Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[hexfix93]

http://www.cadenzarecording.com/papers/Digitaldistortion.pdf

Which masters on here know about this and accomidate for it in their sessions? i noticed that once my buddy pointed me to the rme mettering stuff, once i isntalled it and looked at my mixes, i was like whoa.. i am getting overages, so now i turn my master fade down more to make sure it never happens..

but once you smash a mix to hell with a l2, or l3, you really get overages even though it says 0.3 on the meters max in wavelab, my rme meters say its going over... so yeah.. you should really check this out. you may not hear a difference on your dacs, but you will once its burned and you listen to it in a car or on your home stereo..

And nika aldrich's book "digital audio explained" to me is as good as bob katz's "mastering audio".. they are both teaching me a lot...

[Zilla]

hexfix93 wrote on Fri, 14 October 2005 13:30

.. i am getting overages, so now i turn my master fade down more to make sure it never happens..

This is a common gain-structure misconception.  Yes, turning down the master fader reduces the output of a buss, but the buss input may well still be in overload condition.  The way to achieve a more proper gain-structure would be to leave the master fader at unity and lower the channel faders.  By optimizing the gain structures of individual elements, you will free the whole signal path of unwanted overs/clipping.  This is true for both digital and analog scenarios.

[dave-G]

I'm not sure about other DAWs, but in Pro Tools HD (TDM systems), the master fader is not analogous to a master or buss fader on an analog console in the way Scott describes -- bringing down the master fader changes the coefficient used to to make the mixed ("summed") result..  It is mathematically and sonically identical to lowering all the individual faders that feed the master.

-dave G(reenberg)

[Zilla]

dave-G wrote on Fri, 14 October 2005 14:12

...in Pro Tools HD (TDM systems)...bringing down the master fader...is mathematically and sonically identical to lowering all the individual faders that feed the master.

I performed quick test and can confirm that lowering a MF in PT does appear to be mathematically equivalent to lowering the channel faders.  That is, it does avoid clipping.  Thanks for pointing that out, Dave.  However, I cannot confirm that it is sonically identical.  Somehow the effect of a MF's coefficient acting upon PT signals does leave some sonic fingerprint.  One I would prefer to be without, should I have an option (like removing the MF altogether and just lowering the channel faders).

So, I still say that applying basic, good gain-structure methodology will provide for a higher fidelity product.

[compasspnt]

At the risk of being repetitive, I will take this opportunity to say once again that recording from the first instance at lowered levels (speaking here only of the digital domain) will make everything sound better.  There is no need to go near the red, or even the yellow.  Don't even approach the problem area.

With analogue tape machines, of course, it's another kettle of fish.

[dave-G]

Zilla wrote on Fri, 14 October 2005 18:35

dave-G wrote on Fri, 14 October 2005 14:12 ...in Pro Tools HD (TDM systems)...bringing down the master fader...is mathematically and sonically identical to lowering all the individual faders that feed the master. I performed quick test and can confirm that lowering a MF in PT does appear to be mathematically equivalent to lowering the channel faders.  That is, it does avoid clipping.  Thanks for pointing that out, Dave.  However, I cannot confirm that it is sonically identical.  Somehow the effect of a MF's coefficient acting upon PT signals does leave some sonic fingerprint.  One I would prefer to be without, should I have an option (like removing the MF altogether and just lowering the channel faders). So, I still say that applying basic, good gain-structure methodology will provide for a higher fidelity product.

I completely agree on your last point, and Terry's.  

However, just to clarify... As it has been explained by people with far deeper knowledge than mine, the master fader is merely a GUI representation; If you deleted the master fader,  hit the "all" group and lowered all the individual faders, it is doing the exact same thing as lowering the master fader.  Either action creates the same "equation" (coefficients and all) in the mix engine.  If there is a sonic fingerprint, you'd get it either way.

The only point I try to make here is that there are far better things to worry about when it comes to gain staging and clipping.

-dave G(reenberg)

[chrisj]

You mean intersample peaks? If you're getting this it's because you're trying to be too hot and too bright at the same time. You can get bright, but there's a limit to how much you can cram into PCM audio.

