FLAC vs. VBR MP3: should I hear a difference?

[Dave D]

Hey Folks,

Before I knew anything about FLAC, I ripped my CDs to MP3 (VBR, 160kbps min). Since then, I've read on and on how much FLAC is used here and in other forums. So I thought I would give it an audition before spending all that time to re-rip everything.

I loaded dBpowerAmp (faster and easier than EAC w/FLAC) and ripped just one album: The Cars Greatest Hits.

I was just like a little kid at Christmas, all ready to experience the difference that 4.5x the bitrate (and storage) was going to yield. What a letdown. I went back and forth, back and forth, over a bunch of songs (mainly the beginning of the songs to make it easier to compare side by side), and could not hear any differences!

Maybe you can enlighten me:

-- What kinds of music will benefit most from FLAC vs. VBR MP3? Was this album a poor choice?

-- What frequencies benefit the most (low/mid/high)?

-- Honest now: if I played the two side by side, how hard would it be for you to pick the FLAC version consistently?

-- Anything else?

Kind of bummed. - Dave

[pfarrell]

On Mon, 2005-09-19 at 19:26 -0700, Dave D wrote:
> Before I knew anything about FLAC, I ripped my CDs to MP3 (VBR, 160kbps
> min). Since then, I've read on and on how much FLAC is used here and in
> other forums. So I thought I would give it an audition before spending
> all that time to re-rip everything.

Good idea.

> What a letdown. I went back and forth, back and forth, over a bunch of
> songs (mainly the beginning of the songs to make it easier to compare
> side by side), and could not hear any differences!

Well, it depends on a lot of things. Including the quality of the rest
of your system.

> -- What kinds of music will benefit most from FLAC vs. VBR MP3? Was
> this album a poor choice?

I'm not familiar with that particular album, but lots of
rock was recorded with very low fidelity equipment -- early
Beatles and Stones stuff only had to sound good on car AM
radios. The classic guitar amps, AC30s, Marshals, etc.
were huge distortion machines and had very low
frequency bandwidth. This is nothing in the high frequencies
or very low frequencies in electric rock guitars.

I use acoustic or acoustic and vocal music for testing.
A jazz trio, or a jazz vocalist, classical chamber music,
natural bluegrass, etc.

The key thing is that you have to like the music, and it
has to have quality itself.


> -- What frequencies benefit the most (low/mid/high)?

midrange is there all the action is. There is nothing
down low. High frequencies are critical for reverb
and spacious realism.

> -- Honest now: if I played the two side by side, how hard would it be
> for you to pick the FLAC version consistently?

This gets into the serious theological issues of blind testing.
The reality is that blind testing of good audio systems is
very hard.

If you can't tell, then there is no difference.

But if you can't tell and it is easy, maybe you want to do it
so you don't have to rerip all your files later.
It is trivial to generate MP3 or OggVorbis
from the FLAC files, it is impossible to go
the other way.

And some day, you might get a better system and will
be able to tell.

--
Pat
http://www.pfarrell.com/music/slimserver/slimsoftware.html

[dean]

On Sep 19, 2005, at 7:57 PM, Pat Farrell wrote:

> And some day, you might get a better system and will
> be able to tell.

Of course, your ears aren't going to get any better over time. Enjoy
your hearing while you have it!

[Dave D]

Thanks Pat.

I'm listening through my Labtec :) headphones, since they sound better than my old Advent speakers (might get new speakers later this year). Maybe that makes a difference. The amplifier is an Onkyo TX-860, which was pretty good when I bought it in 1989. Probably far out-performs the speakers.

I just ripped a more recent album, with plenty of piano and vocals: Joshua Kadison - Painted Desert Serenade. I went through each song again (beginning only) and I swear I just can't hear any difference.

It is a bit depressing. Feel like I'm missing out; didn't think my hearing was _that_ bad.

[pfarrell]

On Mon, 2005-09-19 at 20:10 -0700, dean blackketter wrote:
> On Sep 19, 2005, at 7:57 PM, Pat Farrell wrote:
> > And some day, you might get a better system and will
> > be able to tell.
> Of course, your ears aren't going to get any better over time. Enjoy
> your hearing while you have it!

I've seen several credible reports that iPods and the like
combined with high efficiency headphones is causing (enabling?)
folks to seriously damage their hearing.

