Best place to get lossless wav music?

[rippy4500]

I want to get better music and idk where to get it. I use spotify right now, i just use it because its free and it has all of the music i want. But the highest quality i can set (very high, i think its 320kbps but im not sure) is the highest i can go and there is no lossless option. So whats the best equivalent to spotify that has lossless wav format?

[Spuriae]

Apple Music is the cheapest lossless streaming, but doesn't support lossless in Windows. Qobuz is generally the highest audio quality. Others are a mixed bag.

[startrek03]

In most EU countries it would be legal to make private copies of CDs but since you're living in the US that isn't an option for you.

[rippy4500]

45 minutes ago, Nimrodor said:

Apple Music is the cheapest lossless streaming, but doesn't support lossless in Windows. Qobuz is generally the highest audio quality. Others are a mixed bag.

i might try qobuz, 10 dollars per month sounds pretty good. Does it have most of the songs that spotify does?

[OfficialTechSpace]

5 hours ago, rippy4500 said:

i might try qobuz, 10 dollars per month sounds pretty good. Does it have most of the songs that spotify does?

Qobuz is good, but has a rather limited library by comparison to an alternative like Amazon Music or Tidal. Tidal has the best user interface, Amazon Music has the best hi-res selection. Qobuz is bottlenecked by its library, Tidal is bottlenecked by lossy MQA encoding in the 'HiFi Plus' tier, and Amazon Music is bottlenecked by the user interface. Tidal does offer a lossless CD-quality tier which streams in FLAC, but that's about as far as it goes. Although, unless you "need" access to 24-bit files (most don't), Tidal HiFi is my recommendation.

 

Tidal: up to 16-bit/44.1kHz in HiFi tier, $9.99/mo

Amazon Music: up to 24-bit/192kHz in HD tier, $9.99/mo

 

Also, I think you may be mistaken as to what lossless is. There are many lossless codecs, WAV is only one of them and not the only component. There are also variable bitrate lossless codecs like ALAC, and different bit-depths which offer varying levels of dynamic range. WAV is just a less efficient format, and takes up more space than a codec like FLAC while offering no benefit. This is why music collectors (like myself) choose to store our libraries largely in FLAC. I personally have close to 10,000 songs, totaling 273GB.

[cmndr]

Quick question, does this even matter to you?

https://www.npr.org/sections/therecord/2015/06/02/411473508/how-well-can-you-hear-audio-quality

 

If you're not able to differentiate, then you're asking the wrong question.

If you're using headphones... funds are usually best spent on better headphones.
if you're using speakers... speakers and/or room treatment.

 

For what it's worth Spotify 320Kbps is reasonably solid. Amazon DOES have atmos based sound though, which is nice for surround sound set ups.

 

On 5/6/2022 at 11:26 PM, OfficialTechSpace said:

Although, unless you "need" access to 24-bit files (most don't)

 

Tidal: up to 16-bit/44.1kHz in HiFi tier, $9.99/mo

Amazon Music: up to 24-bit/192kHz in HD tier, $9.99/mo

 

Some bits to add to this - those specs listed don't really matter unless you're trying to MAKE music. Listening it's "good enough".

The difference between 16 vs 24 bits is how much dynamic range is supported without loss of quality. I believe 16 bit is good up to 96dB. In order for a person to actually perceive that full dynamic range of a 16 bit stream you'd need music that itself has a high dynamic range (so 96dB between the loudest and softest stuff). For the sake of argument, let's say you can't hear below 20dB (half as loud as a soft whisper) or so (unless you're in an anechoic chamber). This would mean you'd need to have the peaks in the music hit 116dB. This is just above the threshold that gets classified as "DANGER PERMANENT HEARING DAMAGE" - it's really unpleasant to play that loud. For what it's worth, dithering is generally considered "good enough" to get 16 bit audio to handle a dynamic range of 120dB (so DEFINITELY into permanent hearing loss territory) which is VERY close to the level where electrical noise in even a well made, very expensive piece of gear will start to dominate. Most music is NOT mixed to have this much dynamic range.