You should also know that the guys doing your original request of going out to an analog chain and doing EQ and some of the gain boost before resampling are going to avoid this problem, because if you're not clobbering the signal with limiting AFTER resampling, a correctly designed ADC will already  eliminate slew angles that would cause intersample 'overs'. That's because that type of distortion is 'illegal' content for PCM audio.

Interestingly, you can take it and go to DSD quite happily with no problem from any combination of samples in PCM format, and I've been working on the problem of going from DSD to PCM and back (at least in theory- got no DSD authoring *snif*) But you can't take just any samples and put them on, say, a CD. As you've discovered.

...do I really need to add that I deal with this stuff when it comes up?

[hexfix93]

my theory is that the more you push or pull the recorded audio, the worse the sound becomes. and you will get more math errors by using the channel strip faders than just turning down the master fader..

honestly you should record in at the levels where you want the audio to sit in the mix, this way you will not loose sound quality as much... mixing while recording is your best bet for sound quality, but i think messing with the master fader will be less destructive to your sound that messing with each channel..

[zetterstroem]

so you think losing resolution in the A/D conversion wich is generally about 20bits S/N is better than adjusting a fader that has 32bits rsolution or more??

[hexfix93]

to my ears, yes... i've been working with cubase, and i think my mix sounds better when the mix is hot and i must turn down the master fader...

[Ronny]

hexfix93 wrote on Sat, 15 October 2005 07:22

to my ears, yes... i've been working with cubase, and i think my mix sounds better when the mix is hot and i must turn down the master fader...

When things are operating correctly in 32 float, there is no sonic difference between the mixer faders and the master fader. You can mix hot on the channels and attenuate the master, or put the master on zero and make sure the ch faders don't clip. The only place where you should hear any difference is when you are outputting the master section above -0dBFs. I don't do a lot of DAW mixing, but on digi consoles, I often have a fader or two above -0dBFs and as long as I attenuate the master section output by the amount that the ch faders are over, no problems. In a typical DAW, which digi consoles don't exhibit, is gain increase when summing multi-tracks. You can have all faders below -0dBFs, but "the sum" of the tracks places the input of the master section above -0dBFs, so "you must" attenuate the master fader. Either way, if things are operating correctly the system still reads the data as ones and zeros at either channel or master gain structures. The real important thing is, no matter where your final mix faders are at, you must not have the master section outputting above -0dBFs.

[bobkatz]

Ronny wrote on Sat, 15 October 2005 13:50

hexfix93 wrote on Sat, 15 October 2005 07:22 to my ears, yes... i've been working with cubase, and i think my mix sounds better when the mix is hot and i must turn down the master fader... When things are operating correctly in 32 float, there is no sonic difference between the mixer faders and the master fader.

Ronny is 100% right. So it would be interesting to examine how hexfix93 reached his conclusions...


BK

[Jack Schitt]

compasspnt wrote on Fri, 14 October 2005 19:31

At the risk of being repetitive, I will take this opportunity to say once again that recording from the first instance at lowered levels (speaking here only of the digital domain) will make everything sound better.  There is no need to go near the red, or even the yellow.  Don't even approach the problem area. With analogue tape machines, of course, it's another kettle of fish.

Is how far under 0db a platform dependant decision or is there a good rule of thumb in your mind where you like to see the peaks fall?

[masterhse]

Zilla wrote on Fri, 14 October 2005 18:35

However, I cannot confirm that it is sonically identical.  Somehow the effect of a MF's coefficient acting upon PT signals does leave some sonic fingerprint.  One I would prefer to be without, should I have an option (like removing the MF altogether and just lowering the channel faders).

Could this be due to some sort of quantization error? If it's truly mathematically equivalent how can it not sound the same?

[chrisj]

masterhse wrote on Sat, 15 October 2005 21:10

Could this be due to some sort of quantization error? If it's truly mathematically equivalent how can it not sound the same?

You'd have to be able to audit the code. A single bad variable that's int when it's supposed to be 32 bit float could do this. There's C programming errors that can do this, like doing math with constants not expressed as floats- divide by 3, get int output cast to float, divide by 3.0 and you're OK...

Anyway, rather than theorize on how it could sound different with the math the same, it might be a case of the math being broken.

