android
Go Back   abi>>forums > Software & Media > Software

Reply
 
Thread Tools Search this Thread Display Modes
  #41  
Old 03-29-2012, 08:30 PM
McDougal's Avatar
McDougal McDougal is offline
Ultra Senior Member
 
Join Date: Aug 2009
Posts: 2,375
Default

Quote:
Originally Posted by saratoga View Post
Arguments like this are bit tiresome though since they're endlessly rehashed, and of course the kind of people who bring them up are precisely the kind of people who do not care enough to do research or read a sticky.

I don't know that we should do the HA thing and require that people post ABX logs when complaining about audio codecs, but maybe a policy of politely linking people to a FAQ or sticky and then asking them not to claim what they can hear without proper blind listening tests would make sense? Its probably more friendly to new and casual readers then arguing with them anyway.
Agreed....
__________________
iPod Touch 4G 32GB; Google Nexus 7 32GB
JVC HAS600B;Pioneer SE-M390;Audio-Technica ATH-P5

Retired Devices:
Sansa Fuze 8+16GB Rockboxed
Reply With Quote

Advertisement [Remove Advertisement]

  #42  
Old 03-29-2012, 09:14 PM
WalkGood's Avatar
WalkGood WalkGood is offline
ABI Invaluable
 
Join Date: Aug 2006
Location: FL
Posts: 12,577
Default

Quote:
Originally Posted by saratoga View Post
... politely linking people to a FAQ or sticky and then asking them not to claim what they can hear without proper blind listening tests would make sense? ...
This was the purpose of this sticky, it just seems to have created a stir ...
__________________
WalkGood,
Ramón

abi >> | Forum Rules | Glossary | Why Rockbox | FLAC or MP3? | irc
Reply With Quote

  #43  
Old 03-29-2012, 09:41 PM
The DarkSide's Avatar
The DarkSide The DarkSide is offline
Ultra Senior Member
 
Join Date: Feb 2008
Location: Qns Vlg, NYC
Posts: 16,061
Default

Quote:
Originally Posted by WalkGood View Post
This was the purpose of this sticky, it just seems to have created a stir ...
Well, you have to admit one thing - controversy stirs up ratings!!! I give this thread 5 STARS!!!
Reply With Quote

  #44  
Old 03-30-2012, 12:48 AM
h1a8 h1a8 is offline
Member
 
Join Date: Apr 2008
Location: USA
Posts: 479
Default

Quote:
Originally Posted by skip252 View Post
Why mess with the carefully tuned, quality based LAME presets by using CBR 320 and a lowpass filter? I'd just encode to -V0 using LAME. The developers have already taken in to account what's needed to produce the best quality encodes without the addition of any custom switches.

In fact using CBR causes LAME to use lower quality settings.
Code:
VBR
Encoding settings: -m j -V 0 -q 0 -lowpass 22.1 --vbr-new -b 32
Code:
CBR
Encoding settings: -m j -V 4 -q 3 -lowpass 20.5
That would only be after I had ABXed and determined that I needed a setting that high to reach transparency. Despite what some are saying here it's highly unlikely that going with the recommended settings will produce a level of artifacts that will be detected under normal circumstances.
Razorlame has three settings: none, speed, and quality. I chose quality. So will V0 still have better quality settings? To me the graph of cbr seems transparent than the graph of v0. Both I can't abx successfully from the source so either actually would be fine. Thus mayble v0 is the smarter choice since it has a smaller file size.

Do you think a v0 lowpass filter of 16khz would result in a better quality file than ordinary v0?

Which program you use (especially the one where you posted the settings to)?
Reply With Quote

  #45  
Old 03-30-2012, 02:31 AM
skip252 skip252 is offline
Administrator
 
Join Date: Dec 2007
Location: Chicago
Posts: 5,444
Default

I've never used Razorlame so I'm can't tell which setting would produce results that won't change the LAME settings. If I had to guess it would be "none". That would probably only hold true if Razorlame doesn't have other settings that some other encoders like CDex have that set minimum and maximum bitrate and quality settings. If it does that could interfere with the LAME tunings that produce the best results.

