Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: How can I get into a wav/flac file to change the sample rate? (Read 5363 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Re: How can I get into a wav/flac file to change the sample rate?

Reply #26 –
Quote
  8-bit as in what?  Sample formats of old PC games or actual consoles like the NES that had very limited hardware synthesizer capabilities?

Care to describe this lively sound that 16-bits per sample can't do?

Don’t judge my words. 8bit is a metaphor.
And, you think 16bit is capable of a “lively sound” , that depends what’s your lowest acceptable point for “lively sound”. Let’s reformat a cd ripped file into 15bit, 14bit,13bit...  and on , to find your lowest acceptable bit depth. It also depends on perception level of hearing which may differ between individuals. For me, I can hear a difference between 16bit and 24bit ,blindfolded.
The declining precision change (16,15,14,13bit)is continuous. which means ,24bit isn’t the “best” since there’s 32bit ,but is closer than 16bit to its analogue source. Engineers from the past knew those shortcomings of 16bit and tried to fix them, such as by “pre-emphasis” and “dithering”.
I didn’t say 16bit is totally bad or unhearable, it can sound pretty darn good. But with today’s technology we can improve it one step further.

Re: How can I get into a wav/flac file to change the sample rate?

Reply #27 –
Quote from: zjxgjp on
For me, I can hear a difference between 16bit and 24bit ,blindfolded.
Well, that means you can easily comply with the terms of service then. Here is a tool: https://www.foobar2000.org/components/view/foo_abx
Post logs here.

Re: How can I get into a wav/flac file to change the sample rate?

Reply #28 –
Quote from: zjxgjp on
Don’t judge my words. 8bit is a metaphor.
And, you think 16bit is capable of a “lively sound” , that depends what’s your lowest acceptable point for “lively sound”. Let’s reformat a cd ripped file into 15bit, 14bit,13bit...  and on , to find your lowest acceptable bit depth. It also depends on perception level of hearing which may differ between individuals. For me, I can hear a difference between 16bit and 24bit ,blindfolded.
The declining precision change (16,15,14,13bit)is continuous. which means ,24bit isn’t the “best” since there’s 32bit ,but is closer than 16bit to its analogue source. Engineers from the past knew those shortcomings of 16bit and tried to fix them, such as by “pre-emphasis” and “dithering”.
I didn’t say 16bit is totally bad or unhearable, it can sound pretty darn good. But with today’s technology we can improve it one step further.

The only place I can hear a difference is the 8-bit one.  You're welcome to double-blind these as the source was a 32-bit floating point file.

Re: How can I get into a wav/flac file to change the sample rate?

Reply #30 –
Quote from: Porcus on
Quote from: Chibisteven on
as the source was a 32-bit floating point file.
... and when you compare that test setup to the suggestion of #13 in this thread ...

The source is created from foobar2000's MIDI component and that does like to go over 0 dB which could be bad if you dump it as integers using the built-in converter component, the reason there's a 32-bit floating point file here in the first place.  LOL  Once the volume levels are adjusted, you can go straight to 16-bit integers.  I picked a MIDI that required no volume adjustments to prove that it all sounds the same unless it's an 8-bit recording.  32-bit floating point is proof that digital anything can go way over anything analog does as it's near infinite in both directions (loudness and quietness), so it's useful in the mixing, editing, and processing that occurs in DAWs.

Re: How can I get into a wav/flac file to change the sample rate?

Reply #31 –
Quote
. For me, I can hear a difference between 16bit and 24bit ,blindfolded.
Hydrogen audio requires  proper, level matched, statistically significant blind listening tests.  (That doesn't mean blindfolded.  ;)   It means you don't know which file/format you're listening to.)

But, a blind listening test doesn't tell you which one is better, it only means there's an audible difference.     Sometimes it's obvious, such as the quantization noise you get from downsampling to 8-bits or the loss of highs you get from downsampling to 8kHz.    Other times it's not so obvious, but usually we are comparing to an "original".

Usually we are downsampling or converting to MP3 so we know we are throwing-away data and comparing to the higher-resolution original.   Upsampling (or decompressing) doesn't restore the lost data.   If you hear a difference after upsampling it can be considered degradation, but some people may prefer the altered sound.

If you hear a difference after upsampling it could be an improvement or degradation depending on your preference, but "high fidelity" is about accurately reproducing the original sound.     So like I said before, upsampling shouldn't change the sound at all.   

