Can Lossy Audio Sound As Good As Lossless?

[peter312]

I've been experimenting with lossy codecs on my laptop. I've been comparing different file formats to see if I can tell the difference in a blind test. I've been comparing:

• WAV Files
• FLAC Files
• AAC Files (using Nero's new encoder)
• MP3 Files (using the latest stable Lame encoder)

I find MP3 files the easiest to spot and some of the lower bit rate AAC files (ie less than 320kbps). It gets a lot harder to spot AAC files encoded up to 465kbps and well, I don't think it should be possible to spot FLAC files.

So here's a thought - if you were handed an audioCD with say 3 differentpieces of musicon it, with each having been originally encodedin say half a dozen of the highest qualitylossless and lossy formats (to make 18 tracks in all), would you be able to spot the "unmolested" tracks?

Obviously the tracks would have to in a random order and the test would have to be done on Hi-Fi equipment - you can work out which ones were MP3s and AAC files with appropriate software on a PC.

Cheers

Peter :dude:

[hifiwigwam]

pheww, TBH I am not sure I would want to try. MP3 and WMA is quite an easy spot as the midband seem to be missing. In fact most stuff sounds as though it has been recorded in a washing machine. As for the others I really don't know,

[meninblack]

We've had a go at this already. I circulated a disc with .wav, .flac and 320 kbps .mp3 versions of the same song to a number of volunteers - 3 tracks of each type, in random order.

About half the volunteers could pick out the .mp3's, and it tended to be those with the most revealing/expensive systems. No-one could tell .flac from .wav. There's a thread on here somewhere...

[peter312]

With MP3, even at higher bitrates (eg 320kbps), I find the music sounds like there's some "presence" missing - a bit like having a photograph of something, rather than the item itself.

I find AAC is better in this regard. At the highest rates the Nero encoder can do,I find it harder to spot them.

I still can't get my head around FLAC files. They seem to be about 60% to 70% of the size of the WAV files and are supposed tosound the same. It's funny how our preconceptions can get in the way!

Compared to the WAV file I have of a particular song, the other versions are smaller as follows (% size of WAV file):

• FLAC 62%
• AAC CBR (465kbps) 33%
• AAC CBR (320kbps) 23%
• AAC VBR (q1) 27%

It's back to preconceptions - if I accept that the FLAC version is identical to the WAV version, the AAC files are still 37% to 53% of the size of the FLAC file. How can it sound the same when in all but one of the examples, more than half of the information has been removed?

Makes you think.

Cheers

Peter

[peter312]

meninblack wrote:

We've had a go at this already. I circulated a disc with .wav, .flac and 320 kbps .mp3 versions of the same song to a number of volunteers - 3 tracks of each type, in random order.

About half the volunteers could pick out the .mp3's, and it tended to be those with the most revealing/expensive systems. No-one could tell .flac from .wav. There's a thread on here somewhere...

Ah, a case of me reinventing the wheel, me thinks! I'm not surprised about the MP3s.

Thanks for the info mib!

[meninblack]

Think of it like this: the lossless codecs - such as flac - only remove null information, ie they replace strings of zeroes with shorter, more compact code which instructs the decoder to re-insert the zeroes. This limits the amount of compression available, but ensures that no "real" musical information is lost - hence the term "lossless."

Lossy codecs, on the other hand, attempt to remove musical information which is "masked" by other, louder sounds. Of course, the extent of this masking has to be calculated, and some algorithms are better at calculating this than others. If you try to apply too much compression, you will start to remove the "one-quarter-audible" sounds, then the "one-half-audible" ones, and so on.

Mp3 is an old codec, and not terribly good at calculating which sounds are actually inaudible, especially at high compression rates. Some of the newer ones - AAC, ogg vorbis - are reckoned to be a lot better, hence their effects are a lot less audible.

[Geordie]

I think I am the only one I know on any of these forums who has tried Musepack.

This is a lossy codec which is far superior in every way to MP3. A completely audibly perfect file is achievable at bitrates of no more than about 230 kbit VBR. Its also faster than MP3 to compress and actually supports gapless.

The downside is not many third party devices support it - although... the squeezebox does

I would challenge ANYONE with ANY standard of kit to be able to reliably tell the difference between those and WAVs. I am certain they couldnt.

For those of us who dont wish to throw lots of money at, or, IMO, waste hard disk space with lossless formats then these are a good alternative.

You can fit as much music onto a 120 Gb hard drive with musepack as you could with FLAC and a 300 - 400 Gb hard drive with no difference in quality.

However, i DO use FLAC for archiving because why not use something that is in fact an exact copy of the original when uncompressed.

The way I always think of good lossy rips is they are like comparing raw bitmaps with the highest quality JPEG files - essentially no difference unless you are going to analyse it under a magnifying glass. I think a lot of audiophiles have very big egos when it comes to lossy files and some of the claims are outrageous.

[Tons of Fun]

I can safetly say that I cannot tell the difference between WAV, 256kb MP3 and VBR WMA files on my iRiver with Shure E4's.............



..................on the jubilee line at rush hour. Therin lies the point of lossy codecs.

[peter312]

peter312 wrote:

With MP3, even at higher bitrates (eg 320kbps), I find the music sounds like there's some "presence" missing - a bit like having a photograph of something, rather than the item itself.

