Skip to main content

Topic: Re: 32bit WAV to 32bit FLAC, what are my options? (Read 3378 times) previous topic - next topic - Topic derived from Re: 32bit WAV to 32bi...

0 Members and 1 Guest are viewing this topic.
  • krafty
  • [*][*][*][*]
Re: 32bit WAV to 32bit FLAC, what are my options?
Hi folks!

I currently have found a LP rip from a LP I have got.
Inspecting the needle drop, not the only the guy did a great job, but his equipment superseded CD and other LP versions I have from the same recording

It is a "definitive" master.

However, this guy made one stupid mistake. He put it out as WavPack 32 bit 192kHz, and my hardware can't play that.

What are my options now?

Given the fact that I know how to use SoX perfectly well, I ask:

Do I do a 24/192 conversion to make them compatible with my hardware and minimize the problems of downsampling? (Dither applied).
Do I do a 24/96 conversion to achieve the same thing? Or will it have more practicable impact on the downsampling?  (Dither applied).
Or just bite the bullet and do a 16/44.1 conversion to red book material down sample because no one will practically ever ABX something like this, even with the down sampling/missing samples and quantization errors?  (Dither applied).

Thank you for your answers!
  • Last Edit: 06 August, 2017, 11:29:07 PM by krafty

  • saratoga
  • [*][*][*][*][*]
Re: Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #1
I would check that the gain is set correctly and then convert to 16 bit 48/44.1kHz.  You can dither if you want but it won't make a difference since the source material being recorded has less than 16 bits dynamic range. 

  • krafty
  • [*][*][*][*]
Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #2
Thanks.

I'd like to put the experiment of the 24-192 conversion here.
Funny enough, WavPack encoder assigned 32-bit, but SoX is saying the precision is 25-bit.
So in this conversion, I think I lost just 1-bit, and didn't resample a thing. Right?
This is the least to make it compatible to my hardware. Now it can go to FLAC et all.
But won't I lose simply too much samples going from 192kHz to 48/44.1kHz?

Code: [Select]
D:\wav\hunting>sox -S -V 32.wav -b 24 hunting-24-192.wav
sox:      SoX v14.4.2
sox INFO formats: detected file format type `wav'
sox WARN wav: wave header missing extended part of fmt chunk

Input File     : '32.wav'
Channels       : 2
Sample Rate    : 192000
Precision      : 25-bit
Duration       : 00:37:26.51 = 431330027 samples ~ 168488 CDDA sectors
File Size      : 3.45G
Bit Rate       : 12.3M
Sample Encoding: 32-bit Floating Point PCM
Endian Type    : little
Reverse Nibbles: no
Reverse Bits   : no


Output File    : 'hunting-24-192.wav'
Channels       : 2
Sample Rate    : 192000
Precision      : 24-bit
Duration       : 00:37:26.51 = 431330027 samples ~ 168488 CDDA sectors
Sample Encoding: 24-bit Signed Integer PCM
Endian Type    : little
Reverse Nibbles: no
Reverse Bits   : no
Comment        : 'Processed by SoX'

sox INFO sox: effects chain: input       192000Hz  2 channels
sox INFO sox: effects chain: output      192000Hz  2 channels
In:100%  00:37:26.51 [00:00:00.00] Out:431M  [      |      ] Hd:2.5 Clip:0
Done.

D:\wav\hunting>

  • saratoga
  • [*][*][*][*][*]
Re: Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #3
Funny enough, WavPack encoder assigned 32-bit, but SoX is saying the precision is 25-bit.

Probably because its floating point.

But won't I lose simply too much samples going from 192kHz to 48/44.1kHz?

No.

  • Wombat
  • [*][*][*][*][*]
Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #4
It is a "definitive" master.
It is blasphemy to take any byte from the "definitive" master coming from vinyl. Get better hardware.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

  • DVDdoug
  • [*][*][*][*][*]
Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #5
1 bit out of 24-bits is somewhere below -140dB.        1 bit out of 16-bits is -96dB (that's 1/65,535).

Just to give you an idea of how quiet/small that is, try an experiment - Open your file in Audacity* (or your favorite audio editor).     At about the 30-second or 1 minute mark, drop the volume to -90dB for 5 or 10 seconds.   Then play-back at any volume you like (but don't adjust the volume after playback starts) and listen to what -90dB "sounds like".      Or, you might find it "interesting" to adjust to something less-drastic like -60 or -70dB.

