Hey Guys,

It's my first post. Probably a dumb post but after looking for the answers to my questions for about an hour I decided to register and post...

I discovered this guide: Chris Myden's MP3 Guide. He suggests using EAC V0.9 beta 4 and LAME 3.90.3. He recommended using these versions even though newer versions were available. That's understandable... "newer" doesn't always mean "better."

However, I can't tell when his guide was last updated. Are EAC V0.9 beta 4 and LAME 3.90.3 still considered a safe bet.

My goal is to get a small collection of CDs (about 300) ripped to MP3 format. I just want the best quality that mp3 format has to offer. I understand that FLAC would provide better quality and is lossless, but it takes up a lot of space. MP3s ripped to CBR 320 or VBR near 320 is perfectly fine as far as size goes. Again, just hoping for the best quality MP3s I can make.

Thanks for any advice you have to offer. If I missed a recent thread that discusses (in newb terms) the pros and cons of EAC V0.9 beta 4 and LAME 3.90.3 vs. the more recent versions of EAC and LAME please post the url.

Thanks!

Hi. Welcome to HA.

Chris Myden's MP3 guides are very old and outdated. It is probably futile to discuss about that. I discovered his site and THEN came here.

I recommend that you read the articles in the wiki. If you have any questions after that, please feel free to post here.

I also suggest that you conduct a blind ABX test to determine if you really need 320kbps/-V 0 encoding. Bitrates that high are generally regarded as being overkill in that you can encode using lower bitrates while still retaining perceptual quality. Remember that having the storage space is not really a good reason for using high bitrate encoding especially if your ears don't need it.

Edit: spelling

KC02's advice is excellent: the topic you've raised is fodder for heated discussion in these parts - but [only you] can answer the question of whether you can or cannot hear a difference between a file ripped to, say 320Kb/s vs. CD.

Here is how to settle it for yourself:

Rip a track from your favorite CD, one you know well, and save it as a lossless .wav file.

Rip a series of tracks from the same CD and save them as a 320 / 256 / 192 / 160 / 128Kb/s MP3s and then convert each to .wav files afterwards with a freeware application such as www.audacity.sourceforge.net . Use Audacity’s ‘File / Open’ - followed by - ‘File / Export As .Wav’ menus to accomplish the task. I could be wrong, but I believe that most, if not all, ABX applications require that both files being compared must be of the same file format (.wav in this case).

Now download a free ABX application such as WinABX http://www.kikeg.arrakis.es/winabx and load the .wav files into the application, starting with 320Kb/s [converted to .wav] vs. raw, uncompressed .wav, and listen away. If you have a PC with a really decent sound card and either a killer set of headphones or a line out to a DAC / amplification setup of some sort - that would be great.

Keep comparing lower bitrate [converted to .wav], files to the raw uncompressed .wav file until an audible difference is detectable to your ears. Listen to, among other things, the shimmer of high-hats and the sharpness of drum-cracks. When all is said and done, and you’ve selected the bitrate that you feel is ‘transparent to your ears - select the next bitrate up from that as your future-default ripping bitrate – a small data-storage price to pay for “peace of mind”, and then rip away at your CD collection, comfortable with the knowledge that you cannot discern any degradation of material.

I suspect that you’ll be blown away by how low of a bitrate you’ll feel is transparent! It’s bloody stunning how good the MP3 algorithm really is…

With regards to LAME and EAC; just use the latest 'stable release' versions, not the bleeding-edge Alfa releases:

In the case of LAME its here (3.97 - Sept 2006)
http://lame.sourceforge.net/index.php

And EAC (PB 0.99PB4 - January 2008) is here
http://www.exactaudiocopy.de/en/index.php/resources/download

The version of EAC is debatable, there are also three older versions you might want to check out – I’m sure that some here have far more knowledge with regards to that topic than myself.

Andrew D.
www.cdnav.com

ABXing MP3s directly is very easy with Foobar2000's ABX plugin - no need to convert to WAV or anything else.

