If you want some testing, download samples that are provided here
and encode them with your encoder of choice!
http://www.hydrogenaudio.org/index.php?act=ST&f=16&t=4601&
If you want to go further have a deep look into ff123s site
http://www.ff123.net/
When you look thru hydrogenaudio.org there you will find
parts of music that sound different to the original with
whatever setting they are encoded into mp3! Also abxed
from others...
Some of these test-samples are found by me btw. 😉
and encode them with your encoder of choice!
http://www.hydrogenaudio.org/index.php?act=ST&f=16&t=4601&
If you want to go further have a deep look into ff123s site
http://www.ff123.net/
When you look thru hydrogenaudio.org there you will find
parts of music that sound different to the original with
whatever setting they are encoded into mp3! Also abxed
from others...
Some of these test-samples are found by me btw. 😉
My main gripe with MP3 is the loss of fine detail. The little things in complex recordings seem to just disappear.
Unless you know the recording, or A/B it, you tend not to notice, but something is missing. Even >192kHz. I doubt that I could line up 128 vs 192 vs 320 in the correct order, but on the right recordings I can tell the difference from a CD every time.
I've noticed treble "jangly-ness" but I think this may be encoder-specific. I'm not convinced it is a problem with MP3.
MP3 is a great format for many things. It is unfortunately just inferior to CD.
Unless you know the recording, or A/B it, you tend not to notice, but something is missing. Even >192kHz. I doubt that I could line up 128 vs 192 vs 320 in the correct order, but on the right recordings I can tell the difference from a CD every time.
I've noticed treble "jangly-ness" but I think this may be encoder-specific. I'm not convinced it is a problem with MP3.
MP3 is a great format for many things. It is unfortunately just inferior to CD.
Like I said, of course there are differences. But if you don't know WHICH of the two sources you are hearing, and maybe were not in the least biased (in other words: trying to find out which one is the mp3), and simply pick the one you liked best, mp3 would probably win...
Maybe this is due to distortions and "wrong" taste of people, but then why do a lot of high enders prefer LP records over CD, I don't think because of the cleaner sound...
And people like Wombat and the others from hydrogenaudio, who actually use ABX comparison programs, they also use "special" audio clips, which are known to produce problems (I have to admit that I did not hear the "fatboy.wav" clip before, and yes, one can de3finitely pick the mp3 even at ABR 192 kbit/s, with OGG at 192 kbit/s it get's better, but still not completely right)
But I would say about 98% of the average material (at least on my HD, but I don't listen to much electronic sounds like Fatboy Slim) those "special situations" won't occur.
It's like a real, real classical music fan, who has listenend to the same composition played by different artists over and over again, when he goes to a concert, of course he can pick out faults by some musicians, but the rest of the audience can't. And he probably could not hear any "faults" in a track by Metallica or something else, encoded to mp3, simply because he doesn't know it well.
But yeah, the frontlines are somewhat hardened in this discussion, so I won't pursue this any longer (done that enough in this thread already...) 😉
Bye,
Arndt
Maybe this is due to distortions and "wrong" taste of people, but then why do a lot of high enders prefer LP records over CD, I don't think because of the cleaner sound...
And people like Wombat and the others from hydrogenaudio, who actually use ABX comparison programs, they also use "special" audio clips, which are known to produce problems (I have to admit that I did not hear the "fatboy.wav" clip before, and yes, one can de3finitely pick the mp3 even at ABR 192 kbit/s, with OGG at 192 kbit/s it get's better, but still not completely right)
But I would say about 98% of the average material (at least on my HD, but I don't listen to much electronic sounds like Fatboy Slim) those "special situations" won't occur.
It's like a real, real classical music fan, who has listenend to the same composition played by different artists over and over again, when he goes to a concert, of course he can pick out faults by some musicians, but the rest of the audience can't. And he probably could not hear any "faults" in a track by Metallica or something else, encoded to mp3, simply because he doesn't know it well.
But yeah, the frontlines are somewhat hardened in this discussion, so I won't pursue this any longer (done that enough in this thread already...) 😉
Bye,
Arndt
Ogg Vorbis IS the only way to go.
It really does leave MP3s behind...
I encode at 180kbps average, and find it very hard to pick the difference between it and the original.
One of the best things about Ogg is there is only one encoder avaliable, so you also know that other Ogg encodings are going to be of good quality.
Not like MP3s where a lot of 192kbps MP3s off the net are comparable to 128kbps because they used a bad encoder.
http://www.vorbis.com/download.psp
It really does leave MP3s behind...
