Digital Lossless Music Store?

[norritt]

Hey everyone,

 

just wanted to hear if anyone can recommend a good, DRM-free plattform to purchase FLAC encoded music?

[manikyath]

pretty sure bandcamp supports flac.

[MaximusX]

pretty sure bandcamp supports flac.

Yes all music on bandcamp flac.

[KaminKevCrew]

Or HDTracks, if you like. Although their selection can be fairly limited, depending on what you like to listen to.

[GoldSrc]

You got some suggestions in the previous comments.

Some people will hate me but "high resolution" audio or loss-less audio is pointless for your average listening.

[KaminKevCrew]

You got some suggestions in the previous comments.

Some people will hate me but "high resolution" audio or loss-less audio is pointless for your average listening.

Anything above 44.1 is pointless.

[GoldSrc]

Anything above 44.1 is pointless.

I know, funny thing is I used to think that a higher sampling rate made a difference not too long ago.

[norritt]

@manikyath: Thanks for the hint to bandcamp, they have a couple of nice tracks there also I think it's good to support the artists directly.

 

 

Or HDTracks, if you like. Although their selection can be fairly limited, depending on what you like to listen to.

 

Also didn't know that one yet. Yes the limited offers are a big problem for the FLAC stuff, especially I mostly stick to older music. Most of the new (mainstream) stuff coming out is terrible and if its any good its 99% a rip-off of a 80s/90s band that originally did it better.

 

Genere varies a very incomplete list very roughly gouped by genre:

 

Hard-Rock/Metal:

- Black Sabbath (especially recodings of non Ozzy periods - like the albums Tyr, Heaven and Hell or Headless Cross, but also the classics like Iron Man or Paranoid ofc)

- Judas Priest

- Led Zepplin

- Iron Maiden

- Hammerfall

- Masterplan

...

But also some newish mainstream stuff like:

- Nickelback

- Some Linkin Park recordings (the non Rap/HipHop stuff)

- The Offspring

- Nightwish (Tarja era)

- Evanescence

- Poets of The Fall/Old Gods of Asgard

 

 

(Mainsream) Pop Rock mostly 80s stuff like

- Queen

- Roxette

- Toto

- Joe Cocker

- Michael Jackson

- Some Mike Oldfield stuff

...

also a bit of the electronic synthe-corner

- Jean-Michel Jarre

- Chris Huelsbeck

...

 

Newer mainstream Pop:

- Anastacia

- Alicia Keys

- The Corrs

 

Folk-Rock, Celtic

- Blackmores Night

- Loreena McKennitt

....

and probably tons of stuff I forgot to mention.

 

I know, funny thing is I used to think that a higher sampling rate made a difference not too long ago.

Actually it does but not for the better. My on board-sound provides settings all the way up to 24Bit 192kHz, but anything above 16Bit/44.1kHz will actually sound worse. Its kinda self-explanatory if you play Music recorded at 44.1kHz at 192 kHz the sound card has to make up (interpolate) 75% of the samples only 25% you hear was recorded the rest is pulled out of thin air using statistics.

[KaminKevCrew]

@manikyathIts kinda self-explanatory if you play Music recorded at 44.1kHz at 192 kHz the sound card has to make up (interpolate) 75% of the samples only 25% you hear was recorded the rest is pulled out of thin air using statistics.

That's not actually how that works...

[norritt]

@KaminKevCrew If you have 44100 samples per second and upsample this to 192000 samples per second you have to somehow interpolate avg. ~3.4 samples between two samples on the original recording, thats the only way it can work - you can't automagically get the missing samples that aren't there in the source material back.

[KaminKevCrew]

@KaminKevCrew If you have 44100 samples per second and upsample this to 192000 samples per second you have to somehow interpolate avg. ~3.4 samples between two samples on the original recording, thats the only way it can work - you can't automagically get the missing samples that aren't there in the source material back.

Again, that's not how that works.

[norritt]

Again, that's not how that works.

Yes this is exactly how it works, if you don't believe me just check out the wikipedia article.

[Guest]

@KaminKevCrew If you have 44100 samples per second and upsample this to 192000 samples per second you have to somehow interpolate avg. ~3.4 samples between two samples on the original recording, thats the only way it can work - you can't automagically get the missing samples that aren't there in the source material back.

So don't play it at 192kHz. No sample rate conversion necessary.

[norritt]

@SSL: I don't. I was picking up GoldSrc's comment and trying to explain why upsampling is pointless or worse counterproductive in this case.

[Guest]

@SSL: I don't. I was picking up GoldSrc's comment and trying to explain why upsampling is pointless or worse counterproductive in this case.

 

But it isn't making it up out of thin air.

 

Sound is just sine waves, every possible combination of sine waves can be reconstructed from a bandwidth limted source based on Nyquist sampling theorem. Upsampling is pointless, and requires resampling, but audibly it will not be appreicably different.

[Patrick3D]

PonoMusic store, not a great selection but they continue to grow their catalog.

[KaminKevCrew]

Yes this is exactly how it works, if you don't believe me just check out the wikipedia article.

SSL's comment is why I was saying thats not how it works.

But it isn't making it up out of thin air.

Sound is just sine waves, every possible combination of sine waves can be reconstructed from a bandwidth limted source based on Nyquist sampling theorem. Upsampling is pointless, and requires resampling, but audibly it will not be appreicably different.

