Bitrate query

[0ddb0d]

Hi,
Not a very auspicious debut, but yesterday I emailed my entry for the "A Scary Thing" songfight and I've just realised it was mistakenly encoded at 256k bitrate, rather than 128k.
Is it okay to re-submit it at 128K?
Cheers

[jast]

Opinion of someone who doesn't run this site follows. Frankly, I would just leave it as it is (and remember to adjust the bitrate next time you submit anything). In fact, some submissions in the past few fights have been at even higher bitrates, some of them by regular fighters. That leads me to believe that the 128 kbps thing mentioned in the FAQ isn't really enforced much.
Personally, I think 128 kbps is the absolute minimum for decent sounding mp3 files. Most music (especially with many essy sounds, e.g. cymbals all over the place) needs 192 kbps to not sound ugly. If disk space (and traffic) was an issue I'd suggest that Song Fight switch to Ogg Vorbis. The quality it gives you at level 2 (average bitrate 96 kbps) is quite amazing when compared to mp3 at 128 kbps CBR.

[0ddb0d]

Hi Jast
thanks for the info.
I'll just leave it as is then, unless otherwise instucted.
Cheers

[jb]

Leave it. Although, you should know that when people send in 8mb mp3s that sound like you pooped on a shoe and called it music, that's when the voodoo dolls come out in the Song Fight! management back-room.

Not that you did. I haven't listened to the fight yet.

JB

[Billy's Little Trip]

I submit at 192 bps and have never had a complaint. But if you don't tag your MP3 with at least the band name, song title, genre and year, you will get complaints.

[roymond]

Ditto that (@ 192). Also @ metadata tags. There are some surprisingly established acts who apparently don't appreciate the promotional value of tags. I add artwork and lyric as well. It's great to have the lyric in the iPhone while it's playing. I think Tonetripper and Boltoph put their photo in.

[Billy's Little Trip]

There's only one reason why I don't add lyrics while tagging my MP3 for a Song Fight title fight. I don't type them out until I am posting them in the lyrics thread. They are a scribbled mess on paper until that point.

[0ddb0d]

But if you don't tag your MP3 with at least the band name, song title, genre and year, you will get complaints.

I'm pretty sure I managed to tag all of those so hopefully I won't get my ear chewed off!
Cheers

I've never tried adding the lyrics and artwork to an MP3. Not sure if my current tagging software covers that.
I'll check it out.

[roymond]

I've never tried adding the lyrics and artwork to an MP3. Not sure if my current tagging software covers that.

I know it's the anti-christ, but iTunes makes it easy, as do many other tools.

[Spud]

At least it's not the good old days:

p.s. this also explains the quality (or lack thereof) of some of the older recordings in the archive. Not all of them, of course, but some of them.

[Billy's Little Trip]

Yikes! 64 kbps and mono?

[Spud]

Yeah, it used to be about the song writing, not the production.

[Billy's Little Trip]

So, then write a book.

[fluffy]

Bandwidth and space are cheap.

A switch to ogg would be a major pain in the ass for pretty much everyone (especially those of us who like to listen to fights on an iPod or whatever). Also I've found that ogg doesn't really give that much savings compared to LAME's excellent VBR psychoacoustic model. The only real justification for ogg is purely political, and why the hell should politics that only a tiny fraction of people care about influence an entire site populated mostly by people who don't give a shit? Leave that crap to Wikipedia.

Really you should be encoding with LAME or WinLAME or the like, with the settings -b 32 -V 5 -q 5. That's a pretty nice tradeoff between quality and filesize (my songs tend to average somewhere between 128 and 192Kbps), and while it's not going to be absolutely CD-quality it's also going to be a lot better than iTunes at 128CBR (iTunes' mp3 encoder is actually quite shitty). The only downside is that some older mp3 players are retarded about VBR but really you shouldn't be running Winamp 1.0 ANYWAY.

[Spud]

If disk space (and traffic) was an issue I'd suggest that Song Fight switch to Ogg Vorbis.

