VBR new versus VBR old

2009-01-19 01:22:46

Hi, sorry if this thread is a bit basic for you guys. I'm not quite an expert but have a reasonably good understanding of LAME and VBR.

I got my first MP3 player over 4 years ago (Winter 2004) and before I started ripping my CD collection I did a bit of research into what software was best to use.

This led me to EAC and consequently LAME. I can't remember exactly what versions I was using but I know EAC was using VBR old by default with alt preset standard as the command line option.

Anyway, I never really looked into updating to newer versions of EAC or LAME because I was always happy enough with what I had.

Just recently I decided to get the latest version of EAC and also discovered that LAME as a new VBR algorithm by default now.

I've done a bit of reading up on it and from what I've read VBR new V2 is roughly the same as 'alt preset standard' correct?

I decided to do a quick comparison of an album rip using different settings. I noticed the total size of the album's 11 tracks were as follows:

My 2004 rip using VBR-old at alt preset standard:  105MB

Then using VBR new with LAME 3.98.2:

V2: 83.2MB
V1: 94.6MB
V0: 109MB

This surprised me a bit as I would have expected the V2 VBR new rip to be similar in size to my old 2004 rip.

And the V0 rip is only 4MB larger!

Can anyone explain this for me? Is the new algorithm just more efficient or something?

Sorry is this is a stupid thread, I've searched for recent comparisons between vbr-old and vbr-new but can't find any good ones.

Thanks

VBR new versus VBR old

Reply #1 – 2009-01-19 04:28:46

You are correct, -V 2 is basically the new equivalent to --alt-preset standard.  The reason why the bitrate have decreased is that Lame has become more efficient over the years.  This means that Lame can produce the same results as before but use less bits doing it.  I take it that you switched over from Lame 3.90.3?  If so, the bitrate decrease is rather normal.  From my experience, Lame 3.90.3 always used excessively high bitrates.  Lame 3.97 did the same thing as it would produce files using lower bitrates than 3.90.3.

Now, if you want to check things out, you can always conduct your own blind ABX test to see if -V 2 is right for you.  I have seen many members on this board actually step down to -V 3 which produces file at around 175kbps VBR.  They claim that Lame is transparent (ie CD quality) at those levels and I would have to agree.

VBR new versus VBR old

Reply #2 – 2009-01-19 05:29:58

I, also, encourage the ABX test to compare --vbr-new to --vbr-old.

As I prefer lossless over lossy formats I am interested in what to use for distributing recordings online. I usually went with whatever seemed safest (like --alt-preset standard) but eventually stuck with 128 CBR since nobody noticed problems with the distros (and I listen to the originals).

What I did notice was using VBR over different versions the huge encoding speed increase of --vbr-new. It made me wonder if it was sacrificing some quality, but so far there is no evidence that it is inferior in any way.

VBR new versus VBR old

Reply #3 – 2009-01-19 08:57:53

It's also worth noting that the presets in older lame versions (like 3.90.3) had a minimum bitrate of 128, whereas in current versions V2 can drop below 128 as needed - down to 32kbps in instances of digital silence.  This can accumulate into a decent reduction in filesize over the length of an album.

VBR new versus VBR old

Reply #4 – 2009-01-19 11:57:58

True, but with digital silence the 3.90.3 the minimum bitrate was 32kbpbs  (sorry to split hairs here)

VBR new versus VBR old

Reply #5 – 2009-01-19 14:09:12

Thanks for all you help and info guys.

Im not sure what version of LAME I was previously using. It was on an old PC and I don't have it anymore. I first installed and used EAC and LAME in Oct/Nov 2004 and I'd say I just downloaded the latest version of LAME which would have probably been 3.96 or 3.96.1 at the time based on the release log ( http://lame.cvs.sourceforge.net/*checkout*...ml/history.html ).

I haven't done an ABX test but the brief comparison I did between my old rips and the new V2, V1 and V0 rips didn’t yield any noticeable differences for me. So V2 is probably adequate.