transparency with NeroAACenc

[audyovydeo]

I'll soon be throwing my birthday party and I've decided to once & for all encode my entire CD collection to m4a audio.

Obviously this is something I'll be doing only once, so I want to know what settings are going to be near-transparent.

Starting from a "reference" of mp3 cbr at 128kbps, I set a hypothetical target cbr of 256kbps for neroaac, then varied the -q settings until I reached a similar file size.

However, I'm not convinced that this preserves the full sound quality, especially the deep bass part.

Any suggestions are welcome

thanks
audyovydeo

[ramicio]

I ripped my whole collection to 192 kbps AAC once. I was disappointed. So I just went to FLAC. Bass is the easiest to replicate, it's the higher frequencies that suffer.

[SeeMoreDigital]

Starting from a "reference" of mp3 cbr at 128kbps, I set a hypothetical target cbr of 256kbps for neroaac, then varied the -q settings until I reached a similar file size. In my opinion, 128Kbps was always too low for generating MP3 audio streams (regardless of the MP3 encoder). Personally I always used 320Kbps!

I now have all my audio CD's backed-up to AAC at either 256Kbps or 320Kbps. And to FLAC :)

[TinTime]

Obviously this is something I'll be doing only once...

Unless space is at an absolute premium then I vote for FLAC too, especially in light of the above statement. Once you've done that it's easy enough to encode to any other format for whatever reason, and you don't have to while away hours trying to ABX different -q settings to see what's transparent for you.

[Blue_MiSfit]

When you say you're not convinced that this preserves the full sound quality - do you mean you're just generally worried, or do you mean you've noticed artifacts?

I'd be quite shocked if anyone could tell a substantial difference between ~256kbps stereo AAC and uncompressed, for most sources, on standard / decent audio gear anyway. If you've got an insane audiophile rig then all bets are off, and you're used to the placebo effect anyway ;)

If you don't mind settling for FLAC level compression (and having to transcode every time you want to toss something on a portable device / flash drive for the car, whatever), then just go with that, and you'll never have to think about it again.

Otherwise, pick a Q value and move on.

Don't worry about bass. It's very easy to keep. Heck, last night I did a quick transcode of a Battlestar Galactica episode to watch on my iPhone while I was doing laundry. I did the audio at 96kbps CBR, and was surprised by how good this bass heavy show sounded in my car.

Derek

[audyovydeo]

Thanks to all for your inputs.

... and you don't have to while away hours trying to ABX different -q settings ...

Alas !!! I recognise a fellow sufferer ...
and we've all done that for x264, so long ago, haven't we ;-)


... 128Kbps was always too low for generating MP3 audio streams ...

I absolutely agree, but it was *some* songs to play back on mp3player while on public transport : quality was never the point.
(I used the venerable BladeEnc player upto now).


1. do you mean you're just generally worried, or do you mean you've noticed artifacts?
2. you're used to the placebo effect anyway
3. you'll never have to think about it again.
4. Otherwise, pick a Q value and move on.


- both : it depends on sources, and I haven't done extensive *listening* tests. But if a song sounds bad on my thinkpad's earphones, I shiver at the thought of playing it back on my rig & have people dance to it.
2. no, I'm not a quality fanatic. Just want to do a tedious job once.
3. ditto
4. I think I'll settle for Q 0.75


I've never tried FLAC. Anyone got a cmdline for windows encoder ?

thanks again
audyovydeo

[audyovydeo]

Otherwise, pick a Q value and move on.



quick rant update :

let's say I pick Q 1. I end up with :

44,461,034 Cornershop_10_Spectral_Mornings_Q1lc.m4a
23,153,111 Cornershop_10_Spectral_Mornings_Q1he.m4a
12,107,466 Cornershop_10_Spectral_Mornings_Q1hev2.m4a

meaning I get nearly 1/4 the file size just by forcing the profile.
Notice neroaac defaults to LC for Q 1.
The natural question is : where's the catch. What quality loss can there be between Q1 @ LC and Q1 @ HEv2 ?

and this is without even listening to the files : if there is such a big difference, what has been lost ?

ideas welcome.

cheers
a/v

[GodofaGap]

You mean you don't know what HE stands for? HE does SBR (half the sampling rate and tries to reconstruct it afterwards) HEv2 does parametric stereo on top of that. You don't want to use these at high bitrates: these are tricks to make artifacts less audible at low bitrates (e.g. ~64 kbps).

With no offense you seem to be rather clueless here and you should do proper ABX testing, or use the encoders defaults (q 0.5) which is generally considered safe, or go lossless.

[Gser]

lossless or q 0.7.

[TinTime]

I've never tried FLAC. Anyone got a cmdline for windows encoder ?

A simple command line for highest compression would be:
flac.exe -V --best input.wav
This would produce a file called input.flac

But a better option as you're ripping your CDs would be to use a ripper that supports FLAC so that they get tagged as you rip. I use EAC (http://www.exactaudiocopy.de).

I then use foobar2000 to transcode all my music from FLAC to MP3 (LAME) for use in my portable, so I keep all my music in two formats.

The advantage of this is I've always got the FLAC masters to go back to. For example I apply track replaygain during the transcode to MP3 because on my portable I want everything to be the same volume, but I don't want that for my home listening. If I change my mind and decide I do want the original volumes on my portable I can transcode again. If I want to drop the bitrate to get more onto my portable it's then again it's easy to do another transcode. In the future if I buy a new device that supports AAC then I can convert all my FLACs to AAC. If I'd just ripped straight to AAC to start with then I wouldn't have the master to go back to for any other transcoding needs.