He seems to acknowledge your point, fluffy. Traffic would go WAY down.

[fluffy]

Heh, yeah.

[jast]

A switch to ogg would be a major pain in the ass for pretty much everyone (especially those of us who like to listen to fights on an iPod or whatever).

Quite true.

The only real justification for ogg is purely political

Actually, it isn't, because mp3 is just about the only audio codec where you have to pay for providing files that use it (if you make a certain amount of money, so that's not all too relevant for SF). Not even the AAC folks do that.

Really you should be encoding with LAME or WinLAME or the like, with the settings -b 32 -V 5 -q 5.

That's a weird combination of flags: -V 5 is the default, -b conflicts with -q. If anything, you should specify -h (= -V 2, more quality at the same bitrates) and perhaps -q 5.

That's a pretty nice tradeoff between quality and filesize (my songs tend to average somewhere between 128 and 192Kbps)

Yes, and my point was that Ogg Vorbis VBR sounds pretty good even at lower average bitrates (particularly because it doesn't do the ringy sibilance stuff mp3 does). However. We now all know that there are no problems with traffic anyway, so the point is moot and your argument of major PITA wins the case for mp3.

[Billy's Little Trip]

I know this has been discussed in the past, but I never seem to come to that conclusion I need. Why can't it be easy, like when I order my lunch. Yeah, ummm, I'll have a double double with cheese, large fries and a chocolate milkshake.

The one thing that still sits in the back of my head is should I start mixing down with 24 bit instead of 16? From what I've read, CD quality is 16bit/44.1khz, so why would I need 24/96? It all ends up at 16/44.1 when I burn a CD, right. 24/96 uses up a 3 times the space. I know space is cheap these days, but is it worth it when you don't even need it? Maybe 24 bit with a 44.1khz sample rate is the ticket.

This leads me to my next thought about sample rates. Some say that analog vinyl records sound better than a digital CD. Even though the average human can only hear up to 20khz, could it be because a record is at 50khz, where a CD at 44.1khz is really only 22khz because digital is only half of it's peak rating. Still above human hearing. So that would mean that 50 khz is over twice what the human ear can hear where a CD is only 2 khz over human hearing ability. Is there something to this? Head room, more volume with better dynamics? So if I mix down at 96khz, I'll end up with 48khz, almost the same as analog.

Ok, too much coffee, I'm kind of firing off a lot of thoughts here.

[fluffy]

Really you should be encoding with LAME or WinLAME or the like, with the settings -b 32 -V 5 -q 5.

That's a weird combination of flags: -V 5 is the default, -b conflicts with -q. If anything, you should specify -h (= -V 2, more quality at the same bitrates) and perhaps -q 5.

Those are just the settings I've used for years but I know that sometimes LAME changes the semantics of its settings. But my understanding was that -q specified how much precision to use in the DCTs and -b specified the absolute minimum size of the frame, but it sounds like -h replaced -q or something. And, they've often changed default settings so it's nice to be explicit about those things.

I used to care a lot about getting the absolute best size/quality tradeoff but that was many years (and LAME versions) ago, so yeah I guess I need to update my knowledge. Or just start using one of the --alt-preset settings.

The only real justification for ogg is purely political

Actually, it isn't, because mp3 is just about the only audio codec where you have to pay for providing files that use it (if you make a certain amount of money, so that's not all too relevant for SF). Not even the AAC folks do that.

I'd consider that a political point as far as 99.9% of the Internet is concerned. The only ones who have to worry about that are commercial mp3 providers like Amazon and CDBaby.

That's a pretty nice tradeoff between quality and filesize (my songs tend to average somewhere between 128 and 192Kbps)

Yes, and my point was that Ogg Vorbis VBR sounds pretty good even at lower average bitrates (particularly because it doesn't do the ringy sibilance stuff mp3 does). However. We now all know that there are no problems with traffic anyway, so the point is moot and your argument of major PITA wins the case for mp3.

Sure, but AAC also doesn't have the sibilance problem, and it's supported pretty much everywhere mp3 is. It would still make things a major PITA for many people though (for example, I'd have to adapt my automatic download-and-tag stuff yet again).