Thanks to all of you for the responses!

Maybe one last newb question before I go hit the wiki...

Maybe I don't need 320Kb/s. Tell me what you think. I currently have a pair of Sennheiser HD595 headphones with a custom cord and a decent headphone amp. While I don't pretend to have a "golden ear", I am picky when it comes to sound quality. And, if I had a friend over who claimed to have a golden ear, I would like to think that I'm pumping out some good stuff. Also, my reasoning for wanting some overkill is for future proofing for MY particular situation. I am currently a graduate student and semi-broke . However, when I'm finished with school I will definitely be getting some more audiophile grade equipment in my future .

While ABX testing sounds like a good idea now, I hope to not have to rip again in the near future to discover that newer equipment reveals more flaws from compression. I'm sure I'll do some ABX testing, but I hope to not get too wrapped up in that and just trust some of the experts here.

I suppose the only way to even come close to archiving for future-proofing as I describe would be to use FLAC. However, based on the comments here such as: "Bitrates that high are generally regarded as being overkill" and "I suspect that you’ll be blown away by how low of a bitrate you’ll feel is transparent! It’s bloody stunning how good the MP3 algorithm really is…" I suppose FLAC is waaaay overkill.

I'm guessing I'm missing something here.

EDIT: Eli just posted... Is that what most of you guys do... Rip to FLAC for archiving and then encode to MP3 as needed?

@ digital - a lot of your info was just what I was looking for. Thanks!

Time to check out the wiki. THANKS!

You have several options if you want to avoid the very large filesize you get with a lossless codec while accepting pretty large file size.
It's all ideology though and if you ask 5 people you probably get 10 answers with all of them being correct in the one or other sense.

IMO your first choice qualitywise is to use a lossy variant of a lossless codec. Use for instance lossyWAV + FLAC or wavPack lossy. At a bitrate of ~500 kbps you get an extreme quality with a high safety margin. At a bitrate of ~400 kbps you get a quality which usually is overkill too but there may be rare samples where you get at the edge of transparency. At a bitrate of ~300 kbps things usually still are transparent but there will be samples with slight deviations from the original. These deviations usually sound like a tiny amount of added hiss. The choice is yours, with a bitrate around 400 kbps probably being the sweetspot even for perfectionists.
Compare this against the ~900 kbps usually needed by lossless codecs when encoding pop/rock music.
The theoretical advantage of this approach against codecs like mp3 is that you have a simple clean signal path - the technical representation of the music signal doesn't change whereas with mp3 there's a formation of frequency subbands and their transformation from the time domain representation into the frequency domain - and back when decoding.