Another interesting experiment is to take a 2nd random recording, reduce the volume to a very-low level and then mix it with your original recording.   That's kinda' like dither.... Dither is mixed-in noise, normally added before bit-reduction.   Personally, I've never wanted to add noise to a vinyl record.  ;)



* In Audacity, use the Amplify effect, and you'll have to do it in two steps because Audacity can only adjust-down -50dB in one step. 
  • Last Edit: 07 August, 2017, 12:33:27 PM by DVDdoug

  • krafty
  • [*][*][*][*]
Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #6
It is blasphemy to take any byte from the "definitive" master coming from vinyl. Get better hardware.

Wombat, I was just saying that from all the tested recordings, this one achieved the most pleaseant sound to my ears. Of course, people will find the many CD versions better, or not. What I said was in the subjective field. You're right. Making that statement as a rule would be a blasphemy. I don't have turn table, but I have the vinyls.

DVDoug, thanks for this information I will make this experiment.

Turns out that -50dB deems to completely DEAD SILENCE. Let alone -40dB added on top of it.
The second experiment I am not sure if I understood:

1) Take an excerpt out of it...
2) Create a second track and reduce it to a very low level volume, but still tiny detectable.
3) Then mix the two tracks to "hear" the final mess... is that right?
Updating: I did the experiment by doing a -20dB second track. Then mixed the two. I CAN'T really hear a difference, but yes, theoretically, audio/noise has been added.
Perhaps I will hear this different if dB is louder.


In your opinion:
(1) Would you go with 48000 Hz in downsampling because of it's divisible or go 44.100? Or it would never matter?
(2) What is the switch to prevent SoX to automatically dither such conversion? (I know, being a bit lazy now...)

Thank you.
  • Last Edit: 07 August, 2017, 04:23:45 PM by krafty

  • saratoga
  • [*][*][*][*][*]
Re: Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #7
Quote
(1) Would you go with 48000 Hz in downsampling because of it's divisible or go 44.100? Or it would never matter?

Doesn't matter that it is divisible.  If you're listening to a lot of CD audio, 44.1k is a fine choice just so everything can be the same.  If not, I'd probably do 48k. 

  • krafty
  • [*][*][*][*]
Re: Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #8
Yes, most of my rips, over 500 of them are 44.1 so I think 44.1k will be fine so everything can be the same, and if there is a need for a safeguard copy on CD, it's already properly done for it.

  • krafty
  • [*][*][*][*]
Re: Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #9
I'd probably do 48k. 

I'm interested to know the reasons why you would go 48kHz. I mean, your technical preference or why would you not choose 44.1. It can be picky reasons with no sense or some sense, but I still would like to hear them. Thanks.

  • Porcus
  • [*][*][*][*][*]
Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #10
I'd probably do 48k. 

I'm interested to know the reasons why you would go 48kHz. I mean, your technical preference or why would you not choose 44.1. It can be picky reasons with no sense or some sense, but I still would like to hear them. Thanks.

There was an "If not" there.
There could be good reasons for choosing the most prevalent sampling rate in your collection, as a DAC could make a noise in the moment it has to change sampling rate from one track to another.  And back in the day, some sound cards would resample everything to 48 too. (Of course, if you think you can hear above 20 kHz, then going 48 is a cheap insurance against those losses. The big waste is keeping the two octaves of noise above that, at a dynamic range more than twice the medium digitized.)

And make sure to normalize volume. Whoever records to 24 or 32 bits should stay way off the 0 dB mark upon recording - indeed, that is the reason for using so many bits.

Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #11
I'd probably do 48k. 

I'm interested to know the reasons why you would go 48kHz.

Yes, let's focus on that word Reasons. Not the usual spew from the web. Not the usual twaddle from high end propagandists.

Quote
Quote
I mean, your technical preference or why would you not choose 44.1. It can be picky reasons with no sense or some sense, but I still would like to hear them. Thanks.

There was an "If not" there.
There could be good reasons for choosing the most prevalent sampling rate in your collection, as a DAC could make a noise in the moment it has to change sampling rate from one track to another.

But might not, especially if it is between tracks and muted.

Quote
And back in the day, some sound cards would resample everything to 48 too.

To further clarify the above content, the usual artifacts of a halfwasy-decent job of resampling between 44 and 48 are by defintiion about 20 KHz, or more realisitcally, 16 KHz.

Quote
(Of course, if you think you can hear above 20 kHz, then going 48 is a cheap insurance against those losses. The big waste is keeping the two octaves of noise above that, at a dynamic range more than twice the medium digitized.)

Depending on how it is done, the big waste could be stuff that is always there whether you need it or not.

let's face it, in terms of modern hardware, we are swimming in surplus resources.  Name it and we probably already have 10x or more of it than we absolutely need in the digital domain.

