[stonesfan129]

Regular plain vanilla first generation MP3 (pre-LAME)

Have you done any listening tests using their AAC encoder (Fraunhofer FDK-AAC)? I didn't think it sounded as good as the iTunes one at the lower bitrates and I have to wonder if it has some kind of low-pass filtering going on. At the highest VBR setting (-m5) it says they don't do that, but I swear it doesn't sound as good as the iTunes one at that bitrate (~200KBs). Maybe I'm just imagining things.

 

[bigshot]

I use iTunes to encode AAC. Never used anything else.

 

[gregorio]

I was really surprised because I always remembered 128Kbs MP3 files sounding terrible like I was underwater. Maybe LAME has just gotten that much better over the years, who knows?

The people on Hydrogen Audio know. This is because some of the developers and those who test LAME hang out there and by extension, some of us know because we also frequent or occasionally visit. LAME has certainly developed and gotten better over the years, as testers discovered recordings that caused audible artefacts when encoded and the algorithms were tweaked to reduce/eliminate them.

What you're experiencing could well be a bit of both. Both a bit of mis-remembering and a bit of the actual improvements made to LAME. Are you sure the 128kbps MP3 files you remember sounding terrible were actually LAME encoded? 15-20 years ago, the difference between standard MP3 encoding and LAME encoding was often relatively obvious with many recordings.

Have you done any listening tests using their AAC encoder (Fraunhofer FDK-AAC)? I didn't think it sounded as good as the iTunes one at the lower bitrates and I have to wonder if it has some kind of low-pass filtering going on.

Not recently but many years ago the Fraunhofer AAC encoder certainly could be somewhat deficient compared to Apple's. I doubt, though don't know for sure, that it's as deficient today.

G

 

[Gosod]

I have to listen to some musical artworks in mp3 128!

 

[Àedhàn Cassiel]

Also with MP3, the encoder you use makes a difference. An old encoder won't produce files that sound as good as an up to date one.

I'll admit I could be fooling myself, but I'm still under the impression I can tell the difference between 320 mp3 and FLAC with some files. Very specifically, what I notice is that when a lot of instruments start going at once, the instruments that aren't the "focal point" of the song aren't as clear. Like, drums, keys, two rhythm guitars, and a vocalist can be going at once and the mp3 and FLAC seem indistinguishable to me, but then a lead guitar can come in alongside them all and now the keys and rhythm guitars sound warbly to me. When I listen 'for enjoyment' I really enjoy listening to polyrhythms and the interplay between the rhythms going on different instruments, and I like a lot of music with multiple instruments spewing chaos all at once for that reason (prog metal, math rock, etc.) I did also subjectively notice that newer albums can be indistinguishable anyway where older ones more likely aren't.

I'm assuming the encoder is the most likely explanation for this (I'll be taking the blind test as soon as it's sent to me). That said, how can I tell good from bad encodings if I'm downloading older music from something like Soulseek? Is there a time period for the album release that works as a useful heuristic? And if I want to compress my own FLAC collection to save disc space, what's the best thing to use?

 

[bigshot]

Generally, there are two types of degradation with compressed audio. The first kind is high end roll off. At lower rates, the encoder applies a low pass filter to reduce ultra high frequencies which take up more bandwidth. The higher the data rate, the higher the filter until at a certain point, you should be getting every frequency humans can hear. The other kind of degradation is artifacting. Certain sounds are difficult to encode- massed voices, applause, complex orchestral string textures, etc. If there isn't enough bandwidth to render these sounds, they make a digital splat or gurgle. At very low rates, the digital distortion is easy to hear... as it gets higher, they become less and less frequent until the track achieves complete transparency. That happens at different points with different codecs.

Generally, I've found that people who claim to hear differences in soundstage or clarity not related to artifacting don't test blind. I think those two descriptions are pretty safe to chalk up to bias.

With downloads from sharing it's impossible to know what the data rate or encoders are. Some people are idiots who take 128 MP3s and transcode them to FLAC and then upload that. The download stores state what they use... Apple is AAC 256 VBR and Amazon is MP3 LAME 256 VBR I believe.

Hope this helps!

 

[Blackwoof]

I don't see much talk about Opus 1.3 which sounds fully tranparent at 128kbps?.

 

[bigshot]

From what I'm told, Opus is pretty much the same as AAC. AAC is transparent at 128 for most music. AAC 128 VBR is transparent for most everything.

 

[stonesfan129]

My FiiO M6 doesn't support Opus. I mainly just convert everything to 256k Apple AAC which I cannot distinguish from lossless on it anyway.

 

[castleofargh]

I don't see much talk about Opus 1.3 which sounds fully tranparent at 128kbps?.

It's a topic on mp3, but clearly it has become the grandad of lossy codecs. Anybody trying to achieve even more compression than high bitrate mp3 while expecting good transparency, will have to move on to AAC for compatibility, and opus for performance at really low rates. IMO mp3 has been on the way out for a long time, and having main streaming services and YouTube using other formats is the official burial of mp3.
On a positive side, our old mp3s will still be compatible with almost anything for many years to come, so there is no need to panic if we have some. But I wouldn't advise to keep encoding in mp3 nowadays.

 

[Blackwoof]

From what I'm told, Opus is pretty much the same as AAC. AAC is transparent at 128 for most music. AAC 128 VBR is transparent for most everything.

From my experience Opus at 128 can handle Industrial & harsh experimental like a champ, While AAC/Vorbis struggle hard. Only 1 song from my archive needed to encoded at 384kbps.

Werid having transparency on a Merzbow album at 128kbps.

 

[housekrl]

I have never been able to tell the difference between Flac and 320mp3. But then again, my hearing isn't the greatest. Which is also why I don't care about measurements.

 

[Blackwoof]

I have never been able to tell the difference between Flac and 320mp3. But then again, my hearing isn't the greatest. Which is also why I don't care about measurements.

With some samples/music i can tell 256/320 that also sound fine with 170Kbps musepack??.

 

[bigshot]

It isn't just the data rate. The codec matters too. Frauenhofer MP3 is different than LAME, and both of those are different than AAC. Each one has its own level of transparency. There are also certain sounds that are harder to encode than others. But it is possible to reach total transparency with the big modern codecs.

 

[Blackwoof]

It isn't just the data rate. The codec matters too. Frauenhofer MP3 is different than LAME, and both of those are different than AAC. Each one has its own level of transparency. There are also certain sounds that are harder to encode than others. But it is possible to reach total transparency with the big modern codecs.

MP3 has hard limits with pre echo/hard attacks, But AAC/Ogg can overcome that by using bitrates above 320kbps. But Vorbis is just as good as AAC at 128kbps from my A/B tests but stereo suffers at 80 ~ 96kbps unlike on AAC. Oddly i have like 5+ samples from my ambient stuff where Lame/Vorbis >>> AAC at 128kbps, So i just stick with Q4.4 and >192kbps above for more demading stuff.