And it is usually permanent. Take care of your hearing. Turn it down.

It is interesting that most (or at least a huge number of) audiophiles
are middle aged men, most of whom have listened to too much loud
rock and have hurt their hearing.

You can learn to listen, and can learn to pick up subtle things
that make the difference between codecs, or amps, or even speakers.



--
Pat
http://www.pfarrell.com/music/slimserver/slimsoftware.html

[pfarrell]

On Mon, 2005-09-19 at 20:17 -0700, Dave D wrote:
> I'm listening through my Labtec :) headphones,

Be warned, headphones have all sorts of non-linearities.

> since they sound better
> than my old Advent speakers (might get new speakers later this year).

Large Advents? I had four of them. Still have two. Altho they were kinda
like my gradfather's ax, in that I replaced the tweeters and the woofers
in each. I bought the first ones in 1971 or so.

Large Advents have a serious problem with linearity. They sound great
when played fairly loud with an amplifier that delivers a lot of watts.
But they do not sound good at low levels. You need to crank them
up to near live sound levels.

And you have to deliver at least 100 real watts per channel to them.
A lot of receivers are rated 70 or so watts, but can't deliver it
continually, and the Advents have to have it constantly.
Advents sound a lot better with 200 or more watts per channel,
which is a bit weird, as you rarely listen to more than about 5 to 10
watts at normal listening levels. But the transient stuff with
Advents are brutal loads.

I retired mine because my wife was tired of looking at them.
She was actually tired of looking at them a decade before I
replaced them.

> It is a bit depressing. Feel like I'm missing out; didn't think my
> hearing was _that_ bad.

Try this, re-encode them at something far lower fidelity, like 120 CBR.
See if you can tell then. Or even worse.

For a lot of music, moderate VBR can sound pretty decent.

What you need to listen to are subtle things like the sense
of acoustic space, the way that the piano sounds decay slowly,
or the snap and sizzle of hit hats. Or brushes on a snare in a
jazz trio.

When it is right, it sounds more real.

--
Pat
http://www.pfarrell.com/music/slimserver/slimsoftware.html

[bernt]

Don't forget that FLAC also is a very good backup of your cd.

[seanadams]

YEMV but as far as I can tell, recordings with more distortion/noise reveal MP3's artifacts most readily. This is assuming of course that you know the CD *very* well, otherwise you can't tell what's from the source vs the compression. I heard this for myself before I knew much about how MP3 works, but it makes even more sense now.

It's analogous to video compression really. On a DVD, when the image is static and simple, it looks very crisp and detailed. However during a big explosion and/or when the camera is panning over a complex scene, that is when you will see the artifacts.

Ignoring for a moment how the mp3 encoding process decides what to keep and what to discard, the file itself is at the end of the day a small number of frequency components representing each tiny unit of time (a frame). A really clean studio recording will have simpler waveforms that can be more accurately reconstructed from a small set of frequency components.

However when you add background noise, audience noise, disortion from tubes and guitar amps etc you get an extremely complex signal. Also drums and cymbals especially can have a wide freuency content and fast transitions which mp3 has trouble with. You generally won't hear the most artifacts in the main, loudest element of the program, because that's where mp3's psychoacoustic model can get the most perceived quality per bit.

The best example I can think of is on one of the DMB live albums (sorry forget which one) during the into to "two step". They're getting into the intro and the audience is clapping in unison. On each clap, with mp3 the clapping sounds like an "envelope" of noise but on the CD it actually sounds like lots of hands coming together. The thing is you'd have to really be familiar with the CD in order to know that you're missing something. But this one example where I can repeatably tell the difference blind (at 160K anyway).

A good experiment would be to make a few encodings starting at a very low bit rate, say 32kbps. What the artifacts sound like and what kind of complexity causes them will then be obvious. From there work up to higher bit rates. Above 128kbps it will be get tougher quickly, especially with VBR, as this allows any complex section to use the max 320kbps when needed.

Also frankly, don't overlook the bit between your ears either. Knowing that you're getting all the data is important too! Sucks to be listening to something that you only have in mp3 and wonddering if it might sound better otherwise.

[Robin Bowes]

Dave D wrote:
>
> -- Anything else?
>

Dave,

Don't overlook the fact that your equipment might not be good enough to
show the difference.

