AnalogTV (@_AnalogTV_) header
AnalogTV (@_AnalogTV_) profile picture

A whole television, in software. 📺 Real analog TV from physics: camera tubes, composite, VHS, RF ghosts & authentic CRT phosphor. No fake effects. https://nitter.cf/t.co/PsxWghEeSR

Joined April 2026
Photo by AnalogTV (@_AnalogTV_)Photo by AnalogTV (@_AnalogTV_)Photo by AnalogTV (@_AnalogTV_)Photo by AnalogTV (@_AnalogTV_)Photo by AnalogTV (@_AnalogTV_)Photo by AnalogTV (@_AnalogTV_)Photo by AnalogTV (@_AnalogTV_)Photo by AnalogTV (@_AnalogTV_)Photo by AnalogTV (@_AnalogTV_)Photo by AnalogTV (@_AnalogTV_)
Every VCR had a tracking control, and every rental tape eventually needed it. Turn it the wrong way and a band of noise crawled through the picture. Turn it back and it went away. VHS records each field as a long diagonal stripe, laid down by a head on a spinning drum. To play it back, the head has to run straight along the middle of the stripe that was recorded, and to know where those stripes are the deck reads a control track, a row of timing pulses along the edge of the tape. Tracking set the timing offset between those pulses and the drum. Tapes recorded on someone else's machine had their stripes in a slightly different place, so the head ran partly off the track. Where it strayed far enough it read the neighbouring stripe instead, which was recorded with its head tilted the other way and comes back as almost nothing. That is the noise band. Later decks did it for you. A Sony patent from 1987 describes the idea: the signal coming off the tape is strongest when the head is exactly on the track, so the deck adjusts the timing until that signal peaks. A reader on our pause post last month remembered being surprised when that arrived on home machines, having been a professional feature first. The clip is a Sydney tram on George St with the tracking off, then set correctly.
1
1
24
1,304
Cinema runs at 24 frames a second. American television ran at about 30, as 60 fields. Nothing divides evenly, so film on NTSC used a trick called 3:2 pulldown: one film frame is held for three fields, the next for two, then three, then two. Four film frames fill five video frames. It works, but the frames are held for uneven lengths of time, so smooth movement picks up a faint, regular hitch, most obvious on slow pans like this one. Watch the buildings slide past and you can feel the rhythm. European television ran at 25 pictures a second, which is close enough to 24 that broadcasters simply played the film slightly fast. No uneven frames, so no judder, but everything ran about 4 percent quicker. A two hour film lost almost five minutes, and the soundtrack went up in pitch by a bit less than a semitone. The clip is a slow pan over the Danube, first as film transferred to NTSC, then shot directly on video.
20
29
4
239
20,988
Tune an old set between stations and you got snow, a hissing field of black and white grain. A popular claim says about one percent of it was left over from the Big Bang. Nobody seems to have measured that on an ordinary aerial, and physicists who have argued about it think the share was smaller still. Nearly all of the snow was made inside your own set. Every electronic component makes a little noise just from the heat of its atoms jiggling. The input amplifier in the tuner picks up its own share plus whatever the aerial collects, and with no station to swamp it, the set's automatic gain control turns everything up until that noise fills the screen. The same thing happens gradually as a signal gets weaker. The picture doesn't switch off, it sinks into the snow, which is what the clip shows: Sydney Harbour at night with the signal getting weaker in steps. A completely empty channel showed snow in black and white, even on a colour set. That's because a colour set only decodes colour while it can find the colour burst on every line, and pure noise has no burst, so a circuit called the colour killer switched the colour off. In the clip there is still a station underneath the noise, burst and all, so the colour stays and the snow is coloured too.
