17

The following video for Kraftwerk's Die Roboter is very interesting because it features some effects which I think might require a framebuffer to achieve. Having a framebuffer in 1978 would to my knowledge be a very big deal.

I'm interested in how the following effects were achieved:

  1. First of all we see vertical mirroring of the video multiple times such as at 04:20 in the video. What's interesting is when the effect is enabled you can see a drop in resolution. This makes me suspect that it could have been an early digital framebuffer that helped with this effect. I'm not sure how else it could have been done.
    Vertical split
    Click to enlarge
    Same test with original video

    What's interesting about this is how perfect the mirroring is. I took the top half of the image and vertically flipped it using GIMP. I then moved it down in difference mode until I lined it up. Given that some people were talking about things such a recording studio monitors or using an actual mirror I was surprised how perfect the mirroring effect is.
    Vertical split
    Click to enlarge

  2. At 3:02 we can see what looks like standard video that's been slowed down. What you'll notice about this part of the video is that the vertical resolution has been halved. Did they use a framebuffer for this or maybe it's possible with a professional video tape recorder to play the same fields multiple times and then use it for both the odd and even fields?
    Slow motion/Freeze frame
    Click to enlarge

  3. At 04:49 you can see horizontal mirroring. This one would just need a line buffer but is that how they did it?
    Horizontal split
    Click to enlarge

10
  • 5
    Looking at the video, I think that these effects could be created by using delay circuitry. It is a broadcast from a German station, I am pretty sure that they would have state of the art tools at that time.
    – chthon
    Oct 14, 2021 at 6:27
  • 2
    Vertical mirroring looks like it was done in the good old fashioned way, using a mirror. The horizontal mirroring looks like it is just their lookalike mannequins and again, a mirror.
    – Alan B
    Oct 14, 2021 at 7:39
  • 6
    As the posted answers indicate: analog all the way, most likely. From today's perspective in the 21st century we've all but forgotten how much got done with analog techniques before the rise of the transistor, or even more, the IC. These days analog is mostly a specialty of those working with microwave or higher frequencies, or the last inch of the circuit board between your DAC/ADC and your antenna. But analog ruled for decades before that and achieved great things ...
    – davidbak
    Oct 14, 2021 at 21:30
  • 1
    @alex-hajnal I wanted to thank you for your edits to this question. I think it made a big difference especially the screenshots from the video. Oct 15, 2021 at 5:03
  • 1
    @another-dave I at the time of posting thought it used early digital framebuffer hardware which would be part of a computer system. It seems like there's a good chance that I'm wrong on that but that was reason for posting this here. Oct 16, 2021 at 4:53

4 Answers 4

23

All of these effects are being created using beam splitters to combine two images of the original recording.

The lower resolution is a product of the second image, the lower in the horizontal split for instance, being viewed through a mirror and splitter. A single tape of the original recording is played on a studio monitor, with one side of the splitter aimed directly at it, in this case, the upper image. The second, lower, image has to be presented to the side of the splitter, and this is arranged with mirrors. This leads not only to a (tiny) loss of resolution but other optical effects due to phase differences and timing due to the longer optical path.

The reason for this seemingly over-complicated system, as opposed to two monitors firing at the splitter, is that one has to be reversed, genlocked and phase adjusted. The former are easy enough, but the latter not quite so much.

Most studios already had a beam splitter system for titling or (more rarely) stereoscopic recordings. In the latter case, the two lenses are both facing forward so one has to be turned through 90 degrees to enter the splitter from the side, so most used additional optics to try to make the two paths optically equal. Using two separate recordings rarely worked out well, in contrast to film where the resolution was high enough for this to work. Titling doesn't really need this complexity due to the high contrast, but to the modern eye the softness of the results are obvious, including in this video.