With the cost of storage these days I can't think of any compelling reason not to rip first to FLAC. My 4,719 music tracks only take up 118 gig.

[audyovydeo]

With no offense you seem to be rather clueless here


no offense taken : I *am* rather clueless in this domain, not least because time's a-lacking to delve into the subject as much as I'd like.
I've just tried FLAC on my samples, and I'm dumping the idea of lossless : storage may not be a problem, but is a constraint.

I'll probably settle for Q 0.75 to 1.

Thanks to all for your inputs.

cheers
a/v

[tebasuna51]

1) LC, recommended for bitrates 48 Kb/s per channel, and greater.

2) HE = LC + SBR: low frequencies encoded in LC mode, high frequencies encoded with SBR (Spectral Band Replication) mode. Recommended for bitrates less than 48 Kb/s per channel. Used by default for quality Q=0.31 and less.

3) HEv2 = LC + SBR + PS: only for stereo audio, mono channel encoded like HE, channel differences encoded with PS (Parametric Stereo) mode. Recommended for very low bitrates.

For transparency use only LC mode.

[Dracaena]

It is best to avoid forcing a profile. Let the encoder decide which to use, it knows best.
Typically it will only use HE at low bitrates, and HEv2 at very low bitrates, like 48kbps and under.
For best quality/filesize ratio use -q mode. You should only use ABR/CBR if you really want a specific filesize, or you need CBR for some sort of streaming application.

Personally I would say 0.75 is extreme overkill. If you're going to go that high you may as well just use flac.
-q 0.3 would be pretty much transparent for *most* people on *most* material. Only small proportion of people ABX regular music encoded at -q 0.5.

If that's not enough you may as well go flac. But you should go flac anyway because then you'll never have to rip your collection again - just encode for portables when neede.

[audyovydeo]

If you're going to go that high you may as well just use flac.



thanks, but :

Cornershop_10_Spectral_Mornings
wav 152659172 Bytes
neroaac q 0.75 32180353 Bytes
flac q 8 116684574 Bytes

life is a compromise, and music backup is a subset of life, so it's a compromise too.

cheers
a/v

[ramicio]

Compromise is a losing attitude.

[audyovydeo]

Compromise is a losing attitude.

... like anyone doing lossy compression codecs for audio and video knows ... ;-)

I'll drink to your attitude !

cheers
a/v

[Dark Eiri]

128 kbps for mp3 is just not acceptable for music. anything below 256 kbps for mp3 will sound awful for most things if you have "trained" ears. for AAC I would go VBR 0.7 or even 0.95 / 1.0

[IgorC]

The issue is a placebo effect. If listener already know
beforehand that is 128 kbps it doesnt matter if it's state of art and transparent codec at this rate. He will rate it low. Now if listener is informed that is 256 kbps than he will rate it quite high even this is old encoder with low quality.

LC-AAC was defined as transparent for 128 kbps. And it was verified several times in past. Now if it wasn't true it would turn in legal issues for MPEG, ITU and Fraunhofer.

I would recommend to try the latest Apple AAC encoder in True-VBR with high quality mode at 128 kbps. http://tmkk.pv.land.to/qtaacenc/
This is what can be more close to reference quality of AAC standard.
The output of Apple AAC with high quality settings (--tvbr 65 --highest) at 128 kbps is transparent for 90-95% of all people. This statement was confirmed by some well know developers (Nero's too) ...several times. Yes, several.

And maybe some real audiophiles (not audiofools) will need somehow higher bitrate (up to 192 kbps)

Also it might be usefull to read about ABX and how ironious the objective opinion (placebo) can be.

http://www.hydrogenaudio.org/forums/index.php?showtopic=16295
http://www.hydrogenaudio.org/forums/index.php?showtopic=11442

After reading those and fully understand it you maybe will be in conditions to perform some blind ABX tests.

An application for ABX tests. http://ff123.net/abchr/abchr.html

[ramicio]

I went from AAC@192kbps to FLAC for the reason that to me it wasn't "transparent" over headhones, even over speakers. But yea, 128 kbps is plenty for the majority of the dumb-eared population, this is why it's not a "legal problem" for MPEG, etc.

[IgorC]

But yea, 128 kbps is plenty for the majority of the dumb-eared population, this is why it's not a "legal problem" for MPEG, etc.

The verification was performed by professionals of the area .
http://mp3-tech.org/programmer/docs/w2006.zip

[nurbs]

I went from AAC@192kbps to FLAC for the reason that to me it wasn't "transparent" over headhones, even over speakers.
Yes, and you also believe your vinyl rips only sound right when you do them at 24bit/192kHz. That's exactly what IgorC is talking about, placebo effect.

[shon3i]

@IgorC, nice explanation, but as you know today is much more audiofools as you call it. is Apple AAC now recommend over Nero and CT? When aim high quality?

For transparency use only LC mode. Well, i know there is much researches on AES and EBU and other audio societies, that many times proven that SBR is low bitrate tool only, but CT's high bitrate SBR definitely is not to throw, i mean i use it many times and i in cases where for example LC-AAC @ 128kbps not reach transparency, HE-AAC shows more respect and it's closer to original. But that is my opinions.

[IgorC]

is Apple AAC now recommend over Nero and CT? When aim high quality?

There never was a recommended encoder for AAC.


Well, i know there is much researches on AES and EBU and other audio societies, that many times proven that SBR is low bitrate tool only, but CT's high bitrate SBR definitely is not to throw, i mean i use it many times and i in cases where for example LC-AAC @ 128kbps not reach transparency, HE-AAC shows more respect and it's closer to original. But that is my opinions.

Can you upload the samples where HE-AAC is better than LC-AAC at 128 kbps?
Thank you.