jya@ieee.org> writes: > > >>... that given the present bandwidth limits on a US TV channel, 441 line >>images would yield higher resolution than the standard 512 lines. >>[...] >> >>Disclaimer: If I made a fundamental error in my thinking, I hope to be >>quickly corrected. Corrections from all are welcome. I will consider >>Alan Herriman's judgment definitive, but I may argue with others. :-) > > > You need to research "The Kell Factor." The Kell Factor reduces the > vertical detail in any raster system by some amount, typically 70% for > interlaced systems. There are many factors included in the Kell Factor > and I recommend the "Television Engineering Handbook" edited by K. Blair > Benson (now deceased) and Jerry Whitaker. > > In NTSC there are usually 480 active scan lines (some are not visible at > the top and bottom, but let's ignore that for now). If we apply the > typical Kell Factor of 0.7 to 480 lines, we get 336 lines or 168 line > pairs. > > Your picture bandwidth is also low at 4 MHz. Typically the figure 4.2 > MHz is used, but the real limit is the FM audio at 4.5 MHz. I seem to > recall that the broadcast limit is that video must be 30 dB down by 4.5 > MHz, but my references are not at hand.I know that quoted vertical resolution of cameras is almost always lower than what I would expect -- I sneakily withheld that bit in my roll as jolly troll -- but my understanding of the Kell factor (as best I can remember what Otto Schade tried to tell me) is that it is related to sinc rolloff because of the sampling, and can be (in some mysterious way) compensated. I chose 4.0 MHz as the effective upper video limit knowing that a bit more was possible but knowing also that considerably less is common. In the days when I fixed TV s and had access to RCA's library of repair manuals, many of the IF alignment procedures were lucky to achieve 3.5. Some cheap sets from Madman Muntz had IF bandwidths around 2.3 MHz. The sad thing is not that anyone would perpetrate such a fraud on the public, but that viewers didn't care. Back then, sets had manual fine tuning, and few bothered to use it. (It used to drive me crazy!) Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●March 17, 20042004-03-17
In article <4058698A.4010303@ieee.org>, Jerry Avins <jya@ieee.org> wrote:>... that given the present bandwidth limits on a US TV channel, 441 line >images would yield higher resolution than the standard 512 lines.Scanned CRT's are not square pixel devices. In the chroma-subsampled interlaced NTSC format, the vertical and horizontal resolutions are not directly comparable, especially if there is any color or motion in the image sequences. Also, the human eye-brain combination may not have the same vertical/horizontal sensitivity to changes in certain kinds of motion and texture. And history, politics and business issues had as much or more to do with the choice in line format as any technical issues. IMHO. YMMV. -- Ron Nicholson rhn AT nicholson DOT com http://www.nicholson.com/rhn/ #include <canonical.disclaimer> // only my own opinions, etc.
Reply by ●March 17, 20042004-03-17
"Jerry Avins" <jya@ieee.org> wrote in message news:40588a79$0$2832>> Do you always favor vertical resolution? You could trade away more > horizontal for vertical by using 735 lines. It seems to me that trading > to equalize horizontal and vertical resolution would give best > performance. The fly in the ointment is visible line structure, but > that's an artifact of inadequate design. > > > Jerry > -- > Engineering is the art of making what you want from things you can get. > �����������������������������������������������������������������������Jerry, Someone said: "Engineering is the art of making what you want from things you can get." I find it hard to imagine that visible line structure is an artifact of inadequate design - and, further, allows the suggestion that even coarser vertical lines would be preferable. Line doublers often use 1/4 1/2 1/4 interpolation in the vertical because it's economical and a pretty good approach. But, you're going to pay even for that. Are you suggesting that there's a simple and zero cost method to dither in the vertical? We *could have* had a system with fewer lines - but then what really? What scheme would have made those fewer lines acceptable in reference to the economics of solving the problem that fewer lines introduces? No, I don't favor better vertical resolution at the expense of horizontal. As it stands, the real experts can tell if video is "soft" in the horizontal - it's harder for me. So, I don't imagine that reducing the horizontal resolution is such a good idea. Nor do I imagine that reducing the vertical resolution is a good idea in view of all the commercial effort that continues in reducing the line structure. I think the psychooptics, the Kell factor, is the key to this discussion. I don't think it can be ignored. Equal resolution only works in the math but doesn't seem to work well in video practice. That's what I believe; I could be wrong. You say the choice made for the standard was different than what some proposed. How might one substantiate the claim that the choice made was a bad one? Equal resolution eludes me in making the case. Maybe I should read that book...! Fred
Reply by ●March 17, 20042004-03-17
Ronald H. Nicholson Jr. wrote:> In article <4058698A.4010303@ieee.org>, Jerry Avins <jya@ieee.org> wrote: > >>... that given the present bandwidth limits on a US TV channel, 441 line >>images would yield higher resolution than the standard 512 lines. > > > Scanned CRT's are not square pixel devices. In the chroma-subsampled > interlaced NTSC format, the vertical and horizontal resolutions are not > directly comparable, especially if there is any color or motion in the > image sequences. Also, the human eye-brain combination may not have > the same vertical/horizontal sensitivity to changes in certain kinds of > motion and texture. And history, politics and business issues had as > much or more to do with the choice in line format as any technical issues. > > > IMHO. YMMV.I believe your last sentence. I'm skeptical about asymmetric eye/brain response being significant, though. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
