Does 192kbps suck compared to 990kbps or am I stoned?

[twilyth]

I've been listening to the Lady Gaga Fame Monster album lately. I really like most of the songs. But I wanted a change of pace so I go to Pandora One (192kbps) and pull up the Gaga station I had set up and they played a song I had just listened to.

Damn, what a difference. The song was Paper Gangsta. It has some really punchy keyboards but on Pandora, they sounded fuzzy and muffled. They didn't have the same kick and weren't nearly as sharp.

Is it possible for me to hear such a difference? I believe it's noticeable and real. I wasn't expecting Pandora to be inferior. It was just coincidence that I happened to play the songs so close to each other. And after a while I adjusted to Pandora and it sounded normal. So maybe the difference is as huge as it seemed at first. IDK.

[palo]

It's not the bitrate, it's just pandora encoding these like monkeys.

A few months back, there was a thread on headfi with a blind test. It was a piece of classical music with wild dynamics. One take was encoded with a lossy codec and the other was "lossless". The difference was REALLY subtle and if I've had then the shitty gear I have at the moment I'm sure I would not had been able to tell the difference.

So, yeah, to sum it up, it's not how big your bitrate is, it's how you encode it.

And you're stoned.

[palo]

Just noticed :

I've been listening to the Lady Gaga Fame Monster album lately.

WTF is wrong with you man? (poop)(poop)

[twilyth]

OK, thanks. I should have mentioned that the tracks I listen to on my PC are lossless, but I guess I figured people would get that from the bitrate. I really shouldn't make those sorts of assumption though especially when I'm not all that familiar with a subject.

But don't the codec and bitrate issues go hand in hand? What I mean is, that 192kbps is about 8bit at 22kHz - right? Ok, so that is 1/4 of what the cd bitrate is (16bit x 44.1khz), but even so, isn't it a lot better than most mp3's? So how fucked up does the codec have to be to make that sort of bitrate sound like shit?

Never mind. I think I just answered my own question. No matter how good the codec, you're talking about trying to squeeze all the quality of lossless into 1/4 of the space. That means compromises. The only issue is whether those compromises are really annoying for the listener, and I have to admit, it's not that bad considering the amount of compression involved.

Still, they want me to fork over $36 per year. That's about 1/3 of a Netflix streaming subscription and I'm damned fucking sure that Pandora doesn't use even a tenth of the bandwidth Netflix does. Probably not even 1/100th. So from that perspective, 192kbps really is bullshit.

edit: sorry. forgot about your dissin' my lack of musical taste. (slap) I really like almost all of the tunes and none really bother me. I can listen to both disks from beginning to end with no problem. I don't say that about many bands. Gossip's Music for Men is another I really like. I know it makes me look like a dweeb, but it is what it is. (toast)

[theJesus]

But don't the codec and bitrate issues go hand in hand? What I mean is, that 192kbps is about 8bit at 22kHz - right? Ok, so that is 1/4 of what the cd bitrate is (16bit x 44.1khz), but even so, isn't it a lot better than most mp3's? So how fucked up does the codec have to be to make that sort of bitrate sound like shit?
No, it's still 16bit, 44.1khz. Remember that the frequency is cut in half for each channel, so what you're saying makes it sound like everything above 11khz would be cut off, which is simply not the case. With lower bitrates, there is still a little bit of a cutoff, but not that dramatic unless you go really low (192kbps is generally accepted as the lowest "acceptable" bitrate). If you look at a spectrum analyzer and compare the same track in lossless to it's lossy-encoded version, you should be able to see the compression artifacts manifest as blockiness where it is smooth in the lossless version. Similar to jpeg compression artifacts, etc.
Never mind. I think I just answered my own question. No matter how good the codec, you're talking about trying to squeeze all the quality of lossless into 1/4 of the space. That means compromises. The only issue is whether those compromises are really annoying for the listener, and I have to admit, it's not that bad considering the amount of compression involved.
Continuing my rambling above: Compression artifacts also occur from transcoding one lossy source to another. As an example, taking a 192kbps MP3 and re-encoding it to the same bitrate will not yield an identical file, but will instead lower the quality. The degree to which this effect would be noticeable in that example is rather questionable however. Mostly, transcoding is bad because some people actually try to take a low bitrate file and re-encode to a higher bitrate, expecting it to improve the quality. In reality, this would again lower the quality.

Anyways, if you want to read more about the subject, check this (http://blowfish.be/eac/Spectral/spectral.html) out.

Also: http://wiki.hydrogenaudio.org/index.php?title=Transcoding

[DanTheBanjoman]

Just noticed :



WTF is wrong with you man? (poop)(poop)

You quoted the thing that's wrong.

[twilyth]

No, it's still 16bit, 44.1khz. Remember that the frequency is cut in half for each channel, so what you're saying makes it sound like everything above 11khz would be cut off, which is simply not the case. With lower bitrates, there is still a little bit of a cutoff, but not that dramatic unless you go really low (192kbps is generally accepted as the lowest "acceptable" bitrate). If you look at a spectrum analyzer and compare the same track in lossless to it's lossy-encoded version, you should be able to see the compression artifacts manifest as blockiness where it is smooth in the lossless version. Similar to jpeg compression artifacts, etc.

