Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Compression regret (Read 3464 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Compression regret

Has anyone ever gone through the effort of doing an ABX test, picking the codec and bitrate they need for transparency, encoded a bunch of music, and finding that when they changed equipment, that transparency was not achieved?

I'm wondering whether suddenly switching to a set up with "better" equipment would cause a loss of transparency?

I know a lot of high end audiophiles talk out their ass when it comes to this stuff, but I have friends who swear up and down that they can hear a difference between a 320K MP3 and a FLAC and won't do an ABX test, because you can only hear the difference on their $5,000 tube amp, DAC and speaker combo.

Compression regret

Reply #2 –
Usually abx is easiest with ordinary headphones, so I'd recommend you determine transparency with those.

There is also no real reason to think transparency is harder to achieve with better equipment. Depending on how you look at it the reverse may even be true.

Compression regret

Reply #3 –
FWIW, I've been demanding evidence from people claiming expensive gear makes lossy artifacts easier to hear for years now.  Despite being told some will be provided on more than one occasion, no evidence has been forthcoming

I agree with Saratoga, though, headphones are your best bet.  They needn't be expensive.

Compression regret

Reply #4 –
320K MP3s are perfectly detectable in an ABX trial, but in my experience only CBR MP3.  LAME though, so still better than most MP3s more than 10 years old.  Bear in mind that whether it is the equipment or not, it is very possible for your perceptions to change and what you don't hear today may be glaring and intrusive next year.  Or not

As to the replies so far, who's talking out of their asses now?  If you haven't done the test you shouldn't be claiming what the results would be.

Compression regret

Reply #5 –
Quote
but I have friends who swear up and down that they can hear a difference between a 320K MP3 and a FLAC and won't do an ABX test, because you can only hear the difference on their $5,000 tube amp, DAC and speaker combo.
And, what's stopping them from doing an ABX test with their high-end equipment?

Compression regret

Reply #6 –
Quote from: lithopsian on
320K MP3s are perfectly detectable in an ABX trial, but in my experience only CBR MP3.


I'm rather sure you don't get to just claim that here without some evidence.

Compression regret

Reply #7 –
Quote from: DVDdoug on
Quote
but I have friends who swear up and down that they can hear a difference between a 320K MP3 and a FLAC and won't do an ABX test, because you can only hear the difference on their $5,000 tube amp, DAC and speaker combo.
And, what's stopping them from doing an ABX test with their high-end equipment?



I would assume pride.  You spend a ton of money on AV equipment and you need to keep justifying in your head that all that money was worth it.

I guess they could do an ABX test with another human being playing the music and hiding what's playing.

I found through the years, when it comes to equipment cost, the difference in perceived sound quality between low end and midrange is HUGE.  The difference between mid-range and high end is almost non existent, at least to me.

Compression regret

Reply #8 –
Quote from: apastuszak on
Has anyone ever gone through the effort of doing an ABX test, picking the codec and bitrate they need for transparency, encoded a bunch of music, and finding that when they changed equipment, that transparency was not achieved?
<snip>


So keep the 320K MP3 and the FLAC files. The total space will be about 75% of the space used by the .WAV files. Disk drives are so cheap there is no reason NOT to keep both. With a 3T drive for $90 it works out to about 2-3 cents per CD.

G²

Compression regret

Reply #9 –
Quote from: Glenn Gundlach on
Quote from: apastuszak on
Has anyone ever gone through the effort of doing an ABX test, picking the codec and bitrate they need for transparency, encoded a bunch of music, and finding that when they changed equipment, that transparency was not achieved?
<snip>


So keep the 320K MP3 and the FLAC files. The total space will be about 75% of the space used by the .WAV files. Disk drives are so cheap there is no reason NOT to keep both. With a 3T drive for $90 it works out to about 2-3 cents per CD.

G²


Double whatever your HD needs are for backups.  Remember the rule "If your file doesn't exist in two places it does not exist."

The DAM Book also has a great rule called the 3-2-1 rule.  3 copies of every file you want to keep on two different media, and one offsite copy.

My 3-2-1 strategy:

3 - The original file, the backup on the enternal hard drive, and the copy on Crashplan
2 - 2 different kinds of media, HD and Crashplan
1 - 1 offsite copy on Crashplan

For the truly paranoid, Crashplan lets you create your own encryption keys, so even Crashplan can't get to your data.  I'm sure other online backup services exist also that offer simialr protection.  Crashplan is just who I use currently.

Compression regret

Reply #10 –
I've gone through the opposite. I've ripped my music to a lossy format at a much higher bitrate than I need for transparency and then found that I was format locked unless I was willing to tolerate generational losses. In reality I seldom need more than a modern VBR codec (Vorbis, AAC, Opus) at around 96kbps to achieve transparency for me and everyone in my family. About a decade ago all my music was ripped and stored as a combination of mp3 at 192kbps or higher and Vorbis at 160kbps or higher. This began to cause me issues with the rise of support for only mp3 in certain audio devices such as car stereo receivers. I also found after doing some ABX work that I was basically wasting space with anything over 96kbps. On my hard drives I didn't mind that much, but the sting was felt when flash based media players became the norm. I quickly found that I could double the amount of music I could store on such devices by transcoding to Vorbis at quality 2, but with all my ripped source material already lossy I was trapped, unable to reduce the bitrate to what I wanted without taking a sometimes audible hit in quality.

