Mixing to low levels and boosting Master output to unity(meters) the-same as mixing to unity(meters) in DAW?

[abbey road d enfer]

Of course, the input or output data cannot have 1500dB DR, it is limited by the fact that at the moment where the sample is converted it is constrained to 24 bit (approx. 140dB DR).
That's where the distinction must be made between precision, resolution and accuracy.
DR expresses the ratio between the smallest and highest numeric numbers, but does nor assume the same precision for both.
The big diference is that, with floating point, the "epsilon error" is the same for every sample that's within the DR.
In fixed point, the epsilon error percentage increases inversely proportional to sample value.

 

[john12ax7]

Of course, the input or output data cannot have 1500dB DR, it is limited by the fact that at the moment where the sample is converted it is constrained to 24 bit (approx. 140dB DR).

I think you are missing the point I was making. Stated another way,, 32 float has an effective number of bits of ~25 bits. This limit exists in the math itself and has nothing to do with input or output data and sampling conversion.

 

[tommia]

Nice thread! Keep it comming.

 

[abbey road d enfer]

I think you are missing the point I was making.

I don't think so.

Stated another way,, 32 float has an effective number of bits of ~25 bits.

Actually 24.

This limit exists in the math itself and has nothing to do with input or output data and sampling conversion.

I maintain the limit appears only when creating or extracting a sample. Internally the 1500+ DR is a mathematical fact.

 

[JohnRoberts]

I don't think so.

Actually 24.

I maintain the limit appears only when creating or extracting a sample. Internally the 1500+ DR is a mathematical fact.

Not to confuse this with TMI but human audition arguably has different short term and long term dynamic range*** due to protective muscle contraction in the inner/middle ear that reduce sound transmission via small bones in response to loud noises. Human hearing is broadly specified as something like a 130dB range but don't expect to hear both 0dB (perception threshold) and 130 dB (very loud) at the same time.

JR

*** I can't find a number for reduced dynamic range around loud sounds, and that protection mechanism is not instantaneous so loud transient sounds can do damage.

 

[abbey road d enfer]

Acoustic reflex - Wikipedia gives some rough estimation.

 

[john12ax7]

Actually 24.

No, it's actually 25. 32 bit floating point is equivalent to 25 bit fixed point in the set of numbers they can represent at the same scaling.

I maintain the limit appears only when creating or extracting a sample. Internally the 1500+ DR is a mathematical fact.

Floating point is like a magnifier. You can zoom in on very weak signals, and zoom out on very strong signals. But you can't look at both simultaneously.

You can prove this to yourself in Matlab or your favorite analysis tool. Take two signals 500 dB apart. Add them together and do an FFT. Can you resolve both? No you can't. The limit for 32 float in this type of application is about 150 dB or about 25 ENOB.

 

[abbey road d enfer]

No, it's actually 25. 32 bit floating point is equivalent to 25 bit fixed point in the set of numbers they can represent at the same scaling.

OK, it's 24bits + sign.

Floating point is like a magnifier. You can zoom in on very weak signals, and zoom out on very strong signals.

Exactly.

But you can't look at both simultaneously.

Did I say I could? However the ALU in the DAW can manage to find the exact magnifying factor that will give 24/25 bit resolution, which fixed point cannot.
In the context of this thread, recording processing a track at -150dBfs would allow perfect recovery.
More precisely, if I had a track that would stand at -150dBfs for any reason (like overcompressing it or massive negative EQ), I could put it back without loss at a level compatible with a reasonable mix.

 

[RoadrunnerOZ]

Hi,

Is this true, that in the digital realm of the DAW that creating a mix to lets say -20 on the master output(meters) and using gain to boost it to unity on the master output produces exactly the-same results as mixing to unity on the master output(meters) to begin with? I am using Logic DAW.

The reason I ask is because I have a few mixes where I mixed it way too low and I was told that in the digital realm, once it is recorded properly into your DAW, normalising, gain adjustments produces the-same results.

I did a test and I noticed that when mixing to unity, my final mix seems to be slightly harsher that when on the-same mix, I was mixing to low master level(ie. -20) and boosting gain at the master fader to reach unity. I adjusted the listening levels to be the-same in the comparison. I am still unsure as to whether it has become harsher because it is not the-same mix as I may have adjusted each channel not equally?

Another question is, if I was to raise all my channels by 5 on the faders in a mix, is it correct to say, that 5 raised on each channel fader regardless of where it starts produces the same loudness?
For example;
Kick is at -20 and raised to -15 on the channel fader.
Snare is at -5 and raised to unity 0 on the channel fader.
With all the other channels raised by 5 in a context of a mix, would that produce the-same audio quality as if I mixed that loud to begin. ie. bass, mids, treble relatively equal? I presume that things like harmonics from plugins would not matter in this situation as I am only raising the output of the result of each channel.

Thank you.

If you are using post-fader plugins on your channels then your sound will change by lifting (or dropping) your levels on the same channels. Similarly if you use post-fader plugins on your mix bus. I normally have my individual channel levels metering close to but well under 0, route drums, synths, guitars, Vox, backing Vox etc to sub-groups so the group faders are used as global controls, then these are routed to the main mix bus. I can then control what comes into the mix bus without altering lots of channel faders and having to change all the compression settings etc. on each channel with post-fade inserts.

 

[JohnRoberts]

Acoustic reflex - Wikipedia gives some rough estimation.

Thanks, the time constants associated with our meat compress/limiter are used in dynamics processing to conceal gain step changes after loud events, etc.

JR