MP3 vs. WAV

[DizRotus]

The free download of Beethoven symphonies in MP3 format, which Malcolm kindly brought to the Forum's attention, prompted me to question the differences between the MP3 and WAV formats.

I understand that MP3 files are compressed and smaller than their WAV counterparts. Intuitively, the WAV format would seem to offer greater fidelity. Is it possible for the conversion of the MP3 to WAV to improve the fidelity? It seems like a lifting by boot straps kind of thing.

I burned #6 & #7 onto CDs in MP3 as well as WAV files. Both symphonies fit on a single CD as MP3s, not so as WAV files. I will compare the MP3 CD with the WAV CDs to see if I can discern any differences.

I'm assuming that others who frequent the Forum know the technical differences and relative merits of the two formats.

[smuttynose]

I'm not gonna pretend to be a computer expert, but I think that if you convert a mp3 file into a WAV, or AIFF, which is the non compressed version, you are not going to help matters much. mp3 makes the sound file take up less space in the digital domain by taking a sampling every so often during the song. There is only a limited amount of data to work with in the mp3 file format. To expand this into a larger bit rate file means that the computer software has to extrapolate data (maybe that should be interpolate) in order to fill in the blanks.

For example:

a CD has raw digital data at something like 1440 kbps, mp3 downloads are normally 128 kbps. In order to convert the 128 kbps mp3 into a sound file to burn to CD, the software must try to expand what information it has to fill out the gaps. Of course, this is done in a scale that makes it almost imperceptable to the listener. Unless you have a tube amp and some Klipschorns, then you may as well forget about it.

I listen to tunes in my truck via an iPod, mostly converted at 192 kbps, and sometimes I listen to the original CD, WOW, what a difference. And that is via a vehicle stereo system, unmodified. With all the assorted road and vehicle noises.

But then again, volume sometimes overrides quality. As in the case of my iPod. I hit the road and have a few hundred CDs of music in my pocket. And I haven't been able to figure out how to fit my horns into my truck...yet!

[scriven]

I will pretend to be a computer expert. (I have been in the industry for 30+ years.) Wav files are recorded at 44100 samples per second. This is also the standard for a Redbook CD. Assuming a stereo signal that is 44100 samples, 16 bits per sample on 2 channels that is 1.4 million bits per second. (This method is also known as pcm or pulse code modulation and is a loss less format.) MP3 sacrifices quality for saving space on the recording media. I have taken a wav file, converted it to a 320kbs MP3, which is the highest quality MP3, then did an A/B comparison to the original. Even with the lower compression in the 320kbs MP3 the difference was obvious. If you are after high fidelity stick with wav files or some version of loss-less compression. (That is a compression algorithm that allows you to correctly recreate the original from the compressed version.)

DizRotus,

Your intuition is good; MP3 is not as High Fi as wav. I am guessing you heard no difference between the wav and MP3 cds you burned from the MP3 files. Try the experiment differently, burn an MP3 cd from a Redbook cd and compare them. I encourage you to try different compression rates so you can see for yourself what the differences sound like.

The advantage to wave is good sound, for MP3 it is space savings.

Smuttynose, (by the way, nice moniker!) you are correct that you cannot expand an MP3 back to the wav format to recover the information. The conversion algorithm will have to "make up stuff" to create the file.

You don't need tube amps and Khorns to hear the difference.

Note to people who know a lot about audio but not computers: The type of compression that we are discussing here is not what is used to reduce the dynamic range in audio that can be undone by an expander. This is an algorithm running in a computer that reduces the space used to store the signal. The dual use of the term compression has caused confusion in other threads.

[smuttynose]

Mark,

Thanks for the info. Is AIFF the same as WAV, as far as sample rate is concerned? Also, how great an impact does the encoding hardware make when converting a CD or LP recording into mp3 or burning to a CDR. For example, that built in to a computer vs a PCI card such as Emagic Audiowerk or such? Thanks.

ps. There is a nice little brewery in NH with that name.

[scriven]

"Is AIFF the same as WAV, as far as sample rate is concerned?"

AIFF is a loss-less format used by Apple (and a few others). It is not the same format as wav but it is functionally similar. It supports several sample rates, sizes and channels. For more information than you ever really wanted to know about it click

here! (Be careful! They have come up with AIFF-C, which does support compression. I am not sure if it is loss-less or not.)

"Also, how great an impact does the encoding hardware make when converting a CD or LP recording into mp3 or burning to a CDR. For example, that built in to a computer vs a PCI card such as Emagic Audiowerk or such?"

I am not sure exactly what you are asking here. You brought up two very different situations.

First if your source is a cd drive in the computer, or sending the computer a digital signal, encoding hardware makes no difference. You are just reading bits from the drive, applying the MP3 algorithm to them and recording the results.