[hexfix93]

well i don't know if this is true for sure or not, but from what i have gathered and read, the more gain changes, up or down, it doesnt matter if its decrease or increase, you get more math errors, more data loss, and it adds up. if its just the main fader that gets the gain decrease instead of all the channel sliders, it makes more sense to me that becasue more gain decreases from all the tracks would add up more errors than just turning down the main one...

let me know if i am way off, i could be, that is just what i think so far from what i have read. because the more db you change over all, the more damaged the sound becomes..

and seriously, when i leave my channel faders alone, or turn em up a bit, and crop the main/master, it sounds louder, and more present, like there is more volume to the sounds, when i turn down all the channels, and turn the master up to 0 and mix that way, i think the sound sounds more distant.. more plastic like... just from my ears in cubase, it could be just cubase too.. i dunno how they do the math for their mixing pre master fader...

i havent done any science to prove or disprove this, its just what i hear...

[bobkatz]

Denny W. wrote on Sat, 15 October 2005 20:40

compasspnt wrote on Fri, 14 October 2005 19:31 At the risk of being repetitive, I will take this opportunity to say once again that recording from the first instance at lowered levels (speaking here only of the digital domain) will make everything sound better.  There is no need to go near the red, or even the yellow.  Don't even approach the problem area. With analogue tape machines, of course, it's another kettle of fish. Is how far under 0db a platform dependant decision or is there a good rule of thumb in your mind where you like to see the peaks fall?

You can play it safe on all platforms and deliver a 24 recording that is between -10 and -3 dBFS on the highest peak with no problems or loss of resolution or sound, and it may very well sound better at the end of the game. Especially if the mastering engineer uses any analog processing or the recording goes through any further analog stages priot to the release medium. If it is mastered totally digitally, if the files are floating point, and  if the mastering engineer uses a floating point workstation and reduces level digitally prior to conversion to fixed point and dithers to 24 bits at that time... then probably the point of "level" is academic.

BK

[HansP]

let me mention that a wave editor (e,g, audition) will show intersample overs in the waveform display, and this should eradicate any doubt about the accuracy of the particular meter tools. you can see it precisely and then you can listen to it at once, or note the cue points and listen on the test CD player how it translates.

[bobkatz]

HansP wrote on Sun, 16 October 2005 09:26

let me mention that a wave editor (e,g, audition) will show intersample overs in the waveform display

Really? For this to happen the editor has to have considerably more intelligence than my SADiE...  The waveforms are drawn from the true digital level. And even if a waveform is upsampled in your editor (that would be the only way to get intersample peaks in a waveform) how can you judge by the waveform if the intersample peak is +0.3 dB or +1.5 dB? In my opinion, only an oversampling meter can accurately judge those peaks, and furthermore, even that is open to interpretation, as the nature of the filtering that the meter uses may not produce the same degree of peaks as the filters in an MP3, broadcast, or outboarrd sample rate converter. In the end, it is an approximation of the intersample peaks that will be produced.

BK

[Ronny]

bobkatz wrote on Sun, 16 October 2005 09:39

HansP wrote on Sun, 16 October 2005 09:26 let me mention that a wave editor (e,g, audition) will show intersample overs in the waveform display Really? For this to happen the editor has to have considerably more intelligence than my SADiE...  The waveforms are drawn from the true digital level. And even if a waveform is upsampled in your editor (that would be the only way to get intersample peaks in a waveform) how can you judge by the waveform if the intersample peak is +0.3 dB or +1.5 dB? In my opinion, only an oversampling meter can accurately judge those peaks, and furthermore, even that is open to interpretation, as the nature of the filtering that the meter uses may not produce the same degree of peaks as the filters in an MP3, broadcast, or outboarrd sample rate converter. In the end, it is an approximation of the intersample peaks that will be produced. BK

I wouldn't underestimate Audition. It does a lot of stuff that progs 10 times the cost do. It's not only a viable product, but is a great entry level DAW prog and I recommend it to the newbies on my recording groups when they are looking for their first DAW editing and multi-track program.