Nonetheless there's great quality with mp3, vorbis, aac, mpc.
The main advantage of mp3 is its universal usability. Other than that qualitywise you're a little bit better off with the other codecs (in terms of quality/kbps and robust quailty even in extreme musical situations).
The universal usability is a heavy weight though, so it's a good idea to use mp3.
As was said with mp3 you usually get transparent results at a bitrate of roughly 200 kbps (or a bit below - it's a fluid borderline and it's a useless discussion whether Lame -V3 or -V2 or -V1 is the right thing to use). But if you want a security margin for situations the encoder has trouble with, and you don't have to care too much about filesize, it's the right decision IMO to go higher than 200 kbps. With 320 kbps you get the best security margin, but most of the time this is overkill, and the safety margin you get in practice isn't much higher than when you go to something like ~260 kbps on average.
Another question is what encoder to use, and whether you should use CBR, VBR, or ABR in the case of Lame.
At a bitrate like 260 kbps it's an open question whether there has been significant improvement from Lame 3.90.3 until now. My very personal opinion is that at 260 kbps there's no improvement when going up to 3.97 final. Instead there's a certain regression for some musical scenarios which are relevant to me. With current 3.98 developent (we're in a very advanced beta stage) things have changed however and I welcome the 3.98 development very much. As for VBR/CBR/ABR the only thing that counts is quality, and there's no intrinsic advantage for instance towards VBR (many people think like that). In fact even CBR does use a variable audio data bitrate. With 3.90.3 up to 3.97 when it's upto a bitrate like 260 kbps I'd definitely not use VBR as there are samples (relevant to me) that are seriously worse than the corresponding CBR or ABR results, due to an amplification of weaknesses in the psy model. With Lame 3.90 up to 3.97 when using very high bitrate ABR is the best way to go IMO. I once used Lame 3.90.3 a lot and was happy with ABR @ 270 kbps.
With 3.98 things have changed. The decisive weaknesses of VBR have gone, so it's an open question whether ABR is still to be preferred over VBR. In the case of VBR I suggest you use -V0 with your quality demand in mind. -V0 is overkill most of the time, but in difficult situations it's welcome for a perfectionist.
You can also consider using a FhG mp3 encoder. Lame has the better reputation here on HA, and because it's a personal enthusiatic development of nice people like Robert Hegemann and Gabriel Bouvigne (hope 'Bouvigne' is correct), it's a very sympathic thing especially as there's still vital development. But looking at mere quality using FhG with, say, CBR at 256 kbps, is a good thing as well. You have access to an FhG encoder when using for instance Windows Media Player, or if you use dbpowerAmp, or the free command line (surround) encoder available at all4mp3.
It's not a bad idea playing around with these alternatives, but if you want to avoid that my advice is to use Lame 3.90.3 ABR 270, if only a final version is what you want to use, or if you don't care too much about that, use Lame 3.98b8 ABR 270 or -V0. There's a thread on Lame3.98b8 here (and there are threads on earlier beta versions) so you may want to look there.

Hmmm, my post seems to have disappeared. Yes, alot of people here archive to FLAC and then transcode to a lossy codec of choice. In a way, you are right, FLAC is overkill. However, if you ever decide that you want a new/different codec you can always go back to your FLAC archive and transcode to the new codec, which will be as good as using the original CD as a source.

Again, I suggest dBpoweramp, FLAC archive, and MP3 at a lower bitrate, but you will have to listen and decide for yourself.

Eli -

I'd like to give dBpoweramp a look. I suppose the "reference" version is optimal. Is there a guide or owners manual that describes in detail how to set-up and ripp to FLAC and LAME MP3?

Also, is R13 planned for release in October? Sounds like there has been talk about R13 for about a year. If I buy R12 now I suppose I'd have to buy R13 when it is released since R13 would be considered a major release.

Thanks

Im not associated with dBpoweramp and don't have any inside knowledge on a release date. Spoon has said June-August-ish, though subject to change. There will be an upgrade fee, but less then the full purchase price of Reference. I personally use and suggest Reference and feel its well worth the money. You can buy R12 now and use the R13 alpha until release. Its very stable and has PerfectMeta. You can use the multicoder DSP to encode to FLAC and MP3 at the same time, though I like to encode to flac with track-gain (replay gain per track) and transcode to MP3 normalized with the track gain values. This does alter the MP3 audio, but you still have the lossless FLAC as an archive.

Thanks... Is there a guide? I've been to the dBpoweramp forum. It must be hiding from me

I have the same size CD collection as you and had the same goals: transparent quality encodes of everything. So I will comment.

First thing I want to point out. I have ripped my entire collection three times now. You do not want to have to do this. Trust me. Given the extreme annoyance and waste of time, even when you have it down to a science and multitask, if you have the storage space, archive to lossless. Then you can convert to whatever format you want and you are set for the rest of your life.

In the past, I was satisfied with 256k CBR LAME. This was guaranteed to work on everything and sounded transparent to me, and widely claimed to be as such. However, it wasn't lossless and still took up a significant amount of space, especially on my iPod, so I looked for a better solution.

