Axiom Home Page Forums General Discussion The Water Cooler FLAC vs MP3 (Stereo, Joint Stereo, Dual Channel)

Ok, I know that the FLAC vs MP3 thing has been done to death but in comparing things for myself I found that not all MP3 rips of the same bitrate are created equal. So I’m posting this in hopes to add something to the discussion that doesn’t come up here that often.

Before deciding how to rip my CDs I did some testing and ended up getting an education on MP3 ripping. Initially I used Media Monkey to rip a track to FLAC, MP3 128 and 320. I listened to the FLAC rip to get a baseline and then set them to random play. Wow was there a difference. I could pick out both MP3 tracks instantly. No way do I have “golden ears” so I set about to find out what happened.

Turns out, something I didn’t realize, that there are other factors than source quality and bitrate which determine the quality of the MP3 file you get. Apparently different encoders produce slightly different results but what I was noticing was the difference in the method of stereo encoding.

I did some listening tests with my Sennheiser HD600 headphones hooked up to my Denon 2807 with the music being played from my NAS via SoundBlaster X-Fi Extreme audio on a MSI P7N Diamond motherboard hooked up to the Denon with coax (SPDIF) sending 24 bit 96KHz PCM.

I chose 3 tracks:

1. Dire Straits : Calling Elvis (On Every Street, Warner Remasters)
2. Antonio Vivaldi : The Four Seasons, Spring Concerto No. 1 in E Major, (HDC Classics)
3. After the Fire : Der Kommissar (80s Pop Hits, Sony Music)

-----------------------------------------------------------------------------------
Don’t want the boring details you can skip to the conclusion at the end
-----------------------------------------------------------------------------------

I used Media Monkey to encode each track to the following with the file size listed for “Calling Elvis“:

1. FLAC - 38.8MB
2. MP3 Stereo 128 - 5.9MB
3. MP3 Stereo 320 - 14.8MB
4. MP3 Joint Stereo 128 - 5.9MB
5. MP3 Joint Stereo 320 - 14.8MB
6. MP3 Dual Channel 128 - 5.9MB
7. MP3 Dual Channel 320 - 14.9 MB

For the first test I listened to “Der Kommissar” figuring it was the worst recorded track (turned out it’s not a bad recording). First I listened to the FLAC to get a baseline and then hit random play and tried to figure out if it was the FLAC or an MP3. As each MP3 played I eliminated it from the playlist and continued. I didn’t try to guess but only called the track an MP3 if I was sure it wasn’t the FLAC version. Here were the results:

1. FLAC - NA
2. MP3 Stereo 128 - couldn’t tell
3. MP3 Stereo 320 - couldn’t tell
4. MP3 Joint Stereo 128 - MP3
5. MP3 Joint Stereo 320 - MP3
6. MP3 Dual Channel 128 - MP3
7. MP3 Dual Channel 320 - MP3

Numbers 4-7 were very easy to spot. Unsure on 1-3 so I didn’t make a call.

For the next tests I listened to the FLAC version of “Der Kommissar” and “Spring Concerto” using blind A/B comparisons of the FLAC vs each other rip with the following results in this order:

“Der Kommissar”/”Spring Concerto”

1. FLAC - NA/NA
2. MP3 Stereo 128 - MP3/couldn’t tell
3. MP3 Stereo 320 - couldn’t tell/couldn’t tell
4. MP3 Joint Stereo 128 - MP3/MP3
5. MP3 Joint Stereo 320 - MP3/couldn’t tell
6. MP3 Dual Channel 128 - MP3/MP3
7. MP3 Dual Channel 320 - MP3/MP3

For the next test I A/B compared “Der Kommissar and “Spring Concerto” FLACs vs each other rip listening for any differences knowing which track I was listening to with the following results in this order:

“Der Kommissar”/”Spring Concerto”

1. FLAC - NA/NA
2. MP3 Stereo 128 - MP3/MP3
3. MP3 Stereo 320 - MP3/MP3
4. MP3 Joint Stereo 128 - MP3/MP3
5. MP3 Joint Stereo 320 - MP3/MP3
6. MP3 Dual Channel 128 - MP3/MP3
7. MP3 Dual Channel 320 - MP3/MP3