I tried a few external DACs with my SB1 and, when I first tried, I
couldn't hear any difference between the SB1 internal DAC, an Arcam
Delta, and Art DI/O, and a Perpetual Technologies P3-A plus P1-A
correction engine.

I identified that it was my amp that wasn't good enough (a Rotel RA820A)
so, being that way inclined, I got the soldering iron out and made a few
mods - upgraded op amps, replace coupling caps, bypass electrolytics
with film caps, re-routed the signal away from the tone controls, etc.

Once I'd done this, I was easily able to hear differences between DACs.

I see from one of your later posts that you're listening with Labtec
headphones. Well, I don't have any experience of those, but looking at
their website I see they are primarily a computer peripheral
manufacturer and would not expect their headphones to be very good.

Also, your main system - Advent speakers and Onkyo TX-860. Again, I have
no direct experience of those components, but you say yourself that you
believe the Labtech headphones to be better.

Anyway, I hope that's given you food for thought.

R.
--
http://robinbowes.com

If a man speaks in a forest,
and his wife's not there,
is he still wrong?

[Fifer]

Be warned, headphones have all sorts of non-linearities.

I'd argue that in general, pound for pound, decent headphones will provide more linearity and fidelity than speakers. (Obviously I mean pound sterling, although the same might be true for pound weight if your neck muscles are strong enough ... )

In support of much of what has been already said in this thread, mp3 files encoded at a decent bit-rate can sound pretty convincing through average equipment. However, FLAC versions of music with which you are familiar, played through reasonably decent equipment is significantly different (better) to my ear. I convinced myself a wee while back when, by trying to resolve a wireless networking issue by returning SlimServer to default settings, I unknowingly set FLAC files to be sent as MP3. I very quickly noticed that the life and sound-stage had been sucked from my music. It took some time to figure out why though.

FLAC is also perfect as an archiving format too. You can easily transcode from FLAC to any format you choose, but once it's in MP3, it's in MP3 for good unless you want to lose even more detail.

[Robin Bowes]

Fifer wrote:
> I very quickly noticed that the life and sound-stage had been sucked
> from my music.

That describes the problem perfectly. When you lose detail (which is
what happens in mp3 encoding) it isn't always immediately obvious,
especially on lower grade equipment but you often get a sense of the
music being "tiring" to listen to without being able to put a finger on it.

This also applies to the differences between DACs and, to some extent,
differences between cables - even digital interconnects. In the digital
domain, it is jitter that causes most of the problems by messing up the
high-frequency content and losing coherence.

All of the psycho-acoustic cues for "life" and "soundstage" are in the
high-frequency content and if that gets messed with you end up with
"lifeless", "narrow" sound.

R.
--
http://robinbowes.com

If a man speaks in a forest,
and his wife's not there,
is he still wrong?

[Dave D]

Thank you all for the replies while I was sleeping :) Couple of comments:


I've seen several credible reports that iPods and the like combined with high efficiency headphones is causing (enabling?) folks to seriously damage their hearing. And it is usually permanent. Take care of your hearing. Turn it down. It is interesting that most (or at least a huge number of) audiophiles
are middle aged men, most of whom have listened to too much loud
rock and have hurt their hearing.

I have not used headphones much in the past; only recently with the SB2. The sheer convenience of the SB2 has really opened up the music world for me. But now I probably have been listening a bit too loud.


Large Advents have a serious problem with linearity. They sound great when played fairly loud with an amplifier that delivers a lot of watts. But they do not sound good at low levels. You need to crank them up to near live sound levels.

Funny you mention this. They are not large Advents (Prodigy Towers), but I recently re-foamed them and they sound bad now, compared to what I remember before the buzzing(!) There is a whole midrange missing--I can tell this without comparing to anything, so it must be bad. I've been wondering if the foam needs to break in, but I've run them many hours while out of the house and don't notice any difference. And I have also thought that the sound is not nearly as good at low levels, which is where I listen most of the time when not using headphones.

I'm waiting for av123 to offer their new XLS speakers. Then I'll retire these Advents. I was very surprised at the Labtecs. About 8-10 years ago, my brother visited and brought with him several sets. He said they sounded so good that when they went on sale, he couldn't resist picking up more to give away.


