The following came from Speeks page "If you want to make a normal audio CD from your MP3s you can use any CDR program to decode the MP3 files on-the-fly. But there are two reasons why you shouldn't do that:

- on-the-fly decoding can stall and then you get a coaster

- decoding engines in CDR programs are not top quality

So it's best to first decode the MP3 files to wave files with a good quality decoder. And the best decoder is MAD. It features dithering, clipping report, attenuation and amplifying."

What are some of the higher quality decoders? Thanks Jeff

1. - You shouldn't copy texts from Speek's page and don't give him credit, you know that?

2. - Here's a good MP3 decoder test. It's a bit outdated, but it's the best one I know of:
http://privatewww.essex.ac.uk/~djmrob/mp3decoders/

Regards;

Roberto.

your right RJ, I'm sorry. I hope that I did that correctly. Jeff

Eheh. No problem, it's fine with me.

Regards;

Roberto.

i dont see MAD as the "best" decoder there is. the mp3 decoder in winamp is also high quality, and although i can't speak for everyone else, i myself cant find any significant improvement in quality between MAD and the winamp decoder aside from MAD being able to perform clipping attenuation. also, the mp3 decoder in programs like sound forge and wavelab are also very high quality

I agree that the MP3 decoder in winamp is also 100% OK. (Winamp 2.666 and on, of course. 2.22 too).

But I'm always afraid of someone using Winamp to decode and leaving the equalizer, or some kind of DSP plugin, on. That's why I find safer to recomend MAD or mpg123 as decoder for CD burning.

Regards;

Roberto.

I can certainly hear the difference between the Nullsoft mp3 decoder and the MAD and mpg decoder, although I cannot tell the difference between MAD and mpg on my system. Its a personal perference but the Nullsoft is definitely number 3 for me. I would like to see some ratings determined by good listening tests done by people who have high quality gear as opposed to "number testing". Are there any other mp3 decoders to compete with the three mentioned ie ones that can plug into Winamp?

Marcus

That's very strange. These three decoders are 100% ISO compatible - I.E: The only way one of them could be providing different quality is by post processing, which Dibrom explained in this post as not being really recommendable quality-wise.

Are you sure it's not just psychological? Have you tried blind-testing outputs of these decoders?

Regards;

Roberto.

MAD provides dithering support.

this can enhance the accuracy in some cases.

I think the more important feature of MAD is the clipping protection.

Clipping Attenuation:
Dithering Support:
Clipping Protection:
Thanks Jeff

Dithering is a bit difficult to explain shortly. Basically dithering means, that a very small amount of noise is added to the signal to make quantization "smoother".
In this case in mp3 decoding you can add a bit noise so that when the 16 bit signal is formed (decoded) from the MP3-data, the waveform (when zoomed very closely) is a bit smoother.

Clipping protection/attenuation means that the amplitude of the signal is reduced so that it wouldn't hit the 16 bit digital full scale values (-32768 or 32767) many times in row which leads to clipping.

My experience with MAD is that it can be a bit *too* literal... that is, it doesn't correct damaged MP3's effectively. Occasionally I'll d/l something from the Net that plays just fine with WinAMP (and thus, decodes to .wav fine) yet MAD screws it up (cuts off prematurely, or repeats sections of the song, dunno what causes it exactly).

P.S... the only time I really d/l from the Net anymore is when my younger brother (who doesn't care much about sound quality) wants a compilation CD, mebbe a couple times a year... just thought I'd add that .

In my experience MAD handles otherwise obviously hearable sync errors better than Nullsofts decoder. They are less pronounced with MAD

Well MAD manages 24 and 32 bits outputs...
Make a 24 bit file then convert it to 16 (with a precise tool like Batch Converter) and u'll have the best results...

But the point is that MAD decodes at 24-bit accuracy internally whatever output accuracy you choose - so 16-bit MAD output is dithered from 24-bit anyway. The dither in the current MAD plug-in is better than the one I original tested on mp3decoders.org. The advantages of using something else to convert the 24-bit output to 16-bits are very small.