I don't think that it accurately measures intersample peaks though, what the waveform display does is approximate the post reconstruction filter waveform, so that it does show intersample levels above two samples, I would consider this more as a post DAC reference view than an actual intersample peak evaluation, as it takes the amplitude of the waveform and approximates the curve between samples. If two samples are taken where amplitude is above these two samples, the samples will be connected by an arc, instead of a straight line like some of the better DAW progs exhibit. It's quite useful, but not not a real intersample peak indicator that is accurate to the 1/100dB.

[chrisj]

hexfix93 wrote on Sun, 16 October 2005 07:14

and seriously, when i leave my channel faders alone, or turn em up a bit, and crop the main/master, it sounds louder, and more present, like there is more volume to the sounds, when i turn down all the channels, and turn the master up to 0 and mix that way, i think the sound sounds more distant.. more plastic like... just from my ears in cubase, it could be just cubase too.. i dunno how they do the math for their mixing pre master fader...

You could be clipping in the channel faders, and liking it. Small amounts of clipping will add a sort of sparkliness to the sound and if you're stepping on peak levels relative to RMS the 'body' of the sound will seem more solid.

There are better ways to do that, though.

As far as I know Cubase is 32 bit float throughout so I really don't think you could be clipping channel faders- but I haven't seen the code, so I don't know what's up with it. Maybe you're clipping plugins? Are they pre-fader or post? I'd have thought they'd be pre-fader but what do I know?

[hexfix93]

chrisj wrote on Sun, 16 October 2005 12:47

You could be clipping in the channel faders, and liking it. Small amounts of clipping will add a sort of sparkliness to the sound and if you're stepping on peak levels relative to RMS the 'body' of the sound will seem more solid. There are better ways to do that, though. As far as I know Cubase is 32 bit float throughout so I really don't think you could be clipping channel faders- but I haven't seen the code, so I don't know what's up with it. Maybe you're clipping plugins? Are they pre-fader or post? I'd have thought they'd be pre-fader but what do I know?

It doesn't sound like clipping at all to me.. It sounds more like the wavs are just louder and more present.. not distorting, its not sparkle. It sounds closer, as to where when i turn the channel strips down, it tends to sound more distant.. I honestly think its how cubase does its math...

[bobkatz]

hexfix93 wrote on Sun, 16 October 2005 16:23

It doesn't sound like clipping at all to me.. It sounds more like the wavs are just louder and more present.. not distorting, its not sparkle. It sounds closer, as to where when i turn the channel strips down, it tends to sound more distant.. I honestly think its how cubase does its math...

Create two files, one with the input faders at +whatever and the master at -whatever and the other with the reverse. Then put the two files in line with eachother in an EDL and play and switch between them. Can you tell the difference? Do they null?

By any chance are you doing any submasters? If Cubase is using the master as a multiplier coefficient for the submasters and for the input faders instead of as a multiplier on the sum, then you can get double attenuation results when you lower the master 3 dB you get 6...  The old Sonic Solutions suffered from that one. And I just read on another forum in the pro sound web that is the case with Pro Tools. I would have to test it.

Anyway, take a test tone and confirm the output of Cubase is identical when you are comparing the two choices.

BK

[HansP]

bobkatz wrote on Sun, 16 October 2005 15:39

HansP wrote on Sun, 16 October 2005 09:26 let me mention that a wave editor (e,g, audition) will show intersample overs in the waveform display Really? For this to happen the editor has to have considerably more intelligence than my SADiE...  The waveforms are drawn from the true digital level. And even if a waveform is upsampled in your editor (that would be the only way to get intersample peaks in a waveform) how can you judge by the waveform if the intersample peak is +0.3 dB or +1.5 dB? In my opinion, only an oversampling meter can accurately judge those peaks, and furthermore, even that is open to interpretation, as the nature of the filtering that the meter uses may not produce the same degree of peaks as the filters in an MP3, broadcast, or outboarrd sample rate converter. In the end, it is an approximation of the intersample peaks that will be produced. BK

I think it's not upsampling but a spline interpolation, which seems not a bad approximation. did someone already do some short math literature about the deviance? this may be important to people who don't have an upsampling peak meter at the moment, but need at least _something_ to look into the problem. with the magnifier buttons the resolution of audition's display can be increased into positive dB values, and little overs become visible.
thx for reminding that it is an approximation, someone else may help to examine how good or bad it is...