Try this, re-encode them at something far lower fidelity, like 120 CBR. See if you can tell then. Or even worse. For a lot of music, moderate VBR can sound pretty decent. What you need to listen to are subtle things like the sense of acoustic space, the way that the piano sounds decay slowly, or the snap and sizzle of hit hats. Or brushes on a snare in a jazz trio.

You and Sean mentioned something similar. I will try this experiment.


However when you add background noise, audience noise, disortion from tubes and guitar amps etc you get an extremely complex signal. Also drums and cymbals especially can have a wide freuency content and fast transitions which mp3 has trouble with. You generally won't hear the most artifacts in the main, loudest element of the program, because that's where mp3's psychoacoustic model can get the most perceived quality per bit. The best example I can think of is on one of the DMB live albums (sorry forget which one) during the into to "two step". They're getting into the intro and the audience is clapping in unison. On each clap, with mp3 the clapping sounds like an "envelope" of noise but on the CD it actually sounds like lots of hands coming together. The thing is you'd have to really be familiar with the CD in order to know that you're missing something. But this one example where I can repeatably tell the difference blind (at 160K anyway).

Ok, I understand. This is an excellent example because it describes the aspects of the MP3 artifacts, and they certainly are things I should be able to hear. Thank you!


Also, your main system - Advent speakers and Onkyo TX-860. Again, I have no direct experience of those components, but you say yourself that you believe the Labtech headphones to be better.

As I mentioned above, I think the Advents are kaputt. The _specs_ on the TX-860 are better than any low-midrange home theater receivers I've been considering recently from Denon and HK. Have not looked at 2-channel amps. I really have a hard time justifying spending money on separate components (2-channel vs. HT), and I'd like to add HT sometime in the next year or two. I'm going to start with new speakers, since that I _can_ justify.

Thank you again for all your thoughtful replies. - Dave

[Robin Bowes]

Dave D wrote:
> As I mentioned above, I think the Advents are kaputt. The _specs_ on
> the TX-860 are better than any low-midrange home theater receivers I've
> been considering recently from Denon and HK. Have not looked at
> 2-channel amps. I really have a hard time justifying spending money on
> separate components (2-channel vs. HT), and I'd like to add HT sometime
> in the next year or two.

One possible alternative is to look at HT decoders, i.e. boxes that do
all the processing but have no amplifier built in. Then you can use good
quality 2-channel amps, or at least use one good quality 2-channel setup
for the main L+R speakers and make do with something cheaper for the
rear and centre speakers.

R.
--
http://robinbowes.com

If a man speaks in a forest,
and his wife's not there,
is he still wrong?

[m1abrams]

I would argue that the biggest benefit of FLAC is not that it sounds better than mp3. In fact for most people with most equipment in most environments flac will not sound better than decently encoded mp3s.

However I rip ALL my music to FLAC. Because I do not want to EVER rip my music twice. You will not suffer generation loss transcoding FLAC to the codec du jour, however with mp3 and other lossy formats you suffer generation loss when transcoding.

Also FLAC is a backup for my CDs, I actually have my CDs packed neatly in a box that I have not opened in years. I once had over 200 CDs stolen (before I ripped them) and I lost alot of great music, some of which not replaceable (small local bands). That will not happen again.

[pfarrell]

On Tue, 2005-09-20 at 03:20 -0700, Fifer wrote:
> > Be warned, headphones have all sorts of non-linearities.
> >
> I'd argue that in general, pound for pound, decent headphones will
> provide more linearity and fidelity than speakers. (Obviously I mean
> pound sterling, although the same might be true for pound weight if
> your neck muscles are strong enough ... )

You bet. once you get into a couple hundred dollars for headphones,
which is still over a hundred pounds sterling, you can get
pretty good sound. Headphones under $20 are not going to deliver any
fidelity.

But one of the first rules for people with home studios is
to not try to do critical mixing and mastering on headphones,
no matter how high quality. My mixing monitors are Mackie 824s,
which cost a bit over a grand, they were the cheapest ones
I could really hear critical differences on.


--
Pat Farrell PRC recording studio
http://www.pfarrell.com/PRC

[Dave2]

Most players don't give you "gapless" playback from mp3 -- at least they didn't six months ago, when I was deciding which way to go.