Continuing my rambling above: Compression artifacts also occur from transcoding one lossy source to another. As an example, taking a 192kbps MP3 and re-encoding it to the same bitrate will not yield an identical file, but will instead lower the quality. The degree to which this effect would be noticeable in that example is rather questionable however. Mostly, transcoding is bad because some people actually try to take a low bitrate file and re-encode to a higher bitrate, expecting it to improve the quality. In reality, this would again lower the quality.

Anyways, if you want to read more about the subject, check this (http://blowfish.be/eac/Spectral/spectral.html) out.

Also: http://wiki.hydrogenaudio.org/index.php?title=Transcoding
Thanks. I forgot about the frequency doubling (so you get at least 1 stereo sample per second even at the high end - if that's how it works :confused: )

I was trying to back into the numbers and assumed a fixed bitrate with no encoding - which was pretty stupid now that I reflect on your comments.

I think I tend to confuse the sample rates and bits per second since the numerical part is often the same (e.g., 192khz and 192kbps). But I did double check that this time. Not sure where I got the idea that 192kbps was considered better than most mp3 - probably from recalling the old days when a 256Mb mp3 player was the shizznit.

Thanks for the explanations though. I do appreciate that. I like to research things but I like it better when someone explains shit to me in a way I can understand. (toast) I already spend most of my day online and it just gets exhausting after a while, so thanks again.

So is Pandora really using a shitty codec or are they just up against the wall given the fact that they cap their streams at 192kbps? My guess is that they're screwed regardless, but Palo seems to thing they could improve the quality by going with a better codec.

And there's the issue of their being too cheap to use a decent bitrate. (nutlick)

edit: forgot to mention that I don't really do any re-coding of audio files. I sometimes have to do it for video. It's good to know though.
You quoted the thing that's wrong.
http://i181.photobucket.com/albums/x206/twilyth/ccc66f8f.jpg

[hellrazor]

Does 192kbps suck compared to 990kbps

Yes, it does.

or am I stoned?.

Yes, you are.

[theJesus]

So is Pandora really using a shitty codec or are they just up against the wall given the fact that they cap their streams at 192kbps? My guess is that they're screwed regardless, but Palo seems to thing they could improve the quality by going with a better codec.
I'm not sure what Pandora uses, but 192kbps seems to be the norm for most internet radios. I'm pretty sure more than a few still use 128 and lower, but not many go above 192. There are some other factors that may affect the quality besides the bitrate, such as network issues, poor source material, resampling, DSP, etc.

Really though, it's certainly not out of the ordinary to notice a difference between 192kbps and lossless. As I said before, 192 is generally accepted as the minimum acceptable bitrate. If you have decent hearing and fairly nice speakers and/or headphones, then it is very possible you will notice a difference, especially on well-mastered tracks. I believe it's most noticeable in the higher frequencies, since lower bitrates tend to have a "cut-off" point.

I remember finding a really good example recording of some castanets or something; if I can find it again, I'll post it.

[twilyth]

I can't hear over 10khz any more so in my case it probably isn't the high end. :o Getting old sucks.

I see what you're saying about 192kbps being the standard for internet radio, but since Pandora One is supposed to be a "premium" service, you'd think they would want to do more than the minimum people expect.

The only reason I even subscribed was because last.fm stopped working for me - regardless of which browser I tried to use. All of the links that used javascript were malformed and I couldn't find a work around. If I can get last.fm to work again, I'm asking for my money back from Pandora.

I'll try to pay more attention to the actually numbers on audio files from now on. I seem to have a really shitty memory for somethings and this seems to be one of them. I think at this point I'll at least be able to remember that 192kbps (as opposed to 192khz) is the minimum acceptable bitrate.

[theJesus]

I think you might be hearing "pre-echo" which is the exact type of artifact in the castanets.wav example. I found it, but the server seems to be having permissions problems so it won't let you download it. Anyway, if you're in the mood for some reading, you can find out about the effect in varying detail from the following links:

http://wiki.hydrogenaudio.org/index.php?title=Artifact
http://ff123.net/training/training.html
http://ff123.net/preecho.html

Edit: BTW, here is the server that I believe is the original source of the castanets.wav example, just in case anybody else might be able to grab it: http://ccrma.stanford.edu/~bosse/

[twilyth]

I think these are the samples you wanted - at least the castanets file is.

http://lame.sourceforge.net/download/samples/

[theJesus]

Yeah, I believe that is the source file in lossless wavpack format. You can make a transcode to some lossy format like 128 or 192kbps MP3 and probably hear the difference.

[twilyth]

OK. I honestly couldn't tell the difference. If it's obvious to some people, they are definitely in a completely different league.

[theJesus]

Pay close attention to the four "clacks" of the castanets right at the start of the sample, it's easiest to notice there. They should sound very "sharp" and "snappy" etc. But in the 128kbps file, you should be able to notice the pre-echo right before each clack. It's kinda like the "th" sound.

