Print

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

[FenceFurniture]

I'd really appreciate a lay-person's explanation of a few things please. (and if this is in the wrong sub-forum then would a Mod kindly move it please?)

I've certainly heard the difference between 16 Bit/44kHz and 24/176kHz, but I don't understand what the two parts of that actually do, in terms or what the add to the sound.

Bit Depth: My best estimation of Bit Depth is that it will be able to carry 50% more info from 16 to 24, and 100% more from 16 to 32, but how does that affect the sound? Louder? More "stuff" in there able to be carried (so a richer sound)? You can probably already tell I'm a Bit out of my Depth.....

Sample Rate: The explanation I've heard is that more frequent sampling knocks more rough bits off the Sine waves, creating a smoother or more accurate sound. Is that an accurate description?

I know that both parts affect the file size in different ways, and I have also observed that the sound quality improvement is not necessarily automatic just because it might be 24-192. Many (most) of the 24-192 Vinyl Rips that I've heard are just mud compared to a CD rip of the same album. By comparison the CD rips are punchier, and louder, where the VRs are choked and muddy. I understand that there are a great many more components involved in Vinyl ripping (Stylus, cartridge, Tonearm, Turntable, software.....oh yeah, and condition of the Vinyl.....).

Cheers,
FF

[Absinthe]

Here is a fairly technical writeup on sampling theory.

https://en.wikipedia.org/wiki/Nyquist–Shannon_sampling_theorem

I say technical because unless your a math major the theory isn't going to mean much to you.  In short, sampling a continuous waveform with a digital system involves taking snapshot of the analog wave at regular intervals and noting the "value" of the wave at that time.  If you take snapshots at decreasing intervals and plot the value versus time you get an approximation of the waveform. The integral of the sampling function taken at the limit (math nerd way of saying as the number of snapshots per time interval approaches infinity) you come up with the realization that you have to sample any waveform at least twice per cycle to achieve an accurate representation of the wave.  This is the Nyquist theorem and simply stated to accurately represent a wave, you have to sample it at a minimum of twice the frequency of the wave.  The human audio spectrum is typically specified to be 20Hz - 20Khz.  Although few can actually hear that broad of a spectrum, Phillips and Sony adopted these limits and set the sampling frequency for their CD encoders/players at 44Khz which in theory will produce accurate sounds as high as 22Khz.  That gives a little padding on what the "average" human can hear.  Some would argue that 22Khz will not encompass the harmonics of the music and higher sampling rates produce more accurate sound with more "body"  This is why you have Hi-Def recordings at 96Khz and higher.

Bit Depth is the number of digital bits used to represent the "value" of the wave taken at the snapshot interval.  The more bits available, the more accurately the value is represented.  Initially it was set at 14 bits but later increased to 16.  The main effect of higher bit depth is greater signal to noise ratio; that is the more bits used to represent the wave value, the greater the differentiation between the audio portion of the wave and the noise, or the differentiation between the loud and soft portions of a recording (dynamic range). 

You are correct in stating that the technical aspects of the sample do not guarantee a good or poor recording, there are many aspects of audio production that can effect the final outcome.  There are a few audio engineers on the board that can better explain this better than I can but suffice to say there are a lot of CD's out there that are absolute garbage because of the poor audio engineering. 

As for vinyl ripping, the following gives a pretty good perspective on the technical aspects of the results as far as "what I get and what I hear"  The argument on which is better has been raging since digital audio came out in the early 80's 

https://www.audioholics.com/audio-technologies/dynamic-comparison-of-lps-vs-cds-part-4

[mojave]

Bit Depth
Both 24 bit and 16 bit are at 100% volume levels. The difference is in precision, which results in a lower noise floor, not a higher volume level. For example, the number 10 and 10.000 are both identical. However, if you divide 10 by 3 you get 3 if you can't use decimal places. If you divide 10.000 by 3 you get 3.333. The higher the bit depth, the greater the precision when using DSP and volume control.

Sample Rate
Sample rate is basically units of measurement. Wavelengths vary in length based on frequency. The smaller the wavelength, the higher the frequency. The larger the wavelength, the lower the frequency. I'm going to give you some some arbitrary measurements to make it easy. Note these are for illustration and not the actual wavelengths.

1/4" = 24,000 Hz frequency
1/8" = 48,000 Hz frequency
1/16" = 96,000 Hz frequency

1/8" = 48 kHz sample rate
1/16" = 96 kHz sample rate
1/32" = 192 kHz sample rate

If I have a ruler that is marked in 1/8" increments, I can measure everything from 24,000 Hz and below.
If I have a ruler that is marked in 1/16" increments, I can also now measure from 24,000 Hz to 48,000 Hz.
If I have a ruler that is marked in 1/32" increments, I can also now measure from 48,000 Hz to 96,000 Hz.

In other words, having a finer measurement capability doesn't do anything for me if I need to measure 3', 2', 1' 6" or anything else greater than 1/8". Each ruler can accurately can measure 1/8". All sample rates 48 kHz and higher can accurately represent audio up to 24,000 Hz. Higher sample rates allow one to measure and record smaller wavelengths.

You can also think of wavelength as money. The smaller the bill or coin, the smaller the price you are able to pay with the available money.

$20 = 24,000 Hz frequency
$10 = 48,000 Hz frequency
$5 = 96,000 Hz frequency

$10 = 48 kHz sample rate
$5 = 96 kHz sample rate
$1 = 192 kHz sample rate

All wavelengths 24,000 Hz and larger cost at least $20 with increments no smaller than $10.
All wavelengths from 24,000 Hz to 48,000 Hz cost at least $10 with increments no smaller than $5.
All wavelengths from 48,000 Hz to 96,000 Hz cost at least $5 with increments no smaller than $1.