This is a big problem for those many classical albums where one track is supposed to move into the next without any audible pause or "gap." (See, for example, http://www.pretentiousname.com/mp3players/ and http://forums.slimdevices.com/showthread.php?t=14970&highlight=gapless.)

It's irrelevant for most rock albums, which have the standard 2-second gap between tracks.

Dave2

[cliveb]

Most players don't give you "gapless" playback from mp3 -- at least they didn't six months ago, when I was deciding which way to go.
Just to make this perfectly clear, the SB2 does not play MP3 gapless (unless a firmware update has changed things since I tested it).


This is a big problem for those many classical albums where one track is supposed to move into the next without any audible pause or "gap."

It's irrelevant for most rock albums, which have the standard 2-second gap between tracks.
Ahem. There are literally thousands of rock albums where tracks segue; some very famous ones being: Dark Side of the Moon, Abbey Road, Hounds of Love, The Lamb Lies Down on Broadway ... the list goes on and on.

[Fifer]

It's irrelevant for most rock albums, which have the standard 2-second gap between tracks.
Rock, jazz, spoken word, live albums of any genre, electronica and many other types of recording require gapless playback to be heard as the artist intended. It certainly isn't just limited to classical. That's why I use FLAC, OGG Vorbis and a Rio Karma as my portable player. Incidentally, I'm pretty sure that the Karma firmware does do gapless MP3, although I don't use it.

[radish]

Rock, jazz, spoken word, live albums of any genre, electronica and many other types of recording require gapless playback to be heard as the artist intended. It certainly isn't just limited to classical. That's why I use FLAC, OGG Vorbis and a Rio Karma as my portable player. Incidentally, I'm pretty sure that the Karma firmware does do gapless MP3, although I don't use it.

Off topic alert:

The Karma firmware is touted as playing mp3 gapless via either LAME tags if present, or silence detection otherwise. However, I've never been able to make it work properly, and others have reported similar problems. It's gapless for sure, but just like the SB2's FLAC playback, a chunk is sometimes skipped at the transition. Vorbis, on the other hand, is great.

[pfarrell]

On Tue, 2005-09-20 at 10:51 -0700, cliveb wrote:
> > It's irrelevant for most rock albums, which have the standard 2-second
> > gap between tracks.
> Ahem. There are literally thousands of rock albums where tracks segue;
> some very famous ones being: Dark Side of the Moon, Abbey Road, Hounds
> of Love, The Lamb Lies Down on Broadway ... the list goes on and on.

Huge numbers of my favorite rock albums seque.
Probably the best one is Jackson Browne's The Load-Out sequeing
into Stay, but The Nitty Gritty Dirt Band's seque from Uncle
Charlie and His Dog Teddy into Mr Bogangles is also classic

Nearly all of the "art rock" of the early 70s was based
on "concept albums" that sequed like crazy.

Now, since I'm old, most of my favorite rock is what younger folks
probably scoff at.

--
Pat
http://www.pfarrell.com/music/slimserver/slimsoftware.html

[Fifer]

but The Nitty Gritty Dirt Band's seque from Uncle
Charlie and His Dog Teddy into Mr Bogangles is also classic
Thanks Pat, I haven't listened to 'Uncle Charlie' for ages. "Sing 'The Old Rugged Cross' now, chord with this ..."

[Michaelwagner]

Now, since I'm old, most of my favorite rock is what younger folks probably scoff at.
--
Pat
Yeah. When I was a kid, my parents would say "what is that noise?" when I played the beatles or the stones.

Now my kid plays this rap stuff, and I go to say "what is that noise?" and then I think ... I've turned into my parents.

So I hold my mouth closed (and just think it).

Which reminds me ... I was surprised to find out recently that Queen Latifah (the rapper) can actually sing ...

[street_samurai]

I completely agree with what both Pat and Sean wrote.

So I have very little to add except that MP3-encoders can make a hugh difference in the quality of your Mp3. Obviously settings within those encoders can make a big difference as well.

If you are going the Mp3 route, I'd seriously consider ensuring that you used LAME and a standard -preset when you encoded your tracks. LAME has been very widely tested and the presets have equally been tested many times.