The one thing that still sits in the back of my head is should I start mixing down with 24 bit instead of 16? From what I've read, CD quality is 16bit/44.1khz, so why would I need 24/96? It all ends up at 16/44.1 when I burn a CD, right. 24/96 uses up a 3 times the space. I know space is cheap these days, but is it worth it when you don't even need it? Maybe 24 bit with a 44.1khz sample rate is the ticket.

24 is a good idea for a pre-mastering mixdown but there is absolutely no point to having that in a final listening format. The maximum sampling error at 16bit is well below the threshold of human perception at any reasonable volume level (and if the volume is high enough that the sampling error would be perceptible then the actual sound is so loud that you still wouldn't be able to hear it ANYWAY). Also, converting from 96 to 44.1 can actually introduce further weird aliasing problems; if you're intending to target a 44.1KHz medium, you should record at 44.1 or 88.2.

This leads me to my next thought about sample rates. Some say that analog vinyl records sound better than a digital CD. Even though the average human can only hear up to 20khz, could it be because a record is at 50khz, where a CD at 44.1khz is really only 22khz because digital is only half of it's peak rating. Still above human hearing. So that would mean that 50 khz is over twice what the human ear can hear where a CD is only 2 khz over human hearing ability. Is there something to this? Head room, more volume with better dynamics? So if I mix down at 96khz, I'll end up with 48khz, almost the same as analog.

Vinyl sounds better because people expect it to sound better. I actually prefer the sound of CDs, but that's what I'm used to. The main differences between the two formats are really more to do with EQ and dynamics and harmonic distortion and the like. If you were to record a good record player's output to a CD and play them both back at the same level and randomly switch between them, it'd be very hard to tell when the switch occurs, and really the only ongoing perceptible signal difference would be because every time you play vinyl, the signal degrades slightly.

[jast]

The one thing that still sits in the back of my head is should I start mixing down with 24 bit instead of 16?

First of all, you need a damn good sound card for the 24 bit to make any difference. Secondly, well, let's just say I wouldn't care. The only time I'd really consider it useful is for when you apply a lot of digital gain on the signals, because then the additional resolution makes up for the scaling.

Even though the average human can only hear up to 20khz, could it be because a record is at 50khz, where a CD at 44.1khz is really only 22khz because digital is only half of it's peak rating.

That's partially right, but taken together it isn't. Here, have some details.

When people say that the average human can only hear up to 20 KHz (note that many mp3 encoders cut off at 16 KHz), they mean the peak frequencies, i.e. things with short wavelengths that got mixed into your signal. Sampling rates, on the other hand, are about representing an (approximate) analog signal digitally. It's true that a signal sampled at 44 KHz can represent frequencies up to 22 KHz, yes, but just because you can represent a signal with 22 KHz frequency sampled at 44 KHz doesn't mean you can represent it accurately. Remember, a sine wave corresponds to a very smooth curve. 22 KHz sampled at 44 KHz would mean we can only represent it as these data points: max peak, min peak, max peak, min peak, etc. Not very smooth, is it? The lowest frequencies always get the best levels of detail and it kind of detoriates as we go up the scale.

In short, the higher your sampling rate, the better for the accuracy of the high-frequency parts of your audio. I guess this is particularly relevant if you use filters that mess around with high frequencies. Still, I get along just fine with 48 KHz.

[fluffy]

But remember that downconverting from 48 to 44 introduces aliasing artifacts. Whether they're perceptible or not is debatable, however.

[Billy's Little Trip]

So, mixing down at 24 bit depth with an 88.2khz sr would be optimal it sounds.
I think the last time I was experimenting with 24 bit, something didn't like it. It might have been my MP3 converter, I can't remember now, because I gave up the fight out of frustration, lol. I'd make a terrible scientist, I'm too impatient.

Also, I do remember reading once that 48khz is preferred for movie sound tracks. Not sure why that is, but I'm guess it has to do with DVD players or some kind of movie industry standard.