One thing about sample rate is that it all has to be in the same currency. If you are going to use 192 kHz, you can only carry $1 bills. This would be rather cumbersome, but represents the increase in file size required by higher sample rates.

[FenceFurniture]

That's great - thank you both very much! Both explanations make it much easier to understand. Totally understand the resolution now. (the money analogy was the clincher, esp $192k in dollar bills!)

So, part two of the question: I have a friend who is using ResampleV software to increase the sampling rates of various tracks from 44, 88, 96 up to 176 and 192 (all are kHz or course). He says he can hear an increase in quality (we are both early 60s, so upper end is diminishing anyway).

How achievable is this? I listened to just a few tracks, directly comparing 16-44 and 16-176 for the same sections, and I tell you what, I'm struggling to hear any difference - maybe I might with a headphone test. However, when I compare a CD to an SACD of the same music I can usually hear a significant difference in the fuller richer sound of the SACD (not always, but mostly). I presume that is because there is remastering etc going on with SACDs, which obviously does not happen with ResampleV.

He also says that RV does a brilliant job of upscaling 320kbps MP3 tracks to near FLAC quality - but I thought that (just like JPEGS for images) MP3 had missing data, which is how the files are made so very much smaller. It must be estimating what has been taken out and replacing it? I have not heard any of these yet.

[Absinthe]

If I understand you correctly, your friend is using ReampleV to "up-sample" a CD to a higher sample frequency?

If this is the case, resampling a CD only pads the final rip with more samples of the same data.  Resampling only increases the number of times you inspect the existing waveform.  On a CD, the waveform only exists as already sampled data and no matter how many times you sample that data, your not going to get any additional unique data out of it; you just now have additional samples of the same thing.

Up-sampling an MP3 file is never going to get you any closer to a lossless FLAC file so you are correct in your understanding.  Once the information is removed through sampling, the best that can be done is an interpolation of the new up-sampled data based on the sample before and after the new sample.  I do not use this software or any other type of up-sampling utility so I have no idea if this exists or not but I can't see how this would do anything to improve the quality of the recording

The primary reason for the enhanced richness and fullness of SACD's is because they are sampling the master recording at a higher rate with more bits per sample resulting in more information; getting you incrementally closer to the actual analog waveform.  They key here is the sampling of the master recording.  In their simplest form, DAC's simply take a digital value and convert it to a voltage.  The changing voltage is filtered by some sort of low pass filter (to remove artifacts caused by the sampling) and sent to an amplifier where it becomes audible.  The smaller the changes in voltage from the DAC the less intrusive the filtration required becomes and the smoother the resulting audio will be

[Absinthe]

Is this what your friend is using?

https://sourceforge.net/projects/resamplerv/

[FenceFurniture]

Thanks again Absinthe (have you ever tried it?). That all makes good sense.

If I understand you correctly, your friend is using ReampleV to "up-sample" a CD to a higher sample frequency?

Yes, and I do recall that he said he got it from Sourceforge, so that link is probably it. He's unfortunately a little dyslexic, so spelling and subsequent pronunciation is not always fully accurate (hence I thought it was ResampleV, not ResamplerV. He says he has some other software now which will allow the bit depth to increase to 24, but.....he ain't got any more info to store in the extra depth......so that just sounds like booting/padding up a Vinyl Rip to 24-192 with lots of zeroes. You can put 20 litres of water into a 20L container or a 30L container....but it's still only 20L of water.

The smaller the changes in voltage from the DAC the less intrusive the filtration required becomes and the smoother the resulting audio will be

I see. In exactly the same way as the more sides a polygon has, the closer it becomes to a perfect circle. Simple, really!

DING! (did you hear that penny drop?)

The primary reason for the enhanced richness and fullness of SACD's is because they are sampling the master recording at a higher rate with more bits per sample resulting in more information; getting you incrementally closer to the actual analog waveform.  They key here is the sampling of the master recording.

Yes, getting it closer to that perfect circle. Of course equally as important is the quality of the original recording (studio quality, tape quality, microphones, skill of the Engineers etc etc). A SACD of early Miles Davis will only ever be an improvement on the original - can't be turned into late 20th C quality of even a normal 16-44 CD that is beautifully recorded and engineered.

[FenceFurniture]

I guess the ultimate test of something like ResamplerV is to start with a good quality FLAC file, convert it to MP3 320kbps, and then upscale it with RV and compare that to the original FLAC (and the MP3 of course).

[Absinthe]

I guess the ultimate test of something like ResamplerV is to start with a good quality FLAC file, convert it to MP3 320kbps, and then upscale it with RV and compare that to the original FLAC (and the MP3 of course).

I've mistakenly converted music for my car audio from 320MP3 to FLAC and even in the car it was obvious that the FLAC file playing was just an MP3 file in a better suit.  Perhaps ResamplerV has some special sauce that makes it work.  Give it a go and see for yourself; music is so subjective you may not hear much difference

[FenceFurniture]

I don't have any MP3 files, that I don't have in FLAC, so it's of no real interest to me. In fact I only created those MP3 files so I could fit some music on my phone, but have since purchased a DAP so don't need them at all.

[JimH]

I've mistakenly converted music for my car audio from 320MP3 to FLAC and even in the car it was obvious that the FLAC file playing was just an MP3 file in a better suit.  Perhaps ResamplerV has some special sauce that makes it work.  Give it a go and see for yourself; music is so subjective you may not hear much difference

Please don't do that, regardless of what anyone says.  You won't get a better file.

[Awesome Donkey]

Lossy to lossless is a terrible idea and a waste of space. Like Jim said, you won't get a better file at all.

[JimH]

Closing this because it is spreading opinion as if it were fact.