I find AAC is better in this regard. At the highest rates the Nero encoder can do,I find it harder to spot them.

I still can't get my head around FLAC files. They seem to be about 60% to 70% of the size of the WAV files and are supposed tosound the same. It's funny how our preconceptions can get in the way!

Compared to the WAV file I have of a particular song, the other versions are smaller as follows (% size of WAV file):

• FLAC 62%
• AAC CBR (465kbps) 33%
• AAC CBR (320kbps) 23%
• AAC VBR (q1) 27%

It's back to preconceptions - if I accept that the FLAC version is identical to the WAV version, the AAC files are still 37% to 53% of the size of the FLAC file. How can it sound the same when in all but one of the examples, more than half of the information has been removed?

Makes you think.

Cheers

Peter

I've done some more blind testing and come up with some interesting results:

Lame Encoded MP3: Even at 320kbps, I can spot an MP3. I can't, however, differentiate between 320kbps CBR and v0 VBR. However, I prefer the sound of the Lame MP3's to the iTunes AAC files (see below). The sound lacks "air" or "presence".

iTunes encoded MP3: For me, a big no no!

iTunes encoded AAC: Whilst I find this has a better treble extention than Lame MP3, I find it sounds "mushy" and less clean.

Nero encoded AAC: Once we get up to q1 VBR or 465kbps CBR, I find it harder to spot. However, some source material does show it up. I find it better than Lame encoded MP3 where the bitrate is much higher.

Windows Media Player encoded MP3: Another no no.

Windows Media Player encoded WMA: Better than WMP MP3 but still I still prefer Lame MP3.

I'm very impressed with FLAC, now I've got EAC and Foobars working properly. I encode FLAC files with EAC and can convert those that I want to into Lame MP3 with Foobars (for portable use). I've even got my Windows Media Player recognising and playing FLAC, Ogg and AAC files and listing them in the library!

Not too sure why I started this off, but I'm happy with the result!

Cheers

Peter

[julian2002]

to be honest i only care about compression on my portable devices (phone and archos) neither of which are hi-fi so whatever flavour sounds decent at low memory - sweetspot for me is 128bit wma but ymmv. for home use i use flac - storage is cheap (relatively) and my collection is pretty static now anyway.
cheers


julian.

[hifikrazy]

182-Geordie wrote:

I think I am the only one I know on any of these forums who has tried Musepack..

I have tried it. (and pretty much every other codec) I prefer it to Ogg Vorbis. But I dont use lossy for anything other than the occasional mp3 on to my W800i for travel.

[Vincent]

I use a benchmark dac1 and I can tell the different between mp3 and flac, but its ever so slight. It really depends what kind of music you are listening to, if its pop and heavy metal, the jitter may even make it sound better

[julian2002]

wow, necrothread....

just a note on how flac works (iirc).
first the encoder generates a 'best guess' compressed version of your file - this is essentially what mp3, ogg, aac, et al do. what it then does is create a difference table between it's guesses and the original data, usually it's guesses are pretty accurate so the differences are usually very small and efficient to compress using simple methods like bit reduction or huffmann which work well for structured data but poorly for very random files (such as sound). This way it can knock the size of a wav file down by 30-50% without losing any of the original information.
hope that helps some get their head around what flac is doing and why, if you think you can tell the difference between flac and wav you have subjectivitis....

[Chumpy]

Some lossy audio files sound better being because of having less nasty-sounding extra data/some nice sounds seem better being lossless.

[drummerjohn]

I can't be bothered to compress mine. As a music CD is normally around 600MB uncompressed that equates to around 7p in Hard Drive space. You want so save yourself 3 or 4 pence for a music CD that you may have paid £8 upwards for then knock yourself out.

Compression is for portability and a time when disk space was a premium.

[gjm]

^ What he said!

I'm normally able to tell the difference between 320kbps mp3 and flac, even when tracks are played in isolation. Back-to-back, the mp3s sound flat and dull.

[julian2002]

I can't be bothered to compress mine. As a music CD is normally around 600MB uncompressed that equates to around 7p in Hard Drive space. You want so save yourself 3 or 4 pence for a music CD that you may have paid £8 upwards for then knock yourself out.

Compression is for portability and a time when disk space was a premium.

i'd agree with you but for 2 reasons..

1) i've been into server based hi-fi for 5+ years now and drive space wasn't always as cheap as it is now.

2) flac's are easier to tag than raw data. which makes database searching and management far easier.

at the end of the day though it's up to you but more music is always better than less imo.

[drummerjohn]

i'd agree with you but for 2 reasons..

1) i've been into server based hi-fi for 5+ years now and drive space wasn't always as cheap as it is now.

Ummm... I thought that's what I said here...

"Compression is for portability and a time when disk space was a premium."

[browellm]

You know, I've never understood this mentality of "it's so cheap, I store it uncompressed". It's bandied around like some badge of honour by its advocates, and yet it serves no useful purpose. Unless you have some kind of impossibly shit replay system where transcoding the alac/flac to pcm would actually affect the SQ, what possible motive could you have?

I currently have 1.2TB of music on my home server, so with just one backup copy, that's 2.4TB. Would I rather store it uncompressed and pay for 5TB of disk space, and the associated time spent backing up? Do you know how long the initial backup on 2.5TB actually takes?

[Matt J]

1.2tb?! jesus, I have the massive amount of 77gb although quite a bit is mp3.