Analog does NOT have infinite resolution.   Analog resolution is limited by the noise and there is no analog format that can match CD quality digital.    I don't see anybody claiming that analog VHS is better than DVD or Blu-Ray.  There are people who prefer vinyl over CD and there is a difference but most "normal people" realize that CD is better. 

Our ears also don't have infinite resolution.    Those of us who are critical listeners often over-estimate our hearing ability because we may be more picky than the average person, but blind listening tests can be humbling and very-few people have "superman" hearing.   

If you digitize vinyl it can sound exactly like the vinyl.    It's hard to prove because you can't make a record at home but if you make a vinyl record from a CD the record will sound different (there will be more noise and possibly other changes/defects).

One of our forum members did an informal A/B test comparing direct vinyl with analog-to-digital-to-analog and nobody could hear a difference.

Re: How can I get into a wav/flac file to change the sample rate?

Reply #32 –
@zjxgjp :  You've come to the right place if you want to learn and to discern good versus bad practices. But this implies that you have to accept the ways to provide these evidences. The rule 8 of the terms of services and the use of ABX tests is the central part of this community.

Now, let's analyse your claims:
- 32bits is better than 24bits:  Truth is, there is no audio hardware capable of playing 24bits accurately. So 32bits cannot sound any better than 24bits because it cannot be generated by current hardware. (24bits means that it should claim at least 144dB of SNR)

- 16bits is like a "mosaic": Truth is, audio is not digitized in "dots". Sampling theory explains that in order to properly convert digital to analog again, a reconstruction filter needs to be applied and this filter is similar to your "resampling experiment".
This recent thread here on hydrogenaudio can give you some more understanding of the concept, especially the Video by monty https://hydrogenaud.io/index.php?topic=120215.msg990773

- Upsampling/downsampling as a mechanism to improve 16bits noise floor: It is usually a good practice to use a higher bit depth when resampling, in order to preserve the resampled signal as is.  But doing 16bits44Khz->24bits88Khz->24bits44Khz will not really improve the noise floor, even if the 24bit signal isn't exactly like the 16bit signal. There might be a measurable difference (with software), but it is questionable if it will sound any different.

Re: How can I get into a wav/flac file to change the sample rate?

Reply #33 –
You cannot improve the dynamic range through resampling. Try it with an 8-bit recording containing obvious quantization noise. A higher sampling rate can give lower noise floor across a smaller frequency range, but the recording has to be digitized at that rate to begin with. It is possible to use a noise reduction process (multi-band gate), sometimes specifically designed and marketed as a "dequantizer", but the result will be worse, with uniform hiss, which does not draw attention to itself, being replaced by metallic artifacts.

Bit depth only limits the possible noise floor. At audible frequencies 16-bit at 48kHz can provide up to 120 dB of dynamic range, which is greater than most recordings and almost every listening environment. Using a spectrogram, compare the noise floor of a quality recording of your choice to dithered silence.

Resampling of picture data uses a similar process to sound, a sinc with some window to trade sharpness against tolerable ringing. I believe the "demosaicing" methods require a second dimension to guess and extrapolate edges. They couldn't do anything with a single row of pixels, which is what sound is.
 

Re: How can I get into a wav/flac file to change the sample rate?

Reply #34 –
Also, if one wants to do any digital processing of the signal - and anyone who uses the volume slider or ReplayGain does that! - then the player application should handle it. IIRC, foobar2000 works with 32 bits floating-point internally? So there is no need for the end-user to convert to a "better suited for EQing" format. It happens automagically.

And using ReplayGain, fb2k can prevent volume past the digital full scale.

Seems that someone needs to be reminded that
* No, your DAC cannot even utilize the full potential of a 32-bit floating-point signal (nor of 24-bit integer) - the DAC might accept it as input, but there is no DAC that can output even 24 bits. 32 bits is a processing tool.
* Even if your DAC could hypothetically deliver 32-bits, and your amp could, your loudspeakers could not. Even this one in the world's most quiet dead room would not be (ditherless) 32 bits.
* Even if you had loudspeakers that could, your ears could not. Hearing a full 32 bits means hearing whether a lightbulb is on or off when the sound pressure is so loud that it literally kills you. Given that the listener would be permanently deaf already, I'd call this "Overkill".