Date : Wed, 03 Aug 2011 16:20:04
From : heyrick.beebsoc@... (Rick Murray)
Subject: [BeebSoc] Auntie, dear Auntie
On 03/08/2011 13:08, Mick wrote:
>> I remember Mullard used to do Teletext (WST) sets in America, but when
> I vaguely remember seeing the Mullard factory in Hackbridge (Surrey).
They used to make valves as well. My old radio had them.
Had to get rid of it (ages ago) as I know zip about valves, and when you
had ones with the legend worn off and a schematic that had mostly
faded... probably somebody who read the 1950 RSGB manual from cover to
cover could have fixed it. But not me. :-(
I dunno. People these days p*ss&moan when their computer takes more than
a minute to boot up. That's pretty quick compared to the time between
switching on a valve radio, and hearing something come out of it.
Assuming it is all in working order, that is.
> It still needs to be stable enough to be read without missing a bit.
Imagine, if you will, that you only need to give a line or two of text
for subtitling. Well, you have several options:
1. Stretch the data pulses and sequence the information over
subsequent frames. This is equivalent to how people speak more
slowly when shouting.
2. Repeat the same data on three/five frames, average the results,
and use what comes out as the data. This is how a woman has to
talk to a man when he is programming / watching football. [in
other words, let the cat out-let the cat out-let the cat out...]
3. Something else clever I've not thought of. ;-)
If you aren't sending full pages, and you aren't sending a few hundred,
then I'm sure you can fiddle the data to work well, even on video tape.
Aha... Wiki says, for teletext:
The teletext signal is digitally coded as 45-byte packets, so
resulting data rate is 7,175 bits per second per used lines (41 7-bit
'bytes' per line, on each of 25 frames per second).
and for Line 21 captions:
It uses a fixed bandwidth of 960 bit/s.
If we assume 30fps, that works out at 32 *bits* (not bytes) per frame.
32 doesn't divide by 7, so if we guess eight bit bytes, it's a mere four
bytes per second. It looks like the system uses two-byte commands, so
the effective result is roughly two characters per frame. A little slow
(but could be 60 per second, usually enough for subtitling etc), and as
such (compared to a kilobyte per VBI line for teletext!) ought to be
slow enough to pass video tape 'damage'.
> Have you ever tried going into text while playing a VCR? All you get is
> junk although the quality of junk varies from one machine to another.
It is running at a seventeenth of the bitrate. Couple this with teletext
using multiple VBI lines, and you'll just see that number drop.
Reality? Teletext uses lines 6 to 22 (note, closer to or on top of the
'cut point' for VCR head switching!) to provide roughly 16KiB per second
of broadcast data.
Against 60 bytes.
I think you can understand why pressing text results in a garbled mess.
> So what is the difference between line 21 and teletext that wobbly VHS
> can cope with I wonder?
As I said above, teletext uses lines 6 - 22. I think (but it depends
upon the model) that anything below about line 14-15 is likely to get
messed up on a domestic VCR. Also, as you will be aware, there is some
degree of visual degradation in a video recording. It is like MP3 from
raw audio looks and sounds similarly, but it works in part by
compression and in part by chucking data away.
Well, VHS tech doesn't compress, so it can only throw away information.
This is how a 5MHz PAL signal with a 627kHz chroma signal gets itself
squished into a 3MHz capable video tape (with around 400kHz chroma).
As you can imagine, tiny dots don't tend to look so good afterwards.
Pause a VHS tape and play with your vertical hold and you might see the
biggest evil of all (and how VHS got away with discarding a little under
HALF of the signal) - it will be slightly BLURRY.
BLUR is an absolute killer for digital data. How can you tell where one
dot is, or isn't? I would hazard a guess that you would obtain less data
from a blurry recording of a perfect signal, than you'd get from a live
'weak' (and thus noisy) signal. With a weak signal, you might get
erroneous dots, but your timing and dots will be correctly placed.
Those of you who have ever used a capture card with software decoding
(such as the HCCS vision, or anything homerew that was basically a frame
buffer rather than a decoder chip) will know that raw data from a tape
can be highly irregular. I wrote a program to try to 'fix' data in the
HCCS Vision's framebuffer. It only worked in monochrome, attempting to
line up the edges of lines. The jitter on a prerecorded tape was low,
maybe a single pixel. A good signal recorded tape had a jitter of +/- 3
pixels average, and a bad signal tape was +/- 8 pixels, though sometimes
a line would be so far off it wouldn't be recoverable as data would be
missing.
Analogue TVs put up with a *lot* of crap - my PSP floors every single
video capture device I own, but it looks okay on the telly. Go figure.
> I've found a place that has reminded me how to use ATS.
You have the cheese wedge type?
> The trouble is the signal here has been terrible since they put a up new
> metal roof.
