Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
What's that horizontal line in the spectrogram? (our old ~16 kHz buddy)
The old-timers here will be well aware of this phenomenon and I certainly won't claim to have been the first to figure out what it was; a bunch of people here had worked this out long before I was even doing this kind of stuff, but I don't know if anybody's actually explained in any detail why it's at the specific frequency that it usually is. I figured others who also find this kind of thing interesting might benefit from having the numbers laid out clearly. This isn't the kind of thing that will help you capture LaserDiscs correctly, but it will give you some background info on what the hell those big horizontal lines are that appear in the spectrograms of the vast majority of LaserDisc captures (in my experience at least).

Quick note on fields vs frames: a "field" is half of one full frame of picture, containing either the odd-numbered or even-numbered scanlines. The electron beam in a CRT renders these alternately, drawing odd then even then odd and so on (I won't go into field order here because it's not really important to the thing I'm talking about, but you get the idea).

Right, so. Here's an example of a spectrogram for a PCM audio stream captured digitally from a LaserDisc (side 1 of one of my three different copies of THE HUNGER):

[Image: ML102216-filename-cropped.png]

Notice that there's a very obvious straight horizontal line through basically the entire recording, just below 16 kHz:

[Image: ML102216-closeup.png]

As others have quite astutely noticed, this line is created when the person recording the soundtrack does so in the same room as one or more powered-on CRT monitors/TVs.

A CRT TV or monitor contains a flyback transformer that sends a sawtooth signal to the electromagnet, which is what causes the electron beam to bend in order to "draw" scanlines horizontally across the screen. Since the (full) resolution of NTSC video is actually 525 lines (of which only 480 are usually intended to be visible), and the frame rate of NTSC is 30/1.001 whole frames per second, this means that to render one full frame of video, the flyback transformer in an NTSC monitor or TV has to go on and off at 525 x (30/1.001) = 15,734.27 times in one second. In other words, it oscillates back and forth between being magnetised and demagnetised at a rate of about 15,734 Hz. As this happens, the magnetic core very, very slightly deforms then snaps back to its previous shape over and over again due to a phenomenon called magnetostriction, which creates an audible noise at that frequency. Incidentally, magnetostriction is also what makes those retail security tags work: they contain a small magnet and a thin, ferrous metal ribbon that's cut so that magnetostriction will cause it to oscillate at a frequency of around 58 kHz when its subjected to an alternating current electromagnetic field (e.g. the one created by those security things at the doors of a shop). There's a really good video about that here:

Anyway, yeah, so the magic number here is 15,734 Hz (15.73 kHz). Sure enough, that's pretty much exactly where that line seems to sit on Spek's Y axis scale: I quickly cropped in from the top and bottom to see how many pixels off 16 kHz the line is, and it's 10 off 16 kHz in the exported image with 75 pixels in between 14 and 16 kHz meaning that that line is at (65/75) * (16 - 14) + 14 = 15.73 kHz.

But what's that? There are *two* horizontal lines in the above spectrogram? Why, yes! Just beneath the 15.73 kHz line there is a significantly fainter one that *also* runs through the entire recording from start to finish.

All the numbers used above are for NTSC, but there might be some cases where the line will be shunted downwards a little bit. If so, it's probably a decent indicator that somewhere in its lineage, it's been recorded in Europe, or another PAL region. PAL TVs/monitors render 625 lines (576 of which are visible) at an effective rate of 25 fps. So the calculation from before becomes 625 x 25 = 15,625‬ Hz (15.63 kHz). And sure enough, if I use the same method of just cropping the image by a pixel at a time, I find that in that 75-pixel range between 14 and 16 kHz, the lower of the two lines sits 61 pixels beyond 14 kHz, which means it's sitting at (61/75) * (16 - 14) + 14 = 15.63 kHz. So indeed, at some point, this track has been recorded with a PAL monitor running somewhere nearby. The effect of the NTSC resonant frequency however is a lot stronger in this example, which makes sense because it's an NTSC LaserDisc.

Some more recent home video releases deal with this artefact in a fairly unsophisticated way: they sometimes use what I've referred to in the past as the "ice pick lobotomy" method, a term I decided was appropriate because it looks rather like they just smashed a long, sharp implement through the spectrogram somewhere between 15 and 16 kHz, obliterating everything in the chosen range (which is arguably as devastating to the 15-16 kHz frequency band as a lobotomy is to a person's prefrontal cortex). You can clearly see this on TERMINATOR 2: JUDGMENT DAY, for example, as it appears on the now out of print 2008 French Blu-ray Disc:

[Image: T2FR-CDS.png]

For comparison's sake, here's an earlier version of that same track (from an old Japanese DVD):

[Image: T2JP-CDS.png]

(T2 CDS mix spectrogram discussion is in this thread: Thanks to @MrBrown, @Stamper, and @Chewtobacca for info on the 2003 Japanese DVD and 2008 French BD containing this particular track.)
Thanks given by: PDB , Hitcher , TomArrow , Plissken1138
Good read, Pipe. Thanks for writing that up.
New, low or non-posting members: Please do not post or PM me asking where to get something or what the current status of a project is. Such requests are ignored. Instead, stick around and become a participating member. Thank you.

Thanks given by:

Forum Jump:

Users browsing this thread: 1 Guest(s)