Cheers,
David.
http://www.mp3decoders.org/

EDIT:

Can I quote from my own web site - I know the tests are out of date, but for those versions of that software, this still holds true:

Real Jukebox and Media Jukebox are both good choices if you don't want to spend any money, and are quite easy to use. If you already own Easy CD Creator Deluxe 4, and don't mind the last 1/20th second being clipped off your files, then that is a good choice too. For CDs that really matter, I still decode to .wav first.

I am currently considering following Patchwork's suggestion which states:
Well MAD manages 24 and 32 bits outputs...
Make a 24 bit file then convert it to 16 (with a precise tool like Batch Converter) and u'll have the best results...

My question is there any advantage to using Cool Edit along with the MAD decoder? As stated above, "u'll have the best results..." what results might that be? And, where do I find Batch Converter? Thanks Jeff

surely most mp3's are encoded from cd, which uses 16bits. what's the point of using 24bits internally if the source (the cd) and the final result (the wav file or playback) are both 16bit?

or am i missing something obvious here?

see

http://privatewww.essex.ac.uk/~djmrob/mp3decoders/24bit.html

especially the "Do we care?!" section and FAQs 1-3.

Cheers,
David.

Thanks for the link, interesting article.

I can see it now... use L3DEC to decode, convert the HEX files to .WAV using the conv2wav program, load into Cool Edit Pro, downsample to 16-bits and burn to CD... LOL! And probably zero audible improvements...

One thing I'm not too clear on. With 16-bit audio, you have 8 bits representing each stereo channel (so I assume). That allows a number from 0-255 to represent each sample, which doesn't sound like a whole lot. From everything I'm reading, bit-depth seems to have more to do with dynamic range than anything else, is this correct?

The combination LAME with RazorLame should also be mentioned as a clean, free and no frills solution (no dithering, no attenuation).
At least it knows about it's own offset
--
Ge Someone donated to this thread.

You can try the MAD Decoder. I think it supports 16-bit with dither, which is nearly the same method that Cool Edit uses.



No, with 16-bit you have 16-bit for each channel.

About the audible benefits of decoding to 24 bits, see:

http://www.ff123.net/madchallenge.html

"The only feature of MAD which I have not seen demonstrated to my satisfaction is the benefit arising from the dithering to 16 bits. I have tried blind tests of various samples for myself and have so far not been able to demonstrate that I can hear a difference between MAD and the Fraunhofer decoder within Winamp versions 2.666 and above. I administered a blind test on Usenet as well (alt.binaries.sounds.mp3.d), and included Nawhead in that test, but with negative results."

I believe that, as ff123 says, nobody has till today be able to tell a difference between a WinAmp 2.666 or above 16 bit decoded file and a MAD 24 bit decoded file, in a blind test.

As always: the reason people say they can "feel" that MAD is better than FhG is psychological.

Or it's the warm feeling of using an Open Source decoder, not a closed one...

Regards;

Roberto.

So, it looks like it comes down to using two of the top decoders to decode mp3 into what HEX? The two top decoders being L3DEC and/or MAD. Can anyone paste the link for one if not both of these decoders? Thanks Jeff

thanks for that. just read a bunch of stuff about internal accuracy and dither too which all makes sense.

btw, how does the MAD decoder handle the winamp eq? does it do it's own 24-bit internal processing or does it hand the 16-bit dithered version to winamp for eq'ing?

My comments are about burning mp3s to audio CD in a 1-stage process. Last time I checked, RazorLame didn't do this.

Cheers,
David.

P.S.
If I ever finally do update the site, I'll include a new test that shows the decoders scoring 6 AREN'T as good as the ones scoring 7 on this page:
http://privatewww.essex.ac.uk/~djmrob/mp3d.../objective.html
This would cut down the list of recommended decoders dramatically. However, the error (and it IS an error - I couldn't prove it before, but I can now) is at -90dB. Until I publish the test, don't quote me! ;-)

Sorry, but that wasn't obvious as also WinAmp and MAD were discussed.

That would be interesting just to know. As the page states:

In other words it seemed to depend on what's the reference, but we'll know more after you publish your results.

Oops, I did it again.
--
Ge Someone