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: Spectral Band Replication (Read 4955 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Spectral Band Replication

I do not wish to create a flame war on Spectral Band Replication
these are just my views

AS mp3 was out a new codec called mp3pro came out i learnt it had SBR
Spectral Band Replication is where it does'nt encode the high frequencys
In a file but leaves data about it.So when it gets decoded it reconstucts
the high Freq.SBR is also implented in AAC.

what i hate about

1.what is the point of EAC, SBR it reconstructs it so it will be artificial,I know i don't whant to have artificial audio,frequency looks like it has had a lowpass in spectral viewing.
2.not so bad.Well one thing that i hate about SBR is.have the original wave file then encode it into AAC with HE at 128k and below.do A ABX test and you can cleary hear the artifacts ,no it doesn't ring like lame but ,(hard to desribe) sounds smeared,powdery,and dry like the quality of the audio has been stripped down.
3.unfair on codecs like ogg 

You can hear it on a song with bass it sounds smeared and dry?

Spectral Band Replication

Reply #1 –
Quote
1.what is the point of EAC, SBR it reconstructs it so it will be artificial,I know i don't whant to have artificial audio,frequency looks like it has had a lowpass in spectral viewing.


Not sure what you mean here.

Quote
2.not so bad.Well one thing that i hate about SBR is.have the original wave file then encode it into AAC with HE at 128k and below.do A ABX test and you can cleary hear the artifacts ,no it doesn't ring like lame but ,(hard to desribe) sounds smeared,powdery,and dry like the quality of the audio has been stripped down.


HE AAC at 128 kbps? Which program did you use to encode that?

Quote
3.unfair on codecs like ogg


So what?

Quote
You can hear it on a song with bass it sounds smeared and dry?


The aim of HE AAC is not to sound transparent, but to offer an acceptable quality for streaming or where file size must be kept low.

Spectral Band Replication

Reply #2 –
Quote
I know i don't whant to have artificial audio

Then you probably do not want do use ANY lossy compression. The goal of lossy audio compression is to SOUND good to your ears, whatever the tools used.

Spectral Band Replication

Reply #3 –
>>HE AAC at 128 kbps? Which program did you use to encode that?
Pure technically you should be able to obtain it from Nero encoder, when using external frontend, but I never tried to do it myself...

Spectral Band Replication

Reply #4 –
i used the nero encoder but compared to mp3 it sounds dry and smeared.
By artificial i mean cause it has fake high frequency's via reconstruction.By that it is supposed to keep the high frequency's where as spectral viewing shows a lowpass.I would perfer musepack at 3.5 anyday.

Please just trust me and convert a wave file to 96k HE via nero then abx.

Spectral Band Replication

Reply #5 –
at 128, LC-AAC should be perfectly acceptable.  possibly even transparent on simple music.  no need for SBR there.  just my 2 cents...

HE isn't designed for transparency (so ABX isn't all that useful in this case), it's designed to be _bearable_ at bitrates that would otherwise break the encoder.

i don't get what you mean about "fake" high frequencies.  every lossy encoder will trash the input data and use lossy methods to reconstruct it.  SBR just uses a much more radical approach.  sure, a mDCT will allow perfect reconstruction, but that's only if there's no quantization going on.  as soon as you do that, the output is going to be "fake" in the same way that 3 does not equal pi, and as such you're going to come up short when you try to wrap it around a circle.

Spectral Band Replication

Reply #6 –
Are you sure your decoder is supporting HE-AAC and not just LC-AAC?

Spectral Band Replication

Reply #7 –
One example, but I'd like to have some confirmation.

Could someone test this sample? It's very short (3 seconds) and I've also included encoded version.

In my opinion, lame mp3 at 96 kbps sounds good, whereas HE-AAC at the same bitrate is very coarse ("powdery" is maybe the exact term I was looking for). Many violins (or high tonal instruments/signal) suffers from this strange grainy sound with HE-AAC, even at 96 kbps.
Wavpack Hybrid -c4hx6

Spectral Band Replication

Reply #8 –
1) You should not confuse the limitations of our current encoder with a fundamental limitation of HE-AAC. There is still tremendous improvement possible, but this takes time.

2) IMHO 96kbps is not a bitrate where HE-AAC will really shine. I would strongly consider using LC-AAC at this bitrate.

3) The Nero encoder WILL NOT use HE-AAC at 96kbps or more unless forced to.

