Mastering Volume

Back to post

Comments

Rednroll wrote on 4/26/2002, 5:24 PM
Thanks for the further explanation Jeff, I was running out of breath going into further detail about this, but I'm glad you filled in the blanks for me. And for further support to your explanation, this is "why" an "Anti-imaging" filter is used. An anti-imaging filter is nothing more than a Low pass filter, with the cut-off frequency at 20KHz-22Khz. Because as you mentioned, when you put all these samples back together there is the square edges due to the spacing in the quantization levels. Well on your graph paper this looks like straight lines instead of smooth curves as you mentioned. Thus instead of having a smooth curve much like a pure sine wave (ie analog signal) would be, now you have more of a straight edged angled shape, in other words a sqaure wave (because only straight lines are permitted in connecting the dots on your graph paper as you mentioned). Square waves are "distortion" or NOISE or sinewaves lacking that warmth as you mentioned, remember I mentioned this "distortion" in my previous post? So now here's where the use of the low-pass filter (Anti-imaging filter) comes into play. If you take a sinewave and start adding infinite amount of Harmonic frequencies to it, (ie frequencies higher than 20Khz), your sinewave will eventually turn into a square wave at the same fundamental frequency. If you now run that square wave through a low pass filter, it removes all those infinite harmonic frequencies you added to your original sine wave. Thus your square wave will now return back to the sinewave it originally was, and have those rounded edges once again. Thus resembling the original analog signal it originally was. 24 bits as you mentioned is just a closer approximation to the original signal than 16 bits, thus the edges will we smaller square waves, but both still get run through the "Anti-imaging" filter to truly round off the edges, to closer resemble the original analog signal.

Alright, enough school lessons for today, I've got to get back to editing video or something like that I know more about, all this audio talk makes my head spin....lol.
VU-1 wrote on 4/26/2002, 5:51 PM
I've already finished my editing for the day. I'm off to play softball.......

........I hope I don't strike out thinking about all this "Anti-imaging, quantization, XY, bit depth stuff..........

........I wonder if you could digitize the flight path of a softball? Hmmmmm, maybe then you could make it more predictable or even pre-program...... 8=)

........Hmmmmmmmm.

JL
OTR
PipelineAudio wrote on 4/26/2002, 8:37 PM
Well thanks for that.. I remember there were options in cool edit for where you wanted to pack the bits, I was assuming the lab would have the same sort of choices.

I stand corrected

as for mashing it down to 16 bits. I am assuming the lab is using 24 bit stuff mostly and that youd need to get it to 16. Dithering and such
VU-1 wrote on 4/26/2002, 9:50 PM
<<as for mashing it down to 16 bits. I am assuming the lab is using 24 bit stuff mostly and that youd need to get it to 16. Dithering and such<<

Not necessarily. I have done mastering projects that have gone into the DAW at 44/16. It really depends on the mix. Sometimes, I will take a mix directly off of a CD-R (digital xfer) & work on it at that resolution (44/16). Other times, I will convert it to analog & run it thru my tube A/D on the way into the DAW setting the A/D to 88.2/24. It's all about "what does the song need?". There is no set chain to use when mastering a tune.

when you work with an audio file in anything other than 44/16, yes, it will need to be converted back to 44/16 to go to CD and it is a good idea to apply dithering as the last process in the signal chain in that case. There are different kinds of dither noise & which one you use depends, again, on what works best for that particular song.

How's that for a vague answer?

JL
OTR
PipelineAudio wrote on 4/26/2002, 11:32 PM
I have yet to have anything mastered digitally at least with me there. The guy I use locally always goes back to analog first, so I am way off on the digital mastering stuff I guess. Ill ask though
PipelineAudio wrote on 4/27/2002, 2:11 AM
If you ARE doing digital mastering, knowing that a FEW at least, peaks are gonna be close to digital zero, even if your average level is WAY less, what would you do?

Chance really goofy cymbal and reverb fade outs?
Or just automate out the really high peaks?

I DO remember having a few things mastered in the past that went to a digital only mastering chain, but that it certainly peaked near zero. I didnt go to these places though I would have liked to have seen it. And it was a while ago, about 6 or 7 years, so Im not too sure what stuff they had.