I used foobar2000 for encoding the file that produced that info. The information itself was generated by MediaInfo. Both have options for either a portable or full installation. I use portable to be sure they don't conflict with any installed programs. When I have done a full installation they didn't cause any problems but using portable installations whenever possible is a holdover from when I've installed programs in the past that caused conflicts.

foobar2000 has a Converter component that comes in the base installer but you need to chose it for installation during the setup. You drop a copy of LAME into the foobar2000 folder so foobar2000 can find it by default. Then drop your files or folders into a foobar2000 playlist. Right click the files you want to encode and use the converter dialog box to chose your LAME VBR setting. The default is -V2 but you can change that to -V0 by either adding a new setting or editing the preset. If you just use the slider to chose -V0 you'll be encoding with no changes to the LAME presets.

You can use the Custom setting to change the switches used. If you're familiar using the older encoder front ends you'll recognize what you see once you open the custom settings.

From what I've read at Hydrogenaudio and from knowledgeable staff members here changing the -Vx quality based presets by adding a highpass filter isn't advised. LAME is a lot smarter than it once was and has been set to create the best quality files using it's presets.

A lot of the tweaking that people used to do to try to produce better results isn't needed now. It's possible to get worse results by interfering with the current LAME tunings by adding switches that cause it to work is a less than optimum fashion. I leave everything at the settings that come from the presets and haven't found any reason to complain about the results.

You've mentioned looking at graphs a couple of times to gauge results. Since I can't hear with my eyes I never bother with that. If I have reason to doubt the quality of a particular encode I use the ABX Comparator in foobar2000 to see if I can detect a difference from the lossless source.

I've seen some spectrograms of files that looked as though the files should might not be the best quality. After a few ABX sessions I learned that my ears were a much better judge of sound quality than my eyes.
Reply With Quote

  #46  
Old 03-30-2012, 11:08 AM
saratoga saratoga is offline
Rockbox Developer / Moderator
 
Join Date: Apr 2007
Posts: 3,646
Default

Quote:
Originally Posted by h1a8 View Post
Razorlame has three settings: none, speed, and quality. I chose quality. So will V0 still have better quality settings?
Razorlame shows you the lame settings its going to use at the bottom of the settings screen:

http://www.dors.de/razorlame/lameoptions.png

So if you're not sure, you can click that setting and see exactly what its changing. My guess is that its changing the "q" settings in lame, which basically just tell it how much time to spend on huffman coding and thus don't directly impact quality (and are largely ignored in modern lame versions anyway).
__________________
Interested in Google's Summer of Code ? PM me.
Reply With Quote

  #47  
Old 04-01-2012, 03:01 PM
McDougal's Avatar
McDougal McDougal is offline
Ultra Senior Member
 
Join Date: Aug 2009
Posts: 2,375
Default

Ok guys, I got a 4GB microSD card for my phone's music player.

My entire music collection is approx. 13GBs (as shown on pic in previous post), and I'd like to squeeze all my songs on the 4GB card.

I'll only be listening to this on $30 headphones, so sound quality isn't the concern.

My phone, a TracFone LG 500G, can play MP3 and AAC files. I was wondering, what's the best format and bitrate for me to convert to? Or is it even possible to get all my songs on there without going below 128kbps?
__________________
iPod Touch 4G 32GB; Google Nexus 7 32GB
JVC HAS600B;Pioneer SE-M390;Audio-Technica ATH-P5

Retired Devices:
Sansa Fuze 8+16GB Rockboxed
Reply With Quote

  #48  
Old 04-01-2012, 03:17 PM
saratoga saratoga is offline
Rockbox Developer / Moderator
 
Join Date: Apr 2007
Posts: 3,646
Default

Quote:
Originally Posted by McDougal View Post
My phone, a TracFone LG 500G, can play MP3 and AAC files. I was wondering, what's the best format and bitrate for me to convert to? Or is it even possible to get all my songs on there without going below 128kbps?
For very low bitrates, AAC works quite well. As for what bitrate you need to use, since you haven't said the total length of the files you want to encode to be 4GB I don't think anyone can answer that.
__________________
Interested in Google's Summer of Code ? PM me.
Reply With Quote

  #49  
Old 04-01-2012, 03:24 PM
McDougal's Avatar
McDougal McDougal is offline
Ultra Senior Member
 
