DC offset?

  • Thread starter Thread starter Deege
  • Start date Start date
D

Deege

New member
I noticed a long time ago that when I recorded something, the wave wouldnt be centered....like, thered be a wave that ends way up high (when actually looking at the waveform) and it gradually goes back to the center, and if i select it and go to Amplify and "center wave" it makes it look normal, and doesnt seem to change the sound at all. So why exactly do recordings do that? In preferences, it automatically adjusts DC offest while it records and I do that all the time now, but I really just don't understand what that is.....
dcoffset.gif
 
Yeah, that gif you posted shows a DC offset problem. It's caused by the soundcard apparently, and Cool claims to be able to correct it by enabling "DC Bias Adjust" in Transform/Amplitude/Amplify.

The thing I don't understand is when the waveform's centred okay, but it's not symmetrical, and the waveform has loads of information above the center line for example, but not much below. What causes that? It sounds okay, but what does the irregularity in the symmetry of the waveform indicate?
 
.....

that's what I don't understand. LOL when centered the wavs that were recorded like that, they would go up and down like a zigzag...but sounded fine....odd.
 
My understanding is that's caused by poor Analog-Digital conversion. A better convertor and clock would help if that is happening alot. DC offset is actually an issue with analog gear also. It's one of those little reasons why cheap gear can suck.

I (think) that the assymetrical stuff is basically errors in the conversion but I'm not sure.
 
Hi TRK - that conversion idea is possible, because sometimes when it happens, it happens along the length of the entire waveform. However, sometimes it just happens on some sections of the track, as if the waveform is reflecting some kind of variation in the sound - maybe me moving around in front of the mic or something.
 
From the Cooledit Pro Help system... Thorne

Many sound cards record with a slight DC offset, which means that direct current is introduced into the signal by the card, causing the center of the waveform to be off from the zero point in the waveform display. This is seen as the waveform being “shifted” above or below the center line in the Wave Display, and can cause a click or pop to be played at the beginning and the end of the file.
 
From the authority - Bob Katz (from his site www.digido.com)...

If there is severe DC offset in program material then it may be desirable to fix it. DC offset shows itself by pops or big clicks if you stop and start the CD or tape. But if the DC offset is mild, or imperceptible, it's probably better to leave it alone. Especially since you're starting with 16 bit material, and after any calculation, including DC offset removal , you're going to have to dither it back down to 16 again. So many of these programs that extract audio from 16-bit CDs and then allow you to manipulate it, do so in a haphazard or unsophisticated manner from the point of view of DSP quality. And no matter how well they do it, double 16-bit dither is somewhat costly to sonic transparency.

In addition, the quality of your calculation can be costly to the sound, adding distortion or grunge, especially with a high-pass filter (which is needed to remove DC offset). High pass filters require more DSP power to do well than simple equalizers. But if the DC disease is bad, and you use a high quality calculation, the cure can be better than the disease. This advice goes for any calculation, for example, gain change, equalization, to sample rate conversion, and DC offset.

Some high-quality programs will perform a double precision (48 bit) dithered high-pass filter, for example. Ask the designers of your program if that's what they are doing, to maintain the highest quality audio.

There is another method of removing DC offset, that does not require redithering. This method is "scanning". The program searches for and measures the DC offset in the material, and instead of filtering, simply subtracts or adds an exact quantum, returning the centerline back to 0. This is theoretically much more transparent-sounding than the filtering above. But usually the problem of DC offset is that it is not constant, and moves, due to A/D converters changing with temperature, and other factors. So, scanning may not do a perfect job of removing it. Not too many programs have a scanning option. It should be more available. If your program is truly performing scanning and subtraction then the calculation should be very transparent.

Your job is to listen carefully, measure if possible, and also find out from your software vendor the real details under the hood.

Best wishes,

Bob Katz
 
You must log in or register to reply here.
Back
Top