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: 32000Hz MP3 sounds worse than 44100Hz (Read 2597 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

32000Hz MP3 sounds worse than 44100Hz

Hello. I noticed that 32000Hz 96kbps stereo MP3 sounds worse than 44100Hz 96kbps stereo MP3 with this sample (Erase/Rewind by The Cardigans, I currently listened encodings of only this and Sonic Boom from Sonic CD). I'd expect the opposite to happen because 32kHz allows for longer MDCT blocks (in real-world time) and they are much shorter than ideal in 44100Hz MP3s. They both use the same cutoff point and they both should have their sfb21 empty according to this (it was 15kHz). I attached the samples and the ABX report (between two lossies to prove that they are audibly different). Why could this be happened?

Re: 32000Hz MP3 sounds worse than 44100Hz

Reply #1 –
I can't tell the difference, but I don't claim my hearing is any great shakes.
It's your privilege to disagree, but that doesn't make you right and me wrong.

Re: 32000Hz MP3 sounds worse than 44100Hz

Reply #2 –
I detect it at third and fourth snares, maybe that helps.

Re: 32000Hz MP3 sounds worse than 44100Hz

Reply #3 –
Both sound bad but the 32 kHz sounds worse. I remember from the old days that at least LAME was only tuned for 44.1 kHz and everything else was said to perform worse.
Are these encoded with the same tool? I wonder because the lengths don't match and they have an offset error.

Re: 32000Hz MP3 sounds worse than 44100Hz

Reply #4 –
Quote from: Case on
Both sound bad but the 32 kHz sounds worse. I remember from the old days that at least LAME was only tuned for 44.1 kHz and everything else was said to perform worse.
Are these encoded with the same tool? I wonder because the lengths don't match and they have an offset error.

@Case Both are encoded with FhG MP3Enc.

I should say that I now listened them with headphones and I detect much more artifacts and interestingly don't detect the artifacts that I used to ABX them more strongly and can't decide which one is better when I use headphones. What about you?

And, what is an offset error?

Re: 32000Hz MP3 sounds worse than 44100Hz

Reply #5 –
I always use headphones when testing something. Speakers hide details. The 32 kHz variant has more artifacting going on in the background, and not just in any specific instrument. Both sound like they have unintentional stutter but the 32 kHz version even more.

Offset error is a timing difference in the audio data. An example ASCII presentation of a single character offset difference:
Track 1: \_/\_/\__/
Track 2: /\_/\_/\__
It's not audible on its own of course, but when comparing two tracks their timings should be identical. Otherwise you may mistake hearing parts of different instruments as a quality difference. Also it can ruin ABX, you may for example simply test a part that is completely silent in one track and offset error lets you hear last or initial samples of the signal in the other track.

That FhG encoder would never by my choice because of the length and offset issues alone. MP3 has valid methods to ensure such issues don't happen. Both LAME and Helix for example are able to not mess things up.

Re: 32000Hz MP3 sounds worse than 44100Hz

Reply #6 –
Quote from: Case on
I always use headphones when testing something. Speakers hide details. The 32 kHz variant has more artifacting going on in the background, and not just in any specific instrument. Both sound like they have unintentional stutter but the 32 kHz version even more.

Offset error is a timing difference in the audio data. An example ASCII presentation of a single character offset difference:
Track 1: \_/\_/\__/
Track 2: /\_/\_/\__
It's not audible on its own of course, but when comparing two tracks their timings should be identical. Otherwise you may mistake hearing parts of different instruments as a quality difference. Also it can ruin ABX, you may for example simply test a part that is completely silent in one track and offset error lets you hear last or initial samples of the signal in the other track.

That FhG encoder would never by my choice because of the length and offset issues alone. MP3 has valid methods to ensure such issues don't happen. Both LAME and Helix for example are able to not mess things up.

I understand. I also heard a lot of artifacts in 112, 128, and 144 kbps encodings of this track (same encoder) when I used headphones and I think this means the beginning of this track is a killer sample (I didn't try 160 and beyond because MP3Enc doesn't use joint stereo with these bitrates).