https://en.wikipedia.org/wiki/Audio_bit_depth

https://en.wikipedia.org/wiki/Dynamic_range#Audio

Somewhat similar story with the frequency support. People usually can't hear above 20KHz (most adults it's somewhere in the 15-20KHz range and it gets lower with age). You can represent a wave perfectly by sampling at 2 intervals per Hz and then fitting a curve to it. For a 20KHz signal you need 40,000 samples (so 40KHz). CDs go up a bit higher as they want some overhead for handling a high pass filter over the sound.
https://en.wikipedia.org/wiki/44,100_Hz
https://en.wikipedia.org/wiki/Nyquist–Shannon_sampling_theorem

 

[8tg]

This kinda comes down to if you’re really intent on getting into this side of audio. Because tidal is a good option for streaming, apple has a decent bit of cd quality stuff, Amazon music is really good for high resolution stuff, but my suggestion is that if you want to get into this I recommend starting to buy your own music from a variety of sources, buy and rip CDs, and make sure you have hardware to really utilize it.

 

The whole “you can’t tell the difference” argument imo is mainly perpetuated at some level from streaming audio providers who want to keep people on low resolution audio to keep costs down. That’s my conspiracy theory though. It is economical both in cost and storage and bandwidth requirements for YouTube to use 128kbps audio or Spotify to top out at 320kbps ogg, but it’s really noticeable if you start to pay attention to it.

 

I have a dedicated player, a Walkman a55, some really nice in ear monitors, and a library of 2000ish songs I’ve purchased in some way or another just to get decent quality. Bandcamp is a great source for buying music and getting flacs out of it, prostudio masters is also a good option for stuff from major labels that won’t be hosted on bandcamp, and then the final source is physical media, CDs, when all else fails.

 

But this is expensive, if you really listen to a lot of music and can appreciate higher quality audio this is the way to go. You don’t gotta worry about your music being delisted from a service some day because the label decides they don’t want to license out their music anymore, you own the files and they’re not going anywhere.

A lot of albums will be anywhere from free, maybe a dollar, to 20$ or more if you’re buying really high resolution files.

 

But again that’s not something I would recommend to most people. A more sensible option is to just use your phone and some decent headphones or iems, for CD quality Bluetooth is fine, but above CD quality is where you’d want a wired connection as the bandwidth of Bluetooth can become a limitation.

Id start there, find some albums you like in a streaming service for high resolution audio, play it off your phone, and then look into just downloading music from a variety of sources and playing it off your phone, and if you really want to commit to it that’s the point where (after headphones/iems) look into building a larger, owned library of music.

 

 

[8tg]

And to note, I speak on personal experiences, as someone listening to everything from 320kbps MP3’s to DSD256 and everything in between. But listening from a dedicated player, on 1000$ earbuds. This is what I notice, your mileage may vary depending on what you’re listening to, what you’re listening with, and what you’re listening from, and what media you go with.

It was when I got a pair of moondrop starfields, 100$ iems, that I suddenly noticed the errors in MP3 compression on a 2nd Gen iPod touch, that was the point where I wanted more. 
You may not notice or care, your hardware may not really show those flaws, in which case this whole ordeal is probably just something to avoid.

 

Semi related, another comparison I like to make in these types of communities where more people know PC stuff than anything, is to compare it to pc gaming:

Spotify is the music equivalent of playing current games at 720p low. It’s passable, you can enjoy the game that way. CD quality is 1080p 60hz high settings, the way it’s meant to be seen by the developers, the view that’s easily accessible to the masses with regular computers. 
96/192khz 24 bit flacs are like 4k, you’ve got the 3090 ti, you’ve got the fantastic monitor, you’re cranking the settings in everything to get a fantastic experience. But it’s not like it’s vital over the 1080p experience, it’s just nice to have those small changes.

DSD256/512 and stupidly high sample rate 32 bit flacs are where you’re going beyond that, you’re doing LTT tier “8k gaming” nonsense, this is a hardware commitment and costs more per album than most peoples earbuds do. It’s 100% excess and super fun but it’s for like 1 guy to do for the lulz with throwaway money, nowhere near something anyone not super dedicated should ever do.

 

Analog formats are like playing games on OG hardware, no emulation, you scoff at the idea of playing N64 emulators so you buy a whole ass N64 and CRT television and all the games on their original cartridges. It’s more for the fun of the physical items than it is the quality of anything. 

