Spotify to launch 'Hi-Fi' CD Quality Tier.

[antcollinet]

I think it's here in optimum_bandwidth() in lame.c:

    const band_pass_t freq_map[] = {
        {8, 2000},
        {16, 3700},
        {24, 3900},
        {32, 5500},
        {40, 7000},
        {48, 7500},
        {56, 10000},
        {64, 11000},
        {80, 13500},
        {96, 15100},
        {112, 15600},
        {128, 17000},
        {160, 17500},
        {192, 18600},
        {224, 19400},
        {256, 19700},
        {320, 20500}
    };

This defines the start of the stop-band. Later it is rounded up to one of the 32 band-buckets that lame is using. For example for 17000:

17000 * 2 / 44100 = 0.771
23 / 31 = 0.742
24 / 31 = 0.774

therefore stop-band starts at:
24 / 31 * 44100 / 2 = 17071

and transition-band starts at:
(24 - 0.75) / 31 * 44100 / 2 = 16538

To verify:

]$ lame -b 128 test.wav test.mp3
LAME 3.100 64bits (http://lame.sf.net)
Using polyphase lowpass filter, transition band: 16538 Hz - 17071 Hz
...

I love this place.

I'm not sure there are many other places than ASR, where someone will quote C source code to answer a question that is NOT about coding.

Thanks - fantastic.

 

[pablolie]

These discussions always go down the rabbit hole of bits per second.

Let's make it clear:

(1) Start with a GREAT, FAMILIAR recording, if you are even remotely trying to make this a successful experiment.
(2) Even better if the recording has sharp percussion, ideally something like castanets. Every time someone claims they hear "fuller bass" when listening to "uncompressed" shows they don't know what they are talking about.
(3) Realize 90% of recordings do not merit anything over 256k, because they are already compromised in some way.
(4) If you are not very familiar with the original version of a great recording, you are wasting your time.
(5) If you can't enjoy great music and recordings at 320k, you are sabotaging yourself.

Spotify have said they have run experiments to see if users are more engaged with "uncompressed", and that the results were that it doesn't make a difference. And I believe them, once I measured the bandwidth and saw it was higher than usual, in fact clocking in at 20/96 or so for about 30 minutes. I didn't measure because I heard a difference, it was just a coincidence I was measuring the bandwidth consumption of different apps at the time.

With a familiar great recording like Kevin Mahogany's "Pride and Joy" I *can* accurately tell apart 256k from CD 100% of the time, 320k goes down to 80%ile except with one song, which I nail every time because I know exactly which part to pay attention to. But as a rule, I hate the exercise and I'd rather never do it, it's silly and obsessive. It's like saying that, because 350hp in a car is fun, 5000hp must be even better. The law of ridiculous overkill has not yet registered for many audio fans. Guess what else is compressed? Your very ears. Ask your dog or cat.

 

[Sal1950]

I love this place.
I'm not sure there are many other places than ASR, where someone will quote C source code to answer a question that is NOT about coding.

ROTFLMAO

 

[Jeromeof]

ROTFLMAO

I work as a Software Architect / Engineer and I remember years ago some manager wanting me to document everything we were doing on a very technical project and I told him the source code was the ultimate documentation. He was not very impressed

 

[Timcognito]

These discussions always go down the rabbit hole of bits per second.

Let's make it clear:

(1) Start with a GREAT, FAMILIAR recording, if you are even remotely trying to make this a successful experiment.
(2) Even better if the recording has sharp percussion, ideally something like castanets. Every time someone claims they hear "fuller bass" when listening to "uncompressed" shows they don't know what they are talking about.
(3) Realize 90% of recordings do not merit anything over 256k, because they are already compromised in some way.
(4) If you are not very familiar with the original version of a great recording, you are wasting your time.
(5) If you can't enjoy great music and recordings at 320k, you are sabotaging yourself.

Spotify have said they have run experiments to see if users are more engaged with "uncompressed", and that the results were that it doesn't make a difference. And I believe them, once I measured the bandwidth and saw it was higher than usual, in fact clocking in at 20/96 or so for about 30 minutes. I didn't measure because I heard a difference, it was just a coincidence I was measuring the bandwidth consumption of different apps at the time.

With a familiar great recording like Kevin Mahogany's "Pride and Joy" I *can* accurately tell apart 256k from CD 100% of the time, 320k goes down to 80%ile except with one song, which I nail every time because I know exactly which part to pay attention to. But as a rule, I hate the exercise and I'd rather never do it, it's silly and obsessive. It's like saying that, because 350hp in a car is fun, 5000hp must be even better. The law of ridiculous overkill has not yet registered for many audio fans. Guess what else is compressed? Your very ears. Ask your dog or cat.

Well put. I too can hear differences but only with single or dual instrument closely miked recordings. Best done with headphones excluding the room effects. But that kind of reinforces there are no differences except in the meticulous recording techniques and environments where dramatic differences can be found

 

[gags]

I work as a Software Architect / Engineer and I remember years ago some manager wanting me to document everything we were doing on a very technical project and I told him the source code was the ultimate documentation. He was not very impressed

Actually, it’s unit tests.

Now back to work!