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: Has anyone compared FhG and fdk (and qaac?) encoders yet? (Read 14107 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Has anyone compared FhG and fdk (and qaac?) encoders yet?

And by "anyone" I mean someone with the ears and equipment to tell them apart reliably.  Even if I had a Windows machine to test on I don't think I'm up to it, but just from what I have done, the free fdk-aac encoder seems to be extremely good.  My first shock was getting used to how good modern codecs are compared to the mp3 I'm used to (probably CBR and certainly not lame) from about ten years ago.  When I finally got down to 64kbps before hearing anything weird I was starting to think I'd done something terribly wrong.

FDK at 64kbps was acceptable for casual listening, although I could hear defects, while OGG (aoTuv beta 6) and FAAC were pretty bad at that rate.  OGG was much better at 96 and really good at 128, while FAAC needed just a bit more to sound clear to me.  Seems to me that the FDK encoder is the best of those, especially below 100kbps.  I though aoTuv was tuned especially for low bitrates but maybe that's just compared to the original OGG encoder?  Just as a final throwaway, I'm probably not surprising anyone if I say that the Nero AAC encoder doesn't sound quite as good to me as the FDK one at comparable file sizes, although both still amaze me compared to my expectations from ancient mp3s.

Has anyone compared FhG and fdk (and qaac?) encoders yet?

Reply #1 –
Hello! Glad you saw the light, many people still have prejudices towards lossy codecs in 2014 because of what they remember from late 90s/early 2000s about poorly encoded files made by Blade, Xing-old, Plugger etc. and think anything below 192 kbps is "shit". If you are pleasantly surprised by AAC at 64 kbps, try Opus at that bitrate. For my ears at least, it definitely beats 128-160 kbps mp3s made by Blade, Xing, l3enc etc. and in fact sounds quite "hi-fi" down to 48 kbps. You can play it back by Foobar2000 and encode by LameXP if you want a more friendly interface than commandline.

Samples:

Original sample (I intentionally took it from a dynamic 80s song full of percussion so possible artifacts will not be masked by brickwall dynamic compression of modern CDs):
http://www.sendspace.com/file/hpfa1y

Opus ~64 kbps VBR:
http://www.sendspace.com/file/nojd5h

BladeEnc mp3 128 kbps:
http://www.sendspace.com/file/ea532g

Back to original question fdk aac encoder = FhG aac encoder, it is the same thing. It is a good encoder, but the reason why FAAC is doing so badly is because it is a quite bad aac encoder. AFAIK it is about equal to LAME mp3s in quality, perhaps a bit lower, which is striking considering aac has many technological advantages over mp3.

EDIT - Oh, and to prevent any beating over the head with TOS8, here are my ABX Opus 64 kbps vs Blade results, and yes, I found the Opus version better:
foo_abx 1.3.4 report
foobar2000 v1.3
2014/02/28 15:14:09

File A: C:\Users\pc\Desktop\sample.opus
File B: C:\Users\pc\Downloads\blade\sample.mp3

15:14:09 : Test started.
15:14:27 : 01/01  50.0%
15:14:58 : 02/02  25.0%
15:15:17 : 02/03  50.0%
15:15:38 : 03/04  31.3%
15:15:54 : 04/05  18.8%
15:16:17 : 05/06  10.9%
15:16:34 : 06/07  6.3%
15:17:13 : 07/08  3.5%
15:17:47 : 08/09  2.0%
15:18:25 : 09/10  1.1%
15:18:49 : 10/11  0.6%
15:19:05 : 11/12  0.3%
15:19:21 : 12/13  0.2%
15:19:35 : 13/14  0.1%
15:19:52 : 14/15  0.0%
15:19:54 : Test finished.

----------
Total: 14/15 (0.0%)

It is both funny and sad how people ate up Microsoft's claims about 64 kbps wma being equal to 128 kbps mp3, yet many people seem to react to Opus in the "meh we already have mp3 and 64 kbps will always sound shit" way.

Has anyone compared FhG and fdk (and qaac?) encoders yet?

Reply #2 –
Oh, and here are LAME 64 kbps mp3 (ABR) and to horrify people, 64 kbps Blade samples  (both stereo just like the Opus sample) :

64 kbps LAME, if you think that is bad, listen to the next sample...
http://www.sendspace.com/file/k8jl0d

64 kbps Blade mp3 - total audio collapse
http://www.sendspace.com/file/m7q5k3

Has anyone compared FhG and fdk (and qaac?) encoders yet?