Join Date: Aug 2009
Posts: 2,375
Default

Quote:
Originally Posted by saratoga View Post
For very low bitrates, AAC works quite well. As for what bitrate you need to use, since you haven't said the total length of the files you want to encode to be 4GB I don't think anyone can answer that.
What do you mean by total length?
__________________
iPod Touch 4G 32GB; Google Nexus 7 32GB
JVC HAS600B;Pioneer SE-M390;Audio-Technica ATH-P5

Retired Devices:
Sansa Fuze 8+16GB Rockboxed
Reply With Quote

  #50  
Old 04-01-2012, 04:53 PM
saratoga saratoga is offline
Rockbox Developer / Moderator
 
Join Date: Apr 2007
Posts: 3,646
Default

Quote:
Originally Posted by McDougal View Post
What do you mean by total length?
Without knowing how long the audio you want to fit into 4GB is, its hard to say what the bitrate needs to be.
__________________
Interested in Google's Summer of Code ? PM me.
Reply With Quote

  #51  
Old 04-29-2012, 04:11 AM
robdean robdean is offline
Junior Member
 
Join Date: Dec 2011
Posts: 17
Default

I've come late to this thread so would just like to hark back a page or two to the issue of the 'soundstage'.

I'm a former audio engineer, and an aficionado of lame -v0. In blind testing I cannot reliably pick out -v0 mp3 from 44.1/16 WAV.

However to be fair to those who feel that the soundstage is degraded by lossy compression, their complaint makes a great deal more sense than some contributors to this thread give them credit for.

The 'soundstage' is the sense of realistic space, ambience, and positioning of instruments: the miraculous 3-D illusion which is created by stereo speakers. Low frequencies contribute relatively little to this illusion, because the point of origin of low-frequency sounds is much harder to pinpoint perceptually even in real life than that of high frequency sounds. Low frequencies, of course, require proportionately less bit rate then high-frequency sounds.

A convincing 'soundstage' is all to do with high frequencies, phase relationships between the channels, and accurate reproduction of high harmonic frequencies. These are the elements most likely to be degraded by lossy audio compression, partly because they are simply the least compressible elements of the signal, and partly because it is very difficult for the codec to separate the perceptually significant (but subtle) from that which is insignificant.

A quite 'dry' recording of a single cello might sound tolerable even at 128kbps. A wide stereo recording of two 12 string guitars in a bright reverberant room, accompanied by a jazz drummer using brushes and lots of subtle hi-hat and cymbal work will demand a high bit rate because of all the complex high frequency detail, not just in the instruments but especially in the resulting soundstage, full of complex reflections and interactions.

Given a recording specially selected or created to be a challenge to psychoacoustic compression, it's possible (though far from certain) that even an old cloth ears like myself could prevail in blind tests. What I don't personally believe is that I'm somehow getting this 'goodness' subliminally in the overwhelming majority of recordings in which I couldn't possibly distinguish in blind testing between the lossy and be uncompressed reproduction. However, whilst we do live in a world of superstition and craziness, the 'soundstage' argument does have credibility.
Reply With Quote

  #52  
Old 04-29-2012, 10:06 AM
Satellite_6's Avatar
Satellite_6 Satellite_6 is offline
Senior Member
 
Join Date: Feb 2009
Location: VA, USA
Posts: 871
Default

"If “Probability that you were guessing” does not reach 5% or lower, you were probably guessing and therefore cannot hear a difference."

You were probably guessing if the p value was 50%+, so this is misleading.

I would also like to point out that this is the standard in scientific papers and stuff, seems extreme for the layman just trying to prove to her/himself that s/he can tell the difference. i.e. if I did an ABX properly and got a p value of 10% there is a 90% chance that I could tell the difference. . . that seems pretty darn good to me.

Well I like to argue or "debate" about this general subject sometimes, and I think debate is a good thing, better than the censorship at both hydrogenaudio and Head-Fi. That's why I continue to post here.

The above post is interesting, I've just never heard such a thing.