The reason why older MP3s that you find (and somteimes newer ones as well) sound terrible is because they were ripped improperly with poor Mp3 codecs and poorly thoughtout settings.

-Very- few people can tell the difference between a properly encoded Mp3 and a FLAC. You'll need a golden system AND golden ears. As Sean pointed out, this can heavily depend on the source material. MP3 presets are optimized to be a general as possible but some types of music will fall outside that optimal range.

Having said that, I have neither (a golden system nor golden ears) but I use FLAC because a) hard disk space is very cheap right now. b) I know i'll never need to re-rip all my CDs again if improved lossly codecs come out. c) its nice to -know- that you aren't missing out on any of that amazing sound.

ss.

[Mark_H]

-Very- few people can tell the difference between a properly encoded Mp3 and a FLAC. You'll need a golden system AND golden ears. As Sean pointed out, this can heavily depend on the source material. MP3 presets are optimized to be a general as possible but some types of music will fall outside that optimal range.


Just call me Midas! :D

Seriously, with any half decent equipment the difference is audible. During casual listening, or with high background noise, it might not be so apparent but if you adopt the audiophile position; sat still, lights out, concentrating on nothing but the music you will hear the difference...

Mark

[Mitch Harding]

Is that even true between FLAC and, say, 360 CBR LAME mp3?

On 9/30/05, Mark_H <Mark_H.1w6zyn (AT) no-mx (DOT) forums.slimdevices.com> wrote:
>
>
> street_samurai Wrote:
> >
> > -Very- few people can tell the difference between a properly encoded
> > Mp3 and a FLAC. You'll need a golden system AND golden ears. As Sean
> > pointed out, this can heavily depend on the source material. MP3
> > presets are optimized to be a general as possible but some types of
> > music will fall outside that optimal range.
> >
>
> Just call me Midas! :D
>
> Seriously, with any half decent equipment the difference is audible.
> During casual listening, or with high background noise, it might not be
> so apparent but if you adopt the audiophile position; sat still, lights
> out, concentrating on nothing but the music you will hear the
> difference...
>
> Mark
>
>
> --
> Mark_H
>

[pfarrell]

On Fri, 2005-09-30 at 17:23 -0500, Mitch Harding wrote:
> Is that even true between FLAC and, say, 360 CBR LAME mp3?

It depends. How good are your ears? how good is your system?
VBR is nearly always better than CBR.

But why take my word for it? Form your own opinion.
All the Flac software is free. The SB2 handles flac
internally, but even SB1s work with the slimserver.
WinAmp has a flac plugin, I'm sure other stuff does.

So, make your own tests. See if you can tell.
you can tell no difference, then there is no difference.

If you can hear a difference, then decide if it is worth doing.

Now, the downside to Flac is minimal. The files are a little bigger
than your "360 CBR LAME mp3" but disks are nearly free. The
difference in setup/rip time is very low.

And the upside is that you never have to touch your physical
CDs ever again. You can just transcode from Flac to anything
else you want. Its lossless. its perfect.


--
Pat
http://www.pfarrell.com/music/slimserver/slimsoftware.html

[Mark_H]

I have to disagree on the VBR vs CBR. I find VBR easier to spot than CBR rips. And yes FLAC outperforms CBR. I guess at some higher bitrate it might be increasingly difficult to tell, but 320 is not high enough.

As Pat says, test for yourself. I've been listening to music for 25 years and an audiophile for 15 of those and my system could be described as golden, so perhaps I'm just more attuned to how it should sound and can immediately spot when it's not "right".

It would be interesting to do the tests away from my main hi-fi though to see if I could still spot the differences on unfamiliar systems. I can certainly hear the problem on my iPod with 320CBR.

Mark

[Fifer]

-Very- few people can tell the difference between a properly encoded Mp3 and a FLAC. You'll need a golden system AND golden ears.
I have to disagree. Being in my mid-forties, I certainly don't have golden ears (if I ever did) and my system is reasonable but by no means high end. I can usually tell the difference and I'm sure most folks with normal hearing could. Spotting the signature artefacts of compression is I suppose, a wee bit like looking at those 'magic eye' pictures or spotting the ad-break warning marker on the corner of the TV screen. Most folk don't see them, but once spotted you can often pick them out. I'm not knocking high bit rate compression as it most cases, it's very good but the differences can often be heard.