After getting a MacBook, the decision became obvious and easy. I ripped everything to Apple Lossless. Since time had gone by and MP3 was no longer "required" (due to AAC support in my car head unit, iPod and computer), I switched to 256k AAC. After again getting annoyed by the large amount of space used on my iPod and MacBook, I tested and ultimately switched to 128k AAC, which is where I stand now.

In the future, when my next MacBook has a 320GB hard drive and my next iPod has 120GB flash memory, I may just use the lossless files. But until then, I am prepared for any situation or desire.

As you're honest enough to profess to being a skint student, why don't you just download Foobar2000 for nothing and use that to archive all of your CDs to lossless FLAC. If you set Foobar2000 up correctly, it should auto-tag the tracks for you as it goes. If you have a reasonably fast PC, this shouldn't take you more than a couple of minutes per CD to do. You now have a lossless archive to experiment with as far as conversion to MP3 goes and never need to touch the CDs again.

The version of the LAME MP3 encoder provided with Foobar2000 when used in VBR mode may well be transparent to you all the way from -V3 up to -V0 to be honest. You can only be sure by ABX-ing as suggested above. -V3 led to a typical average bitrate of around 165Kbps when I encoded my 150 CD collection with it last year. Most of the tracks still peaked at 320Kbps as and when the encoder deemed necessary. That's the beauty of VBR encoding.

You may want to consider getting Foobar2000 to add ReplayGain parameters to the files at one stage or the other. This doesn't affect the stored music content in any way, it just stores a calculated gain offset figure for a ReplayGain-aware player to use (optionally) to ensure that all of your encoded CDs plays back at uniform apparent loudness regardless of the level of the original recording. It also helps to reduce the occasional clipping problems that can occur when playing back lossy encoded material sourced from "hot" CD recordings.

Total cost to you, zero.

Cheers, Slipstreem.

If you have continued reading further within Chris Myden's site (or even the ubernet site), you'd stumble upon an FAQ regarding LAME 3.90.3 and 3.97. IIRC, it read that it doesn't matter whether 3.90.3 or 3.97 you'd use (but not those in between).

3.90.3 APS has it's merits especially with regards to achieving transparency most of the time (but it's slower and bloats the bitrate compared to 3.97 -V2).

Personally, I'm still using 3.97 -V2 via Audiograbber (however ancient it may be).

Regarding your question, LAME 3.90.3 is still safe... while 3.97 is safer and more efficient (speed- and bitrate-wise)

Cheers!

Rio

EDIT: Regarding the 1st paragraph of my post, I might add that Chris Myden also suggests using 3.97, but their so-called AQS script and the ubernet people still uses 3.90.3. The people there also regard HA as the authority in audio, so you can't go wrong here in HA. Welcome!

Beginning with 3.97, LAME has some serious problems with CBR. I already pointed out here
http://www.hydrogenaudio.org/forums/index....st&p=537861
Unfortunately, some bastards removed my entire hosted space without notice so I lost the referenced samples. I don't really care about lame development so I didn't keep the files on my drive.
But if you read the responses, I guess you will get a better picture. The reported problem exists with 3.97 final version as well.
I even reported the problem with 3.97 in its early beta phase (years ago) but I got ignored, so this error propagates (and will probably propagate) on. I wonder if any of the developers care enough to correct the issue.
The last "CBR-safe" version I know of is 3.96.1.
But if you have a mediocre hearing (like majority of people in that topic demonstrated), you don't even need to care...

@Martel

I had followed that thread re: 3.97 CBR. Personally, I use FhG ACM (from WMP 11) for my CBR needs.

I must admit, comparing forced 128 CBR between the latest FhG ACM and LAME 3.97, FhG wins in my ears.

I read his entire guide. It just didn't mention any of the most recent versions of LAME (ie 3.98b8) which I realize isn't an "official" release, but I thought the members here might still consider 3.98bx superior. While I'd still like to get more opinions regarding 3.98bx, it doesn't matter as much to me as it did when I first made my post... I'm convinced, now, that I should initially rip to FLAC and then I can do some trial and error with the different versions of LAME.