I want to do some blind tests this summer if I get around to it. . .
__________________
satelliteaudioblog.blogspot.com
Reply With Quote

  #53  
Old 04-29-2012, 05:09 PM
Enigmatic Enigmatic is offline
Senior Member
 
Join Date: Mar 2008
Posts: 737
Default

Quote:
Originally Posted by Satellite_6 View Post
"If “Probability that you were guessing” does not reach 5% or lower, you were probably guessing and therefore cannot hear a difference."
Yes, I should have written it better. But the p-value to get is still 0.05 or lower. Is it 0.05 or lower? Then one can legitimately claim to be able to hear a difference. Is it, say, 0.07 or 0.06? Then one cannot legitimately claim to be able to hear a difference. It is what it is.

robdean wrote about soundstage. If the soundstage is affected—or any another aspect of the sound quality is affected—that means the two files will sound different from each other. If the two files sound different from each other, this difference should be audible in an ABX test. And if it is not audible in an ABX test? Perhaps too much importance is given to soundstage.
Reply With Quote

  #54  
Old 04-29-2012, 05:56 PM
robdean robdean is offline
Junior Member
 
Join Date: Dec 2011
Posts: 17
Default

I'm 100% aboard with ABX testing and with scepticism as to the benefits of lossless playback.

My comments about the soundstage were motivated by some posters' apparent misconceptions or uncertainties of what the soundstage is, and of the challenge its reproduction represents.


A way to reframe the concept is to see it as very much to do with the challenge you take on when you move from mono reproduction to stereo reproduction. A key technique in lossy compression of a stereo signal is to isolate what is different between the two channels from that which the two channels have in common, and that difference is a lot to do with the soundstage. Thus when psychoacoustic compression works badly the soundstage will suffer. You can force this by encoding at low bit rate. In my experience, whilst both mono and stereo recordings degrade at lower bitrates, there are a host more distracting artefacts in the average degraded stereo recording because the spacial soundstage data goes to pieces in a really conspicuous ugly way.


To be clear, I'm not suggesting this is a real world problem for well encoded high bitrate lossy files, rather that it's not a senseless issue to consider in general. I'd certainly like to see it submitted to objective testing if suggestions persist that -v0 or 320 is perceptibly insufficient, as I'd be fascinated to be proved wrong and perfectly delighted to be proved right!
Reply With Quote

  #55  
Old 04-29-2012, 06:33 PM
saratoga saratoga is offline
Rockbox Developer / Moderator
 
Join Date: Apr 2007
Posts: 3,646
Default

Quote:
Originally Posted by robdean View Post
However to be fair to those who feel that the soundstage is degraded by lossy compression, their complaint makes a great deal more sense than some contributors to this thread give them credit for.
In my experience, its usually its just used as a byword for "i don't like", not actually what you and I would refer to as sound localization.

Quote:
Originally Posted by robdean View Post
The 'soundstage' is the sense of realistic space, ambience, and positioning of instruments: the miraculous 3-D illusion which is created by stereo speakers. Low frequencies contribute relatively little to this illusion, because the point of origin of low-frequency sounds is much harder to pinpoint perceptually even in real life than that of high frequency sounds. Low frequencies, of course, require proportionately less bit rate then high-frequency sounds.

A convincing 'soundstage' is all to do with high frequencies, phase relationships between the channels, and accurate reproduction of high harmonic frequencies.
This isn't really true. For phase relationships, its mostly only the low frequency content that matters. You cannot perceive phase differences between channels much above a Khz since the response time of your brain is much too slow. For higher frequencies, localization is mostly about amplitude differences (e.g. its louder on the left then the right). For this reason, when you look at crossfeed circuits people build for headphone amps, they usually only crossfeed the low frequency content, typically with a cutoff around 500Hz. Thats the part that really matters:

http://gilmore2.chem.northwestern.ed.../meier_prj.htm

Your reasoning about higher frequencies being easier to localize makes sense given how diffraction works, but your ear's localization is no where near good enough for diffraction to come into play, so in practice you tend to do better at lower frequencies where its much easier to compare phase/delay.

