Interlace
A picture has to be refreshed often enough not to flicker, and it has to have enough lines to look sharp. Both cost the same thing: lines per second down the wire. Interlacing buys one of them without paying for the other, by sending half the lines each time, and it charges for that on anything that moves.
A moving bar, drawn line by line, at the instant each line was sent
Each row below is one scan line, and it shows where the bar was when that line was sent. In interlaced mode the even rows carry one instant and the odd rows carry the next, a field later. Freeze it and the two halves separate.
Frozen. The bar is in two places, one field apart, in a single frame.
- the system
- 625 lines, 50 fields/s
- each field carries
- 312.5 lines per field
- the line count is
- odd, so each field ends mid-line and the next one falls between
- down the wire
- 15,625 lines/s
- a whole picture
- 25 frames/s at 50 fields/s
- without interlacing
- 31,250 lines/s, twice as much
- the bar moves
- 0.90% of the width between fields
- measured off the lines above: 0.90%
Why half an odd number is the whole trick
Every interlaced broadcast standard has an odd number of lines. The BBC's first high-definition service had 405, North America settled on 525, most of Europe on 625. That is not three coincidences. Half of an odd number is not a whole number, so each field ends in the middle of a line, and the half-line left over is exactly what drops the second field's lines between the first's. An even count would put the second field back on top of the first and there would be no interlacing at all.
Switch the system above and the readout keeps the half. It would be tidier to print 312 lines a field and nobody would notice, and the machine would have stopped working.
What it buys, and what it costs
A 625-line picture refreshed 50 times a second needs 31,250 lines every second. Interlaced, it needs 15,625, and the screen is still lit 50 times a second so it does not flicker. The saving is exact and it is a half.
The bill arrives on motion. The two halves of one frame were captured a fiftieth of a second apart, so anything moving is in two places inside one frame. Drag the speed up and freeze it: the bar separates into two interleaved bars, which is the comb you see on a paused video of anything shot for broadcast. The readout gives the displacement from the arithmetic and then measures it back off the lines drawn above, because a page that computed one number and drew another would look completely correct.
The date is not 1936
Interlacing is usually dated to 1936, and 1936 is when the BBC's 405-line service opened, which is a different claim. The patent this page cites was filed by Randall C. Ballard at RCA on 19 July 1932 and granted on 28 March 1939. Neither of those is 1936. The page is dated by the document it can show you rather than by the anniversary, and it does not claim Ballard was first: a patent is evidence that somebody wrote an idea down on a date, and priority is a harder question than this page needs to answer.
What is real here, and what is not
This is not a picture of a television
You are reading this on a progressive display, and no arrangement of pixels on it will show you what a cathode ray tube looked like: the phosphor persistence, the beam spot, the way the eye integrates two fields into one picture are all physical and all absent here. What this page can be honest about is the arithmetic. Each row shows where the bar was at the instant that line was sent, which is a fact about the signal rather than about the screen, and the comb that results is the same comb a frame-grab has.
Forty-eight rows, not six hundred and twenty-five
The picture above is drawn with forty-eight lines because that is what fits and stays readable. The line counts in the readout are the real ones and everything derived from them is computed from the real ones; the drawing is a diagram at a convenient scale. Nothing on the page divides by forty-eight.
The line rates are derived, and checked against the standard
There is no table of line frequencies in this page's code. The rate is lines times fields divided by two, and the test reads the real figures out of an archived copy of ITU-R BT.470 and requires the formula to reproduce them: 15,625 Hz for 625 lines at 50 fields, 15,750 Hz for 525 at 60. The 405-line system is not in that Recommendation, which withdrew it, so its 10,125 Hz is the same arithmetic without the same witness, and this sentence is the difference.
Real interlaced video also moves within a field
This page steps time once per field, so all the lines in a field share an instant. A real camera scans continuously, so each line inside a field is also a slightly different moment, which is why a fast horizontal pan can shear as well as comb. The field-to-field jump is much the larger effect and is the one interlacing is responsible for, so it is the one modelled here. The within-field slope is not.
Nothing here is about deinterlacing
Everything a modern player does to make interlaced footage watchable, from throwing away a field to motion-compensated reconstruction, is a large subject and none of it is on this page. What is here is the thing all of it is trying to undo.
Sources
- Randall C. Ballard, Television system, US Patent 2,152,234, assigned to RCA, filed 19 July 1932 and granted 28 March 1939. The claims describe a second series of scanning paths
interlaced with respect to the
first, and call it skip-line scanning. This page's dates come from here. - ITU-R BT.470-6, Conventional television systems, for what was actually broadcast. It states the 2:1 interlace ratio and prints the line frequency of every analogue system, which is what this page's arithmetic is checked against rather than compared with.
- Eight Sprites, on this site, for the other machine here whose limit is a consequence of a number rather than a decision.