Reply #3 –
Why would you ABX two lossy encodes against each other? Doesn't that just prove that at least one of them sounds different from the original? Shouldn't you use ABC/HR instead?

I'm not challenging you, just curious.

Has anyone compared FhG and fdk (and qaac?) encoders yet?

Reply #4 –
I don't know about other methods than ABX, sorry.... but I definitely hear a difference and you can listen for yourself that the Blade 128 kbps encode is worse.

Has anyone compared FhG and fdk (and qaac?) encoders yet?

Reply #5 –
Quote
think anything below 192 kbps is "shit"

That's certainly what I was expecting.

Those files are horrible!  Especially if you play the 20kHz version at 44.1kHz

I don't really need to transcode down to 64kbps, but I'll give Opus a try and see what it sounds like.  Anything that I can just about detect issues with on my computer speakers isn't good enough when I actually play it on other devices.  I'm also trying to avoid AAC if I can.  Ogg compared favourably with Nero AAC and I was happy with that, but I'm not so sure with FDK.  Certainly not so good at 64kbps, but the two seem closer at higher bit rates.  Seems like Ogg quality drops rapidly from about 96kbps downwards.

Has anyone compared FhG and fdk (and qaac?) encoders yet?

Reply #6 –
Horrible? The mp3 samples certainly are, but I would not call the Opus sample at 64 kbps "horrible". And what do you mean by playback of 20 Khz at 44 Khz?

I tried both LC and HE AAC using iTunes and both sounded inferior to Opus for me at 64 kbps.

Has anyone compared FhG and fdk (and qaac?) encoders yet?

Reply #7 –
I don't know what happened there.  When I downloaded before, the first link gave me a file called sample64LAME.mp3 and it was a 24kHz file, which I inadvertently played at near double speed.  Funny!  Even at the right speed it sounds bad:
http://www.sendspace.com/file/k8jl0d

Now your post contains links to a FLAC file, an Opus file, and sample64.mp3 which was there before and sounds awful.  I encoded some opus tracks of my own from samples I'm familiar with, and possibly a little better than the FDK AAC.  Close though.  I wouldn't like to commit myself without listening to more samples with more care.  It's a moot point at the moment, not enough Opus hardware support yet.  Just being able to play it on Linux isn't the purpose of lossy codecs for me.

Seems like these newest codec advances have stepped things up at that 64kbps point whereas they are much closer (to things like Ogg Vorbis, Lame MP3, etc.) at the 100-150kbps I'll probably settle on for transcoding.  I did create a bunch of tracks at Ogg q6 (VBR around 192kbps) which I thought would be "sufficient" for my needs, but eventually came to realise I just couldn't detect anything at all off at that rate even on much better hardware than my computer, so here I am doing more research ...

Has anyone compared FhG and fdk (and qaac?) encoders yet?

Reply #9 –
Quote from: no404error on
Quote from: Neuron on
Back to original question fdk aac encoder = FhG aac encoder, it is the same thing.

?


https://www.google.sk/search?q=fdk+aac+enco...22&ie=UTF-8

It seems to be a FhG codec for Android.

lithopian, how would you describe the quality of the Opus sample I uploaded? I was honestly surprised when I first tried Opus.

The same sample in 48 kbps and 128 kbps Opus.
http://www.sendspace.com/file/yof6ga 48 kbps
http://www.sendspace.com/file/o2k80b 128 kbps

Has anyone compared FhG and fdk (and qaac?) encoders yet?

Reply #10 –
Quote from: Neuron on
https://www.google.sk/search?q=fdk+aac+enco...22&ie=UTF-8

It seems to be a FhG codec for Android.


fdk-aac is an open source library derived from the Android library that was derived from FhG.  It is widely available on Linux platforms.  Obviously it is pretty similar to FhG but I don't think it is identical.  I just wondered if a proper comparison had been made.  I see some scope views that identify some differences but not necessarily anything that would be audible.  Given that the FhG codec in Winamp rates very highly, I was curious whether FDK is likely to be comparable.

Quote from: Neuron on
lithopian, how would you describe the quality of the Opus sample I uploaded? I was honestly surprised when I first tried Opus.

The same sample in 48 kbps and 128 kbps Opus.
http://www.sendspace.com/file/yof6ga 48 kbps
http://www.sendspace.com/file/o2k80b 128 kbps


The 64kbps sounds OK.  The 48kbps sounds OK too, also obviously not quite so good.  I could listen to it.

