[oceter]

Hi, i hope some of you audio lovers will give ma an answer to few questions. First i have to mention that i'm reading a lot about it from this site and other's but don't have a straight clean answer to few things.
 
1. The digital signal coded in 01 system (binar) gives a more square soundwave, but when its coming from dac it is normalized to smooth linear wave yes ? So if it happens, the dension of sample rate and bit depth will determine how precise it will be transformed into linear wave due to original analog wave.
But the Nyquist theory about half of sample rate determines that if we want the highest audible frequency we need about 40 000 hz to hear 20 000 hz. Am i thinkig good that it's beacause of that transformation into digital code and that high frequencies are faster, so they need higher sampling rate to catch more curve angle ? So if 44.1 khz gives as enough dense sampling to reproduce good sound, what will higher sampling rates give ? We don't hear more than 20 000, so what's up with it ? Is the lost of the non audible for human frequencies will change the reproduction of analog signal from digital and distort, change the 0-20 000 frequencies ?
2. The bit depth give's us even more dense code to transform into analog, so more precise. But if it gives better dynamic range (higher in db) what does it mean ? If we have higher dynamic range it will play more maximum and minimum decybel sounds or it will just reproduce the same sounds, frequencies but they will more vary in loudness. I mean for example the sound of 1 db and 96 db with better dynamic range will be now 1db and 100db.
3. Is neutral sound described also as flat of frequencies is more realistic due to original recording ?
 
 
About vinyls, does the vinyl record have linear waves or is it also changed in some way as digital ? If not, it should have more realistic reproduce.

 

[MindsMirror]

1. It sounds like you understand the basic concept. In theory, digital audio represents an impulse train, not a square wave. It is very difficult for a DAC to produce an impulse, so they may use the "sample and hold" method instead, which produces a square wave. This is also called linear zero-order interpolation. The steps in a square wave contain very high frequencies, so another filter is applied to filter those frequencies which are above half the sample rate.
Higher sample rates won't really give you anything audible except the ability to use a softer filter. If you have a 44.1KHz sample rate, the filter needs to pass frequencies up to 20KHz and cut frequencies above 22.05KHz, which is only a 0.1025 octave transition band. If you sample at 96KHz, you have a 2.4 octave transition band. It is easier to build a filter which has a wide transition band, so while 44.1KHz sample rate is enough in theory to perfectly represent the audio, the 96KHz sample rate might still give better results. 44.1KHz is usually enough though, that sample rate was chosen for CDs because it has a sufficiently wide transition band.
 
2. Changing the bit depth changes the noise floor. The maximum level is 0dBFS, and if you have 96dB of dynamic range, the noise floor will be at -96dB. If you have a sound at -96dBFS and a sound at -1dBFS, the -96dBFS sound will be mostly covered up by the noise floor. If you increased the dynamic range to 102dB by adding one more bit, the -1dBFS and -96dBFS sounds would still be in the same place, but now the -96dBFS sound is not covered up as much by noise because the noise floor is lowered to -102dB.
 
3. I'm not sure what you're asking here.
 
Vinyl is a continuous analog wave, but that does not mean that it is better or more realistic than digital. Digital audio becomes a continuous analog wave when you listen to it, so to compare vinyl and digital, you have to compare the analog waves. Vinyl has lower dynamic range, worse frequency response, worse crosstalk and distortion, surface noise, and can have dust and scratches which cause pops and clicks.

 

[oceter]

Do i understand correctly that noise floor can cover the quietest sound ? So the dynamic range only changes that ?

 

[RRod]

1) A DAC using sample-and-hold both has to correct for the high frequency content and for the sinc drop-off in amplitude as frequency increases in order to reproduce the correct waveform. Oversampling helps with both these things, though these days most common DACs are also using sigma-delta.
 
