IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Removing frequencies for low bitrate transcoding
subinbar
post Oct 14 2012, 06:08
Post #1





Group: Members
Posts: 29
Joined: 10-June 11
Member No.: 91381



Would it be effective to remove certain frequencies from an audio file in order to make it more compressible?

I am currently encoding some of my movies to a portable format, and I'm looking to save space where I can. I'd like to transcode the audio tracks at 32-48kbs with QuickTime HEv1 AAC. (Of course I'd like to use Opus, but compatibility is basically non existent.)

I've done some listening tests with QuickTime vs FhG HE, and my conclusion is that I prefer QuickTime. QuickTime seems to include more high frequencies, while FhG seems to eliminate more. However sometimes I believe QuickTime includes too many (or too high) frequencies that are not able to fit into such small bandwidth, resulting in that "underwater" flanging type sound, similar to low bitrate mp3's. It usually happens with sudden high sounds, such as crashing or cymbals in music.

At these low bitrates it looks like QuickTime applies a hard low-pass at about 15khz.

I'm wondering if I can identify the specific frequency range(s) that these artifacts occur in, and remove or lessen this range, if it will help lessen these resulting artifacts and lessen the complexity of the audio. Similar to de-noising a video clip.

Is this a sound theory, or would removing these frequencies do nothing to reduce the complexity of the audio, and therefore the bitrate requirements? Am I better off reducing the bit-depth instead? In my sound tests, this seemed less effective and made the audio sound too dull.

Thanks.
Go to the top of the page
+Quote Post
saratoga
post Oct 14 2012, 07:37
Post #2





Group: Members
Posts: 4866
Joined: 2-September 02
Member No.: 3264



Lossy files don't have a bit depth, so decreasing that isn't an option.

Low passing does help to some extent, and is pretty common to reduce bitrate with mp3. However aac he is a little weird because of the tricks it uses to reconstruct higher frequencies. I guess try it and see if it helps.
Go to the top of the page
+Quote Post
subinbar
post Oct 14 2012, 08:18
Post #3





Group: Members
Posts: 29
Joined: 10-June 11
Member No.: 91381



Well after testing a few scenarios, it seems as though the frequency removal would have be done by the encoder itself. Removing frequencies and setting a lowpass before conversion seems to give QuickTime less information and the resulting transcode sounds worse.

Removing frequencies after conversion was somewhat effective - I was able to carve out some nasty frequencies and clean up the sound a bit. However this doesn't really help unless the encoder can perform the actions itself.

Maybe some day encoders will be advanced enough to allow custom instructions on which frequencies to keep.
Go to the top of the page
+Quote Post
lvqcl
post Oct 14 2012, 10:03
Post #4





Group: Developer
Posts: 3339
Joined: 2-December 07
Member No.: 49183



I encoded a test signal (sweep sine) at 32 kbps. It's easy to see that both FhG and Apple AAC encode only frequencies lower than 5500...6000 Hz. Frequencies higher than this value are reconstructed.

This post has been edited by lvqcl: Oct 14 2012, 10:22
Go to the top of the page
+Quote Post
subinbar
post Oct 14 2012, 10:25
Post #5





Group: Members
Posts: 29
Joined: 10-June 11
Member No.: 91381



QUOTE (lvqcl @ Oct 14 2012, 11:03) *
I encoded a test signal (sweep sine) at 32 kbps. It's easy to see that both FhG and Apple AAC encode only frequencies lower than 5500...6000 Hz. Frequencies higher than this value are reconstructed.


Very interesting... thanks.
Go to the top of the page
+Quote Post
slks
post Oct 14 2012, 21:41
Post #6





Group: Members
Posts: 389
Joined: 31-March 06
From: Houston, Texas
Member No.: 29046



Indeed - what complicates this with HE-AAC is that it uses SBR, spectral band replication. The frequencies above a certain threshold (6 kHz sounds right) aren't encoded normally, they're just reconstructed on playback.

If you tell an HE-AAC encoder to lowpass at 8 kHz for example - you wouldn't see a difference, since it's already lowpassed at 6 khz, and everything above that is reconstructed in the decoding stage.

SBR certainly doesn't sound identical to the original signal. That's why it only kicks in when you get to "High Efficiency AAC" - at 48 kb/s we're not even dreaming of encodes sounding identical, but which encodes sound better. And I think we can all agree that HE-AAC wins over lowpassed MP3 or LC-AAC, hands down, at these bit rates.


--------------------
http://www.last.fm/user/sls/
Go to the top of the page
+Quote Post
C.R.Helmrich
post Oct 14 2012, 22:26
Post #7





Group: Developer
Posts: 686
Joined: 6-December 08
From: Erlangen Germany
Member No.: 64012



QUOTE (subinbar @ Oct 14 2012, 07:08) *
Would it be effective to remove certain frequencies from an audio file in order to make it more compressible?

FhG's encoder already removes the frequencies it can't encode well, or reconstructs them parametrically using SBR. I assume Apple's encoder does the same. So the answer is no.

What you can do is to remove stereo Information, e.g. by downmixing to mono. But even that you don't need to do if you're using HE-AAC v2. In which configuration did you test FhG's encoder? VBR 1 or CBR? HE or HEv2?

Chris


--------------------
If I don't reply to your reply, it means I agree with you.
Go to the top of the page
+Quote Post
subinbar
post Oct 15 2012, 01:33
Post #8





Group: Members
Posts: 29
Joined: 10-June 11
Member No.: 91381



QUOTE (C.R.Helmrich @ Oct 14 2012, 23:26) *
QUOTE (subinbar @ Oct 14 2012, 07:08) *
Would it be effective to remove certain frequencies from an audio file in order to make it more compressible?

FhG's encoder already removes the frequencies it can't encode well, or reconstructs them parametrically using SBR. I assume Apple's encoder does the same. So the answer is no.

What you can do is to remove stereo Information, e.g. by downmixing to mono. But even that you don't need to do if you're using HE-AAC v2. In which configuration did you test FhG's encoder? VBR 1 or CBR? HE or HEv2?

Chris


VBR + HE v1. I haven't done much with HEv2; was never really impressed by it.
Go to the top of the page
+Quote Post
C.R.Helmrich
post Oct 15 2012, 09:26
Post #9





Group: Developer
Posts: 686
Joined: 6-December 08
From: Erlangen Germany
Member No.: 64012



QUOTE (subinbar @ Oct 15 2012, 02:33) *
VBR + HE v1. I haven't done much with HEv2; was never really impressed by it.

VBR 1 is fixed to HEv2. If you don't like the stereo artifacts, I suggest you try CBR at 44 or 48 kbps. At least one of those should give you "true" non-parametric stereo.

Chris


--------------------
If I don't reply to your reply, it means I agree with you.
Go to the top of the page
+Quote Post
subinbar
post Oct 15 2012, 09:40
Post #10





Group: Members
Posts: 29
Joined: 10-June 11
Member No.: 91381



QUOTE (C.R.Helmrich @ Oct 15 2012, 10:26) *
QUOTE (subinbar @ Oct 15 2012, 02:33) *
VBR + HE v1. I haven't done much with HEv2; was never really impressed by it.

VBR 1 is fixed to HEv2. If you don't like the stereo artifacts, I suggest you try CBR at 44 or 48 kbps. At least one of those should give you "true" non-parametric stereo.

Chris


Oh, I thought since the bitrate varied it was VBR... I will give that a shot and see how it sounds. Thanks!
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 30th July 2014 - 22:47