[norritt]

But it isn't making it up out of thin air.

 

Sound is just sine waves, every possible combination of sine waves can be reconstructed from a bandwidth limted source based on Nyquist sampling theorem.

Yes correct but Nyquist comes into play later when transforming the discrete signal back to continous one that is emitted by your speakers. But to apply this you still need samples to reconstruct the analogue signal from the digital one. As a general rule of thumb more samples are better, since this allows the continous signal to be determined more precisely. Ofc at a certain point more samples are useless. In practice you have to make sure that the sampling theorem f_sample >= 2*(f_max - f_min) is fulfilled.

 

Since the human hearing ranges from 20Hz to about 20kHz, you'll need a sample rate of at least 39960Hz (or a bit more if you assume some humans have better hearing beyond that range). More is not needed. I'm also pretty sure there is a story behind why the sample rate ended up at exactly 44100Hz.

 

Edit: Just found the above mentioned story also on wikipedia:

 

Firstly, because the hearing range of human ears is roughly 20 Hz to 20,000 Hz, and via the Nyquist–Shannon sampling theorem the sampling frequency must be greater than twice the maximum frequency one wishes to reproduce, the sampling rate therefore had to be greater than 40 kHz. In addition to this, signals must be low-pass filtered before sampling, otherwise aliasing occurs, and, while an ideal low-pass filter would perfectly pass frequencies below 20 kHz (without attenuating them) and perfectly cut off frequencies above 20 kHz, in practice a transition band is necessary, where frequencies are partly attenuated. The wider this transition band is, the easier and more economical it is to make an anti-aliasing filter. The 44.1 kHz sampling frequency allows for a 2.05 kHz transition band.

 

Anyways the upsampling has nothing to do with the DA-conversion but it just works on the discrete signal. And there it just interpolates values that aren't there. There are different techniques to *create* those samples range from simply copying the previous sample n times, over calculating mean values to some fancy filters. But all of those try to find sample values that were *probably* there.

 

@Patrick3D thanks for the hint on PonoMusic!

[GoldSrc]

Actually it does but not for the better. My on board-sound provides settings all the way up to 24Bit 192kHz, but anything above 16Bit/44.1kHz will actually sound worse. Its kinda self-explanatory if you play Music recorded at 44.1kHz at 192 kHz the sound card has to make up (interpolate) 75% of the samples only 25% you hear was recorded the rest is pulled out of thin air using statistics.

Still, for music is pointless to go over 44.1/16, I only know of a few applications where sample rates over 44.1KHz is needed.

The frequency range that a higher sample rate gives you is pointless for us humans, most of us are not going to hear anything beyond 20KHz.

 

I don't know if you already watched this, but this cleared up a lot of questions I had about "high-resolution" audio.

[Guest]

As a general rule of thumb more samples are better, since this allows the continous signal to be determined more precisely.

 

NO IT DOES NOT.

 

As long as the signal is bandwidth limited to half of the sampling frequency, the original signal can be reconstructed perfectly. That means with no loss or approximation. Adding more samplings simply increases the bandwith, not resolution. Get it?

[norritt]

NO IT DOES NOT.

 

As long as the signal is bandwidth limited to half of the sampling frequency, the original signal can be reconstructed perfectly. That means with no loss or approximation. Adding more samplings simply increases the bandwith, not resolution. Get it?

Did you read my post that is exactly what I said, let me quote myself:

 

Ofc at a certain point more samples are useless. In practice you have to make sure that the sampling theorem f_sample >= 2*(f_max - f_min) is fulfilled.

 

[Guest]

Did you read my post that is exactly what I said, let me quote myself:

 

Did you read your post? Let me quote you:

 

As a general rule of thumb more samples are better, since this allows the continous signal to be determined more precisely.

 

False!

 

[norritt]

 

Did you read your post? Let me quote you:

 

 

False!

 

Not it is true:

 

2kHz will allow better reconstruction than 1kHz

4Khz will allow better reconstruction than 2Khz

8kHz will be better than 4kHz and so on

 

So a higher sampling rate is better than a lower one until you hit 2 times the frequency spectrum you want to reconstruct. And this is exactly what I said.

 

But again all of this has nothing to do with upsampling. What you're describing here is taking a digital signal and reconstructing an analog signal.

Upsampling takes a digital signal and reconstructs missing samples. The ouput is again a digital signal, that has more sample than the original signal. This has *nothing* to do with DA-conversion or the Nyquist theorem. And all of these additional sample values are not result of a perfect reconstruction but just an interpolation.

[Guest]

Not it is true:

 

2kHz will allow better reconstruction than 1kHz

4Khz will allow better reconstruction than 2Khz

8kHz will be better than 4kHz and so on

 

You have idea what you're talking about, apparently.

 

Define "better". If you mean higher frequencies, then yes. If mean more "resolution" or "accuracy", then no. You cannot construct any frequency at all above the nyquist frequency; that results in aliasing. Below that frequency, you have perfect reproduction. It can't get any better than that.

[norritt]

You have idea what you're talking about, apparently.

 

Define "better". If you mean higher frequencies, then yes. If mean more "resolution" or "accuracy", then no. You cannot construct any frequency at all above the nyquist frequency; that results in aliasing. Below that frequency, you have perfect reproduction. It can't get any better than that.

 

Better means you're loosing no frequencies, just take some music and recode it with 8kHz sample rate, you'll hear the difference.