2) Extra bits means less rounding error, basically. A simplified way to see the PCM signal is:
original signal + rounding error
When the original signal is loud relative to the rounding error, you don't notice the latter. But for soft signals, the rounding error can start to become noticeable. Without dither, the error manifests as harmonic distortion (everything starts to look like a square wave as you get low). Dither de-correlates the error from the signal and makes the error sound like hiss. More bits means less rounding error across the board (though there are some minutiae between float and integer formats).
 
3) A flat frequency response is the standard for mixing and mastering material via speakers. Since headphones are directly on the head, they can't put out a flat response and give what you would expect to hear on the speakers. See:
http://www.innerfidelity.com/content/headphone-measurements-explained-frequency-response-part-one

 

[oceter]

2) Soft signal is the quiet sound yes ? SNR is telling us how this distortion is loud in contrast to original signal ? As i read now, the dynamic range is the difference beetwen the loudest and quietest sound, so the low bitrate not only changes the noise but also give more difference beetwen those two, for example the lowest volume played note and loudest on recording (i think this part is wrong, and i should understand that the higher dynamic range is covering by loudness the noise floor which is quiet)
 
 
Edit. The components used in DAP's also can affect frequency respose characteristic (i mean boosted bass, cutted highs etc) ?
These are the general distortions made by equipment, not only harmonic. It's what makes changes in sound reproduction of different amp's and dac's ?

 

[RRod]

  2) Soft signal is the quiet sound yes ? SNR is telling us how this distortion is loud in contrast to original signal ? As i read now, the dynamic range is the difference beetwen the loudest and quietest sound, so the low bitrate not only changes the noise but also give more difference beetwen those two, for example the lowest volume played note and loudest on recording (i think this part is wrong, and i should understand that the higher dynamic range is covering by loudness the noise floor which is quiet)
 
 
Edit. The components used in DAP's also can affect frequency respose characteristic (i mean boosted bass, cutted highs etc) ?
These are the general distortions made by equipment, not only harmonic. It's what makes changes in sound reproduction of different amp's and dac's ?

 
It's just different ways of thinking of the same thing. If you think about lower bit-rates just being truncations of the higher bit rates, you can see the equivalence. For instance, think of the implications of making 16 bits from 24 by lopping off the last 8 bits.
 
DAPs can have all kinds of deliberate coloration in the DAC and amp stages. High output impedance from the amp, for instance, can audibly affect the response of low-impedance dynamic headphones.

 

[MindsMirror]

  2) Soft signal is the quiet sound yes ? SNR is telling us how this distortion is loud in contrast to original signal ? As i read now, the dynamic range is the difference beetwen the loudest and quietest sound, so the low bitrate not only changes the noise but also give more difference beetwen those two, for example the lowest volume played note and loudest on recording (i think this part is wrong, and i should understand that the higher dynamic range is covering by loudness the noise floor which is quiet)

I think you are misunderstanding something. If you have a recording with 96dB of dynamic range which has a quiet sound at -90dB, making a second recording with 102dB of dynamic range will not change that sound, it will still be at -90dB. The 102dB dynamic range would allow you to record an even quieter sound that would have been covered up by the -96dB noise floor. It won't make the -90dB sound quieter.

 

[dprimary]

I believe you need at least 40,001Hz to get 20,000Hz. You have to have greater then 2 times sampling rate to reconstruct. Think back to graphing curves in school. You have on one axis the sample   rate (40,001Hz) it does not vary. You graph the curve for 20k using only two points, the math only allows one possible answer to go through the two points. Now you can increase the sample rate to 80,002 and you will have four points all four fall on the curve the same you have not increased resolution, only your bandwidth has increased since you can now go up to 40,000Hz.
 
The bit depth is the how many levels up and down those points on the curve can have. More levels equal better resolution. 16 bit is over 65,000 possible points per sample. Take a sheet of paper draw line down the middle and make 32,000 dots going straight up and 32,000 going straight down. It is pretty hard to find a pencil that makes dots that small. Maybe I will use a laser printer 600dot per inch 11" paper I'm close. Now draw a nice sine wave with your analog pencil. Look at your "digital" dot size to your pencil line. Maybe you want to trade in that blunt #2 to for nice hard and really sharp pencil. 
 