I will continue judging encodings and I will also try Fastenc (I wonder about its stereo seperation bug, I never listened to it with headphones), MP3 Surround, and Helix. I hope I'll be able to do an ABC-HR test (it's not as easy as it seems for me). For now, I think the minimum bitrate that is required for me to hear very few or no artifacts is 112 or 96 kbps for MP3.

I also want to add that Adobe Flash is able to make Flash Player play MP3Enc-encoded MP3's gaplessly.

Re: 32000Hz MP3 sounds worse than 44100Hz

Reply #7 –
I have the album, I just tested the track with Helix and LAME. LAME beats your FhG encodes as it doesn't introduce the weird stuttering nonsense, but Helix gives the best quality. None of these sound good at such low bitrate. MP3 just doesn't have the tools to encode this when you set the bitrate so low.

Re: 32000Hz MP3 sounds worse than 44100Hz

Reply #8 –
I don't know why @Klymins doesn't just use 128 kbps at 44.1 KHz?  Some PC games I have (Half-Life 2 and it's episodes) and SimCity 4 Deluxe use that exact bitrate.

Re: 32000Hz MP3 sounds worse than 44100Hz

Reply #9 –
Quote from: Chibisteven on
I don't know why @Klymins doesn't just use 128 kbps at 44.1 KHz?  Some PC games I have (Half-Life 2 and it's episodes) and SimCity 4 Deluxe use that exact bitrate.
From what can be inferred from his previous posts,  the gentleman appears to be quite the distinguished connoisseur of exotic CODECs* - or something like that.

* Sorry, I've been re-watching The Bleak Old Shop of Stuff 🎩 lately. 
• Listen to the music, not the media it's on
• The older, the 'lossier'

Re: 32000Hz MP3 sounds worse than 44100Hz

Reply #10 –
Comparisons are meaningless without original lossless source, and info what each file have undergo, which encoding/transcoding steps.

Re: 32000Hz MP3 sounds worse than 44100Hz

Reply #11 –
Quote from: Chibisteven on
I don't know why @Klymins doesn't just use 128 kbps at 44.1 KHz?  Some PC games I have (Half-Life 2 and it's episodes) and SimCity 4 Deluxe use that exact bitrate.

I use this exact configuration for most stuff except for Flash.

Re: 32000Hz MP3 sounds worse than 44100Hz

Reply #12 –
You can try lame with --lowpass 17 -q4  as an alternative method.

Re: 32000Hz MP3 sounds worse than 44100Hz

Reply #13 –
@shadowking Why cutoff at 17kHz? Isn't it too high even if sfb21 had a scalefactor? And, I read that its average bitrate is about 160kbps; does this mean LAME performs worse than MP3Enc 128kbps when q5 is used?

(Plus, its VBR bitrate calculating algorithm gives nonsense as can be seen with the silent or trebleless moments...)

Re: 32000Hz MP3 sounds worse than 44100Hz

Reply #14 –
Quote from: Klymins on
@shadowking Why cutoff at 17kHz? Isn't it too high even if sfb21 had a scalefactor? And, I read that its average bitrate is about 160kbps; does this mean LAME performs worse than MP3Enc 128kbps when q5 is used?

(Plus, its VBR bitrate calculating algorithm gives nonsense as can be seen with the silent or trebleless moments...)

Its mainly useful for cbr / abr 44khz. (e.g  lame -b96 -q5 --lowpass 17) You could use a lower lowpass like 16. 
 

Re: 32000Hz MP3 sounds worse than 44100Hz

Reply #15 –
Quote from: shadowking on
Quote from: Klymins on
@shadowking Why cutoff at 17kHz? Isn't it too high even if sfb21 had a scalefactor? And, I read that its average bitrate is about 160kbps; does this mean LAME performs worse than MP3Enc 128kbps when q5 is used?

(Plus, its VBR bitrate calculating algorithm gives nonsense as can be seen with the silent or trebleless moments...)

Its mainly useful for cbr / abr 44khz. (e.g  lame -b96 -q5 --lowpass 17) You could use a lower lowpass like 16. 

You know, some encoders use the -q swich for indicating the VBR quality, and I'm sorry for confusing LAME with them.