When going from an LP your signal must go through an analog to digital converter. My experience is that most of the A/D converters on sound cards are not that good. One problem that they cannot overcome is that there is a LOT of RF inside a computer cabinet where they live. If you are serious about retaining the fidelity use an external A/D converter. Of course, if you are converting to MP3 then you are not optimizing for quality!

I don't use any of the equipment you asked about so I am not in a position to advise you on it.

[Erukian]

This is a fun experiment. Take a song. Convert it from WAV to MP3 @ 192kbps

Now make 4 copies of each filetype giving you 10 copies, 5 of each.

You can do this two ways, first try to have someone (not you) mix them up randomly and burn them to a CD. That person also takes note of a cheat-sheet of which ones are from MP3 and which are from WAV. Then you put on your listening cap on, perk out your ears, and try to discern the differences between the songs, rating each song as either MP3 or WAV. Then look at the cheat sheat and see how you did when your all done.

You can also do this by yourself by taking the 10 audio files, making the first 5 WAV and the last 5 MP3.

If you can put your CD-player on random, cover up the LCD and when the song is just about done playing, pause it before it's over, restart it if you want to listen to it again then rate it as MP3 or WAV. Make sure it's paused before it's over so you don't lose the track # of what you just listened to so when you take a peek to see how you did, you'll discover if you can discern any differences.

Blind listening tests are fun because they give the listener an idea of what they actually hear vs what they want to hear because someone with golden ears told them they should hear the difference. It doesnt mean though however that their ears cant train to pick up the artifacting in MP3. Just like a musican can train his ear to pick out flat/sharp notes.

-Joe

[wmilas]

I'll pretend to be a puter expert. I'm the author of Kexis which as pretty much lapsed and Flac has taken over.

AIFF-C Is a mess. Don't use it.

If you are looking for good, lossless compression take a look at shorten, kexis (both long in the tooth now) or better yet, Flac. Flac is being supported by alot of high end hardware now, and its the best software lossless compression codec by far.

Lets define a few things for those of you who may be confused.

As stated above Wav file are just PCM audio with a header noting what KIND of PCM audio. Please note that Wav is NOT limited to 44.1khz samples at 16 bit stereo. The whole idea is that the wav file tells the software/hardware thats processing it what kind of PCM data is packaged along with it. It might be 44.1/16/2, it might be 48/24/4 (48 khz 24 bit 4 channels). PCM is the defacto standard format for digital modeling of sound... for instance your CD transport sends 44.1/16/2 PCM data to your pre/pro via a toslink or digital coax connection. So think of Wav as the container that holds the pcm data.

PCM data is the father of all data types. Everything must be turned back into PCM data to be played digitally.

MP3 is vastly different. Also note that there are MANY different types of mp3. MP3 is a lossy format that throws away data as it takes a PCM stream and converts it to a mp3 datafile. As almsot all of you here are aware, human hearing is more sensative at certain wavelengths. mp3 uses a psychoacustical model that filters out "lesser" wavelengths and sounds that are hidden. Its really complicated to explain and whole books are written on it. Lets suffice to say that the tighter the mp3 compression, the larger the amount of original PCM information that is thrown out.

Some quick math. 44.1k samples per second at 16 bits x 2 channels = is 1411200bits/second (b/s) or ~1378 kilobits/second (kb/s).

Mp3 is usually compressed at 96, 128, 192, 256, or 320 kilobits/second. when you see 128k mp3 thats kilo BITS not kilobytes. As you can see 128 is ALOT smaller that 1378.. its less than 10% of the size. ALOT of information has to get thrown out. As you can imagine when a 128k mp3 file is converted back to PCM to be played, it is not nearly the same PCM that the original was. Its a close aoproximation, but its not the same.

There have been improvements to mp3 over the years. The msot noteable is VBR (Variable Bit Rate) mp3. The idea here is pretty simple. A 128k mp3 guarantees that there are going to be 128 kilobits per second no matter what. 10 seconds of mp3 128k = 1280 kilobits in file size. VBR mp3s adhere to the total file size constraint BUT the bitrate INSIDE the stream changes as long as the average bitrate through the whole stream works out to the bitrate stated.

Example. Lets say you encode a classical track at mp3 vbr 128. There may be long segments where one channel, or both are completely silent. The encoder will use 0, or very few bits to encode this. Think of this as a bucket. All the bits that were saved because there was little, no, or easy sound to be encoded get stashed into this bucket. Later on when there is alot of sound there are extra bits saved in the bucket and the encoder can now use more than 128k to encode the "harder" parts. This results in a stream that sounds better but is still the same total size. Note that is still a lossy stream and the resultant PCM stream when played is NOT the same as the original.