[MRCL]

Just noticed :



WTF is wrong with you man? (poop)(poop)

Admit, Lady Gaga has a better voice than most broomsticks with breasts on stage these days, and catchy tunes.

[FordGT90Concept]

It depends on where the increase in bitrate is coming from. For example, 24-bit, 96 KHz, 5.1 music (13,824 Kbps) is a huge difference from standard CD audio which is 16-bit, 44.1 KHz, stereo music (1411.2 Kbps). The difference between 1411.2 Kbps CD Audio and 128 Kbps lossy MP3 is virtually none though.

Internet music might only be 22.05 KHz too...

[Wile E]

It depends on where the increase in bitrate is coming from. For example, 24-bit, 96 KHz, 5.1 music (13,824 Kbps) is a huge difference from standard CD audio which is 16-bit, 44.1 KHz, stereo music (1411.2 Kbps). The difference between 1411.2 Kbps CD Audio and 128 Kbps lossy MP3 is virtually none though.

Internet music might only be 22.05 KHz too...

That is total and utter bullshit.

[Reventon]

Going from 128 to about 320 makes a huge difference. From 320 to ~1000 makes as much of a difference.

[theJesus]

Going from 128 to about 320 makes a huge difference. From 320 to ~1000 makes as much of a difference.
128-320 is definitely a noticeable difference, even with inexpensive setups. However, I don't going from 320 to lossless isn't as noticeable unless you've got a good chunk of change invested in your gear (and decent ears too).

[FordGT90Concept]

That is total and utter bullshit.
I took a CD, extracted the WAV and converted to MP3 128 and MP3 320 and couldn't hear any differences. I challenge you to do the same and see if you can pick out any differences between them. There was a big discussion about this on TPU.

And don't feed me the shitty speakers line because 13,824 Kbps would sound the same as 1411.2 Kbps if that were the case. :p

[Wile E]

I took a CD, extracted the WAV and converted to MP3 128 and MP3 320 and couldn't hear any differences. I challenge you to do the same and see if you can pick out any differences between them. There was a big discussion about this on TPU.

And don't feed me the shitty speakers line because 13,824 Kbps would sound the same as 1411.2 Kbps if that were the case. :p

I can hear the difference on all but the most shittily mastered CDs. You sir, are fucking deaf, stupid, have really shitty equipment, or all of the above.

Give me some A/B tests with decent material, and I bet I get most of them right. I bet theJesus would as well.

[FordGT90Concept]

There's no way to eliminate the possibility of cheating which renders the test void before it starts. :p

[twilyth]

I'll hold the cash until you 2 sort it out.

[Reventon]

128-320 is definitely a noticeable difference, even with inexpensive setups. However, I don't going from 320 to lossless is as noticeable unless you've got a good chunk of change invested in your gear (and decent ears too).

I meant *not as much. I have 320 music and 1K+ music on my computer. It's noticeable, just not a lot. Even with studio monitors.

[Steevo]

Almost everything less than 256 is crap to listen to unless you like the cheapo Sony Wal-Mart sound. Even if you can't hear that great, the sharpness of a cymbal, highhat, and or the brushes on some good percussion is trashed. The intricacy of vocals are junk, it sounds like they are singing through a sweater. The low end bass that can be felt is usually tossed and so some of the feeling is lost. A bass guitar like on RHCP Scar Tissue or Soul to Squeeze will sound like shit, the fine string noise that lets you know that a real awesome dude is fingering it like two lezzies in a 69.

[twilyth]

That's why I canceled my Pandora One subscription. I had to threaten to do a charge back (and actually started one) before they would give me my money back.

They're charging me $36/year and the best they can do is 192? Fuck that shit and fuck Pandora. I'll listen to last.fm for free.

The problem with last.fm is that you need a plugin to decode the javscript tags. So if you try to click on on the tag for say 'darkwave', it doesn't work without the plug-in. And they don't bother to tell you this anywhere.

[btarunr]

The difference between 1411.2 Kbps CD Audio and 128 Kbps lossy MP3 is virtually none though.

http://img.techpowerup.org/110922/bta8763jhd.jpg

[FordGT90Concept]

I finally figured out how I could do a test: embed the audio resources into an executable. But, I can't be arsed to do it. Not to mention, it would be a big download and rather illegal by having licensed music in it. I suppose I could ask a publisher for a song to use but what's the likelihood of that getting approved (answer: not at all)?

[Magibeg]

I finally figured out how I could do a test: embed the audio resources into an executable. But, I can't be arsed to do it. Not to mention, it would be a big download and rather illegal by having licensed music in it. I suppose I could ask a publisher for a song to use but what's the likelihood of that getting approved (answer: not at all)?

I dunno ford.... I know i can tell the difference on my sound system pretty easily. The contrast between 128 and lossless is so different even my parents can tell.

Father: Why doesn't this song sound as good as the last one?

That was a pretty good blind test.

[El_Mayo]

I can DEFINITELY hear a difference between 128 and 320
Anything less than 192 sounds lame to me

However between 320 and 990+ (FLAC) I couldn't hear a difference
but I only have entry level speakers/headphones so I might not be one to judge