The real questions relate to what maximizes our enjoyment of life?

Quote
And make sure to normalize volume. Whoever records to 24 or 32 bits should stay way off the 0 dB mark upon recording - indeed, that is the reason for using so many bits.

Normalize to what level?  The right answer is not FS because it is that last dB between -1 and FS where most of the baddies happen. 

The right answer might be more like -3 dB if you are THAT worried about inter-sample overages.

  • Last Edit: 08 August, 2017, 08:02:33 AM by Arnold B. Krueger

  • krafty
  • [*][*][*][*]
Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #12
Quote
if you think you can hear above 20 kHz

No, I can't. Probably 15 kHz maximum on a sine tone. I'm 41.

Usually the hardware desk Blu-Ray player which is connected to the Yamaha receiver is a BP-630 from LG. It does a good job with FLAC, up to 24/192 no problem. There is always a less than a second gap when switching tracks (noticeable specially with mixed songs), but I never heard any problems when switching tracks. My workstation still has a Sound Blaster Live! with EMU-101k chip (this one.) - probably the best thing I found after the Ensoniq card they used (one of the best and did not resample to 48kHz - here.)

So, actually I never encountered any problems with FLAC 24/96 or 192 in the first two mentioned hardware. I could probably have issues with the Ensoniq chip, doing 48kHz (actually I remember I couldn't play natively that content or above).

I think I don't really have any real reasons to worry about 44.1kHz or 48kHz, do I?


  • Wombat
  • [*][*][*][*][*]
Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #13
I think I don't really have any real reasons to worry about 44.1kHz or 48kHz, do I?
There is really nothing to worry about even with SoX standard settings to resample to 44.1kHz.
It uses a decent 1kHz wide transition band that nobody ever proved to be audible with music AFAIK.
Like mentioned before it may be some hardware is faulty with one samplingrate or the other.
Dither with a vinyl source indeed may be a waste so "sox --no-dither" should be safe.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

  • krafty
  • [*][*][*][*]
Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #14
I used the -D parameter, which is the same, according to the manpage.
I'm glad. I redid the vinyls at 44.1kHz. No problem. It's all good. No dither.
  • Last Edit: 09 August, 2017, 08:20:47 PM by krafty

Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #15
Wouldn't it be safer to let Sox dither as it does by default, just as a best practice? Even if the original vinyl noise is louder than the potential quantization noise, it would also be louder than the dither.

  • krafty
  • [*][*][*][*]
Re: Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #16
Wouldn't it be safer to let Sox dither as it does by default, just as a best practice? Even if the original vinyl noise is louder than the potential quantization noise, it would also be louder than the dither.

I think what the guys tried to explain above was that, if the source is a vinyl rip which the dumbass put out as an unnecessary 32/192 source, there is no benefit of dithering. I'm sure dithering will be useful in some ways, but from what I understood it's completely useless by converting to red book material, as it can even add noise to it, due to the nature of the original encoding source.

  • Porcus
  • [*][*][*][*][*]
Re: Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #17
The right answer might be more like -3 dB if you are THAT worried about inter-sample overages.
Are inter-sample peaks any issue upon (proper) downsampling?

  • saratoga
  • [*][*][*][*][*]
Re: Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #18
The right answer might be more like -3 dB if you are THAT worried about inter-sample overages.
Are inter-sample peaks any issue upon (proper) downsampling?

I don't think down vs. up makes any difference.  If the track is peak normalized, and you resample such that a new sample is created at a point that wasn't in the original waveform, there is a chance it can be above 0 dB.

  • Wombat
  • [*][*][*][*][*]
Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #19
Are inter-sample peaks any issue upon (proper) downsampling?
Audacity cowboys like us can easily see the clipped samples added after resampling.
For these you can be certain the source already had samples that go above 0 when checked with some "true peak detection".
A simple volume drop often prevents them to show them directly in tools like Audacity.
On resampling like saratoga mentioned they may show up on newly created points.
I still wait for a sample sounding more clipped when resampled.
Some DACs also seem to have more problems with nearly clipped samples.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

Re: Re: 32bit WAV to 32bit FLAC, what are my options?
Reply #20
The right answer might be more like -3 dB if you are THAT worried about inter-sample overages.
Are inter-sample peaks any issue upon (proper) downsampling?

Yes. Many seemingly-innocent operations that change the waveform such as minimum phase filtering and linear phase filtering can cause inter-sample overs.

I'm still looking for the results of  a set of DBTs showing which kinds of situations are audible and which are not.