Ok lets move onto our next format. Ogg Vorbis. While many of you more sophisticated users may know or have heard of this format, alot of you may not know of it. If you use mp3's alot you should drop what you are doing and look into Ogg Vorbis now. Ogg is a freeware replacement for mp3. It came about because, technically, mp3s are patented. Ogg was created as an alternative. Its now MUCH more than that. Ogg compresses tighter than the best mp3 formats AND it was developed with sound quality in mind. Its psychoacustical model is much better than mp3 and when you compress, you don't give its a k/s number.. you give it a Quality setting. The quality setting translates to an aproximate b/s number, but there is no hard and fast rule. As the code is developed over time (although its a very mature code base.. and hundreds of hardware and software players support it now) using the same quality setting will result in smaller files but still sound as you expect. So basicly you find a quality setting you are happy with, encode at that, then forget about it. From an audiophile standpoint its much better than picking a bitrate like mp3 and hoping it sounds like you expect.

If you like mp3's look into Ogg. Right now.

The last format I'd like to talk about is FLAC. It stands for Freeware Lossless Audio Codec. Its a lossless codec that takes a PCM stream of any bitrate, sample frequency and number of channels and compresses it without losing any information. When you decode that flac stream you get the exact same PCM stream that went in. Everytime. It differs from a more general compressor like zip, rar, lharc, gzip, bzip2 in that the compressor is tailored to work on sound information and does a MUCH better job. You typically see 30 to 60% compression rates. Big band tends to compress horribly, pop/rock pretty good, classical great. Obviously the file sizes when compressed are larger than Ogg/Mp3 but they are not nearly as large as Wav files, and they are bit for bit identical when decompressed to the original PCM. If you are ripping your cd collection to hard drive to keep archived THIS is the format to use. You can then use the FLAC files and make a Ogg files from them for your portable player needs. You have the best of both worlds. A compressed exact PCM track and the ability to use that PCM data to make lossy (ogg, mp3 ect) tracks for other uses.

I hope that explains the wonderfor world of codecs to you Class is dismissed for the day.

[scriven]

wmilas,

First, welcome to the forum!

You pretend pretty well! Not being an Apple user or working on codex apps I am not all that familiar with AIFF and AIFF-C. I have heard that AIFF is acceptable, I am speaking of sound quality, but have heard some rumblings about AIFF-C having problems. You seem to have verified that AIFF-C has problems. Could you elaborate that?

[rplace]

Hopefully I am not hijacking the thread too muchbut it sounds like you guys are just the crew to answer a question that has been bugging me.

I currently have every track from every CD I own ripped to mp3 or WMF and on my dedicated music servers (read computers) hard drive. I do this so I dont have to open jewel cases and load up the CD player. I have a turtle beach Audiotron (great device) with digital output to my Pre/Pro. This gives me nothing but digital until it reaches the DACs of my Pre/Pro. It is working great in a living open on 3 sides with a two story ceiling and KSB 3.1 speaker for background music.

I want to more or less duplicate this setup in my more refined listening room where I already here the limitations of mp3 files played on my system.

I do not care one single bit about the HDD space it takes up. I am willing to buy the necessary disc(s) to get the best possible sound. I want to re-rip all my music to the best sounding format but dont really know what that is.

Questions:

1. Should I use Wav, FLAC, something else?

2. What is good software to accomplish this? Something that will use CDDB or the like to catalogue them by Artist, Album, Track, etc.

Thanks a ton!

Rich

[Petrol]

^^^ or for lunch break at least, lol

[meuge]

The best ripping software is CDEx (google it - it's open-source and free). As long as you provide any (not even a valid one) email address in the CDDB options, you get CDDB access.

There are a variety of naming/ID3 tagging options that are variable-based and can be customized to your preferences. I recommend using the following string:

%1 - %2\%3 - %4

which will make "Wish You Were Here" look like:

\Pink Floyd - Wish You Were Here\03 - Wish You Were Here.wav

As far as storage medium, I would say that WAV is the way to go. FLAC is lossless and takes up 30-40% less space, but to be honest, it's a bit of a hassle to use, given that it's not built into CDEx and the proprietary software could use an overhaul. Frankly, the saved space is probably not worth the effort. CDEx will rip to WAV in 2(!) clicks of a mouse, so it's very fast.

As far as disc space, I would say that an average pop/rock music CD will take up 400-450MB, and an average classical music CD will take up 500-550MB.