Try it on your Digibox.
[an hour later]
Bloody hell.
Actually, give up.
I didn't check *every* channel, but the best (only!) I found was this:
http://www.heyrick.co.uk/random/teletext_astra28.jpeg [80KiB]
which doesn't bode well.
If you can access RTE1 you might get something? Teletext data isn't
scrambled (obvious reasons).
If you can find somebody with a sat box tuned to Astra 19.2E, you are in
more luck. Quite a few channels still use teletext, only with a caveat -
only one (BBC World) is in English.
http://www.heyrick.co.uk/random/teletext_astra19.jpeg [372KiB]
> Saves wearing the tin foil inside the hat though :-D
I thought the hats were supposed to be made of tin foil? Have I been
doing it wrong all these years?
<paranoid>
Maybe THEY know you wear a tin foil hat, and the metal
roof is to concentrate the transmissions to penetrate your hat. </paranoid>
> Here's hoping the new communal ariel starts working soon.....
<paranoid>
...need I draw the conclusion here?
New metal roof? New antenna? Holy crap, time to line the walls with foil!
</paranoid>
> So long as the data gets from A <> B who cares? Don't answer that. ;-)
;-)
> I'm almost tempted to pull it out of the box and attempt to register
> just to see if it can.
Is it still for sale anywhere?
Which is yours? Mono type 1, mono type 2, or colour?
> It seems it may have a life in store for it after all.
How's your ARM code getting along? Did you understand a word of my
description of how R15 = PC and how R14 is used instead of dumping stuff
on the stack a la 6502? Or the fact that you have a bunch of registers
to play with and no specific accumulator (just, like, pick a register)? ;-)
>> http://en.wikipedia.org/wiki/IBM_5100
>> Prices ranged from $11,000 (16k model) to $20,000 (64k).
> Eek!
Yeah, but on the other hand I think in those days that sort of money
would have bought you a five bedroom townhouse. Memory was EXPENSIVE.
On the other hand, you wouldn't get some dimwit MP leaving it on the
seat of a train.
I *love* the tape streamer storage.
Talking of pricey - I *think* it was Maplin, but it might have been RS,
had the Heathkit HERO robot kit (for a couple of grand, IIRC). Myself
and my fellow geeks would drool over it.
Here's a fun little story for ya:
We managed to (sort of just) build a line-following robot out of bits in
the physics lad - some meccano, some low power motors, optosensor, a
bunch of 4000 series CMOS chips and some relays... It was a binary
system that had four distinct motor speeds and used two optosensors to
run the track in between them. When the thing sensed drift, it sped up
(or slowed down) a wheel until there was no drift, then it set the wheel
to middle speed. It had a tendency to 'hunt' between extremes. Probably
something that a microcontroller could solve, but it was pretty good for
a logic circuit.
It was, of course, horribly over-engineered. I remember about five
layers of veroboard with maye ten chips per layer, and a maze of wires
all around. This was done with the specific intention of making the
teacher cry. We actually put together, over the course of a weekend, a
truth table for the controller that used some 40 logic chips (and, thus,
something like 150-180 logic gates [*]).
We got it to do some nifty things, like if one motor was changing speed,
the other motor's relay would click even if staying at the same speed
(to prevent the thing spinning with one wheel unpowered) - though truth
be told, the lead acid power pack (yay RS for all the nasty crap that
thing put up with!) was heavy enough that inertia would probably have
prevented a spin. It's a good thing we had lots of vero space free, we
had to fit a shedload of electrolytics to balance out the motor
switching that was run off the same power source. Even WITH a power
diode and running the CMOS through a 5.1V Zener, the ripple was enough
to make an oscilloscope trace of the power rails when "if use" look like
a heartbeat! Yikes. But LOTS AND LOTS of caps sorted that out. By LOTS
I'm talking numerous chunkies per board, plus one smaller per chip (in
addition to the ceramic caps we fitted in the beginning).
How I wish I'd photographed it. I believe the physics lab kept it for a
while as a demonstration of why you shouldn't spend an entire weekend
doing your physics homework. :-) But I would imagine it's no longer
around. Hell, do kids even get to play with logic chips these days?
Best wishes,
Rick.
* - as a fan of WarGames, I wired in many little red LEDs on useful
signals. I also popped in a few 555 timers (hello Acorn!) just to
flash other LEDs at different frequencies for no purpose other than
"it looked cool". Run the thing at night with a small bulb on its
underside, turn off the lights, it got instantly promoted to the
status of "it looks awesome".
We taped a line around the great hall and let it do a few circles,
until the resident teacher doing night duty discovered a massive
oval of gaffer tape stuck to a polished stone floor. Cue lots of
detention. Again. And cleaning it up with a toothbrush. But, damn,
huge room, no lights, it was worth it.