1
1
21
791
Copy a rented movie from one VCR to another in the late 80s and the copy would slowly fade almost to black, come back, and fade again, every few seconds. That was Macrovision. The trick relied on a difference between televisions and video recorders. A VCR recording a signal uses automatic gain control to set its recording level, judging how strong the signal is from its peaks. Macrovision added extra pulses in the blanking interval between pictures, the part of the signal you never see, made to look like sync followed by a burst of super-white. The recorder saw what looked like a far too strong signal and turned its recording level right down, then slowly recovered, then got hit again. The copy ended up recorded at a level that kept sagging and recovering. A television mostly doesn't look at those lines, so the original tape played normally on the set. It was only the copy that suffered. The clip shows a copy made from a protected tape, then an unprotected one.
9
2
33
3,093
Every line on a CRT is drawn by the set's own horizontal oscillator, firing about 15,600 times a second on PAL. The broadcast sends a sync pulse at the start of each line, but the oscillator doesn't simply fire on every pulse, because a burst of noise that looked like one would knock it straight off. Instead a circuit compared the average timing of the incoming pulses with the oscillator and nudged it with a small correcting voltage. Manuals called it flywheel sync, since the oscillator kept its own momentum and was only steered. The horizontal hold knob set where the oscillator ran when left alone. If it was close enough the correction pulled it into step. If it had drifted too far, each line started a little later than the one before, the whole picture leaned over, and the dark diagonal bar you see is the gap between lines that should be hidden off the edge of the screen. Further off still, the picture breaks into several strips. Repair guides used this to diagnose the set. A picture that slides sideways but can be caught for a moment means the sync isn't reaching the oscillator. One that breaks into strips points at the oscillator itself. The clip is the Barangaroo skyline with the hold off, then locked.
2
13
595
Worn tapes had white streaks, but a circuit could hide some of them. Play an old VHS tape and you get short white flashes, a line at a time, flicking across the picture. Those are dropouts: tiny spots where the magnetic coating has worn off, flaked away or picked up dust, so for a moment the head reads nothing. VHS stores the picture as an FM signal. When the signal vanishes, the demodulator has nothing to follow and throws out a burst of nonsense, which tends to show as a bright streak. Almost every VCR had a circuit to hide this, called a dropout compensator. It watched the strength of the signal coming off the tape, and when it fell away it swapped in the previous line of the picture, held in a delay line, for as long as the dropout lasted. Neighbouring lines of a picture are usually nearly identical, so most of the time you never noticed. You saw dropouts when there were too many of them, or when the patch landed on a sharp edge where one line really was different from the next. The clip shows the same badly dropped-out tape with the compensator off, then on.
22
842
Why were interference specks and streaks dark on most TVs but white on old British and French sets? A car with a bad ignition, a hairdryer, a drill: anything that threw out electrical sparks put specks and streaks across the picture. What colour they were depended on where you lived. A TV signal is a radio carrier whose strength carries the picture. Most systems, including PAL and NTSC, use negative modulation: full carrier means the black end, and white sits near the bottom. A spark adds energy to the carrier, so the set reads it as darker. On those sets interference shows as dark specks, which is a large part of why negative modulation won out. Dark flecks are much easier to ignore than bright ones. Britain's original 405-line service and the French systems did it the other way around, with full carrier meaning white. On those sets the same spark came out as a bright white fleck. If you remember sparkling white streaks every time a moped went past, you were probably watching 405 lines in Britain or French TV. The clip runs the same burst of interference through a simulated 625-line set and then a 405-line set using the AnalogTV app. We did a post on impulse noise back in August which drew the bursts white, this is only right for those positive-modulation systems. Thanks to those that pointed this out.
3
16
870
A VHS tape can't carry much fine detail, so VCRs had a circuit to put some of the crispness back. It went by names like detail corrector, aperture corrector or just sharpness. It works by adding a scaled copy of the signal's own rate of change back into the signal. Where the picture is flat, nothing happens. At an edge the correction spikes, so the dark side gets darker just before the edge and the bright side brighter just after it. The result looks sharper than it really is, with a pale halo hugging the bright side of every dark outline. What it can't do is restore detail the tape never recorded. It exaggerates what survived, overshoot included, which is why VHS could look crisp and wrong at the same time. The corrector works on whatever edges it is given, so a passing tram's lettering and the edges of its windows pick up the same halos as the buildings, and noise picks them up too, which is why heavy sharpening made a picture look busier as well as sharper. The clip is a George St tram on VHS with the sharpening up, then the same tape with it off.