How did this work? the albums came back sounding great. Is there some sort of headroom available somewhere?
PipelineAudio wrote on 4/27/2002, 2:13 AM
Almost any device you use now is going to have more than 16 bits, so whether you know it or not, its gonna get mashed down to 16 bits right?
PipelineAudio wrote on 4/30/2002, 1:28 AM
Ok, finally got a chance to talk to my mastering guy.

He said do it the way I said BUT not for the reason I thought.

He said some of the equipment did have extra processing bits for more overhead but those wouldnt go thru the whole chain and were just for handling certain functions without clipping before going out to other digital devices.

The jist of it was to keep the peaks up near zero. To keep the peaks below -6 was to guarantee a maximum of 15 bits of performance, and for the most part even fewer. Especially in my case as I dont run any sort of 2 buss compressor on my stuff.

He said far better for him to have to turn it down a tiny bit than to loose bits during the analog to digital conversion.

I asked "isnt it bad to be changing volumes, even in the digital realm? "

He said " you dont have a problem when I turn it UP do you ? "

I wonder though does it make as much difference if you mix " in the box " ?
Rednroll wrote on 4/30/2002, 10:00 AM
and as I said, peaking at -6dB is a good idea, because you only lose 1 bit of resolution, yet this gives you plenty of room to make EQ adjustments and multi-band compression. Thus giving you the headroom to work with, without having to reduce the input. Peaking the same mix at 0dB or at -6dB gives you exactly the same thing, except you lose the "1 BIT" of resolution on the -6dB mix, but you GAIN the needed headroom, thus elliminate the risk of distorting peaks. Reducing the input when going from digital to analog is a definite NO NO, because you ADD noise, because the input stage of the analog device is no longer working at it's optimal 0dB level. Did your mastering guy mention that fact to you? Now staying in the digital realm, and adjusting the input stage, this no longer applies, but when you lower the input stage, you now just achieved the same thing that my method started with. If you lower the input, then this is your first step in the chain of signal processing and you thus, lowered your bit resolution. Did your mastering guy mention anything about that fact either?

I'm not saying your master guys method is wrong, but it's definitely NOT the holy scripture of mastering.

I personally think my method gives the best of both worlds if the mastering is done all digital or all analog.

PipelineAudio wrote on 4/30/2002, 11:01 AM
Im talking about when youre mastering digitally. There wouldnt be any reduction in gain at any converter.

Mastering in analog from a digital source, who cares? You got headroom for days.

I had no question on the analog side, but was trying to clear up that confused mess I made a couple of posts ago when it came to staying digital.

"If you lower the input, then this is your first step in the chain of signal processing and you thus, lowered your bit resolution"

you know, I need to ask again. I would think somehow youd now be working at a higher resolution now and turning down wouldnt be so bad, but Im not sure.
Rednroll wrote on 4/30/2002, 12:04 PM
Here's a test you can try out for yourself. Open up Sound Forge 6.0, since it has the nice plugin chainer now. Open up a 16 bit waveform and work in all 16 bit processing of plugins also. Normalize the waveform to 0dB peak. I'm working in 16 bit, to better show this example, 24 bit will take more plugins to reduce bit resolution. Open up an EQ plugin. I used the Waves Q2 paragraphic EQ in my test, but any will work that has an input and an output stage. Leave the EQ's flat. Chain multiple Flat EQ's one after the other. Use the first 3 EQ's in your chain to reduce the input stages. Reduce a minimum of -70dB...I reduced to -72dB in my setup. Now, with the next set of eq's chain them together and increase the outputs until they add up to +72dB. So now, you've effectively reduced the inputs, and then in the following plugin chains increased the output back to 0dB, effectively where the wave originally started. Play back, the plugin chain. What do you hear? You should hear a grainy sounding 0dB peaking playback. Thus, this indicates that the prior input stages reduced your bit resolution you were working in, followed by the output stages that increased this lower resolution processing back to the 0dB the original wave started at.

This is a drastic test, but proves the point that the input reduces your resolution although you started at 0dB maximum resolution.

Not bad for a "dumbazz video guy" huh?