Does urbernet still exist? If so, it must be top-secret. I can't find it. I assumed it was down. Maybe when you say "...the Urbernet people still..." you are referring to alumni (for lack of a better term) members.

After spending several days with EAC, I decided to trial dBpoweramp R13. I'm not going to rant like some people and complain about EAC (like I've heard on some other forums). EAC is fine but I will say that dB R13 is more of a pleasure to use.

Anyway, thanks a lot for the comments, guys. I'm learning alot.

halb27, it is always a pleasure to read your posts and I couldn't agree more to almost everything you write. One should note though that many files you want to encode to mp3 have NOT been obtained by ripping from CD (or vinyl). In that case (for example if the source has been an FM broadcast) you have a very good chance that your encode is actually a re-encode of the MPEG file the radio station used for broadcasting. In that case your chances to get transparency are NOT high (maybe at 320k or -V0). This scenario is a real-world one but usually doesn't play any role here on HA or for lame development...



Absolutely true. Lame 3.98b6/8 is much better than 3.97, especially for encoding VBR with --vbr-new... But for a comparison of 3.98b8 with 3.90.3 we need to compare CBR (and maybe ABR) modes as that are the modes where 3.90.3 was shining. And here it is where I think that 3.90.3 might be very competitive. For recent versions (even for 3.98b8) of lame I got the impression that developers and listeners/testers neglected the high-bitrate scenarios while optimizing medium bitrates like 128kbps a lot (at 128kbps I think that newer lame versions have an edge). For newer lame versions some instruuments do not sound good and especially "metallic noise" doesn't sound metallic (drums, etc.) even at 320kbps while the original WAV files present the full "metallicity" in full glory. I am not sure about lame 3.90.3 at that point (I never compared) but this older version MIGHT be better in this higher bitrate case...



Yes. From my own experience I have to say that I would not use the FhG encoder at low or medium bitrates as lame is better there but at high bitrates the FhG encoder is a good alternative. See the sentences about "metallicity" above. The FhG-encoder at high bitrates (-br 256 / -br 320 / -br 0 -m 1) does a good job here. The FhG-encoder can also encode in VBR and it is screamingly fast for CBR and VBR (I encoded 10h of audio in just 30 minutes on my single-core 2GHz system with the FhG-encoder from all4mp3.com just very recently).

In the case of an FM btoadcast, whether or not it had previously been an mp3 is the least of the problem. FM broadcasting applies low-pass filtering, adds significant noise and distortion, and quite possibly adds dynamic range compression. After all of that I doubt that you could tell the difference between reencoding and not reencoding no matter what bitrate you encode to.

What application are you using to rip and encode with the FhG-encoder? Is there a guide for someone fairly new to this stuff (like me) to get set-up/started?

Sure special scenarios need special care, and it's true: I always have encodings from a CD in mind.

This may or may not be considered as good advice by those more experienced than myself with lossy encoding, but I generally stick with LAME in VBR for 'special case' source material such as FM radio and just choose a -V value that gives a lowpass filter to approximately match the bandwidth of the source material. There's no audio content broadcast past around 15 to 16kHz anyway, so it seems pointless wasting bits on encoding noise.

I'm not sure what the quality is like with FM broadcast material elsewhere in the world, but it's already so hideously distorted and mangled here in the UK by the time it gets to me that VBR at -V5 doesn't seem to lose me anything that hasn't been destroyed at the transmitting end already.

Cheers, Slipstreem.

For the encoding step from WAV --> MP3 I always prefer the commandline (more fine-grained control - I do not trust GUI's doing it for me with maybe some "hidden optimizations" silently applied by the shell...), for ripping to WAV you can use lots of programs... good for ripping from CD would be EAC ... for example

as for the guide - no idea ATM - but these programs are often self-explaining....

I'm sorry to getting you confused there. Presently I have no idea if ubernet still exists. I merely stated the FAQ.

Anyway, you're getting nearer to your objective. That's good news.