I encode at 180kbps average, and find it very hard to pick the difference between it and the original.
One of the best things about Ogg is there is only one encoder avaliable, so you also know that other Ogg encodings are going to be of good quality.
Not like MP3s where a lot of 192kbps MP3s off the net are comparable to 128kbps because they used a bad encoder.
http://www.vorbis.com/download.psp
First of all I have to say quality depends on the material you encode - some things sound nice and rich even at 128, some need no less than 192-256, in general electronic music encodes better than rock or classics(IMO classics on MP3 = plain wrong, POPs encodes best at 128... thats pops). Frequency responce of MP3 at 160-192+ is fairly good BUT phasing is worse - you get "muddy" sound with so so stereoscene and sound source placement(english is not my native language so explaining this is rather hard). MP3's are listener dependent - some say 128 is OK some say - 192, No way give me 256+ - thats where psycho-acoustic model plays its role - it for more or less "generic" ears, I've read an article about human ears - there are so many parameters that can differ from ear to ear, take for example acoustic masking - there are 3 types - when 2 sounds differ in loudness, delay and frequency...
As for the audiophiles there are lossless codecs - monkey audio etc, but dont expect 10:1 or even 5:1 compression ratios...
As for the audiophiles there are lossless codecs - monkey audio etc, but dont expect 10:1 or even 5:1 compression ratios...
MWP said:
i second that!
basically it will give a better representation of the original than mp3 with an equal file size. and if still isn't up to your standard, increase the quality level till it is.
Just finished a test...
in which I was a participant. It used aac+, along with some pre-processing. The general test was using three samples of the same material, one "natural", one aac at a pretty low bitrate and the final with same bitrate with pre-processing.
The testing objective was "which do you like better" - and this objective is different than most of the other testing I have been involved in along the way. The stuff in the past was mostly performed by people who had trained thier ears to listen for artifacts and such. Nothing quite so bad as plucked strings at LBR's. I remember a test with Lucent's PAC a few years ago that was almost unlistenable - this was much better, and points to the improvements that have happened in the last 3 or 4 years.
The end result? The average person actually preferred the encoded and processed sample. It is still being studied to determine what about that sample made them prefer it over the "natural" sample. I know for my ears, the pre-processing was critical to overall performance of the codec, regardless of the bitrate used.
in which I was a participant. It used aac+, along with some pre-processing. The general test was using three samples of the same material, one "natural", one aac at a pretty low bitrate and the final with same bitrate with pre-processing.
The testing objective was "which do you like better" - and this objective is different than most of the other testing I have been involved in along the way. The stuff in the past was mostly performed by people who had trained thier ears to listen for artifacts and such. Nothing quite so bad as plucked strings at LBR's. I remember a test with Lucent's PAC a few years ago that was almost unlistenable - this was much better, and points to the improvements that have happened in the last 3 or 4 years.
The end result? The average person actually preferred the encoded and processed sample. It is still being studied to determine what about that sample made them prefer it over the "natural" sample. I know for my ears, the pre-processing was critical to overall performance of the codec, regardless of the bitrate used.
I'm just finishing composing a CD to play with different MP3 bit rates and encoding schemes.
I took the song "Honey Pie" from Lavay Smith and her Red Hot Skillet Lickers. See:
http://www.lavaysmith.com/
Which is contemporay big band jazz and is recorded well and has a variety of instruments which should show up the shortcomings of MP3.
I started by ripping the track to an AIFF file with no compression. This will be track 1 and my reference. I then ripped the same track (from the original CD) tyo MP3 using a fixed bit rate at 128,160,192, and 320 Kbps. I then ripped the same track using a variable bit rate at the above average bit rates.
Finally I encoded the same track using AAC at the same bit rates.
I'll report my findings after I listen for a while. BTW I have a very high resolution system based around electrostatic loudspeakers.
Any interest in this disk? I can post an ISO image if there is enough interest.
Sheldon
I took the song "Honey Pie" from Lavay Smith and her Red Hot Skillet Lickers. See:
http://www.lavaysmith.com/
Which is contemporay big band jazz and is recorded well and has a variety of instruments which should show up the shortcomings of MP3.
I started by ripping the track to an AIFF file with no compression. This will be track 1 and my reference. I then ripped the same track (from the original CD) tyo MP3 using a fixed bit rate at 128,160,192, and 320 Kbps. I then ripped the same track using a variable bit rate at the above average bit rates.
