DC Offset

  • Thread starter Thread starter Rokket
  • Start date Start date
Rokket

Rokket

Trailing Behind Again
OK, again, I will say the more I learn the stupider (if that's a word) I feel.

Just exactly is DC Offset, and when would I be concerned about something like that?
 
Hey bud,

DC Offset is where the center line, or "rest voltage" of your signal - when there is absolutely no noise - is anything other than 0 volts (the center line on your waveform display). Typically offset will show as the flat line of silence being shifted slightly upwards off the center line.

This is usually not a super-duper huge problem (to put it in technical terms :D), but it does create two potential issues and is therefore undesireable:

1.) The more the center line/rest voltage is shifted, up off the 0V center line, the less room in dynamic range you have on that side of the line to fit that side of the waveform in before it'll clip at the top.

2.) That DC voltage, when pumped to a speaker, is in itself inaudible, but it will cause a shift in position (in or out) of the speaker's rest position. Just as the offset decreases the amount of dynamic "room" for your waveform to fi, it will in a similar way decrease the amount of room your speaker will have to move in that same direction before you clip the speaker itself. Additionally, this change in speaker position can cause some change in the response of the speaker itself simply because it is not operating from it's normal position.

G.
 
If it helps, I haven't seen a serious problem with DC offset in probably 10 or 12 years.

Well, maybe with SoundBlasters...
 
Massive Master said:
If it helps, I haven't seen a serious problem with DC offset in probably 10 or 12 years.

Well, maybe with SoundBlasters...
Click to expand...
Let me introduce you to my Toshiba laptop someday, John ;). Not that I actually use it this way for anything other than quick and dirty scratch tracks or location dialog stuff, but if I were to record through the built-in card without filtering the offset, my rest voltage would actually be at positive -25dB or so. Absolutely horrid.

G.
 
There is no impedance for DC, thus it can produce threatening currents to the speakers. A DC of about 1% won't be a real problem, but if they become really big, you better get rid of it as soon as possible.
 
LogicDeLuxe said:
There is no impedance for DC, thus it can produce threatening currents to the speakers. A DC of about 1% won't be a real problem, but if they become really big, you better get rid of it as soon as possible.
Click to expand...

A typical speaker's DC resistance isn't a whole lot different from its impedance. Say 6 ohms. But while the AC signal is variable, that DC is constant, and could cause overheating.

It isn't terribly likely that DC offset in the digital domain is gonna make it to a speaker though. Somewhere along the line, there will be a capacitor or servo (maybe a transformer!) that will kill it, before it hits the power amp output. Nevertheless, it's a flaw that chews up headroom, and should be addressed. If you use a low-cut in your chain somewhere, even 10Hz, that will do the trick.
 
Aha!!
Now I can put a name to the problem with sound files that I've been getting consistently from one of our clients - I can see the offset in the waveform (though I don't hear it), and his stuff tends to clip at levels that don't seem that excessive...
...the quality of the sound is not great, but by the time it's encoded to an .mp3, and listened to by client on computer speakers, it's passable (barely) for most auditions....but short of a hardware replacement, is there any other advice that I can offer to improve the quality of what he's able to send? He's a talented guy, but his sound quality is costing him work...
 
whyseye said:
Aha!!
Now I can put a name to the problem with sound files that I've been getting consistently from one of our clients - I can see the offset in the waveform (though I don't hear it)
...
short of a hardware replacement, is there any other advice that I can offer to improve the quality of what he's able to send? He's a talented guy, but his sound quality is costing him work...
Click to expand...
First what you want to make sure of, whyseye, is that it is true DC offset, and not just natural asymmetry in the waveform. By this I mean make sure the rest line is indeed shifted, and it's not just that the centerline is not shifted but that one side of the wave is larger than the other. The latter one is not uncommon and is quite natural, whereas offset is not.

If it is indeed offset, most DAW software applications include a DC offset button or checkbox somewhere that will automatically rectify the offset. If not, then have your man do what mshilarious suggested, and throw a high-pass filter set to about 10Hz on the offending track(s). This usually works just as well at filtering out DC offset. In rare cases you may have to pump the filter setting up to 15Hz or so to get rid of everything offensive - this is when the offset is not true DC, but rather a low-frequency modulation of more than just of few Hertzeseses - but usually 10Hz will be plenty.

G.
 
Thanks G-
Now I'm looking forward to receiving his next audition, to see what works!
 
SouthSIDE Glen said:
First what you want to make sure of, whyseye, is that it is true DC offset, and not just natural asymmetry in the waveform. By this I mean make sure the rest line is indeed shifted, and it's not just that the centerline is not shifted but that one side of the wave is larger than the other. The latter one is not uncommon and is quite natural, whereas offset is not.
Click to expand...




Glen, what causes the wave to be larger on one side when no dc offset is present?
 
quickmill said:
Glen, what causes the wave to be larger on one side when no dc offset is present?
Click to expand...
Because that is the actual action of the soundwave/instrument itself. For example, when you hit the skin of a drum, the stick will push the skin down a certain distance. When the skin rebounds, it will come back out, but not quite as far as it was pushed down. The first half of that cycle, therefore will create more sound pressure than the second half. The diaphragm of the microphone will react in kind, and a larger voltage will be created at the peak of the upswing of the cycle than the peak of the second half downswing. This is natural, and is not limited to just drums, though that is often where it is the most apparent.