[rikitikitavi]

On 5/9/2022 at 12:30 AM, cmndr said:

https://www.npr.org/sections/therecord/2015/06/02/411473508/how-well-can-you-hear-audio-quality

Afaik this is a BS article, and the provided samples are most probably cherry picked.

 

Yes, hardware should matter first, however, the recording quality can be a bottleneck.

Compressing is not just cutting off very top and bottom frequencies, but depending on quality - introducing distortions, and doing manipulations that might not be picked up by some, but recognized by others. All music/genre dependent of course.

 

... also if someone wants so pay premium for a BS improvement - nothing new, on this forum especially lmao

[cmndr]

I want to preface with "320Kbps ought to be enough for anyone" mostly applies to stereo encoding. I do think that there's a benefit for multi-channel encoding and that DOES require more data. 5 channels is naively 2.5x that data (assuming away clever algorithms) and something like atmos can easily be 10x.

 

Amazon has high bitrates AND (some, not much) multi channel encoded music. By this metric they're a winner. I'd still make a decision based on what content a service has and its price, though.

 

1 hour ago, rikitikitavi said:

Afaik this is a BS article, and the provided samples are most probably cherry picked.

It gets tossed around audiophile forums every now and then. The general take is that most people CAN'T reliably detect the difference between a well-encoded 320Kbps file and a lossless file and that those who do can only find it in narrow circumstances and by rapidly switching between the two.

If you're using speakers the top most frequencies are in the "stochastic zone" where you just moving your head a bit will affect the frequencies hitting your ears more than the encoding (hurray room reflections). There's also a good chance that "noise" from other sources will dominate (things picked up in recording and producing, electrical noise, etc.), speakers, room, etc. not to disregard the fact that most "lossless" music is sampled at 44KHz or so and has a  filter applied at 20KHz, basically throwing away nearly all the frequencies above 20KHz.

I don't think NPR (generally considered a reputable publication) has any hidden agenda.

 

For what it's worth I just went through and did a few and kept alternating between 320Kbps and lossless as my selections. Small n because I'm lazy. So I can tell a difference between 128kbps and the others WHEN I go out of my way to look for it. And I'm someone who can hear above 16KHz. I mostly tried to look for hints of distortion or clipping in the high frequencies.
 

1 hour ago, rikitikitavi said:

Yes, hardware should matter first, however, the recording quality can be a bottleneck.

Compressing is not just cutting off very top and bottom frequencies, but depending on quality - introducing distortions, and doing manipulations that might not be picked up by some, but recognized by others. All music/genre dependent of course.

 

... also if someone wants so pay premium for a BS improvement - nothing new, on this forum especially lmao

RECORDING (and production) quality is certainly a bottleneck for a lot of audio. If you're producing music, keep a high quality copy in case you need to modify the source later.

 

1 hour ago, rikitikitavi said:

Compressing is not just cutting off very top and bottom frequencies, but depending on quality - introducing distortions, and doing manipulations that might not be picked up by some, but recognized by others. All music/genre dependent of course.

 

... also if someone wants so pay premium for a BS improvement - nothing new, on this forum especially lmao

 

https://nautil.us/the-math-trick-behind-mp3s-jpegs-and-homer-simpsons-face-1273/

The compression is usually more apparent at the top frequencies. That's kind of how fast fourier transforms work.
128Kbps mp3 has a cutoff around 16KHz (near the range where most adults can't hear anyway, plus a range where music isn't usually played at).
192Kbps bumps that up to around 19KHz.
320Kbps has its drop off around 20KHz... again at a level which is imperceptible to adults because of damage to the hair cells in the ears.

There will be artifacts throughout the entire sound wave generated, but these artifacts are... again usually (though not always) outside of the range of human perception for a sufficiently high bit rate. Basically a peak (or a dip) in sound pressure intensity for a VERY short spurt in time (1/20,000th of a second) isn't readily perceptible and it won't activate hair cells in the ear (most of which are dead).

Most lossy compression algorithms are (loosely speaking) summing up a bunch of different, simple waves to get a final wave that's VERY close to the original. Not having super granular frequencies added to the mix (what having higher bit rates enables) inherently means that there's going to be information loss at the high frequencies.
 