"Doctor it hurts when I push here." Then don't push there!

Spectral Band Replication

Reply #9 –
1/ I know it
2/ I agree. But LC-AAC at 96 kbps is not very good, and has other problems. With some samples, HE profile seems to be better.
3/ "Will"? but currently, 96 kbps CBR with HE-AAC is an option with Nero AAC for long time, still available with the latest one.

N.B. Some people are looking for 128 kbps and more HE-AAC encoding. I saw it on doom9 recently. This example (limited of course) might give a proof that HE-AAC is not a heavenly encoder, and has some flaws.
Wavpack Hybrid -c4hx6

Spectral Band Replication

Reply #10 –
It is not used by default, and must be explicitly enabled. The profiles don't use it, either. So it won't use it unless you force it to, which is *exactly* what I said.

HE-AAC at 128kbps makes no sense at all for stereo, those people are wrong.

Of course multichannel encodes are another matter.

Spectral Band Replication

Reply #11 –
By "forced" I meant something else (an option not allowed by current GUI). Sorry.

The message is doom9.org is here (next to last message of the page).


EDIT: link corrected.
Wavpack Hybrid -c4hx6

Spectral Band Replication

Reply #12 –
Quote
1.what is the point of EAC, SBR it reconstructs it so it will be artificial,I know i don't whant to have artificial audio,frequency looks like it has had a lowpass in spectral viewing.

I may be able to live with SBR artifacts, but I don't want skips and blips in my audio.

Quote
2.not so bad.Well one thing that i hate about SBR is.have the original wave file then encode it into AAC with HE at 128k and below.do A ABX test and you can cleary hear the artifacts ,no it doesn't ring like lame but ,(hard to desribe) sounds smeared,powdery,and dry like the quality of the audio has been stripped down.

As others have pointed out, SBR targets lower bitrates than 128k. 64k with SBR sounds much better than 64k without. Goal achieved.

Quote
3.unfair on codecs like ogg

Heh, and OGG reduces stereo. Is that fair?
OGG Vorbis is more efficient than MPEG Layer 2. Unfair! At low bitrates you just have to compromise. SBR is just another tool.

Spectral Band Replication

Reply #13 –
Quote
In my opinion, lame mp3 at 96 kbps sounds good, whereas HE-AAC at the same bitrate is very coarse ("powdery" is maybe the exact term I was looking for). Many violins (or high tonal instruments/signal) suffers from this strange grainy sound with HE-AAC, even at 96 kbps.
[a href="index.php?act=findpost&pid=269113"][{POST_SNAPBACK}][/a]


I see you can tell AAC HE sounds dry and powdery.But i just did a abx test with latest nero encoder at LC-128k vs original and it still sounds a bit powdery,dry and smeared though i do remember testing quicktime's encoder and it sounded fine.I can also notice the powdery artifact in mp3pro.

Spectral Band Replication

Reply #14 –
Quote
I can also notice the powdery artifact in mp3pro.
[a href="index.php?act=findpost&pid=269326"][{POST_SNAPBACK}][/a]

I'm sharing the same feeling. I've found exactly the same kind of artifacts with mp3pro and HE-AAC. During Roberto's 64 and 32 kbps listening test, it wasn't hard to guess which encoder was SBR-based, and which one wasn't.
I wonder if this kind of problem will disappear with further tuning, or if it's a structural one, as a reverse side of SBR tool.
Wavpack Hybrid -c4hx6

Spectral Band Replication

Reply #15 –
Double Post

Spectral Band Replication

Reply #16 –
Quote
Quote
I can also notice the powdery artifact in mp3pro.
[a href="index.php?act=findpost&pid=269326"][{POST_SNAPBACK}][/a]

I'm sharing the same feeling. I've found exactly the same kind of artifacts with mp3pro and HE-AAC. During Roberto's 64 and 32 kbps listening test, it wasn't hard to guess which encoder was SBR-based, and which one wasn't.
I wonder if this kind of problem will disappear with further tuning, or if it's a structural one, as a reverse side of SBR tool.
[a href="index.php?act=findpost&pid=269367"][{POST_SNAPBACK}][/a]


I think the artifact is structual because if you listen to a 80k ogg you can just hear it but in nero acc you can hear it even at 128k but when swithching to HE you can hear the artifact way more easilly.musepack and mp3 doesn't seem to have it but wma has some big problems.