I recommend getting a new hard drive, dedicated to music... something humongous. I recommend this:

http://www.newegg.com/Product/Product.asp?Item=N82E16822144391

It's a 300GB drive from Western Digital, which is only $144 after rebates

HOWEVER, for a VERY LIMITED TIME you could grab this steal:

http://shop.outpost.com/product/3492233

This is a 200GB Maxtor drive, for only $70 right now!!!!

[wmilas]

Take a look at:

http://www.borg.com/~jglatt/tech/aiff.htm

Start reading the documentation. Its a complete mess. The core was designed by SGI years ago and its been added to over the years.

The problem is tht it tries to do too much. There are no less than 14 different compression types that I know of.. and there may be more. No one seems to know as you can just increment the bitfield to add your own. No one in there right mind is going to write a codec that supports 14 different types of compression... that are all old and pretty much suck btw.

Then you have the wierd limitation on chunk size, channel formats and anyone seems to be able to add thier own chunks without documenting exactly what they are.

Its a format thats been patched and hobbled over time with no real oversight. It works fine for apple cause they use a very specific subsection of it and ignore everything else.

Does that answer your question or would you like more info?

[STL]

Once a music file has been compressed with a lossy compression (like MP3) then converting it back to a WAV gets you absolutely nothing. The new WAV file will be larger in size, but it won't be any better in quality than what it was encoded in because the "damamge" was already done.

[DizRotus]

----------------

On <?xml:namespace prefix = st1 ns = "urn:schemas-microsoft-com:office:smarttags" />

6/30/2005

11:37:38 AM

rplace wrote: <?xml:namespace prefix = o ns = "urn:schemas-microsoft-com:office:office" />

Hopefully I am not hijacking the thread too muchbut it sounds like you guys are just the crew to answer a question that has been bugging me.

Rich

----------------

Hi-Jack away!

Your question is right on topic. I too want to know about the various options to maximize fidelity from MP3 files; everything from the BBC Beethoven downloads to obscure oldies that I have saved as MP3 files.

From what I've read so far, it seems that converting the MP3 files to WAV files will increase the file size (and storage space needed) without significantly, if at all, improving the fidelity. Is this a "garbage in, garbage out" situation, or am I missing something?

[wmilas]

Questions:

1. Should I use Wav, FLAC, something else?

2. What is good software to accomplish this? Something that will use CDDB or the like to catalogue them by Artist, Album, Track, etc.

1) You should use either Wav or Flac. Flac takes more processing time but uses less space. I personally have a 800 gig raid 5 array serving up all my audio. I have 2 trees. One is a complete set of flac files, the other is a complete set of Ogg files at the quality I prefer for my iAudio player.

2) Depends on your operating system. I'm more familiar with linux, but I'm sure there are dozens of rippers in windows that work well. There are dozen in linux that support CDDB and flac/Wav.

If you are looking for a player to attach to your system and keep the audio digital from your server to your Pre/pro with Wac/Flac/Ogg/mp3 take a look at:

http://www.slimdevices.com/

The squeezebox2 is pretty good. Whats better is that the server software is developed as GPL and cross platform. Its dacs i have no clue about but all you need is a little device to pass the data your the dac in your pre/pro

[wmilas]

The best ripping software is CDEx (google it - it's open-source and free). As long as you provide any (not even a valid one) email address in the CDDB options, you get CDDB access.

Well you are assuming Windows. Alot of people dont use windows. I prefer Grip myself for linux since you can customize it to use any CDDB like service and any encoder front end you like.. plus its scriptable.

[scriven]

"From what I've read so far, it seems that converting the MP3 files to WAV files will increase the file size (and storage space needed) without significantly, if at all, improving the fidelity. Is this a "garbage in, garbage out" situation, or am I missing something?"

Neil,

You got it!

[scriven]

wmilas,

That's sufficient. I had only heard roomers and innuendo without any specifics. Now I know to get my 10' poll out!

[wmilas]

About converting mp3 to wav files...

Remember I said that Wav is just a container for PCM data? Ok. So lets go through the steps.

Take and original PCM file from a CD and encode it to a 128k MP3 file. Now you want to listen to it. The mp3 player takes the mp3 file and decodes it to a raw pcm stream and dumps it to your dac or your transport going out of your computer.

Lets say instead you decide to take that mp3 file and convert it back into a pcm stream and save it on your hard drive. You would use a Wav container to do this. So you end up with a Wav file. But as mentioned before, that PCM stream in that wav container is NOT the same as the original PCM stream because mp3 is lossy. It threw out alot of information when it was originally compressed,. Therefore converting a mp3 back to wav file just yields a really crappy PCM stream encapsulated in a Wav container file that takes up alot of space. There is no reason to do it.

[DizRotus]

It seems that the only reason to convert MP3 files to WAV files is to burn CDs to use in players that will not read and/or play MP3 CDs, i.e., that require WAV files.