DC offset distortion

Thanks for such a detailed reply.

While I was aware of what DC offset was (more/less) I can't say I know what its distortion sounds like compared to other causes (if I were to hear distortion/coloration while mixing, I doubt I'd realize it was caused DC offset by the sound of it).

I do most of my tracking with consoles and transformered pres so perhaps it doesn't end up much in the stuff I'm working on. Said that, I do mix other people's tracks and recently mixed a project that was 98% electronic. Was there DC offset crud I missed? Maybe. Don't know really. But nothing was sounding "wrong" so...yay I guess.

Spectrafoo has a 'remove DC offset' that's helpful when taking measurements, maybe Metric Halo offers this in their other products. Most DAW should have some sort of DC notch plug that you could scatter about liberally.

So the answer is multifold.

The first part is that DC offset itself is not causing the distortion.

If the DC offset plus the audio signal are not affected by the headroom of the circuit, then there is no resulting distortion. If the DC offset is high enough that the common-mode of the audio signal is affected by the headroom, then you'll have distortion.

Thinking about it in frequency-domain terms, you can think of DC as just another tone. If the amplitude of the tone is high enough, you'll get frequency mixing and the result will be harmonics at offsets equal to DC +/- the audio tone.

Either way, the fix has already been mentioned. A high pass filter of some type. You can AC couple or servo (very slow averaging circuit) an analog circuit. Digitally, the DC removal is similar to an averaging circuit but can allow more precision over time.

Often a light, fizzy distortion that can sound somewhat like a quantization error or maybe bit reduction? Enough that you'd know something wasn't right... like a layer of glaze not a full on Borg assimilation. 

Last time I came across any significant amount of DC offset on a project could've been a decade ago? Came in for mix. 

The band tracked in a homebrew kinda setup. I know they had a transformer loaded console... so the mics hit iron but where it starts to get murky is that the console itself? Not in the best of condition. And since nobody in that group was really an "engineer type" - what I remember of the hookup? Lots of guitar cable strung about so probably some gain & impedance mis-matches? Balanced/unbal conversions blah blah. 

That said, almost everything had an offset. Drums? Every track. Guitars? Most. Seemed the only things that didn't have offset were vocals & some other bits where they used outboard preamps. Hmmm.

I remember having a conversation with the group about it, and they were vaguely aware of "an issue" - in that things weren't totally sounding the way they expected... but also didn't have enough knowledge to really diagnose or deal with it themselves.  

Anyway. 

Mixed that one, like most everything else on a desk with all the side dishes. For the most part I did nothing to address the offset. Seemed the act of mixing & processing analog removed whatever was happening and there was no DC offset by the time it reached the 2 track. So this is not a problem? 

Best as I can figure? And I'm guessing here... but I think whatever happened on that one, the DC was baked in at the converter level. Because when I tried processing to remove offset? Headroom changed but the glaze remained. And since headroom was no concern and boy... there's a dozen songs here, hundreds of tracks... and really no sonic benefit? Let 'er rip! 

In this modern era I feel things largely sort themselves out... in that a lot of our digital processing options are smarter about removing & automatically applying corrections if needed. Filtering on the analog side of converters is also better then generations ago. And for the most part, almost everything in my shop is transformer coupled so it rarely has a chance to grab hold. 

Any AC coupling in the signal path, even in the a/d converter input itself will remove DC offset.

I think that from an EE perspective a purposeful "offset" or non-zero DC value (either current or voltage) that is desired is called the "bias". For an opamp or other similar feedback circuit we have something called a "virtual ground" or "virtual short" or "the summing node" and when the circuit is working properly with (negative) feedback applied (the resistors in place in the feedback network, etc.) - an input of 0V should map to an output of 0V (if you have no input signal you want there to be no output signal, including at DC, not just audio). The unwanted DC component is representative of error and called "offset voltage". The unwanted AC portion of the output (with 0V input) is called noise.

We measure offset at the output of an amplifier with no input applied. Maybe, as an example, we measure 100mV DC at the output. If the amplifier had a gain factor of 100 set by the resistor ratio (this would like a 40dB gain setting on your mic preamp), we could come up with an equivalent "input referred offset" by working backwards to answer the question "if we measure 100mv at the output and the gain is 100x, what would be the equivalent input voltage to cause it?". It would be 100mV / 100 = 1mV of "input referred offset" for this example. This "offset" is an error and not the desired "bias" we would want. For the example here with an opamp it is usually caused by mismatch in the transistors inside the opamp circuit, amongst other things.

Now what about Neve? Would it be better to call it an "offset" (an error) or a purposeful "bias"? I think we need to think about the transformer first.
Any transformer used for audio will have a magnetic material for the core (air-core transformers exist but used more for radio-frequency circuits, some DC-DC power converters) since these "ferromagnetic"  materials have much better magnetic properties at audio frequencies compared to air alone (look up "magnetic permeability" if interested). Also, ferromagnetic materials tend to keep the magnetic field concentrated inside the material itself unless overdriven (if the core is "saturated"), minimizing unwanted magnetic stray-fields for the unsaturated case. Transformer windings / cores are usually encased themselves in a metal "shielding" material of high permeability (mu-metal) which can help keep any remaining stray field from interacting with nearby circuitry (or more importantly, for input transformers, stop stray magnetic fields from being picked up as interference).

The permeability of the magnetic material in the core (iron, steel, etc.) is actually a function of the magnetic field intensity and is actually directional (if interested look up transformer "hysteresis") --  from what I can gather a major part of the "Neve" sound is biasing the output transformer with a static magnetic field (caused by the current in the primary coil of the output transformer in the above quoted post). If I understand correctly this forces the AC signal, the audio, reaching the secondary to "hang out" (that's my term for "biasing" here) on one side of the transformer's B-H curve in a particular position, avoiding effects of hysteresis but with a "pleasing" non-linearity that creates musical harmonics - the choice of this position is purposeful, likely based on listening tests. This is apparently done in the 1073 style circuits too and I guess that those musical harmonics are just now called "silk" over at modern RND.

Attachment Deleted

EDIT - everything above is basically what the quoted post said but with more (hopefully not mistaken) detail. Below is probably the main point.

So, this "bias current" to create our "bias magnetic field" originates on the primary side and only requires a "bias voltage" or "offset voltage" on the primary side. The transformer itself still operates on the principle of "induction" only and thus blocks DC offset on the secondary side. In other words, even with the Neve bias scheme, we should not measure any DC voltage ("offset") on the output side of the transformer (the secondary) which would go to your audio interface -- no offset issue with Neve in this sense.

Sorry, I wish I was better at (attempting) to explain this with less words but I guess I'm not. Hopefully this is useful anyway to at least other person :-)

Ha ha, I was just (in possibly the most confusing way possible) trying to distinguish between desirable and undesirable "offset", particularly on the analog side only - I completely agree that this whole thread was supposed to focus on offset that made it past the analog-to-digital conversion process (in the form of "ones and zeroes") and could potentially impact the sound of our plug-ins. I am all too familiar with analog offset (for my day job, not specifically audio) but it had never crossed my mind that plug-in sound quality could really be impacted by it until this thread (and the related one) -- definitely learned something there!
-Chris

Well there you go. Same basic unwanted noise just different explanation or description.

Either way. Garbage.

Speaking of Waves…
I’m reluctant to buy the BB Beauty Beast thing… cos I’ll probably like it for 5 minutes and then hate it.

But OTOH Still can’t believe it took me 8-9 years to throw the H Comp on a snare. It actually sounds like a real comp.