Fair warning, I am only superficially familiar with digital signal processing (roommates were in EE during undergrad and I've skimmed a few lectures)

It's also possible that your dog might interpret the music as weird since dogs have a higher hearing range.

 

128Kbps mp3 (observe cutoff at 16KHz and a bit of banding)
It's also likely (but not readily seen) that there's some distortion at the very lowest of frequencies. It's hard to draw a curve through these properly without throwing A LOT of data at it. Below 20Hz isn't really heard though (mostly felt) so not the end of the world. Also for low frequencies (think 80Hz and lower) people aren't as sensitive to distortion.

[rikitikitavi]

6 hours ago, cmndr said:

I don't think NPR (generally considered a reputable publication) has any hidden agenda.

Well, the author did choose only 6 songs. There is acapella... and rnb and pop/rock which have compressed signature from the get go... That's why I say it looks like cherry picked.

 

2 hours ago, cmndr said:

AAC-LC 320 kbps

TLDR. That's quite high quality and not your typical Spotify.

 

Edit:

 

As I've stated before - depending on the music and the compression - the 'artifacts' might or might not be noticeable. Most 'audiophiles' have a few songs that they know well and can test, they won't just listen to a random song... and/or use some random equipment.

 

'Many years ago me' has a very strong opinion about this, cause he heard the difference between CD and ripped hi res MP3s on his stereo system... 'Today me' is satisfied with iTunes hifi AAC and mid range headphones. However, AAC is way better than MP3...

 

@rippy4500... you can always download a hi res version of a favourite album (not so legally) for the sake of comparison

 

4 hours ago, OfficialTechSpace said:

obody asked your opinion, nor cares to hear it.

Calm down, he asked a legitimate question in the first place.

[cmndr]

3 hours ago, rikitikitavi said:

Well, the author did choose only 6 songs. There is acapella... and rnb and pop/rock which have compressed signature from the get go... That's why I say it looks like cherry picked.

 

TLDR. That's quite high quality and not your typical Spotify.

 

Edit:

 

As I've stated before - depending on the music and the compression - the 'artifacts' might or might not be noticeable. Most 'audiophiles' have a few songs that they know well and can test, they won't just listen to a random song... and/or use some random equipment.

There's other tests out there... they all generally have similar results. At some level if you can't detect a difference across a wide range of common genres, that should be telling.

The default for free is ~96Kbps on Spotify. You can up it to 160Kbps. 160 is the cuttoff for what I'd consider "adequate"

Spotify premium is 320Kbps. Pretty much EVERY premium service offers 320Kbps or higher. I can't speak to the exact encoding but I'm assuming it's "not stupid" since one-off encoding is cheap and decoding isn't spotify/amazon/tidal's problem (and it's probably accelerated on modern hardware anyway).





 

[FliP0x]

Just because humans are not able to hear above 20KHz, it doesn't mean that a higher range does not matter in an audio file.

 

You may not be able to hear frequencies above 20KHz, but you can hear the clarity due to the higher frequency ranges.

 

If you want to give this a go, fire up any DAW of your choice, make a random track and export it at different settings (above 20KHz and below 20KHz). Tell me if you hear a difference.

[Spuriae]

I encourage anyone who thinks there is a significant difference in sound quality between high bitrate AAC/opus and lossless to try running an ABx test between the two (actually blind- no peeking!). Not because there isn't a difference, but because it's a good reminder of how tiny the differences really are. They tend to be difficult to identify even under ideal circumstances.

 

Spotify and a 32 bit flac aren't equivalent to the difference between 720p and 8k; they're more akin to the difference between 4k and 8k – on a laptop screen. Curiously, there are issues significantly more audible than lossy compression, like audio watermarking, that rarely get brought up in audio streaming quality debates.

On 5/9/2022 at 12:30 AM, cmndr said:

Quick question, does this even matter to you?
https://www.npr.org/sections/therecord/2015/06/02/411473508/how-well-can-you-hear-audio-quality

If you're not able to differentiate, then you're asking the wrong question.

The NPR test is a good test of first impression. There is a way to game it using the sample load times though, and the setup isn't well-suited for statistically significant results. Still, it does a good job at illustrating how difficult it is to identify high bitrate compression artifacts.