Wikipedia actually has a fairly good overview:

http://en.wikipedia.org/wiki/Sound_localization

Quote:
Originally Posted by robdean View Post
These are the elements most likely to be degraded by lossy audio compression, partly because they are simply the least compressible elements of the signal, and partly because it is very difficult for the codec to separate the perceptually significant (but subtle) from that which is insignificant.
Yeah, I don't think this is right. The most common artifact by far are temporal effects, particularly in mp3 given how its filterbank works. Stereo coding isn't necessarily easy, but its not so hard as you're thinking. Matching levels between channels is extremely easy (and the quantizer will generally do this by itself). Phase is more complicated, but since its only a small portion of the total signal bandwidth that contributes, you don't necessarily need to encode it efficiently. If 4% of your signal is coded less efficiently, thats not a big loss, so the encoder can easily afford to be more careful with low frequencies.

Quote:
Originally Posted by robdean View Post
A quite 'dry' recording of a single cello might sound tolerable even at 128kbps. A wide stereo recording of two 12 string guitars in a bright reverberant room, accompanied by a jazz drummer using brushes and lots of subtle hi-hat and cymbal work will demand a high bit rate because of all the complex high frequency detail, not just in the instruments but especially in the resulting soundstage, full of complex reflections and interactions.
You're over-thinking this. The problem from the encoder's point of view is to crush redundancy out of the stereo channels, then code them accurately.
The complexity of the sound field in the room is basically irrelevant since you're just recording 2 (or perhaps a few if its a multichannel downmix) points. Your microphone placement handles squeezing 99.99% of the information out of the recording. The encoder just takes those two point measurements, calculates a sum and difference and then compares the entropy in the sum/diff to the L/R representation and goes with whichever one is smaller.

That is not to say that accurate stereo recording (or worse multichannel) is easy. Its not. But thats a problem of acoustics that the recording engineer has to deal with. By the time you've downmixed your recording to stereo and fed it into an mp3 encoder the hard stuff is long since over with.
__________________
Interested in Google's Summer of Code ? PM me.
Reply With Quote

  #56  
Old 05-07-2012, 01:16 AM
h1a8 h1a8 is offline
Member
 
Join Date: Apr 2008
Location: USA
Posts: 479
Default

LOL I don't understand. If one can't hear a difference in a ABX then it is silly to debate about how something sounds better. Just common sense that's all.
Reply With Quote

  #57  
Old 05-08-2012, 04:09 PM
McDougal's Avatar
McDougal McDougal is offline
Ultra Senior Member
 
Join Date: Aug 2009
Posts: 2,375
Default

I just re-ripped some old classic albums I had in yucky 128k mp3s to perfect-quality LAME -V2. What a difference !
__________________
iPod Touch 4G 32GB; Google Nexus 7 32GB
JVC HAS600B;Pioneer SE-M390;Audio-Technica ATH-P5

Retired Devices:
Sansa Fuze 8+16GB Rockboxed
Reply With Quote

  #58  
Old 05-08-2012, 04:13 PM
The DarkSide's Avatar
The DarkSide The DarkSide is offline
Ultra Senior Member
 
Join Date: Feb 2008
Location: Qns Vlg, NYC
Posts: 16,061
Default

Yeah, I notice a slight difference when the files are that low in bitrate. I try to just do V~0, since I'm paranoid about it. And for portable usage I use V~2 myself.
Reply With Quote

  #59  
Old 05-08-2012, 04:16 PM
McDougal's Avatar
McDougal McDougal is offline
Ultra Senior Member
 
Join Date: Aug 2009
Posts: 2,375
Default

Wait, you have a -V2 copy and a -V0 copy of the same stuff? And these are older albums anyway so -V0 would probably be overkill lol.

(trying to get this thread back on topic )
__________________
iPod Touch 4G 32GB; Google Nexus 7 32GB
JVC HAS600B;Pioneer SE-M390;Audio-Technica ATH-P5

Retired Devices:
Sansa Fuze 8+16GB Rockboxed
Reply With Quote

  #60  
Old 05-08-2012, 04:19 PM
The DarkSide's Avatar
The DarkSide The DarkSide is offline
Ultra Senior Member
 
Join Date: Feb 2008
Location: Qns Vlg, NYC
Posts: 16,061
Default

I have the tracks ripped to V~0 on my comp, and I down convert them to V~2 using MediaMonkey when syncing. I'm "down converting", not up. It works out just fine.
Reply With Quote

Reply

Tags
flac vs mp3

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump



All times are GMT -5. The time now is 01:33 PM.