1
4
26
2,014
The Soviet TV so small that families watched it through a tank of water... The KVN-49 was an early Soviet television set with a very small screen. The fix sold alongside it was a lens: a clear plastic or glass vessel, curved on one side, filled with distilled water and clamped in front of the screen on sliding brackets. The details come from the Soviet trade ministry's own commodity dictionary. The standard KVN-49 lens was 350 by 265 millimetres and 75 deep, held about three and a half litres, and magnified 1.5 to 1.8 times with an organic glass vessel and water, or 2 to 2.5 times with silicate glass. You were meant to add two or three drops of phenol so nothing grew in it. Glycerine and medical vaseline oil were the alternatives to water. The dictionary promised an undistorted picture for five to eight people sitting two to five metres away. What it couldn't do was add anything. The lens makes the same picture bigger, scan lines and all, and anyone off to the side saw it bend. Memoirs from the time add a rumour that vodka gave a clearer picture than water. Anyone out there tried it? Certainly drinking the vodka might make the Soviet TV shows more interesting. The clip is a Bratislava fountain on a 625-line black and white set, through the lens and then on the bare screen. This is simulated and we don't quite have a real device to test it on, but we are working on improving how the optics should have theoretically worked. The magnifying mechanism in front of a tube wasn't a uniquely Soviet piece of technology. If you search for them, there are plenty of other examples like: collection.sciencemuseumgrou…
1
1
23
1,569
Politics: Why half of Europe watched SECAM, and why your neighbours' TV came out in black and white? In the 1960s Europe had two new colour systems to choose from: PAL, developed by Telefunken in West Germany, and SECAM, developed in France. A conference in Oslo in 1966 was meant to settle on one standard for the whole continent and broke up without agreement. France went SECAM. So did the Soviet Union, which had adopted it alongside France at the 1965 CCIR meeting in Vienna, and most of the Eastern Bloc followed, though not all of it: Romania went straight to PAL when it started colour. Why is it still argued about? A 1979 book on the standards fight argues that France's main motive was protecting its own television industry. A German study of East Germany's decision in 1969 describes it as a political choice, made knowing perfectly well how popular West German television was, and SECAM was not what West Germany used. A technical argument was also made at the time: SECAM's colour was thought to survive the Soviet Union's very long relay links better. What it meant at home was this. The picture itself, 625 lines and 50 fields, was the same on both sides, so a SECAM set could show a PAL broadcast, but only in black and white, because it could find no colour signal it understood. In the clip (created in the AnalogTV app simulating both standards) a Bratislava tram is shown first as a PAL signal on a SECAM set, then as a SECAM signal on the same set. Czechoslovakia sat right against Austria, which broadcast in PAL.
1
1
1
28
1,560
Watch old news footage or a home video shot after dark and every headlight drags a glowing tail behind it. That was the camera tube, not the film or the tape. Tube cameras build the picture as electric charge on a light-sensitive target, and an electron beam reads that charge off line by line. The beam never quite clears it all in one pass, so part of each picture is still there when the next one is read. Anything bright and moving leaves a fading copy of itself where it just was. This is called lag. The vidicon, the cheap tube in a lot of consumer, industrial and security cameras, was notorious for it. And it got worse in the dark. RCA's own technical notes on the vidicon from 1958 say the lag at low signal currents is generally severe, which is exactly the situation in a dim street with a few bright lights in it. Broadcast tubes like the Plumbicon were designed to discharge faster, which is part of why professional cameras cost what they did. The clip is the Sydney light rail at night through a vidicon camera simulation using the AnalogTV app, then the same shot with no camera tube at all. The app renders the vidicon in black and white, as most of them were.
