DC Offset upon Stereo Sum Limiting

[sonicXpro]

Hi,

 

I am new to this forum, I am a pro audio engineer who has enough of protools and I am working with a macbook pro plus apogee symphony mobile while I am abroad teaching and recording in another country.

 

I just recently switched back to Logic Pro after avoiding it for some years, working on Protools (if I had to) and Nuendo. Now with Logic 8, it seemed that many issues that I had with Logic in the old days have been fixed.

 

Nonetheless, whenever I am trying to master inside the software, I get some DC offset in the bounced stereo file whenever I use a limiter on the output (to limit the stereo mix). It happens with Logic's own limiter (just tried that one out), Waves L1 and the Sonnox Limiter (which I normally use).

 

Is there any explanation, workaround or fix for this issue,

 

thanks a lot,

 

Martin

[Holger Lagerfeldt]

Hi Martin, welcome

 

The long explanation and the solution (hopefully)

True DC offset can only occur naturally OTB. Apogee converters are all prone to DC offset a bit in the A/D, including the Rosetta 800. Granted, we're talking way down, but i'ts not 0% like RME et al. This doesn't mean they sound bad (which they obviously don't) but they're not technically perfect in that regard.

 

If you're getting a post A/D DC offset like syndrome ITB (offset of the waveform) that's a bit puzzling, especially if you're experiencing it no matter what limiter you put on.

 

Let's take it from the top:

 

1) You bounce a stereo mix as WAV with NO limiter on: no DC offset (apparently)

 

2) You bounce the same stereo mix as WAV WITH any limiter on: DC offset

 

Correct? You're not bouncing to mp3? And you're not clipping at some point? (aliasing artifacts when clipping can lead to foldback of frequencies which can lead to massive subfrequencies, looking like DC offset).

 

Your problem description doesn't make sense to me unless there's some kind of voodoo involved, so I'm applying Occam's Razor and guessing you're simply amplifying any offset already present in the original recorded signal. Maybe it's too low to notice but it's there.

 

Nothing to do with the limiter unless your software is locally broken, as there's nothing wrong in my setup with all the same limiters you mention.

 

So go back and check each track for DC offset until you find the culprit. You can remove it either by using the DC offset removal function in the sample editor or manually applying a 20 Hz filter (watch out for clipping).

 

If you can't hear the problem on the individual tracks simply use an analyzer such as the Waves PAZ with the 10 Hz button enabled, and watch out for subs.

 

I hope that helps.

 

The short one

Send me the file and I'll have a look.

[sonicXpro]

Hi and thanks a lot for you quick reply,

 

the audio tracks from that productions were not recorded by me, I think they origin from PT HD and I just received them to mix, so I guess they were tracked through a 96 or 192 I/O.

 

None of the tracks had a "detectable" DC offset.

 

Nonetheless, I just tried out something rather simple which also resulted in a DC offset:

 

I programmed a bar just with the "Klopfgeist" metronome software instrument of Logic and exportet it as a 24bit wave (without any additional plugins and @ -12 dBFS), guess what, the bounced file also featured a DC offset. Seems to me as a problem going from 32bit float to 24bit.

 

Try it out and tell me what you think of it,

 

best regards,

 

Martin

[marcel72]

Is it really DC, or is it the sum of dither artifacts?

[fader8]

Seems to me as a problem going from 32bit float to 24bit.

I haven't experienced this and I use all the limiters you mentioned. How are you detecting that there is offset? Is it noticeable in the waveforms, or is some analysis you're doing with software telling you there is offset?

 

Marcel may have a point. Are you dithering? Also, are there any other plug-ins on your output channel?

[Holger Lagerfeldt]

This doesn't happen here, so please tell us in detail what you're doing to get that result.

[David Nahmani]

I just tried exporting one bar of klopfgeist to a .wav file, re-imported it and opened it in the sample editor, and found a 0.030 % DC Offset. Not sure why that is.

[Holger Lagerfeldt]

I get 0.000% DC offset when bouncing klopfgeist in 24 bit and "no DC offset" when bouncing in 16 bits.

 

I discovered something.. this only happens to bounced 24 bit files (dithered or not). And both from 32 bit float to 24 and 16 bit to 24 (padded).

 

0.002% DC Offset on a full scale digital signal, i.e. a hot mastered song in 24 bit.

 

It doesn't happen to bounced 16 bit files (dithered or not), so I'm not worried.

 

Checked ten Top 20 songs and they all had anything between 0.003% to no DC offset.

[sonicXpro]

Well, here are some reactions:

 

DC Offset detection: Both Nuendo and Logic have a way to detect DC offset ("Statistics" in Nuendo and "Remove DC offset" in Logic's sample editor).

 

After exporting the 16 bit master from Logic, the file is imported into a new project and then checked for DC offset.

 

I somehow feel uncomfortable giving out masters with DC offset, doesn't seem very pro to me. As already mentioned, I worked on Nuendo the last years and never got this funky DC offset.

 

I am still of the opinion that there is some fault in the 32bit float to fixed point conversion in Logic, there is also some clearly visible DC offset when looking at the waveform of the file.

 

Nonetheless, removing it before burning the CD is a possible workaround for now, nonetheless, I am fully of the opinion that this topic should be investigated by the programmers.

 

Best regards and thanks a lot for the commitment,

 

Martin

[Holger Lagerfeldt]

I am still of the opinion that there is some fault in the 32bit float to fixed point conversion in Logic, there is also some clearly visible DC offset when looking at the waveform of the file.

 

Not on 16 bit bounces though:

 

I get [...] "no DC offset" when bouncing in 16 bits.

 

It doesn't happen to bounced 16 bit files (dithered or not)

 

If something happens during 32 bit float then I'm wondering about this:

 

I discovered something.. this only happens to bounced 24 bit files (dithered or not). And both from 32 bit float to 24 and 16 bit to 24 (padded).

 

If the 16 bit file doesn't simply get padded but goes from 16 bit fixed to 32 bit float and then back to 24 bit fixed, then you could be right.

 

Anyway, I passed it on to a Logic QA contact at Apple.