On 5/9/2022 at 12:30 AM, cmndr said:

If you're using headphones... funds are usually best spent on better headphones.

if you're using speakers... speakers and/or room treatment.

I agree that the transducer and enclosure is always the most audible part of the signal path.

However, multiple lossless audio streaming services cost less than $15/month, so it doesn't really cost extra to switch to one. It's really more a feature/library comparison.

On 5/9/2022 at 12:30 AM, cmndr said:

The difference between 16 vs 24 bits is how much dynamic range is supported without loss of quality. I believe 16 bit is good up to 96dB. In order for a person to actually perceive that full dynamic range of a 16 bit stream you'd need music that itself has a high dynamic range (so 96dB between the loudest and softest stuff). For the sake of argument, let's say you can't hear below 20dB (half as loud as a soft whisper) or so (unless you're in an anechoic chamber). This would mean you'd need to have the peaks in the music hit 116dB.

This isn't quite right. You can hear well below the noise floor, since hearing is based on energy detected in frequency bins (similar concept to a Fourier transform) and noise is broadband. For instance, here is a 5 second long file with a 1kHz sine wave played 20dB(rms) below white noise. The sine wave cuts off 3 seconds in so you can confirm that you can actually hear the tone.

Here is a 4kHz sine 35dB(rms) below brown noise, cutting off at 3 seconds again.

I have a 94dBSPL 1kHz sine microphone calibrator and an ear and cheek simulator made from an IEC711 coupler, ITU-T P.57 ear, and a slab of rubber with a hole cut into it. Calibrating my headphones' (HD650) level so that a full scale sine wave measured 92dBSPL, I was able to differentiate between a -89dBFS and -100dBFS 1kHz sine wave in an ABx shootout, but not a -100dBFS and -90dBFS sine wave, implying that my 1kHz hearing threshold at my desk (certainly >20dB background noise, since it has a PC on top) with open back headphones is ~3dBSPL.

 

Theoretically, not accounting for auditory masking, the max rms level would only need to be 99dBSPL for 96dB of dynamic range. This isn't as loud as it seems; after all, the average level may vary significantly throughout a song, by ~10dB between sections. The RMS level of a song (the applicable number for safety) is typically -15dBFS; a 99dBSPL peak song in practice is usually a more tolerable ~85dBSPL average.

 

For a more extreme example, the Interstellar soundtrack contains songs where much of the track is around -50dB while the peaks are just below clipping. In these cases the dynamic range in the quieter sections is "only" <50dB in an undithered 16 bit file. There is a good argument to be made, however, that listening loudly enough to appreciate that much dynamic range is unsafe regardless.

On 5/9/2022 at 12:30 AM, cmndr said:

For what it's worth, dithering is generally considered "good enough" to get 16 bit audio to handle a dynamic range of 120dB (so DEFINITELY into permanent hearing loss territory) which is VERY close to the level where electrical noise in even a well made, very expensive piece of gear will start to dominate. Most music is NOT mixed to have this much dynamic range.

This is all true. Here is a one second test track composed of a full-scale 10kHz sine and a descending sine chirp around 1kHz exactly 96dB quieter than the 10kHz sine. It sounds a bit like a distant car alarm. The chirp is audible at a 107dBSPL playback level in both the 16 bit and the 24 bit file (that's what I tested at - it's certainly still audible a few dB lower, but I'm not interested in checking the exact threshold for safety reasons). I cannot tell the two apart by ear. Perhaps I would if I played it more loudly, but I don't feel safe pushing those limits.
 

 

Here you can see the spectrum graph. Spectral analysis can be iffy with files this short: see the burst of aliasing at the start. In both the 24 and 16 bit files, the faint purple curve at the bottom is the chirp. Note that it is still present despite being below what a 16 bit file would normally be expected to reproduce. Also note the high frequency noise in the 16 bit file (from the dithering). I am unable to hear this.

On 5/9/2022 at 12:30 AM, cmndr said:

Somewhat similar story with the frequency support. People usually can't hear above 20KHz (most adults it's somewhere in the 15-20KHz range and it get slower with age). You can represent a wave perfectly by sampling at 1Hz intervals and then fitting a curve to it. For a 20KHz signal you need 40,000 samples (so 40KHz). CDs go up a bit higher as they want some overhead for handling a high pass filter over the sound.

https://en.wikipedia.org/wiki/44,100_Hz
https://en.wikipedia.org/wiki/Nyquist–Shannon_sampling_theorem

Humans can't hear tones above 20kHz, but we can use temporal cues faster than 1/20kHz to determine timing.