Next I did a blind A/B comparison of the FLAC vs each other rip of “Calling Elvis” with the following results:

1. FLAC - NA
2. MP3 Stereo 128 - MP3
3. MP3 Stereo 320 - couldn’t tell
4. MP3 Joint Stereo 128 - MP3
5. MP3 Joint Stereo 320 - MP3
6. MP3 Dual Channel 128 - MP3
7. MP3 Dual Channel 320 - MP3

I then re-compared the FLAC and MP3 Stereo 320 rip listening for differences and could detect some.
--------------------------------------------------------------------------------------
Conclusion

I have way to much time on my hands. (please add corollaries as you wish).

MP3 Dual Channel - Could easily tell the difference with FLAC at both 128 and 320.

Easily identifiable as MP3 w/o any direct comparison or training on what to listen for. Vocals and upper bass/midrange concentrated in left side with high and very low frequencies playing from both sides. 128 bit very harsh and piercing.

MP3 Joint Stereo - Easily detectable as MP3 only when source had accurate instrument placement.

Stereo separation very muddied at 128 somewhat less so at 320. Both 128 and 320 lacked 3-dimensionality if present in the source. Noticeable hissing on high notes with 128 less so with 320. Crosstalk between left and right side delayed causing an echo effect. Distinct sounds like multiple hands clapping together on the FLAC sound like one muffled hand clapping.

MP3 Stereo 128 bit - Difference between this and FLAC hard to detect without knowing what to listen for.

Lacked some dynamic impact and hissing became noticeable once I learned what to listen for.

MP3 Stereo 320 - Nearly indistinguishable from FLAC

I couldn’t tell the differences until I trained myself to hear them and even then it was tough w/o a direct A/B comparison. I probably couldn’t tell these apart except for the most extreme circumstances.

So one Moral is don’t be like me and automatically use the default settings on your MP3 encoder. For Media Monkey Stereo sounded much better than Joint Stereo and Dual Channel while still resulting in the same size files.

Second moral - Don’t teach yourself to hear the difference between FLAC and MP3 if you don’t want to keep noticing it later.

Next week Same bat-time same bat-channel. Media Monkey flac vs EAC flac. Can the trained ear tell the difference ?

Stay tuned...

Trick question?

The only reason I’ve used EAC over MM for FLAC extraction is the error correction built into EAC. MM is by far easier to setup and somewhat easier for basic ripping.

However, there may be an episode where Media Monkey meets AudialsOne. Can you feel the excitement?

Actually it wasn't a trick question. I am curious to know what all that error correction gets you. MM is dead simple and at 2-3 minutes a disc, fast.

Here’s a pretty good explanation of how EAC works.

http://wiki.hydrogenaudio.org/index.php?title=Exact_Audio_Copy

It also uses Accuraterip to verify the accuracy of your rips.

http://www.accuraterip.com/

In a nutshell EAC takes a lot (and I mean a lot) longer to rip than Media Monkey or any other ripper that I know of. Unlike any other ripper I know of when EAC is done you will know if you have a 100% perfect rip of each track or not and it will leap tall buildings to try and get one.

I started using it because the general consensus was that it’s the most accurate CD ripper available. Since I wasn’t in a hurry to rip everything I have (which isn’t much) and I’m ripping not just to listen but to archive/backup my CD collection I wanted to be as sure as possible I was storing accurate rips.

Based on my personal testing if I were copying just to listen I would use Stereo MP3 at 320 bits. But since I’m archiving I chose to use FLAC so I can recreate the original data if needed. This has already paid off since I dropped my Chill out in Paris case the other day and gouged one disk. Since I have a perfect backup I just burned a new copy to replace the damaged one.

Now I just need to find a way to rip my SACDs before I damage them.

All of this talk of ripping made me read this as "Chili out in Paris" at first.

Being dyslexic you made me look to make sure I didn’t really type it that way.

thanks for this great comparison.

