|
|
|
|
Here's how our crew solved the problem with buzz.
If you can't hear anything, maybe it's time to get your updated version of Microsoft's MediaPlayer that can play a lot of formats, including MP3. The Procedure The sound that desired cleaning, we loaded into the Sonic Solutions DAW. Figure 1 is a snapshot of the Macs screen with running software. Most of the screen occupies editing window, and on the right side we can see two windows more with automated mixers. Although you’ll need a greater close-up onto the waveform, you can see from this point a large number of peaks with a big one at 15625 Hz, which is a horizontal frequency of the European TV PAL standard.
Figure 1 On a Figure 2 you can see a close-up between cca 600Hz and 1900Hz. Vertical red lines mark unwanted frequency. We know that they are unwanted, ‘cause we know how they appear and how these frequencies are built.
Figure 2 Also, we know that step motor, responsible for moving the tape across the video head, is moving the tape 25/50 times a second. We don’t want to elaborate at the moment why with that particular figure. We can say that the picture changes 25 times in the second, and that one change, or one picture, we call frame. Changing the frame must be synchrony with the frequency of the power supply. That’s how TV works. So, according to the power supply frequency, in Europe with PAL system we have 50Hz, e.g. 50 subframes, e.g. 25 frames. In America it’s 60, e.g. 30 frames. (In fact not quite exactly because of the different standard called NTSC, but that’s another story.)Anyway, we have this 50 Hz and we wanna get rid of it. How can we do that? In fact easy. We must apply a bandpass filter, with as narrower slope as we can. We can apply usual high pass filter, but we don’t want to destroy the rest of the lower frequencies. And that won’t solve the problem, cause our camera is very naughty. It produces a great number of harmonics. These harmonics are main frequency in the double. So, if we have 50Hz, the first harmonic is 100Hz, and so on. In our case our main third harmonic was 150 Hz, that produced 300Hz, for 600Hz and so on. On Figure 2 you can see a big red marked peak on 600Hz and the next on 900Hz, and then one marked at 1200Hz… Every single 300 Hz! So for every single disturbing frequency across the spectrum, we must apply a very narrow filter with infinite attenuation, called notch filter. On a Figure 3 you can see a very big magnification of the spectral analysis.
Figure 3 Please notice that you need only 4 Hz of bandwidth, but I used a band of 8 Hz instead, just to be sure. There’s no need to be afraid of the sound distortion, because you can only see it, as you will discover later, but because of the nature of this particular sound, you can’t hear it. In such a way I made a comb filter with 48 different notch filters (Figure 4).
Figure 4 One calls it comb filter, as it looks like one (see the Figure 5)
Figure 5 So, let’s have a look at the buzzed signal for the last time on a Figure 6. You can really see those peaks of buzz. Also, you can see a huge peak at the 15625Hz. This is the horizontal frequency of the PAL video signal that ran away from the video into the audio signal area. This frequency was destroyed with separate notch.
Figure 6 After a few moments of background processing on a Sonic Solutions DAW, we finally have a result - Figure 7.
Figure 7 You can immediately see the difference between upper buzzed signal and lower de-buzzed one. One can get confused as spectral analysis of de-buzzed signal seems to look as a inverted version of buzzed one at a first moment, but that is not the true. As you see in a detail look with a more precise magnification (Figure 8), the frequency that produced buzz is washed out from the spectrum. So, it is a slight degradation of the audio signal, but who cares! The structure of the material doesn’t need to be so god damn precise. You can’t hear the difference. The one and only thing you can hear is missing of the distracting noise of the buzz sound! We could do things more precisely, but there’s no need for that.
Figure 8 At the end we doubled the cleaned material on the second channel (It is a DAW, you know, so you can do it easily!) to produce stereo. As doubled signal doesn’t mean stereo just like that, we moved a part in time on a second channel. We don’t want to hear looping, so we made a long crossfade, and repeat those takes several times. And every single time you hear an underwater sound, there’s the wide stereo from your DVcam. So, this project was done in stereo, but if we’d mix it for Dolby Pro Logic or 5.1, we’d used the same audio, again moved in time, but on the surround channels as well, to produce great atmosphere. |
|
Send mail to digitalmagic@geocities.com with questions or comments about
this web site.
|
