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: Anyone did some transcoding tests? (Read 3060 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Anyone did some transcoding tests?

The trend these days seems to be:

- I like mpc but I won't use it becose I need to transcode for a hardware mp3-only player.

- I also like ogg but I'll wait until 1.0 is released and hardware support comes around.

- I don't like how mp3 sounds , but I use it sometimes for my hardware player...

- I don't know much about AAC, so I don't use it.

The solution would be lossless

But lossless eats up a lot of space... what if  I have 500+ Albums and want to have them all in one place to listen to? So I'm back to mpc --xtreme :confused:

My questions would be:

Is ANY format "suitable" for transcoding? (by suitable I understand with losses of quality that are not so obvious like in mp3-to-anything).

What's the best format for transcoding?

OK. I use only mpc --xtreme anyway. Just curious...

Anyone did some transcoding tests?

Reply #1 –
It's funny you should bring up this topic because I was fooling around with this issue myself today.

If you consider that lossy formats supported by hardware players are transform encoder-based, then a subband encoder like MPC may be the best lossy source format for transcoding.

I performed some VERY preliminary testing and found that a high-bitrate MPC file can be transcoded to LAME --alt-preset standard without much measurable quality loss. Using EAQUAL, I analyzed a LAME aps file, a MPC xtreme file, and a MPC xtreme => LAME aps file. In this case, the transcoded file was measurably worse than either of the two first-generation files. (ODG was -0.10, -0.05 and -0.48, respectively). However, I made a MPC xtreme -nmt 16 -tmn 32 => LAME aps file. The EAQUAL OGD value for this file was -0.15, which seems very good for a second-generation file.

Now, EAQUAL is no panacea...this was just a preliminary test to satiate my curiousity. However, it seems to me that a key value to look at is the NMR (noise-to-mask ratio). The LAME aps and MPC xtreme files had a NMRs of -15.0 and -15.5, respectively. The MPC xtreme => LAME aps file had an NMR of -12.6, displaying reduced resolution and quality. However, the MPC xtreme -nmt 16 -tmn 32 => LAME aps file had an NMR of -14.5, which is pretty decent...and the reason for this is that the original MPC file had an NMR of -21.6.

If you plan on doing some transcoding, I have a feeling that a high-bitrate MPC file is the best lossy source. Use the --nmt and --tmn switches to increase headroom for the transcoding process.

I may have more data for this discussion at a later date.

Anyone did some transcoding tests?

Reply #2 –
Quote
Originally posted by mithrandir
If you plan on doing some transcoding, I have a feeling that a high-bitrate MPC file is the best lossy source. Use the --nmt and --tmn switches to increase headroom for the transcoding process.


Have you tested this using MPC --insane as well?  I'm curious what the results would be...

Anyone did some transcoding tests?

Reply #3 –
Here are some ODG values I generated with a different test WAV (the last minute of Duncan Sheik's "A Body Goes Down", which has some mellow acoustic guitar strumming and voice). It appears that OGG has an edge in transcoding, though it has been said that EAQUAL seems to have some bias toward OGG in its results (something about not being sensitive enough to pick up pre-echos, which are common artifacts of transform encoders). But anyway, these results generally look pretty good, i.e. second-generation files are good quality and comparable to a first-generation 160kbps CBR file.

LAME -b 128 -q2: -0.99
LAME -b 160 -q2: -0.32

LAME aps (181kbps): -0.12
MPC xtreme (210kbps): -0.05
MPC xtreme nmt 16 tmn 32 (252kbps): 0.00
MPC insane (264kbps): -0.07
MPC insane nmt 16 tmn 32 (293kbps): -0.02
OGG -q7 (199kbps): -0.08
OGG -q9 (284kbps): 0.11

MPC xtreme => LAME aps: -0.46
MPC xtreme nmt 16 tmn 32 => LAME aps: -0.36
MPC insane => LAME aps: -0.35
MPC insane nmt 16 tmn 32 => LAME aps: -0.31
OGG -q7 => LAME aps: -0.40
OGG -q9 => LAME aps: -0.18

Anyone did some transcoding tests?

Reply #4 –
Quote
Originally posted by mithrandir
Here are some ODG values I generated with a different test WAV... 

Thanks for the numbers. Although it may be hard to compare different codecs using EAQUAL (as shown by heated debates earlier on this board), it seems fairly useful for comparing similar encodes, or at least it would be "close enough" to the truth to keep me happy

I'd like to do some tests of my own, but I'm not sure how to obtain an EAQUAL executable anymore. Can anyone point me in the right direction? I know mp3-tech was going to host the source in the near future...

Anyone did some transcoding tests?

Reply #5 –
I' wondering what do you mean for "like".
The point is NOT like or dislike. The point is if a format is TRANSPARENT or not, for you.
Did you ever do a blind test?

Anyone did some transcoding tests?

Reply #6 –
Quote
I' wondering what do you mean for "like".


Like or dislike... hmm... means that certain ppl here would choose between let's say ogg -q 6 and mpc --standard because they prefer one format over the other.

The above are both transparent for me... but I choose mpc --xtreme (not even --standard) just to feel safe... (that means I like it better). And by safe I understand that I won't have to reencode everything when RC4 comes out or when I change my audio setup.

BTW: bought a 24/96 card and this made me ditch some of the ogg (~128 kbps though) files I encoded, that were fine on a Sb Live...

So "like" is feeling secure about some lossy format you choose to encode in.