Final mix levels, rendering, resample, dither

SonyEPM wrote on 11/6/2002, 10:09 AM

Subject: Final mix levels, rendering, resample, dither.....HELP !!!
Posted by: Rahman (in the Vegas Video forum originally)
Date: 11/6/2002 9:53:05 AM

I desperately need some guidance on figuring out levels and stuff. I have basically a 3 part question: The first is about master levels, the second about resampling, and the 3rd about best dithering approaches.

(1) I read a mix article that gave this basic guide for levels when mixing
a pop/dance tune:

kick = 0dB
snare = about -2dB (tweak)
cymbols = -25dB
Synths and Keys = -20dB
Bass = -10dB
Gits = -15dB
Vocals = 0dB

These are just a start, but basically when all added up, the article stated that "the master fader peaks should not exceed +3dB". I have followed the basic formula and have been getting pretty good results on most tunes.

However, things get kind of fuzzy as to what to do next. It was suggested to me to
now take the +3dB master fader and reduce it by 3dB, so the max level is 0 dB.
(I was told this was to approximate 0dB on VU meters).

I had been doing this for a number of tunes, until recently I forgot to drop the
level before rendering. When I setup the final mix to be mastered, I noticed the RMS level (ave) was already at my target levels, as compared to commercial stuff. Does this mean that I could have kept master level setting at +3dB ?

(2)
Also, now that I have my 48k sampled, 24-bit mixdown file. what is the best way
to convert to 44.1k ? I have been opening the rendered file in Sound Forge
so I can use the "Anti-alias filter" when re-sampling. Is this really necessary?
Can 44.1k be selected in "render as" menu with as good results as with the anti-alias filter ?

(3) Finally, assuming I now have the 2-trk mixdown file created at 44.1k and 24-bit,
what are the best ways to dither to 16 bit? (i.e. Use Wavelab UV22HR or Waves L1+ IDR, as the final plugin of my mastering chain)? Is vegas better or Sound Forge
for mastering? There is no "render as" selection in Forge, so I'm not sure if the
dithering is actually taking place when i select "process selection" from the plugin
box.

If any of the SF and audio guru's would help me, I'd greatly appreciate it.

(Especially since my family is linig up outside the studio door with torches,
and they don't look very happy)!!!

thanks a bunch

Rahman

Comments

edna6284 wrote on 11/6/2002, 10:30 AM

I've never done mixing like this.

You use your ears when mixing, not your eyes.

Adhering to numbers like this will only make your mixes homogenous and uninteresting.

Rednroll wrote on 11/6/2002, 3:49 PM

I agree that mixing is done with your ears and not with your eyes....but....there needs to be some starting reference point. I usually start with the Kick drum as my starting reference level. Your dB levels don't make much sense to me in the digital world, because you never want to go over 0dB and if you're aiming for +3dB then you're monitoring a bunch of digital distortion. In the analog realm (ie using a VU meter) I would start with raising the fader until my VU meter read -10dB. I then started mixing by ear from there. When I was done, you would find that the VU meter was reading 0dB with occassional peaks going up to +3dB VU....about exactly what you want. In Vegas, which has a peak meter, I raise the level of the kick drum fader so that the master meter is reading around -15dB, then start mixing levels from there. When everything get's mixed in the final mix the reading on the master PEAK meter will be between -6dB to -3dB. You want to keep the final VU level just above that
-6dB reading to give the most bit resolution from your mixes.

2) The anti-aliasing filter should be used when resampling to a lower sampling frequency. Aliasing is when a particular frequency takes on an "Alias" frequency (ie something that it really isn't). In a 48Khz sample rate the highest audible frequency that can be reproduced due to the Nyquest thereom is half of the sampling frequency...which would be 24Khz for a 48Khz sampling rate. The same holds true for a 44.1Khz sampling frequency, that it can reproduce up to 22,500 Hz. So there may be frequencies between that 24Khz and 22,500Hz that are in your 48Khz mixed wave. So what happens to these frequencies?? What happens is that they get "aliased"...which in laymon terms means they take on another frequency which is lower in frequency, therefore reproducing sound that was not intended to be there. So what was once in that 22-24Khz hardly audible area, now becomes highly audible because it's in a more distinquishable area of frequency that your ears are more sensitive too. Adding the aliasing filter will add a low-pass filter at 22Khz for the 44.1Khz sampling frequency, therefore avoiding any "aliasing" of these higher frequencies that can't be reproduced by the 44.1Khz sampling frequency.

