Weekly Watch: BBC Engineering Test Tx

One of my weirder hobbies is digital archaeology. I've gotten oddly good at tracing the provenance of random pieces of media. Pictures and sound follow their own version of Locard's Exchange Principle: Every medium and environment they go through leaves a mark. No part of the signal chain is transparent.

YouTube clips of home recordings of old transmissions are like matryoshka dolls, layer over layer of information on where they come from, and how they got here. This is going to be extremely long-winded and technical, so brace yourselves. Here's the example I'm using.


The first thing you check is the documentation. This is often crap, but a crap in informative ways. This particular clip is labeled "PALplus test" and the caption is in German. Translating things with technical terms in them can get interesting; Google usually at least tries to take context into account, but most machine translators trip over things like "Sendeschluss" and "FuBK" and "Längsspurton".

"Sendeschluss" = lit., 'send-close'. Less literally, 'sign off'. Used for the picture, audio tagline, or video clip broadcast by a station to indicate they are shutting down transmission for the night.

"FuBK" = "Funkanstalten Bildkarte", lit. '(broadcast) station picture card'. The equivalent term in English is 'test card' or 'test pattern', a still (or more rarely, animated) image broadcast by a television station that is specifically engineered to test the technical capability of the receiver. Famous test cards are the RCA-developed Indian head test pattern widely used in the black and white days, and the SMPTE color bars used in NTSC regions.

"Längsspurton" = lit, 'long tracks'. Used here in the context of a VHS tape, it refers to a method of laying down the audio track linearly along the edge of the tape, rather than using the embedded signal-under method of a HiFi VCR, which in less-capable recorders can interfere with the picture.

ARD is sort of a hybrid of the BBC in the UK and PBS in the US, a consortium of public broadcasters supported by a television licensing fee charged to each household, and ZDF (now 2DF, judging by their logo) is basically "BBC2".


The next thing is historical context. This is captioned a "PALplus" test, broadcast in Germany in 1993. PALplus is a variation on the PAL broadcast standard developed to handle 16:9 material, like movies and documentaries. The gist is that a standard 4:3 television would show the material in letterbox format, with black bars at the top and bottom, but that there would be extra data hidden in the letterboxing bars that would enable a 16:9 television to reconstruct the full original picture. The "hidden" data was known to sometimes cause spurious blue and yellow patches in the "black" bars, but that depended on how touchy your receiver was. The picture as presented on YouTube is full-frame 4:3, so there's no way to verify from the given video clip that the original transmission was in PALplus.


YouTube leaves its own mark. While the HD resolution options meet the standards now used worldwide, the SD resolutions are tailored for US sources. (480 is NTSC broadcast/Betamax/VHS SP standard, and compatible with early VGA; 360 and 240 are roughly VHS LP and EP standards; and 144 is roughly equivalent to elderly RealPlayer and QuickTime streaming video.) Standard definition PAL video starts out at 576 lines and counts down from there, so in order to smash it into a US-oriented video resolution, you need to do standards conversion, which is a generally destructive process. The detail in the original picture would have been very sharp and regular; any mottling or unevenness is due to unintelligent downsampling (or, at worst, just discarding lines) from the original video.

The periodic blur - snap into focus - blur cycle is due to the YouTube MPEG compression algorithm not dealing terribly well with fine detail when not being babysat by a competent human. MPEG compression works by dividing the picture into blocks, and looking at those blocks to track how they change over time. If either the block size or the time window are badly-chosen, it gets confused and defaults to going, "Uh, and then... something happens... here...?" and fills in with a sort of average color, hoping you won't notice. A human can go in and fix it, but automatic algorithms are not necessarily that smart.


The video was recorded, as the caption says, on VHS. Terribly. VHS video is easily mangled in predictable ways, almost all of which can be seen on this despite the best efforts of the MPEG encoder to re-mangle everything in its own fashion. The vertical lines are all wibbly in a way characteristic of multi-head helical scan recording methods. The picture lines are recorded on the tape by a rapidly-spinning head assembly that has 2-4 little magnetic chips that actually do the writing; because it's all mechanical, and the tape is somewhat stretchy mylar, it's almost impossible to get the picture lines laid down exactly parallel to one another. TV doesn't use absolute pixel addresses like computer displays do, instead depending on the timing of the signal to line things up. If your timing is just a smidge off, your line will start just a smidge early (displaced to the left) or late (displaced to the right). It's not necessarily obvious, because our brains will ignore/correct a lot, but hard vertical lines show it badly.

Little horizontal white streaks are places where the magnetic coating fell off the tape backing. They're common to all analog magnetic tape formats, which have no error correction. It makes the playback head go 'whoops, no signal, dunno what goes here' and creates a spot where the television essentially gets random static. Different tape formats show the streaks in different patterns; 2" Quad tape, for instance, breaks picture lines up into multiple lines on tape, so when the oxide falls off you get a shower of equally-spaced lines blinking into picture. VHS puts one picture line on one tape line, so you get chunks of static instead.

The weird black stripe with crooked picture bits under it that pops up at the bottom of the picture is another VHS error. It means the timing is screwed up. The VHS player doesn't recognize the black bar as 'hey! this is the bottom of the picture! I'm blanking to give you time to get back up to the top of the screen to restart!'; it just takes X number of lines from wherever you tell it to start, assumes that's a screenful, and sends it to the TV. Tracking adjusts where you start the screenful, and someone here has set it slightly wrong, such that the picture starts a few lines late, sends the blanking interval as a black bar, and puts a few lines of the next screen on the bottom of this one.

Something in the signal chain is rubbish, because the picture has appalling dot crawl. It happens when the horizontal and vertical sync signals interfere with the luminance signal; sync-on-composite is standard on both RCA cables (US), SCART connectors (EU), or S-Video (both), and most VCRs don't understand RGB component signals, so checkerboarding is common.

Interestingly, some of the 'displaced streak' blips in one of the black and white test plates appear to be baked into the original transmission tape. There's another copy of the same test pattern, albeit a different broadcast, that has similar distortion. Either the receiver on the alternate copy is better at coping with it, or the first VHS recording exacerbates it, because it appears less mangled on the cleaner-looking copy. A mid-90s broadcast master would likely be on Betacam, for which that is a plausible kind of picture error. This alternate also doesn't have the problem with dot crawl, which suggests they had a much better SCART/S-Video cable, or a less noisy setup.

There are various pieces of content and picture characteristics that also verify a PAL/BBC source and suggest a date, like the white balance, the noisy reds, and the test patterns, but this is getting long enough as it is. I'll get to that next week.