Next time you look at a kick drum or snare track in your editor, you should notice that the waveform is not symmetrical, but rather kind of "unbalanced" to either the top or the bottom. As long as the volume line returns to the center of the graph when all is said and done, that's fine. But if the waveform at silence "starts" and returns somewhere off the center line, that's artificial offset, which should be taken out.

G.
 
Thanks for the responses, all of you.

I am glad too that my question was able to help someone else in the process.
 
You will also notice an asymetrical waveform on trumpets and other things like that. That happens because the 'wind' coming out of the end of the instrument is blowing in one direction, so it pushes the diaphragm back while it's vibrating it.
 
SouthSIDE Glen said:
Because that is the actual action of the soundwave/instrument itself. For example, when you hit the skin of a drum, the stick will push the skin down a certain distance. When the skin rebounds, it will come back out, but not quite as far as it was pushed down. The first half of that cycle, therefore will create more sound pressure than the second half. The diaphragm of the microphone will react in kind, and a larger voltage will be created at the peak of the upswing of the cycle than the peak of the second half downswing. This is natural, and is not limited to just drums, though that is often where it is the most apparent.

Next time you look at a kick drum or snare track in your editor, you should notice that the waveform is not symmetrical, but rather kind of "unbalanced" to either the top or the bottom. As long as the volume line returns to the center of the graph when all is said and done, that's fine. But if the waveform at silence "starts" and returns somewhere off the center line, that's artificial offset, which should be taken out.

G.
Click to expand...
There are always certain clients whose waveforms are not symmetrical with the center, but do start and end at the center line, and I have no distortion with them....and it's always the same clients whose vocal forms look like that, regardless of where they record - I can tell who it is just by looking at the waveform - but I always assumed that it was just something about their personal voiceprint....
...these other files are clearly off of center by a fair margin, and seem fairly balanced top and bottom, but they click, pop, and clip anyplace there's a spike in level....
Truth be told, I'm really just an actor/director who has had to learn what I can about engineering out of self-defense, and it's been an invaluable tool in my job as a voice director; but being self-taught (and HR-taught) there are still some pretty huge gaps in my recording fundamentals, so I'm really grateful that you've taken the time to break it into bite-sized chunks for my tiny mind....thanks!
 
whyseye said:
There are always certain clients whose waveforms are not symmetrical with the center, but do start and end at the center line, and I have no distortion with them....and it's always the same clients whose vocal forms look like that, regardless of where they record - I can tell who it is just by looking at the waveform - but I always assumed that it was just something about their personal voiceprint....
...these other files are clearly off of center by a fair margin, and seem fairly balanced top and bottom, but they click, pop, and clip anyplace there's a spike in level....
Click to expand...
Those two paragraphs are perfect examples of the difference between simple asymmetry (first paragrpah) and offset (second paragraph).

And let me opine that it's an excellent sign of your abilities that you *do* recognize different sources based just upon how the waveform looks. This is not to say that it's a substitute for using one's ears for mixing, but that you notice those differences in detail is an excellent skill to have which can save you a lot of time and headaches.

Put simply, one cannot tell that everything is right just by looking, but often it's easy to see right off the bat when something is wrong.

G.
 
Yup - I don't think that I'll be giving up using my ears, but the pace that I work at is so fast and furious (it's not unusual for me to be tracking and/or editing 200-300 individual files over the course of a busy day) that I depend a lot on "reading" the files to tell me where I need to pull out mouth noise and breaths, drop in slates, identify clipped or problem files, etc.
I guess you'd liken it to "meatball surgery"....get them cleaned up and stable, then ship 'em out....but for my own personal growth and knowledge, I like understanding the hows and whys of the way things work....and somehow, the more I learn, the less I know....:D
 
A dirty method of "fixing" those asymetrical waveform is a phase shifter (that kind of filter you use for Dolby Surround Prologic encoding). A constant shift is barely audible, but a phase rotator is one of the ugliest things I heard, and I wouldn't recommend one for the world. They are used on most TV and radio stations, usually along with a multiband compressor and a limiter.

About subsonic noise, not only ADC's can produce them, but also effect processors. In the worst cases, I had waveforms which didn't even touch the center line for several seconds. I usually get rid of it with a phase true high pass at 12 Hz.
 
I've just been dealing with a file that had significant offset issues. Rokket, if you're still there: the vocal track you repaired had a fair sized offset when it 1starrived.
Initially I processed the offset using waverepair - it allows for this as an option when normalising. The thing is I don't like to normalize if I can avoid it so I took advice from Glen & used a highpass filter set at 20hz. That worked well & it also partially dealt with some other problems.
Now I have two ways of managing it when it comes up but I'd rather avoid it if I can.
It does seem to be a product of cheaper soundcards. Would that be a fair assumption?
 
rayc said:
It does seem to be a product of cheaper soundcards. Would that be a fair assumption?
Click to expand...
I wouldn't put 100% slam dunk on that assumption - you never know what could happen - but yeah, I'd say 98% of the time a cheap soundcard seems to be the culprit.

Like John said, he hasn't seen offset from another source other than cheap soundcards in 10 years, and I'd have to say my experience is pretty much the same.

Laptops are especially notorious for this, but standard PC game cards can see it too.

G.
 
Followup....
Went back and tried DC offset on one of the old problem files, and just as advertised, the center line returned to center!
The files were still clipped and blown out, but if he brings down his levels, he may actually have workable files for the future - thanks again for all the help in my continuing education.
 
You must log in or register to reply here.
Back
Top