Hey guys.

Something's been bothering me for quite some time now. I listen to web radios a lot. And there's this one radio -snakenetmetalradio- that has good playlist and all but the stream is of bad quality, even though its 128kbps mp3 stereo. The other web radio station I listen to -baranen- has same bitrate and is also stereo and is of far better quality. It is as if one would encode with LAME and the other with BLADE. Or as if one uses -q2 and the other -q9.
I have not done much reading-into for mp3 straming and all, but I have a feeling that something like this is the reason for such difference in quality.
I intend to write to snakenetmetalradio's moderator and staff, but first I want to know how to say it to them. How to explain what's bothering me, and probably a lot of other listeners as my hearing is not above average, as I don't hear any difference between the original and the mp3 from -V 4 and up...
Thanks for all (detailed) explainations.

EDIT: Typo fix.

Today most mp3 encoders are good: lame, fhg, xing, itunes so the encoder quality shouldn't be a big issue and I don't see or hear of people using blade. I'd say transcoding is the offense. Usually from bad sources too. Losts of radio stations are affected, some worse than others. Notice they have multiple streams. Its not hard to figure out that there is some lossy transcoding taking place - mp3 > mp3, although they all give 128k. The sound is really bad to me too the point being ill. This is something that should be pointed out to the owner of the station. I assume most 'normal' people are ringing / preecho deaf, but the rest shouldn't suffer and its similar to the loudness race.

shadowking, thanks for the suggestion. I will write to them and try to explain. Either they do double mp3 encoding, like you said, or they use some -q9 switch, because they need speed. I think the former is the truth.

I haven't used it in over a year, so this may no longer be true, but Winamp's Shoutcast server plugin did indeed default to -q9 (and the sound quality was correspondingly atrocious). Fortunately, it can/could easily be changed in the .ini file, but since it wasn't documented, most people just ran with the default.

Changing the quality setting to 3 - the LAME default - made a huge improvement. I seem to recall that the results weren't quite as good as lame.exe with some problem samples, so Winamp may have been changing other settings, as well, but it was a hell of a lot better than the default of 9.

Thank you too, mixminus1. I have wrote to them trying to explain what might be their problem with both your suggestions. They haven't replied yet, either they're busy or something else. I will post here when and if I get a reply.
Let's hope in time web radio stations will all care more for their sound quality as much as they do for their playlist quality.

That is bad news. Can't the LAME devs map -q9 to -q7 ? This is what still gets me with LAME. With the other encoders it is not possible or not that easy to produce garbage quality. In contrast, The FHG encoder works well without any of these dangerous options.

I thought I'd test samples from these stations with TagTwo v0.8:

SnakeNet sample (10'24", 9.53 Mb): LAME 3.86
Baranen.com sample (9'22", 8.68 Mb): LAME 3.97 alpha


EncSpot says they're VBR files (which they are not, they're both 128 kbps CBR):

SnakeNet sample (10'24", 9.53 Mb): Xing (new) (!) - stereo (!)
Baranen.com sample (9'22", 8.68 Mb): LAME 3.97 alpha - joint stereo


Both streamrips (w. 'GetASFstream') show quite a lot of sync errors..
(stream URLs:
http://205.188.215.231:8016/ and
http://baranen.hopto.org/bazzanen.asx )

Internet radio became the "in" thing while I was in college, and peers attempted to get me involved with their favorite "stations". Back then sites used 24 kbps Real Audio, which probably sounds about as good as using Speex for audio today. I just couldn't get into it and was branded an audio snob (though I had no problems with 80 kbps TwinVQ files). People just couldn't understand that I was unable to hear the signal through the noise. I couldn't make out what anyone was saying unless I dedicated full attention to the stream, which defeated the purpose of "radio". So there's definitely something to the concept of "normal people" vs. "audio snobs".

@Pesmontis

I've not read this topic for some time now, I thought it was dead. I just saw your post today, and I must say your findings are very interresting. Thank you for this info.