Levels and gain staging

[gsHarmony]

 

You're mixing bit depth and resolution. In the colorful world of pictures, the 24-bit refers to the color of the pixel and resolution to the number of pixel (which would be like the bit rate). THe thing is that those 24-bit actually contain 3 different information: the level for 3 different colors (I'm guessing it's 3 groups of 8 bits, one for red, one for green, one for blue... I'm just guessing).

 

 

I think you misunderstood me. I wasn't mixing bit depth and resolution. Bit-depth in the time domain, along with the maximum range defines the resolution. For an image there are two resolutions, one defined by the number of pixels (this is the resolution that is commonly reffered to) and one defined by the maximum color/brightness range divided by the number of quantization levels. So, you are right in saying that 24-bit refers to the color (usually divided up into RGB or luminance/chrominance). However, this 24-bit depth defines the resolution of the color (I wasn't saying it defined the number of pixels). This latter definition of resolution is the DSP definition for A/D converion.

[gsHarmony]

 

In the sound world, those 24-bit are purely there to represent the level of the signal. If the ADC converter receives the maximum voltage it is designed to received, all 24 bits will be equal to 1. If it receives 0 voltage they will be equal to 0. So yes the 24-bits are always used. But it doesn't mean they are effectively used.

 

 

This is exactly right. However, effectively using the bits doesn't change the resolution. The quantization error still ranges from +/- 1/2 LSB regardless of the level.

[gsHarmony]

I think I have figured out the problem here. I think we are all actually arguing the same point, just using different definitions of the same terms, which is causing some confusion.

 

I think what you guys are saying is that using a higher level, the decibel amount of the quantization error becomes a smaller percentage of the overall decibel level of the measured signal, so when sampling at higher levels you get a more accurate waveform. I agree with this completely.

 

The reason I was confused is that you were referring to the accuracy as resolution (I was thinking in terms of the DSP definition, whereas I think everyone else was referring to the accuracy or the percentage error in the waveform. Hence, the confusion over the resolution of an image).

 

So, to conclude, we are all correct. Recording at higher levels reduces the percentage error in the waveform. In DSP terms though, the resolution and the range of the quantization error are constant for all levels.

[guitarbth]

So, the bottom line is... record as hot as possilbe leaving just enough headroom for peaks?

 

Brandon

[Kiwiburger]

I think the bottom line for me now, is that it is important to find a balance between the sweet spot for our analog gear, and the sweet spot for digital. Optimising for one, without considering the other, will result in lower quality sound.

 

Analog gear is (hopefully) designed with a lot of headroom above the nominal 0dB VU. If we set the gain too low, noise becomes a factor. If we set the gain to hot, distortion becomes a factor (maybe a desirable factor - maybe not).

 

Digital processing is very different - bigger numbers mean greater 'bit depth' and a more accurate sound - provided you don't run out of numbers and clip. With fixed bit resolution, it's possible to run out of numbers and clip at 0dB. With floating bit resolution, it's not likely that you could run out of numbers and clip - BUT - external factors are more likely to cause problems.

 

A/D and D/A converters are obviously a very critical point in the whole chain. Having both analog and digital circuitry, it's obviously going to have a combination of the two sets of problems. So chances are - the sweet spot for the converters isn't going to be "as close as possible to 0db FS" for pure numbers. It's more likely to be lower down at the designed 0db VU level. What is that optimimum 0db VU level in dB FS terms? Only the maker should know - so consult the specs for the converter.

 

Clarifying this in my own head has made me re-think some of my practices, and i'll probably be tweaking things a little lower and relaxing a little more.

[where02190]

 

Originally posted by guitarbth So, the bottom line is... record as hot as possilbe leaving just enough headroom for peaks? Brandon

 

 

The bottom line is record your nominal levels to the nominal 0dbu reference point of your converters.

[Billster]

With digital recording, my feeling is that the average level should be a bit lower to take advantage of the wider dynamic range. Any overload will sound ugly, not like analog where a small overload progresses through a mild clip before becomining ugly.

 

Having a lower average in digital doesn't hurt the S/N as it would with analog tape (surface noise, hiss, etc.)

 

Having a lower average when tracking gives you more headroom for the summing of the output level when you mix, again allowing more dynamic range for the final mix.