I decided back then, when hard drive space began getting ridiculously cheap, to ditch my whole ripped library and rerip everything again, this time to FLAC. I now have lossless versions stored, and those are what I play on my LAN. But, this gives me format freedom. On the devices that only supported mp3, I transcoded to mp3 at 128-160kbps VBR, and the devices which support Vorbis get quality 2 transcodes. Since I've had bluetooth audio in our vehicles I've consolidated all my mobile music on my Android phone and so everything is currently Vorbis. But, once Android L is released and I update my phone to that, I will likely transcode my files again to Opus at around 96kbps to gain the quality that affords on the edge cases where Vorbis may have a problem.

Because my entire source library is FLAC, when the time comes, transcoding the whole thing to Opus will require little more than a drag and drop operation. Nice and fast, a single operation, first generation quality every single time. Lossless is really the only way to go my friend.

As for backup copies, I store all my media on a FreeBSD server based ZFS mirrored pool. ZFS provides block level checksumming which combined with the mirror provides fault tolerance against silent corruption. It also takes a daily snapshot for a week, keeps monthly snapshots for a year, and yearly snapshots until I feel like destroying them. I use the ZFS incremental send/receive functions to backup all the datasets to a ZFS external USB drive on my Linux desktop, which I then rotate offsite bi-monthly. So all data is stored on a high reliability fault tolerant filesystem, with 2 copies of each block mirrored between drives from different manufacturers. Then all data is again mirrored to a 3rd fault detecting pool and transported offsite. A regional disaster could still take me out, but if something is bad enough where I live to destroy both sites where my data is stored, I'll have bigger issues to worry about than my music files.
 

Compression regret

Reply #11 –
Quote from: yourlord on
I've gone through the opposite. I've ripped my music to a lossy format at a much higher bitrate than I need for transparency and then found that I was format locked unless I was willing to tolerate generational losses. In reality I seldom need more than a modern VBR codec (Vorbis, AAC, Opus) at around 96kbps to achieve transparency for me and everyone in my family. About a decade ago all my music was ripped and stored as a combination of mp3 at 192kbps or higher and Vorbis at 160kbps or higher. This began to cause me issues with the rise of support for only mp3 in certain audio devices such as car stereo receivers. I also found after doing some ABX work that I was basically wasting space with anything over 96kbps. On my hard drives I didn't mind that much, but the sting was felt when flash based media players became the norm. I quickly found that I could double the amount of music I could store on such devices by transcoding to Vorbis at quality 2, but with all my ripped source material already lossy I was trapped, unable to reduce the bitrate to what I wanted without taking a sometimes audible hit in quality.

I decided back then, when hard drive space began getting ridiculously cheap, to ditch my whole ripped library and rerip everything again, this time to FLAC. I now have lossless versions stored, and those are what I play on my LAN. But, this gives me format freedom. On the devices that only supported mp3, I transcoded to mp3 at 128-160kbps VBR, and the devices which support Vorbis get quality 2 transcodes. Since I've had bluetooth audio in our vehicles I've consolidated all my mobile music on my Android phone and so everything is currently Vorbis. But, once Android L is released and I update my phone to that, I will likely transcode my files again to Opus at around 96kbps to gain the quality that affords on the edge cases where Vorbis may have a problem.

Because my entire source library is FLAC, when the time comes, transcoding the whole thing to Opus will require little more than a drag and drop operation. Nice and fast, a single operation, first generation quality every single time. Lossless is really the only way to go my friend.

As for backup copies, I store all my media on a FreeBSD server based ZFS mirrored pool. ZFS provides block level checksumming which combined with the mirror provides fault tolerance against silent corruption. It also takes a daily snapshot for a week, keeps monthly snapshots for a year, and yearly snapshots until I feel like destroying them. I use the ZFS incremental send/receive functions to backup all the datasets to a ZFS external USB drive on my Linux desktop, which I then rotate offsite bi-monthly. So all data is stored on a high reliability fault tolerant filesystem, with 2 copies of each block mirrored between drives from different manufacturers. Then all data is again mirrored to a 3rd fault detecting pool and transported offsite. A regional disaster could still take me out, but if something is bad enough where I live to destroy both sites where my data is stored, I'll have bigger issues to worry about than my music files.



As a guy who's never used any BSD variants, I know almost next to nothing about ZFS.  It sounds awesome.  Linux is trying to do the same thing with the btrfs filesystem, but it's still not considered stable.

I think I may need to go down the FreeBSD route.  If nothing else, just to increase my skillset.