Finally I encoded the same track using AAC at the same bit rates.
I'll report my findings after I listen for a while. BTW I have a very high resolution system based around electrostatic loudspeakers.
Any interest in this disk? I can post an ISO image if there is enough interest.
Sheldon
Hell yes, there'd be interest. I need a copy of that to hand out to the ignorant- or discover that I'm the ignorant one and then shut up 😉
I'll describe what I hear, but first a bit about my system:
My speakers are DIY electrostatic hybrids with and active crossover. The details can be seen here:
http://quadesl.com/diy_esl1.shtml
They are amazing in terms of tonal neutrality and dynamic punch. (the panel has the same Sd as 28 8" woofers and 3mm of Xmax). The active crossover makes a huge difference in terms of transparency and neutrality.
The source is my Tube based DAC, seen here:
http://quadesl.com/dac.shtml
Comments here:
http://quadesl.com/dac_comments.html
http://www.audioasylum.com/forums/bottlehead/messages/51120.html
I'm also using a modified Counterpoint preamp driving my active crossover. My panel amps are rebuilt Heath W5M's and the bass amp is an Adcom 555.
Here's what I heard:
The original: Lavay Smith is reasonably well recorded with good instrument placement and noticable air around the instruments. There is drums, an acoustic bass, piano, a variety of brass, and female voice. The miking is minimal and non-intrusive, there is a good amount of information in this recording; a rich palette.
AAC: I'm not sure what the difference between AAC and MP3 encoding is, and I'm not sure it's all that universal, so I'm not going to dwell on it here. This compression was quite similar to the MP3 VBR in terms of the artifacts I heard vs. bit rate.
MP3: First of all, the VBR at a given bit rate (average) was noticeably better than the FBR at the same rate. VBR at a given bit rate was about as good as FBR at the next higher rate. However the file size for VBR was about the same as FBR at the next higher rate. So in terms of compression, they are like apples to apples. Here's what I heard: At 128 kbps, the cymbals lost their shimmering "ride" and sounded a lot more like air escaping (or a cheap dome tweeter). Brass instruments didn't sound lifelike; they sounded synthesized like a good midi synth. The air around instruments and the decay of notes is largely gone. There is a hint of it, but it's not right. Female voice takes on a hard metal quality and loses the tonal subtleties, it's like the levels have been more discretized. At 128 kbps the difference is fairly obvious, if I had a helper and did a blind test, I feel pretty confident that I could pick them out every time. At 160 kbps, it's a little less obvious but still a difference. The air is still largely messed up. At 192 kbps, the air starts to return and the brass instruments and sounding more real. Cymbals are still a bit of a problem but less glaringly so. At 320 kbps, things are really looking up, most of the air is back, and the cymbals are sounding a lot more real. Going back and forth, I can tell the difference, but without an a/b I could be happy with that level of quality. The difference is subtle and one of smoothness and "rightness" There is a harshness at all MP3 settings, and it's not a typical bad recording harshness, but one that changes with transient content. It's hard to describe without hearing it.
I would also like to point out that on my okara speakers or my Jordan JX-92 based TLs' this would be a lot less conclusive test. The absolute transparency, low distortion, and neutrality of the ESL's makes differences pop out.
Sheldon
My speakers are DIY electrostatic hybrids with and active crossover. The details can be seen here:
http://quadesl.com/diy_esl1.shtml
They are amazing in terms of tonal neutrality and dynamic punch. (the panel has the same Sd as 28 8" woofers and 3mm of Xmax). The active crossover makes a huge difference in terms of transparency and neutrality.
The source is my Tube based DAC, seen here:
http://quadesl.com/dac.shtml
Comments here:
http://quadesl.com/dac_comments.html
http://www.audioasylum.com/forums/bottlehead/messages/51120.html
I'm also using a modified Counterpoint preamp driving my active crossover. My panel amps are rebuilt Heath W5M's and the bass amp is an Adcom 555.
Here's what I heard:
The original: Lavay Smith is reasonably well recorded with good instrument placement and noticable air around the instruments. There is drums, an acoustic bass, piano, a variety of brass, and female voice. The miking is minimal and non-intrusive, there is a good amount of information in this recording; a rich palette.
AAC: I'm not sure what the difference between AAC and MP3 encoding is, and I'm not sure it's all that universal, so I'm not going to dwell on it here. This compression was quite similar to the MP3 VBR in terms of the artifacts I heard vs. bit rate.