6
  • I added a test I did with the screenshot provided by Alex for the vertical mirroring. I used GIMP to compare the top and the bottom portion of the image and it matched almost exactly. What kind of artifacts would you expect from this process? Oct 15, 2021 at 5:24
  • 2
    @TheMovieMan You might try doing the same with the second video from my answer since it's probably a lot closer to what actually aired. The person who remastered the video may have changed it so that the two sides matched better. Or it could just be solid German engineering on the original beamsplitters. Oct 15, 2021 at 6:33
  • From an optics point of view (@TheMovieMan) setting up the vertical mirroring would be a little fiddly but not actually difficult. Your testing indicates a tiny imperfection in the vertical alignment - the mirror line is off by a fraction of a line, which is why you see horizontal edges in the diff. I assume this is in the original and not in your tests. At some point some banding (pinkish lines on the red background) has been introduced, which is why there are so many horizontal lines.
    – Chris H
    Oct 15, 2021 at 11:45
  • It's interesting that the highlights are different (look at the hair and the girder); the skin is also lighter in the inverted image. This probably means the upper (right way up) image has been through more reflections, each of which would cost about 10% of the light, before being redirected onto the camera. Some correction may have been used, but based on the background, or the correction could have been done when it was digitised
    – Chris H
    Oct 15, 2021 at 11:48
  • @AlexHajnal I did the test with the video link you provided. The results seem the same. So far that's the closest to the original that I've seen yet. I've also attempted to locate a less altered copy of the video and I was unable to do so. Ideally I'd like to watch a 50i copy of the video with minimal compression. Oct 16, 2021 at 5:16
30

It may be helpful to note that this is not a live recording but a pre produced video. These are pretty standard analogue (post production) effects at the time. Including real mirrors, two cameras and insertions. The only digital effect here is the text.

German TV, foremost progressive music shows, of the 1970s was notorious for using cutting edge technology. A sound financial base is one of the advancements of public TV :)

11

Analog techniques of delaying a video signal would have been commonplace by the late 1970s. In fact, I think every color SECAM television set would have needed to include an analog delay that was one scan line long. Conversion between 25 and "30" frame per second video would have required a piece of equipment with hundreds of carefully calibrated delay lines of different lengths, along with circuitry that could switch among them seamlessly. While effects such as described may have been produced optically, technology existed that would have been adaptable to perform such effects directly with an analog video signal.

Speeding up, slowing down, or time-reversal of video playback would have required either transcribing to film or using a video disk system. Such a system was used in the Doctor Who story "Robot" in 1974.

I don't know when frame-buffer-based video manipulation systems started to be available, but a rather powerful one was used in the Doctor Who story "The Leisure Hive", which was broadcast in 1980.

3
  • Every PAL TV except the very earliest had a line delay in the chroma path too; averaging two lines eliminated a "differential phase" artefact seen on early PAL (but still a lot less objectionable than phase errors on NTSC). Searching "Delay line PAL" should bring more info. Oct 16, 2021 at 16:57
  • One-line delays were implemented acoustically in a quartz crystal. But mirroring in such a system would be very difficult and you would notice off-by-one line issues in the resulting image (like horizontal lines not being lines). Vertical mirroring using such delays would be extremely difficult. Oct 16, 2021 at 21:24
  • @MauryMarkowitz: Format conversion between PAL and NTSC requires having delay blocks that are calibrated from 1 to 524 lines and a means of switching smoothly between them. Such equipment was big and expensive, but I think all that would be required for vertical mirroring would be changing the sequence in which the delays are used.
    – supercat
    Oct 18, 2021 at 0:00
9

Two things to note:

  1. The linked-to video is a "remastered" version that has been upsampled to 1080p and cleaned up. How much of what's visible on-screen is original and how much is interpretation by the remasterer is debatable (though it doesn't look like the remastering affected anything you asked about). Compare the following shots of the same frame. The first is the remastered version and the second is (as best I can tell) from the original tape:
Remastered Original
Remastered version Original version
Source: Youtube / RemasteredMemories Source: Youtube / TecnopopEspaña
  1. The pixelization visible in that shot is likely due to the raw footage having been shot using interlaced video. To get a clean freeze-frame image one has to double each scan line1. As an example, here's an interlaced shot of a car wheel in motion that has been produced by instead simply merging successive frames; the "combing" artifacts are obvious:

Interlaced video example
Source: Wikipedia / Mikus

1 At least with 1970's technology. One can do image processing to attempt to combine the two time-offset frames but even with today's technology it's difficult to do good-looking de-interlacing.

1
  • Some cameras of that era could also, given adequate light, use an LCD shutter to shoot 30 discrete frames of video per second.
    – supercat
    Oct 15, 2021 at 14:52

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .