How to get good sounding .mp3? + downsampling question.

[druid]

I posted this question in my thread regarding my latest track 'Abandoned Hope' but no one responded to the thread at all, and the question is important to me so I thought I'd just make a thread about that itself.

I found that the .mp3 I made has filtery sounds in the high frequencies. I've heard this before with the accursed .mp3 format, but I don't know what causes it. I think I recall reading somewhere that if you have higher frequencies in the track, above 18khz maybe?, that it will do this. I have not experienced this with other tracks, but this is the first time I mixed etc at 88khz instead of 44khz, so there would definitely be higher frequencies in there unlike the other times (even though I downsampled, it would still have remnants of the higher frequencies blurred down, like downsampling the resolution of a picture, right?).

Does anyone know how to prevent this? Do you know what I mean?

[mmdejonge]

For a nice MP3 I use:

RazorLame:
http://www.dors.de/razorlame/index.php
It needs Lame encoder:
http://www.rarewares.org/mp3.html

I think it's best practice to render the file (with dither) down to 44.1 / 16 bit and then convert it to an mp3 - though I'm no expert in these matters

[CaseR]

http://www.voxengo.com/r8brain/

[loomchild]

There's a command-line switch for the *Lame mp3 encoder* that allows you to set the lowpass filter threshold and its width.(--lowpass xx where xx is the the frequency in Khz, ie --lowpass 20 will cut everything above 20khz.)

However, the more frequency content you are lossily processing the higher the risk is that it will not sound transparent. This is basically why the mp3 encoder, even on the hi-quality presets, performs filtering before encoding.

The best solution varies depending on what you are after.

1) If you plan to distributing the file(in the KVR café, on your website etc...) the mp3 format is definitely the one to choose. You should then get "Razorlame" and the "Lame 3.96.1" from the links mentioned by a previous poster. Check the recommeded settings from this post from the Hydrogen audio community(community of audiophiles) http://www.hydrogenaudio.org/forums/ind ... opic=28124

You may also consider the Ogg vorbis format(which is freeware, open source and patent free), which is a tad less common but which truly butchers the audio less. It can be played on Macs, Pcs running windows, and-nixes etc... It's main drawback is to be less well supported by hardware players than the mp3 format is.

Head over there to get a list of compiles and settings http://www.hydrogenaudio.org/forums/ind ... opic=15049

2) If you want to archive the file, it might be interesting to consider the lossless encoders. Here is a link to my personal favorite : wavpack
http://www.wavpack.com/
It'll compress the audio(will make you gain around 50% in size) but won't perform any lossy process on it. In other terms, once decoded, the file will be bit for bit identical to the one that you have compressed.


I really recommend to anyone interested in the topic to read the Hydrogen Audio wiki http://wiki.hydrogenaudio.org/index.php?title=Main_Page
which is a real goldmine about ripping, encoding, codecs, players etc...


I hope it helps,


btw, most codecs have a resampling module in their process, but you should consider the resampler from Voxengo mentioned above since it's really efficient.

[herodotus]

I really recommend to anyone interested in the topic to read the Hydrogen Audio wiki http://wiki.hydrogenaudio.org/index.php?title=Main_Page
which is a real goldmine about ripping, encoding, codecs, players etc...

Indeed it is.

Thank you.

[Stupid American Pig]

Lame is the only encoder I would use- the high end sounds better than freunhoffer, or blade IMHO...

[druid]

Thanks for your responses.

I downsample to 44.1kHz 16bit and THEN convert to .mp3. However, I'm convinced that because the master file was 88.2kHz, the higher frequencies that cannot be stored in half the resolution will still have some sound apparent. If you make a picture and halve the size, it blurs the finer detail, so you still get elements of that finer details. I don't know what it's called in the audio world but I'm pretty sure it's called anti-aliasing in the graphics world.

Bicubic filtering etc, that sort of thing.

So I do only mp3-ify a standard CD quality file.

On the topic of .ogg, I use that for my own personal use, but to release tracks on all the online music distributors such as music.download.com, funender, soundclick etc, I must have an .mp3. This is where i run into problems. I personally prefer .ogg files. As for storage of originals, I just keep the .wav files at whatever resolution I mixed it at, along with a CD quality one as well so I don't need to limit and downsample every time I want to compress it or use it.

I will check out some of these compressors and the Hydrogen Audio forums as soon as I can.

[cron]

I know I milk this all the time, but the problem could be that you aren't using joint stereo encoding. At 128 especially, joint stereo is the difference between a good listen and a watery sounding high end. If you have any prejudice against joint stereo (which many do), make sure you at least try it in LAME. It's fair to say that joint stereo doesn't have a great reputation in many other mp3 codecs, but LAME really has it nailed.

edit: BTW, I like CDex for encoding as it comes with everything you need (LAME encoder + presets, and an ogg encoder). You can read about the presets and their history at Hydrogen Audio. I rarely have any need to use different settings.