Thanks Dean - enjoyed reading through that as well (until I realized I ripped all my songs in 320 Joint Stereo).

I did 3 blind listening tests before making that decision and couldn't pick any of them out, so maybe your ears are a bit more golden than you give yourself credit for!

320 Joint... you sure that wasn't 420 joint that you ripped?

Great post Dean; thanks for that.

I did all my CD ripping into least amount of compression (saves time and storage is cheap) in FLAC with Winamp just to have a complete back up, same as Dean. I have a harder time telling the difference between FLAC and MP3 unless the MP3s are under 128K, but I have never done any direct comparisons as Dean did, I just know any MP3s I have heard at less than 128k start sounding hollow and thin, 128K and up they sound like I am listening to the radio, very acceptable.

I double checked the settings in the manual portion of the Media Monkey ripper and 320 was the highest available. Also in the display they are showing that the bitrate is 320. Are there higher settings available? If so how are they accessed?



Ok maybe I’m developing a better ear but my gut feeling is telling me its something else. When you did your test did you use headphones or speakers? I just retested “Der Kommissar” as it had the most pronounced stereo effects and on my M22s I got the following results with a direct A/B comparison knowing which track was playing .

1. FLAC - NA
2. MP3 Stereo 128 - MP3
3. MP3 Stereo 320 - couldn’t tell
4. MP3 Joint Stereo 128 - MP3
5. MP3 Joint Stereo 320 - couldn’t tell
6. MP3 Dual Channel 128 - MP3
7. MP3 Dual Channel 320 - MP3

The delay that was so pronounced it gave away the MP3 on the headphones wasn’t noticeable on the speakers.

I intentionally used the Sennheisers for my tests as I figured they would be the most discriminating “speaker” I have.

"Are there higher settings available?"

Dean, sorry to be so, er, blunt , but... 420 Joint

The irony is that, aside from attending concerts, I've never touched the stuff.

I didn't mean to send you on a wild goose chase.

Duh, I wasn’t even thinking of that reference, my brain isn’t multi-tasking well right now. My ex roommate from university would be very disappointed in me for not catching that.

Unfortunately I like the stuff just a little to much, but my present job frowns upon it, just 7 years 11 months and 11 days plus the time it takes to walk 25 feet out the gate at the base as that’s where the base security’s jurisdiction ends. Not that I’m planning anything.

New Axiom code phrase. "I'm off to rip some music in 420".

My room mate and I called it “Dave” so the uptight types I’ve always worked for wouldn’t know what we were talking about if I was at work.

Did they look at you funny when you said you couldn't wait to smoke Dave when you got home?

Question; when I rip a CD to FLAC with MediaMonkey, is the only difference between compression level, the speed at which the rip will happen? There is no difference in sound quality?

paul

Exactly. When the player decompress the FLAC file (pretty much like a "zip" program), it will play an exact copy of the original CD.

Didn’t quite phrase it that way. Remember for some time my boss' policy is don‘t ask don‘t tell.

Come to think of it we didn’t have to use it that often since when Dave was available he partied almost every night.


It’s by far the biggest difference but higher compression level will also result in slightly smaller compressed files but most people don’t think it’s enough to waste the extra time. Another issue with using higher compression levels with FLAC is that it requires more computing power to uncompress and play on the fly which causes problems for some portable players. But as Efalardeau said no matter the compression level you can still uncompress all the original data non is discarded.

We had Scott. It started as "Who's got doobs" evolved into "Who's Scott Doobs" evolved into Who's Scott Doobalas"

My friend's mom came down into the basement where we were hanging out after induldging and she said. "Who's Scott Doobalas?" We all said in unison, "We do." and then went into a laughing fit.

Where you in “That 70’s Show?”

We weren't in it. We LIVED it!

How large of a FLAC file collection does anyone have?

According to the Media Monkey I have 934 FLAC files totaling 23.9GB which is about 2/3s of my CD collection which really isn’t all that large.

The Monkey also says I have 13,948 mp3 files and counting (did some culling) totaling 72.6GB and counting all of which are grabbed from the internet.

The current FLAC count is at 3,055, which takes almost 89 GB.