 

Here is a study on the audiblity of clicks (stimuli significantly shorter than the period of a 20kHz wave): https://asa.scitation.org/doi/abs/10.1121/1.1912374. Participants were able to distinguish 10µs dropouts (equivalent to a single sample at 100kHz) between stimuli. I don't have a non-oversampling DAC, so I cannot verify the results myself.

 

More importantly, timing differences between channels are a significant cue in positional audio. http://www.socsci.uci.edu/~saberi/saberiandperrott1990b.pdf. Thresholds of audibility extend below 20µs, implying that some audible left-right spatial information may be lost for certain transients at sample rates 48kHz and below.

 

I ran a similar binaural test by creating two files at 192kHz. Each one has a two sample long pulse in each ear. In one file, the left click occurs 2 samples (~10µs) before the right one; in the other file, the right click occurs 2 samples before the left one. The tracks are shorter than a single sample at 44kHz.

 

In blind testing I can tell them apart, though this one took ~50 practice comparisons before I was able to reliably distinguish them. Because the spectral energy of a two sample impulse is low, you may need to turn your volume WAY up to even hear the clicks. The files may also simply not play properly in some environments.

 

Not that this is necessarily worth caring about even if it is audible. I certainly don't mind the soundstage perceptually being a few millimeters offset when listening to music. I can't imagine being able to tell which version of a track is the "correct" one based on this information; only that they are different.

19 hours ago, FliP0x said:

Just because humans are not able to hear above 20KHz, it doesn't mean that a higher range does not matter in an audio file.

 

You may not be able to hear frequencies above 20KHz, but you can hear the clarity due to the higher frequency ranges.

 

If you want to give this a go, fire up any DAW of your choice, make a random track and export it at different settings (above 20KHz and below 20KHz). Tell me if you hear a difference.

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

 

40kHz is the sample rate required to record 20kHz information. The 20kHz export example has a 10kHz audible cutoff.

[cmndr]

11 hours ago, Nimrodor said:

Here is a study on the audiblity of clicks (stimuli significantly shorter than the period of a 20kHz wave): https://asa.scitation.org/doi/abs/10.1121/1.1912374. Participants were able to distinguish 10µs dropouts (equivalent to a single sample at 100kHz) between stimuli. I don't have a non-oversampling DAC, so I cannot verify the results myself.

 

More importantly, timing differences between channels are a significant cue in positional audio. http://www.socsci.uci.edu/~saberi/saberiandperrott1990b.pdf. Thresholds of audibility extend below 20µs, implying that some audible left-right spatial information may be lost for certain transients at sample rates 48kHz and below.

As far as I can discern you're a heck of a lot more knowledgeable than I am... I know "the basics" - enough to not look like an idiot while talking to someone that works in signal processing/sound encoding/DSP for a living (I've done it - my SO used to work for a subsidiary of Sound United [aka Denon/Marantz/Polk/etc.]).

One thing I would put out there - I suspect that HOW someone wears headphones or sits will drown out any benefits from going past ~20KHz in most cases when it comes to positioning. I know as a mater of fact that I move while sitting and that I'm not the most careful when it comes to putting on headphones.

When talking about time to ear differences measured in the microsecond range, you're getting into custom IEM territory. Worst case scenario you're at a point where you perceive something being off by a tiny fraction of an angle.

 

Quote

However, multiple lossless audio streaming services cost less than $15/month, so it doesn't really cost extra to switch to one. It's really more a feature/library comparison.

I touched on that sligthly. I did NOT say "don't pay for premium audio" - just: "consider feature/library first and don't pay extra to jump from high bitrate lossy to lossless"

 

 

 

Edit: this thread just motivated me to change my spotify streaming quality from Automatic to Very High on my desktop. I forgot to change it when I reinstalled my OS.

 

I might be imagining it but the highs sound a bit brighter, I might want to up the rolloff on my EQ.

[The Torrent]

There is absolutely a lot of ways.

 

However many are called piracy so I'd prolly get banned for saying them here.