MP3: First of all, the VBR at a given bit rate (average) was noticeably better than the FBR at the same rate. VBR at a given bit rate was about as good as FBR at the next higher rate. However the file size for VBR was about the same as FBR at the next higher rate. So in terms of compression, they are like apples to apples. Here's what I heard: At 128 kbps, the cymbals lost their shimmering "ride" and sounded a lot more like air escaping (or a cheap dome tweeter). Brass instruments didn't sound lifelike; they sounded synthesized like a good midi synth. The air around instruments and the decay of notes is largely gone. There is a hint of it, but it's not right. Female voice takes on a hard metal quality and loses the tonal subtleties, it's like the levels have been more discretized. At 128 kbps the difference is fairly obvious, if I had a helper and did a blind test, I feel pretty confident that I could pick them out every time. At 160 kbps, it's a little less obvious but still a difference. The air is still largely messed up. At 192 kbps, the air starts to return and the brass instruments and sounding more real. Cymbals are still a bit of a problem but less glaringly so. At 320 kbps, things are really looking up, most of the air is back, and the cymbals are sounding a lot more real. Going back and forth, I can tell the difference, but without an a/b I could be happy with that level of quality. The difference is subtle and one of smoothness and "rightness" There is a harshness at all MP3 settings, and it's not a typical bad recording harshness, but one that changes with transient content. It's hard to describe without hearing it.
I would also like to point out that on my okara speakers or my Jordan JX-92 based TLs' this would be a lot less conclusive test. The absolute transparency, low distortion, and neutrality of the ESL's makes differences pop out.
Sheldon
stokessd said:
As long as the ISO only contains partial tracks, say a minute or so, otherwise those RIAA boys could be down on us like a ton of bricks! 😉
@stokessd
As you are talking about creating aif files i think
you use a MAC for that.
When you use a MAC it is most likely you use a mp3
codec that is based on Fraunhofers mp3.
If that is true you don´t have the best mp3 encoder
for high bitrates. The lame encoder is superior from 160kBit
up. The vbr implementation of fraunhofer is worse even
to the highest quality setting. You don´t need a high-end
system to hear that. I use a simple HD-590 and my soundcard
on the PC. Try to get lame mp3 encoded files and try again.
As you are talking about creating aif files i think
you use a MAC for that.
When you use a MAC it is most likely you use a mp3
codec that is based on Fraunhofers mp3.
If that is true you don´t have the best mp3 encoder
for high bitrates. The lame encoder is superior from 160kBit
up. The vbr implementation of fraunhofer is worse even
to the highest quality setting. You don´t need a high-end
system to hear that. I use a simple HD-590 and my soundcard
on the PC. Try to get lame mp3 encoded files and try again.
When you have the chance to use lame, please try the tested settings from this site together with lame 3.90.3
http://www.hydrogenaudio.org/index.php?showtopic=203&st=0
http://www.hydrogenaudio.org/index.php?showtopic=203&st=0
I don't know what AAC encoding really is. Apple in their typical style isn't saying anything useful on their technical support pages either.
I'm now ripping the same track with Audion which has a lame (3.92) encoder. I can easily pick the fixed bit rates, but I can't seem to be able to set the variable bit rates numerically, just with settings like "insane" and "excellent"
I'll report later when the disk is burned and I'm listening.
With regards to my MP3 disk, I'm a little hesitant to post it given that it's full tracks. I'll be passing it around our Albuquerque Speaker Society though.
Sheldon
I'm now ripping the same track with Audion which has a lame (3.92) encoder. I can easily pick the fixed bit rates, but I can't seem to be able to set the variable bit rates numerically, just with settings like "insane" and "excellent"
I'll report later when the disk is burned and I'm listening.
With regards to my MP3 disk, I'm a little hesitant to post it given that it's full tracks. I'll be passing it around our Albuquerque Speaker Society though.
Sheldon
for all you mac weenies (back off; I'm also a unix and peecee weenie) out there:
There is a really nice interface to LAME (3.92) for iTunes, it's here:
http://www.blacktree.com/apps/
It has the typical aqua interface and is easy to use, and for us control freaks, it has a mini command line to pass the standard LAME parameters to the encoder.
Good stuff.
Sheldon
There is a really nice interface to LAME (3.92) for iTunes, it's here:
http://www.blacktree.com/apps/
It has the typical aqua interface and is easy to use, and for us control freaks, it has a mini command line to pass the standard LAME parameters to the encoder.
Good stuff.
Sheldon
- Status
- Not open for further replies.