I played these samples to my parents today, just for fun.  Nothing, couldn't tell the difference between any of the Opus versions and the original.  They struggled to notice any difference with the Blade MP3!  Amazing ...

Has anyone compared FhG and fdk (and qaac?) encoders yet?

Reply #11 –
Try ABXing the 128 kbps Opus file against the original and the 48 kbps file against the 64 kbps one. You might be surprised. Blade, well this is probably the least bad sounding Blade 128 kbps sample ever, try encoding Metallica or some electronica with Blade mp3 encoder and weep and laugh...

I uploaded a 64 kbps VBR Winamp FhG HE-AAC sample and 64 kbps CBR AAC-LC (I couldn't choose AAC-LC at 64 kbps VBR in the menu).
If fdk aac switches to HE-AAC at low bitrates just like Winamp VBR AAC mode, then you got a HE-AAC file at 64 kbps. It downsamples the file to 22 Khz then synthetizes the higher frequencies from the lower ones, as in real life they are usually related. This is called SBR and brings good sound at low bitrates, but in my opinion Opus handles low bitrates in a more graceful way, as it has band folding that is a form of bandwidth extension but less aggresive than SBR, as SBR basically guesses one half of the audible spectrum, while band folding tries to save band energy in the higher frequencies more thoroughly. Opus always operates at 48 Khz, even at 6 kbps, it just adjusts the bandwidth from 4 Khz at 6-8 kbps, 6 khz at 14-16 kbps, 8 Khz at 22-24 kbps, 12 Khz from 32 kbps to 40 kbps and then finally 20 Khz from 48 kbps until the 510 kbps max bitrate. It is mono until 40 kbps or so, all higher bitrates are stereo if the source is stereo.

Here's the link:

http://www.sendspace.com/filegroup/qHxYDDtL7J2yNFsiBnxB9Q

Has anyone compared FhG and fdk (and qaac?) encoders yet?

Reply #12 –
I don't think the fdkaac encoder in Linux automatically selects HE-AAC or automatically uses SBR at low bitrates, although I haven't checked it thoroughly.

Although I've seen this statement that Opus always operates at 48kHz, it does actually also support 24kHz and some lower rates too, but the libopus library that is by far the best way to decode these files only supports 48kHz.  I've never seen an Opus file at any other rate.

I have seen stereo streams at around 25kbps (spoken word, very good quality really).  At least they claim to have two channels, they are probably tightly coupled in the actual encoding.  Again it may just be a choice of the encoder to use mono for very low bitrates, rather than imposed by the codec.  I know that the bandwidth restrictions can be overridden if the encoder wants to, although it isn't recommended.

Has anyone compared FhG and fdk (and qaac?) encoders yet?

Reply #13 –
Quote
Although I've seen this statement that Opus always operates at 48kHz, it does actually also support 24kHz and some lower rates too, but the libopus library that is by far the best way to decode these files only supports 48kHz. I've never seen an Opus file at any other rate.

I have used libopus before and could set up the decoder to decode at 48Khz, 24Khz, 16Khz, 12Khz and 8Khz. I don't see any reason to decode at any other rate than 48Khz unless your hardware doesn't support a sampling rate as high as 48Khz.

Has anyone compared FhG and fdk (and qaac?) encoders yet?

Reply #14 –
Quote from: LithosZA on
Quote
Although I've seen this statement that Opus always operates at 48kHz, it does actually also support 24kHz and some lower rates too, but the libopus library that is by far the best way to decode these files only supports 48kHz. I've never seen an Opus file at any other rate.

I have used libopus before and could set up the decoder to decode at 48Khz, 24Khz, 16Khz, 12Khz and 8Khz. I don't see any reason to decode at any other rate than 48Khz unless your hardware doesn't support a sampling rate as high as 48Khz.

Agreed.  Plus the opusfile library doesn't even let you pick anything except 48kHz.

Has anyone compared FhG and fdk (and qaac?) encoders yet?

Reply #15 –
Well I've listened to an awful lot of Opus over the last few weeks, mostly at 64kbps.  Mostly acceptable for casual listening, although a few tracks are unpleasant.  That's just random music, not particular difficult pieces.  At 128kbps I can listen to anything except known "judgement" pieces, and I find it hard work to detect individual audible errors.  One stream I've tested with servers 256kbps Opus and that is beyond transparent for me.