On Wed, Sep 22, 2004 at 06:11:24PM +0200, Bernhard Praschinger wrote:
> Hallo
>
> When you have only 2MHz of bandwith, you cannot have a chessboard
> of white/black pixles and in each row 360 white annd 360 Black
> pixles for PAL. The voltage level has to raise and fall. And with
> 2MHZ of bandwith you don't have the slew rate to acomplish that.
Very true, but the biggest difference becomes that unlike the digital
world where you get pixelation from too low of a bandwidth, in the
analog world you'll instead get a smeared grey image, or a smeared
undulating grey image. One can argue which might be better, but that
is what will happen. The VCR simply can't reproduce the
checkerboard, and it becomes a grey smear.
> > > The best thing I have found by now is that the video bandwith
> > > of a VCR is limited to 2MHz. Normal PAL has 5MHz. So
> > > calcualting back this would mean that VCD "samples" every
> > > 0,125us one point, which is close to the halfe amount of
> > > samples that PAL (every 0,067us) has.
> > Except that for a VCR, there's no sampling either.
> So we have only voltage levels, that sound like a VCR is able to
> store a unlimited number of colors.
In theory, yes, an unlimited (well, limited only by the max amount of
voltage difference a TV set can reliably reproduce) number of colors.
Since it's directly storing the analog waveform as an analog
waveform, the only limit would be the ability of the
tape/head/processing circuits to accurately reproduce the appropriate
values.
> Now I would want to kown when I generate a certain color (with a
> pattern generator) with a very exact voltage level. How exact will
> the voltage be when the VCR reads the information from tape ?
Quite likely close, but not exactly the same as what went in.
Although for a constant, unvarying, color image, it will probably do
a pretty fair job. The limits come when you want to generate a
rapidly changing color along a line of the image. Above a certian
rate of change, the VCR circuits simply can't reproduce.
> BTW: That does mean that if we have only one color in the Video line it
> really makes sense to capture at the full resolution ;)
Well, in theory, if we really only have one color across the whole
line, a single color value and a tag saying "repeat this pixel for
the whole line" would suffice.
> > > So that would mean that VCR should be able to recognize and
> > > play back 360 different points per video line.
> > The number you derived is a number that was sometimes quoted for
> > theoretical resolution on a VCR, so if it were a digital device,
> > that would be a close approximation. However, other factors,
> > such as head wear, circuit drift, tape quality, can all impact
> > the amount of signal that is recorded on, and read from, the
> > tape.
> I know that things are always worse than the theortic calculation.
> Can you tell me or point to the information where the have defined
> these parameters ?
Here's what I can lay my hands on right now. There's probably also
info available by googling/yahooing around looking for info also.
This comes from a SAMS book entitled "VCR Troubleshooting & Repair,
Second Edition" by Rober C. Brenner and Gregory R Capelo:
format low high dev lum res
MHz Mhz Mhz TV Lines
VHS 3.4 4.4 1.0 240
SVHS 5.4 7.0 1.6 400
BetaI 3.5 4.8 1.3 250
Beta2/3 3.6 4.8 1.2 240
S-Beta 4.4 5.6 1.2 285
EDBeta 6.8 9.3 2.5 500
Where low is the lowest frequency output by the FM modulator,
corresponding to the sync pulse low point (blacker than black), and
high corresponds to the highest frequency output by the FM modulator,
corresponding to the max white level in the signal. "dev" is the FM
deviation, or high minus low, a larger deviation means the FM
modulator can has a higher bandwidth.
For the actual storage on the tape, the signals end up looking like
this:
VHS
Downconverted color subcarrier centered at .629Mhz, with a bandpass
from .129Mhz to 1.129Mhz. FM signal situated at 3.4 to 4.4MHz, with
the lower sideband lowpass filtered to limit it to starting just
above the 1.129Mhz level of the color subcarrier.
Beta
Downconverted color subcarrier centerd at .688Mhz, with a bandpass of
0.188 to 1.188Mhz. The FM signal situated between 3.5 and 4.8Mhz,
again with the lower sideband filtered to not interfear with the
upper end of the color subcarrier,
In both, the upper sideband of the FM signal is severely filtered
because the heads/tapes simply can not reproduce signals much above
about 5. The chart I'm getting these numbers from shows the -3dB
point of the upper sideband at probably somewhere around 4.8 or
4.9Mhz.
This does not hurt much though, because for an FM signal only one
sideband is necessary to reproduce the signal, but the filtering on
the lower sideband to prevent it from crosstalking with the color
signal further lessens the effective resolution that the machines are
capable of producing.
> > > The other thing is how exact they sample every point (6 or 8 Bit).
> > No sampling, no 6 or 8 bits. They don't work that way (well, at
> > least all the analog one's do not).
> But you will have in the analog world something like bandwith, and with
> a certain bandwith you can store a certain amount of information. It is
> a differnet question how exact you store that information. And read it
> again.
Quite correct, bandwidth is the limiting factor. However, I can't
find in the book a spot where the authors actually quote a real
bandwidth value for either kind of VCR. Somewhere around the 2.0Mhz
that you quoted is a very reasonable number, and given the 4.5Mhz or
5Mhz typically quoted for NTSC/PAL, that means a VCR's only recording
about half the effective resolution of the signal.
> Do you know how much bandwith is "lost" ? That is the information I'm
> seeking.
Unfortunately, the book I have says it's all low pass filtered, but
does not give the filtering amounts. But, consider that a 4.5 or
5Mhz signal goes in, and the machines have about a best case
bandwidth of about 2Mhz, so at least half the input signal is lost in
the process.
> I would say that they are saving money for themselve, because the cheper
> design costs them less. And leaves more money in their own bag.
That is also true. But in 1976, when the first Beta's were costing
several thousand dollars, had they cost several tens of thousands,
it's very likely noone would have bought one at all. It's not always
to put money in the manufacturer's pockets, it's often to cut corners
to produce something that will fall within a price range that
consumers are willing to fork over.
> The is one german saying when you measure things in the analog world
> that says: Wer misst, misst Mist. And it has ALWAYS been correct by now.
However, I speak no German, so the meaning of the saying is sadly
lost on me at this moment.
-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM.
Deadline: Sept. 24. Go here: http://sf.net/ppc_contest.php
_______________________________________________
Mjpeg-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/mjpeg-users