2
12
609
If you lived anywhere near an airport you may have noticed this one. A plane went over and the picture started to pulse, ghosting and brightening and fading again, faster and faster as the plane passed overhead, then settling down once it had gone. The aerial was getting the signal twice, once from the transmitter and once bounced off the plane. A reflection off a building gives a steady ghost. A reflection off an aircraft keeps changing, because the plane is moving and the length of the bounced path changes with it. The two copies of the signal drift in and out of step, adding up one moment and partly cancelling the next, so the picture pumps. As the plane gets closer the path length changes faster and the flutter speeds up. It is well enough known to have its own name, airplane flutter, and to have been measured properly. A 2008 paper in IEEE Transactions on Broadcasting modelled it from the aircraft's size, height and flight path and matched the model against measurements on real TV channels at 60.75 and 208.75 MHz. The clip is Sydney Harbour with a reflection that keeps moving, then the same reflection held still.
1
12
585
AnalogTV retweeted
Replying to @_AnalogTV_
here's some real teletext splash screens from 90s i captured from VHS tapes. they are a bit corrupted due to various factors like signal level or packet speed. with high packet speed it is possible to recover some pages even if recorded broadcast length is less than 15 seconds)
1
1
2
61
A T-120 tape held two hours at standard speed, four at long play and six at extended play, and plenty of us recorded whole seasons in the six-hour mode. It always looked worse, and the reason is primarily to do with the width of the tracks. VHS records each field as a long diagonal stripe laid down by a spinning head. The drum spins at the same speed whatever mode you pick, so slowing the tape to a third just packs the stripes closer together. In six-hour mode each track is about 19 micrometres wide, a third of the standard-speed track. There is no gap between tracks, so a head reading a stripe that narrow is always close to its neighbours. It gets worse because of how the heads were made. A Japanese patent on these machines notes that the extended-play heads were deliberately made wider than the 19 micrometre track, so that slow motion and picture search would work, though this raises the crosstalk from the adjacent track and degrades the picture in normal play. Less tape under the head also means less signal against the tape's own noise. (On a standard 4-head deck, two of them would be optimised for this slower tape speed - though it could also be done on a 2-head deck - often with compromises at the faster speeds). Six-hour recordings lost fine detail too, which our test-pattern post on 31 October showed. On real footage like this, though, what you notice first is the noise and the blotchy colour. The clip plays the same moment at six-hour, four-hour and two-hour speed, fed straight to the screen so the TV's shadow mask doesn't hide the difference.
3
16
836
Why did American TVs have a Tint knob and European ones didn't need one? Colour on NTSC and PAL is carried as the phase of a subcarrier, measured against a short reference burst sent on every line. Any small delay the signal picks up between the studio and your set shifts that phase, and a phase error on this kind of colour is a hue error. Skin goes green or purple, depending which way it went. NTSC sets had a Tint (or Hue) knob for exactly this. You turned it until faces looked right. It is where the old joke that NTSC stands for "Never Twice the Same Colour" comes from. PAL fixed the problem in the signal. It inverts one of the two colour components on every other line, so a steady phase error pushes one line's hue one way and the next line's the other way. The set averages each pair of lines and most of the error cancels. What is left over shows up as a small loss of saturation, which the eye forgives far more easily than a wrong hue. SECAM, the third system, sends its colour a different way altogether and has no phase to go wrong. The clip runs the same moment three times: NTSC with its colour reference 25 degrees off, PAL with the same error, then NTSC set correctly. Measured on the coloured parts of the picture, the NTSC hues move by about 19 degrees and the PAL ones by about 4. When I was a kid it was kinda fun to play with the tint knob on our old Sears TV to get that effect on purpose. You can now play with the knob in the AnalogTV app to get the same experience on your own videos (but only in NTSC mode).
3
2
46
3,951
If you lived among tall buildings with a rooftop aerial, you probably got ghosts: faint copies of the picture sitting a little to the right of the real one. The aerial was picking up the station twice, once straight from the transmitter and again after the signal bounced off a building, a hill (or a gasometer, back in the day), and the set added the two together. The reflection always lands to the right because it arrives later. The picture is drawn left to right, one line at a time, so anything that arrives late is drawn further along the line. That makes a ghost a measuring tape. Radio travels about 300 metres in a microsecond, and a PAL line takes 52 microseconds to cross the visible screen. A ghost sitting a twentieth of the screen width to the right arrived about 2.6 microseconds late, which means the bounced signal travelled roughly 780 metres further than the direct one. With a ruler held to the screen you can work out how much further the bounced signal travelled, which narrows down which building is to blame. The clip is a row of glass towers at Barangaroo with two reflections added to the signal, then the direct signal on its own. I should have posted about ghosts on Halloween, much less spooky to discuss them today. Unless mentioned otherwise, all of the videos in this series were created using the AnalogTV app itself.
