2 Newbie questions: Dithering early, peak -4

[jennajenna]

I am working with sampled piano tracks (Gigastudio) so dense that each one is made up of as many as 8 thinner piano tracks mixed together in Sound Forge. The finished tracks will eventually be brought together in Sonar. It is taking forever because I'm working in 24-96 and each track is 30 minutes long.
For some reason, even though Sound Forge is set to DC adjust while recording the SPDIF signal from the Gigastudio computer, each sub-track fails to generate a reading of -inf when scanned for 'average value' by Forge's statistics function. This is easily corrected after the fact by 'automatically detecting and removing DC offset.'

I am finding it hard to resist adjusting each sub-track after it's been recorded, then each pair after they've been mixed together and on and on.

Can over-dithering a track (by dithering each of its sub-tracks individually) cause damage? Or do sub-tracks with an apparent DC offset not mix well?
Or should I not even be thinking about any of this so early in the project?

If anyone is still reading, or reading at all, one more thing- How do these levels sound for one of the recorded piano tracks (it's a classical-type piece in 24-96): the quiet notes peaking at about -30 db, the overall peak being -4 db. Any thoughts?

 

[lo beam]

I am finding it hard to resist adjusting each sub-track after it's been recorded, then each pair after they've been mixed together and on and on.

Is this after hearing them as a whole or before? I would've thought you'd only need to adjust anything if it wasn't sounding right somehow when you listen to the mixed together tracks.

Can over-dithering a track (by dithering each of its sub-tracks individually) cause damage? Or do sub-tracks with an apparent DC offset not mix well?
Or should I not even be thinking about any of this so early in the project?

You should only ever add dither once - as the last step in the mastering process of your final stereo track, never on individual tracks.

 

[jennajenna]

dither

Thanks for your response lo beam! In response to your first question, I started wondering whether and how to do DC offset right from the start of the recording process. It was during the recording of these first tracks (tracks that are to be mixed together in Sound Forge to form the main tracks that will assemble the song in Sonar) that I got into this whole DC thing. And no, it wasn't because anything sounded wrong at all, nor did anything even look out of place onscreen. It was that I ran SF's statistics option to see how things looked in terms of db level, etc. and noticed that the reading for 'average value' was, say, 900 rather than -inf. The manual says that a reading in this category other than -inf often indicates that there is some DC issue present. So I decided to fix the offset that I could neither see nor hear and sure enough the 'average value' became -inf. But when the first 2 tracks, fresh with 'average values' of -inf, were mixed together, the resulting track's 'average offset' was no longer -inf. So I thought, should I 'automatically detect and correct DC offset' again, and again and again each time I add another track to the mix? (keeping in mind that I am still working on the sub-track level i.e. building up one instrument that will eventually be mixed with other instruments in Sonar).

In response to your second point, please forgive me. I don't know why on earth I even used the word 'dither'. Even as a newbie I do know what it means. I wrote my post very late and I meant to say something like 'Is it possible to over-DC-offset-correct the clips being mixed together to form a track? Where is it coming from- does it adversely affect the track being built (i.e. if 4 tracks with a DC offset are mixed together, does the resulting track have a quadruple DC offset?) Or does it matter at all.
Pardon my rambling.

 

[SouthSIDE Glen]

It was that I ran SF's statistics option to see how things looked in terms of db level, etc. and noticed that the reading for 'average value' was, say, 900 rather than -inf. The manual says that a reading in this category other than -inf often indicates that there is some DC issue present.

While that can be the case, it is not necessarily DC ofset chat can cause such number; sometimes it is a natural phenomenon which needs (and wants) no correction. On many instrument tracks there will be an asymmetry in peak (and average) volumes caused by the nature of the instrument itself.

The most obvious and often-cited examples are on drums. On a kick drum, for example, let's say you mic the kick through the hole in the front skin. When you hit the rear head with the beater, it pushes the skin inwards. The resulting air pressure pushes the diaphragm of the microphone in. Then the drim head recoils and rebounds in the opposite direction. But because of the physics of the materials involved, the head does not flex outward as far as it was hit inward. (It's much like a pendulum that does not swing as far the second time as the first, or a ball that does not bounce as high on the second bounce.) And because the head does not recoil as far outward as it was pushed inwards, the microphone diaphragm also does not respond as far out as it did in. This asymmetry in the pgysical movenemt of the drum head, the air, and therefore the mic capsule translates into an asymmetry in the electrical waveform that is recorded. This is how it's supposed to be.

While it's most obvious on drums, it also happens to a lesser degree on piano, some wind instruments, and other instrumets as well. And when you get that natural asymmetry in the waveform, it can statistically show up in the numbers you're looking at as well.

What you need to do to determine if you have true DC offset or not is to look at just those silent parts of a recorded track, where you have a flat line at or near the "inf" center line. If that line is not riding right on the center line, or if you have natural low-level hiss that is not symmetrically centered on the center line, then you probably have DC offset.

If, however, the zero signal is riding right on the center line, but your waveforms appear to be "lopsided", then you don't have offset, but rather have a natural asymmetry in the nature of the instruments you're recording.

If you do have offset, you can keep it under control by throwin a high pass filter with a cutoof frequency of about 10Hz-20Hz or so on your individual tracks to filter out everything below that (including DC which is a theoretical 0Hz). This will filter out the offset without affecting anything in the audible range.

HTH,

G.

 

[jennajenna]

Thanks SouthSIDE Glen. That was very interesting and helpful. I looked at the flat lines of my silent regions and they are centered. And I am working with piano tracks so it would appear that I have a natural assymetry. Too bad because your high pass filter remedy sounds like fun.