Signal to noise is to ratio your signal level to your noise level. Dynamic range is maximum undistorted signal to minimum  perceivable single down in the noise. You can hear signals down in noise with both analog and digital.

 

[oceter]

 ? Bandwith with more than 44.1 khz will be really hard to notice if not at all. So bit depth only changes only audible noise floor that is covered by music anyway as shown on this video : https://www.youtube.com/watch?v=-Bd99cADk70 yes ?

 

[oceter]

Maybe someone else will want to understand the bit depth effect on our music files lets look ath this best explanation : http://youtu.be/nLEhfieoMq8

 

[castleofargh]

yeah you get the main idea from those videos. I remember posting about something not quite right in lachlian's vid, but I don't remember what it was. still he has a better general understanding than many many "expert" audiophiles.
as you seem to be genuinely interested(that's a fresh change and we'll always be ready to help with what we can), maybe you'll find interest in those vids too http://xiph.org/video/
there is a loooooooot more to learn about how digital processing really works, but those do cover the general understanding very well and don't mix false claims into seemingly sciency stuff.

 

[dprimary]

For the most part those are correct Lachlian's start to get off track about 7 minutes into the video. Increased bit depth reduces your quantization errors which gives you lower noise. In audio SNR and resolution are considered the same high signal to noise is equal to high resolution it is being analog or digital is irrelevant. I can't say that higher bit depth increases the dynamic range of the recording. The dynamic range of the recording is controlled by many other factors before bit depth even become an issue. It does increase your possible dynamic range. Then he starts talking about sound in a vacuum????? 
 
I would have liked him to show the 4bit with dither. People are amazed when you play them a properly dithered 8bit recording. Around 12 bits most people have a hard time telling it is 12bits until you turn it up.

 

[stv014]

  1. It sounds like you understand the basic concept. In theory, digital audio represents an impulse train, not a square wave. It is very difficult for a DAC to produce an impulse, so they may use the "sample and hold" method instead, which produces a square wave. This is also called linear interpolation. The steps in a square wave contain very high frequencies, so another filter is applied to filter those frequencies which are above half the sample rate.

 
Sample and hold is not the same as linear interpolation. If we consider all samples being represented as "infinitely" short impulses (so that in the frequency domain the passband is mirrored between the Nyquist frequency and the sample rate, and then this repeats infinitely at Fs intervals with no roll-off) as having no reconstruction filter, then sample and hold is a filter with a square impulse response, and linear interpolation has triangle impulse response (length = 2 samples), while the impulse response of the ideal reconstruction filter is the sinc (sin(x) / x) function.
 

Higher sample rates won't really give you anything audible except the ability to use a softer filter. If you have a 44.1KHz sample rate, the filter needs to pass frequencies up to 20KHz and cut frequencies above 22.05KHz, which is only a 0.1025 octave transition band. If you sample at 96KHz, you have a 2.4 octave transition band. It is easier to build a filter which has a wide transition band, so while 44.1KHz sample rate is enough in theory to perfectly represent the audio, the 96KHz sample rate might still give better results. 44.1KHz is usually enough though, that sample rate was chosen for CDs because it has a sufficiently wide transition band.

 
The width of the transition band is typically a few kHz for real DAC filters at 44100 Hz sample rate, but they also allow some imaging, and do not cut everything already above 22.05 kHz. For example, it is about 4 kHz for the PCM1792, and 8 kHz for some Realtek HDA codecs I have tested, with 6 dB attenuation at 22.05 kHz and linear phase response in both cases.

 

[oceter]

In audible difference , does bit rate will afect not only signal to noise ratio but also how softer notes are noticeable?

 

[RRod]

  In audible difference , does bit rate will afect not only signal to noise ratio but also how softer notes are noticeable?

 
More bits allows for a bigger difference between softest and loudest. But there isn't any music I know of that actually makes use of a full 16 bits of dynamic range, and much of it uses less than 16.