3) This is a highly debatable area, of too dither? or not to dither? This should be the last thing you do when going from 24bit to 16bit. Which is the best method?? As I said this is highly debatable, in most cases you won't be able to tell the difference and that should be your judgement. If you can tell a big difference between your 24bit and 16 bit files then you're using the wrong dithering method.

I hope this helps....when I finally get up and running on my video editing career I expect to come over to the Vegas "Video" forum and have all my questions answered. :-)

Rahman wrote on 11/7/2002, 9:17 AM

thanks a bunch Red and edna!

Just to be clear in reguards to levels, did you say the final stereo
PEAK reading on the master fader in Vegas should be -6dB to -3dB PEAK?

I have copied wave files of some commercial releases to get an idea of
levels from them. They were recorded pretty hot (like most pop stuff today)
with Vegas and Forge reporting 0.0 dB PEAK, with an Ave RMS level of about
-10 dB.

The style and instrumentation is similar to my composition. So shouldn't I be
trying to somewhat get close to these levels? (Or is that "Mastering" comes
in)?

Also, you mentioned that an "anti-alias" filter is necessary. Especially for
down-sampling. You mentioned that the filter is only a LPF of (in this case) 22k.
If this is so, then are you also saying that it is possible to resample with a 22k LPF (probably a 3-rd party plugin) in line, and get the same result?

OR

How does Vegas handle this? I can't find any reference in HELP as to how
to incorporate "anti-alias" filtering when converting samplerate in Vegas.

Sorry again for the questions, but I want to be clear before
I charge off working and end up dissappointed again?

just a quick note to edna6284:

You are absolutely right. And, actually my mixes probably are "homogenous
and uninteresting". But then 99% of the crap on commercial radio is "homogeneous
and uninteresting" (at least to me).

thanks again for your input.

Rednroll wrote on 11/7/2002, 11:37 AM

I'm sure Vegas takes into account an "Anti-aliasing" filter when going from higher sampling rates to lower ones. I can't say for sure. Maybe instead of worrying about that you should work at 44.1Khz 24 bit files when you mix down instead of 48Khz. Have you heard frequencies between 22Khz and 24Khz? If you have, then please describe them to me what they sound like. I don't subscribe to the the higher sampling rates game. Higher bit resolution Yes, but you can leave that 96Khz sampling frequencies to the people selling you something you don't need. The human hearing is limited to a bandwidth from 20Hz-20Khz and most audio systems are developed to reproduce within this area, so anything above that I feel is unnecessary data and overhead. Also, if you stay within the 44.1Khz sampling frequency you can move audio from one device to another without having to go through an A/D converter (ie digital transfers). I haven't seen a Dat player or external CD burner that supports 96Khz and 128Khz sampling frequencies, nor would I buy one if there was. Many people will argue that the frequencies above 20Khz are "felt" and not heard. Well since you're a video guy, I believe they do similar filters with color frequencies to reduce file size. So I guess I could have the argument that the colors frequencies outside of what the human eye can see are "felt" and not seen.

As for the third party filters. The anti-aliasing filter is a digital filter with a "brickwall" cut-off. Many third party filters will have a DB/Octave slope and may therefore allow frequencies above the cut-off frequency. If you want to use a third party filter, then make sure it is a "brick wall" filter, but as I said, you're just better off working at 44.1Khz and not worrying about it at all.

As for the -3dB to -6dB level. This is the level read on the master Peak meter. Yes, you should shoot for that level and achieve the 0dB Peak /-10dB RMS level achievement to the mastering process. This will give the mastering engineer the headroom they need to do their job, because this is where the levels are achieved at for all the reference CD's you have in the mastering process.