what are some causes of Dc offset?

[balanceman]

sometimes, i get pops and clicks while working on certain tracks. I'm not sure if it is a buffer issue or stopping on a file with some Dc offset...

my whole place is wired from an equitech wall panel and the power seems really good. the noise floor in all my gear is very low and the detail I am getting is great.

what might be some other causes of offset?
equipment?

tube mic power supply?

gremlin?

I should state - the pops and clicks aren't recorded - more when you stop or start playback.
this only happens once in a while...

[DesertDawg]

Originally Posted by balanceman

I should state - the pops and clicks aren't recorded - more when you stop or start playback.
this only happens once in a while...

From your above description, it sounds to me like it's more likely to be a buffer thing, or perhaps even clocking-related (?). Anyway, the two things you're inquring about are easy to assess by process of elimination... 1.) get rid of your DC offset, playback & check it; and 2.) pad your buffer settings back for increased latency & see if you still get the same effect after each process.

I would think you'd have to have a rather extreme amount of DC offset in order for it to cause the pops & clicks as you describe them, and if so, it would be more likely to be constant rather than intermittent/only on startup of playback.

[Bob Olhsson]

I've been told that rounding the numbers rather than dithering or truncating DSP calculations can cause a DC offset.

[DesertDawg]

Originally Posted by Bob Olhsson

I've been told that rounding the numbers rather than dithering or truncating DSP calculations can cause a DC offset.

Bob, how do you "round" without one or the other (either dithering or truncating)? The closest thing I can imagine would be going from 88.2k/32-bit to 44.1k/16 bit without dither - since that's a simpler math conversion on both counts (samplerate AND bitstream rate) - but that's not a very common formatting & subsequent conversion methodology.

[MrHope]

getting pops and clicks on starting and stopping but not recording sounds like DC offset to me. I've noticed that some high feedback settings of certain reverbs can cause it. (Freeverb for example; luckily you can specify some high pass filtering in it). Some poorly programmed VST synths also have extremely high DC offset. Fl3a is an example of one. I have to use a steep highpass filter on it every time. It even puts out DC bias when it's not making sound!

Some DAW programs can be configured to automatically remove DC offset from chunks of audio some of them can also automatically do miniature (10 ms) fade ins and fade outs on audio clips. (CoolEdit 2000 and DDclip, for example). CoolEdit has a function to remove DC offset automatically during recording.

Recently to avoid clicks at the beginning and end of audio tracks I often leave about a 1/3rd of a second of pure digital silence. It's not a track gap, it's actual samples set to 0 amplitude. On track starts sometimes I have to edit the audio to have miniature fade in.

Most of my experience with DC offset however has been from hardware such as minidisc recorders or cheap mixers and such. However, sometimes it just builds up from layering lots of sounds. If you use samples a lot, you might want to use a high quality DC filter on each of the samples to prevent the layering from building up DC. I have found that if your source material doesn't have DC offset then you don't usually have to filter the final mix (which could sound bad).

[Bob Olhsson]

Originally Posted by DesertDawg

Bob, how do you "round" without one or the other (either dithering or truncating)? ...

Rounding is changing the bottom bit according to the value of the exponent. Most DSP routines automagically do this so that tax returns and other financial calculations will come out as expected.

Truncating just leaves the bottom bit alone at whatever value it happens to be.

Dithering randomizes the bottom bit so it sounds like noise rather than like distortion. Rounding sounds less like distortion but according to Jim Johnson of Bell Labs it generates a DC offset.

[Francis Vaughan]

Hmm, if Jim says it create a DC offset it does.

The problem is going to be one of the definition of rounding, and in particular the definition of rounding in the negative half of the cycle. The definition of rounding, and truncation, in the face of negative values always takes care. Rounding down towards zero is fine, except when your zero amplitude is not represented by a numeric zero.

[Bob Olhsson]

Rounding as defined by IEEE standards which is what most if not all calculation routines default to.

[Alexey Lukin]

Originally Posted by Bob Olhsson

Dithering randomizes the bottom bit so it sounds like noise rather than like distortion. Rounding sounds less like distortion but according to Jim Johnson of Bell Labs it generates a DC offset.

I believe that it's truncation that creates DC offset, not rounding. Anyway, this DC offset is quite minimal and should not have any impact on playback. Much more important are effects of dithering (or non-dithering). Truncation and rounding create a quantization noise that sounds like distortion (with truncation it's 6 dB louder than with rounding). Proper dithering decorrelates quantization noise from the signal, which means no more distortion, just even noise.

Returning to the author's question - there can be different causes of DC offset in recordings. One of them is absence of high-pass filtering in mic pre-amps or A/D converters. Many microphones (including phantom-powered condenser microphones) are capable of producing very low frequencies, fractions of a Hertz. If this infrasonic rumble is not filtered out, it gets into the recording and can be interpreted as a time-changing DC offset. Sometimes poorly designed A/D converters generate certain constant-level DC offset. Finally, any non-linear processing of a recording (particularly, distortion or saturation) can introduce a DC offset, if not properly post-filtered.

[Adam Dempsey]

Originally Posted by Alexey Lukin

Proper dithering decorrelates quantization noise from the signal, which means no more distortion, just even noise.

Although containing meaningful signal within the noise, not "just" noise.. to be clear. (off topic...)

[Alexey Lukin]

Of course, it's just noise + original high-resolution signal.

[sonicdefault]

You could use a plug-in to detect DC offset at various stages of your work to see when the problem is occurring. Ozone for example, has an offset meter which shows DC offset reduction as it occurs. Perhaps you should experiment with something like this to see if you can isolate the issue.


-SD

[Adam Dempsey]

The dither is random noise, yes (noise shaped or not), however...

dithering decorrelates quantization noise from the signal, which means no more distortion, just even noise.

ie, the result contains sonic detail (signal). It's audible within the dither and is the whole principle of retaining >16bit resolution.

In any case, I think "much more important are effects of dithering" is like saying "much more important are effects of convertors/other signal path components, et al.. True, but the point here (again) being the old "all other things being equal..."

The classic Sonic system would default in its prefs to rounding on input, if memory serves correctly. It also had a DC filter at something like 18Hz - two of the first things I bypassed.

DC offset? I like transformers.