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: Lossy to Lossy (generation loss) tests? (Read 2342 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Lossy to Lossy (generation loss) tests?

Even among psychoacoustic fans, the common wisdom is "never convert lossy to lossy, you will incur generational loss and it will sound bad". Has this been tested?

I've converted some V0 and 320 MP3s to 96k Opus and they sounded fine in casual listening. I'm wondering if modern-ish codecs (MP3 V0, AAC, and Opus) really have this problem. Back in the day I think it was common to convert lossy DTS / AC3 movie audio to 128k MP3 for sharing online. If the problem exists, I wonder if anything-> Opus has the problem. Opus anecdotally has less generation loss for Opus->Opus repeated.

For actual music listening, I think FLAC->lossy->different lossy (2 lossy hops, 1 lossless hop) is a far more realistic situation than repeated encoding of the same format.

Some tests: http://lists.xiph.org/pipermail/opus/2013-December/002477.html

https://hydrogenaud.io/index.php/topic,20625.msg201958.html#msg201958

Re: Lossy to Lossy (generation loss) tests?

Reply #1
These are ten Opus encodings. original -> 01.opus, 01.opus -> 02.opus, 02.opus -> 03.opus, -> 03.opus -> 04.opus, etc.

Encoding using libopus 1.4 (audio)
-----------------------------------------------------
Input: 48 kHz, 2 channels
Output: 2 channels (2 coupled)
20ms packets, 96 kbit/s VBR

Artist: Lagwagon
Song: Surviving California



Re: Lossy to Lossy (generation loss) tests?

Reply #4
Quote
FLAC->lossy->different lossy
I'd say "too many variables" (formats, bitrates, program material).    It you have a specific conversion in mind you could narrow it down.

The thing is...  You know you should avoid loss-to-lossy conversion (whenever possible).   And if you don't have a choice, you don't have a choice, and it's not worth worrying about something you can't avoid.  ;)

Re: Lossy to Lossy (generation loss) tests?

Reply #5
various lossy formats + various Bluetooth codecs - could be interesting if they have some unexpected weaknesses.

Bluetooth codecs sometimes fail transparency on their own if poorly implemented. As an extreme example, I have heard a _really_ broken AAC encoder in an Android smartphone. Most obvious with 'dense' music, like suicidal black metal: it sounds comparable to LAME -V 6 or even worse.

So if you also try to squeeze the file size as much as possible, who knows, maybe there are some unlucky combinations where the 2nd step suddenly makes a difference. But everyone has got more interesting things to do than to test such a huge n. of combinations, I'm pretty sure.
a fan of AutoEq + Meier Crossfeed

Re: Lossy to Lossy (generation loss) tests?

Reply #6
Quote
FLAC->lossy->different lossy
I'd say "too many variables" (formats, bitrates, program material).    It you have a specific conversion in mind you could narrow it down.

I do.
FLAC16->MP3 LAME V0 -> Opus vs. FLAC16 -> Opus

Quote
The thing is...  You know you should avoid loss-to-lossy conversion (whenever possible).   And if you don't have a choice, you don't have a choice, and it's not worth worrying about something you can't avoid.  ;)
I can get MP3s or FLAC from Bandcamp. The FLAC purism of audiophiles generally boils down to
A. higher perceptual quality (lack of transparency in lossy)
B. source for future lossy codecs (desire to avoid having to convert today's lossy codec to tomorrow, introducing generational loss)

A seems disproven except in rare, but definitely existent, killer samples (vs. V0 MP3 or perhaps 192k+ AAC / Opus). I would feel a lot better about B if I knew that "realistic" lossy-> lossy chains don't introduce much perceptual loss. FLAC->MP3->Opus is about the only lossy conversion I think I realistic in this day. So much content is easier to get in MP3 than FLAC. Some places even charge extra for FLAC.

My phone has 64GB of storage. Opus is the last lossy format I need. By the time a new format that's transparent at 64k is invented, our cheapest phones will have 1TB of storage.

Re: Lossy to Lossy (generation loss) tests?

Reply #7
Major digressions from generation loss:

The FLAC purism of audiophiles
Uh. You haven't met audiophile purists. Who insist that WAVE sounds better. Especially untagged.
I think it was B&W (the loudspeaker brand) who had a music store where you would get AIFF, because of customers who insisted that sounded best.
Yes it is possible to "encode" FLAC as only VERBATIM subframes, and one application does that to cater for the audiophools. Acceptable, to the slightly tainted semi-purists.

The true purists will go full denial when they read things like "encapsulating the audio in a WAVE container".
</rant>

generally boils down to
A. higher perceptual quality (lack of transparency in lossy)
B. source for future lossy codecs (desire to avoid having to convert today's lossy codec to tomorrow, introducing generational loss)
The even filthier among us also appreciate
C. a sane file format, natively gapless and checksummed, and where most applications also honour those properties
(... *shrugs* Android's FLAC implementation *shrugs* ... ).
And after the CUETools database, ripping CDs as lossless also means a chance to repair. Although that usually refers to point A, then ... download and listen: https://hydrogenaud.io/index.php/topic,113978.0.html