4
8
502
The last post was about data being sent in the analog signal, so it's clear why the slow bit rate meant that teletext weather maps looked like they were built from Lego On a basic (Level 1) teletext page there were no pixels to send. Every picture, the weather map included, was built from a set of 64 block characters. The decoder split each character cell into two columns of three blocks, and each block was either lit or dark. Six on/off blocks give 64 patterns, so a cell of graphics cost one character code, exactly like a letter. That comes from how the page reached your set. Teletext rode on spare lines in the blanking gap at the top of each field, and each of those lines carried one packet of 45 bytes: 3 for sync, 2 saying which magazine and row it was, and 40 holding the characters for one row of the page. At the 625-line rate of 6.9375 Mbit/s those 45 bytes take 51.9 microseconds, about the length of the visible part of a 64 microsecond PAL line. One line, one row of text. So a page was 24 rows of 40 characters, the top row being the header with the page number and clock, plus an optional 25th row for the coloured links along the bottom. At two by three blocks per cell, the 23 rows under the header came to 80 blocks across by 69 down. The map area on the page in the picture is 36 by 17 cells, which is 612 bytes. Your set's character generator turned those bytes into blocks itself, which is why every coastline came out as a staircase. The same packet format also carried pages that were never meant to be shown, including general data broadcasting, so a TV signal could deliver data to equipment that never put it on screen. The picture is the app's teletext weather map page with the outline only (in the app, weather symbols are drawn on top), shown on a PAL colour set. The image is from the AnalogTV app's built-in Teletext decoder. If you give it permissions to use Apple Maps (on iOS, ipadOS, macOS and tvOS versions) then it will actually show you live current weather. We tried to get the maps looking as good as possible for each country, but still a way to go for some of them. It's on the backlog.
1
22
848
Four ways analogue TV carried digital data without the picture even noticing.... An analogue TV signal has room in it that the picture never uses, and from the 1970s to the 1990s broadcasters found at least four places to hide digital data. 1. Whole lines in the vertical blanking gap, fast version. Teletext used lines in the gap at the top of each field that are never shown. Each line carried one 45-byte packet at 6.9375 Mbit/s on 625-line systems, a one being a pulse at about two thirds of white level and a zero sitting at black. One line, one row of a page. 2. Whole lines in the gap, slow version. American closed captions went on line 21. Each field carried a pair of bytes, 7 bits of character and a parity bit each, so the captions trickled out two characters per field. The second field of line 21 was later allowed to carry the programme's rating, plus text and extended data services. 3. Beside the picture on a carrier of its own. NICAM stereo sound, in the UK on 6.552 MHz above the vision carrier (5.85 MHz in most of Europe), was a 728 kbit/s data stream in frames of 728 bits, one every millisecond, 704 of them sound or data. 4. Inside every line of the picture. The satellite MAC systems did away with composite colour and sent each line in pieces, one after the other. D2-MAC starts each line with a burst of 105 bits of digital data, a 6-bit line sync word plus 99 bits of sound and data, then the colour information squeezed into a third of the time it would normally take, then the brightness squeezed by a third. With 623 of those bursts a frame, that works out to about 1.5 Mbit/s for sound and data. The picture is one line of D2-MAC on the app's waveform monitor. The yellow burst on the left is the data, coded in three levels (duobinary) rather than two. Then comes the squeezed colour and the brightness. (The flat blue stretch on the far left is an analogue sync pulse the app keeps for its own decoder; real MAC has no sync pulse there, the 6-bit word does that job.)
1
1
19
548