Spectrogram and mp3 quality conflict

I am starting to get into higher quality music, and was using the spectrogram to see what quality of mp3’s I actually have compared to what the file tells me. I came across a song that got me puzzled. The file says it is 192kbs, but when I look at it on the spectrogram, it looks lossless. Doesn’t make sense right? I tried encoding it up to 320kbs to see what would happen. It looked like any normal 320kbs mp3. It is the only one that does it and I tried looking it up to see why, but no luck. I would like someone to download it and try and shed some light for me. Thanks.

Google Drive link: https://drive.google.com/open?id=0B5wlKG89l1uNVU1YYV9xdXFudnM

It is the only one that does it and > I tried looking it up to see why, but no luck> .

What does that mean?

I would have bet you couldn’t tell MP3 compression quality at the higher values with the spectrogram display. Further, MP3 damage is permanent. The music doesn’t get better or even change at higher recompression qualities.

I may not have used a heavy metal song with 300% harmonic distortion for testing. What does additional distortion look like?

Koz

The spectrogram can sometimes give you a clue that it’s lossless but it doesn’t tell the whole story. MP3 doesn’t simply throw-away high frequencies.

By default, MP3 tries to throw-away parts of the sound you can’t hear, but there are options that allow you to keep the highest frequencies (which might not be the best sound quality because it will cause other information to be thrown-away). MP3 isn’t designed to make nice looking spectrograms, it’s designed for good sound. If you didn’t make the 192kbps MP3 yourself, someone may have changed the settings.

The file says it is 192kbs

Is that constant bitrate or variable bitrate? 192kbps and 320kbps can both sound identical to the uncompressed original (depending on the program material). And with a 192kbps VBR file, you’ll usually get better quality than with a 192kbps CBR file.

The file says it is 192kbs, but when I look at it on the spectrogram, it looks lossless. Doesn’t make sense right?

The spectrogram doesn’t tell you the quality and the bitrate doesn’t tell you the quality… Your ears tell you the quality… For example, if the 192kbps file sounds identical to the uncompressed original (in a proper [u]scientific blind listening test[/u]) we can’t say 320kbps is “better”.

It was the only song with a weird spectrogram. I tried looking up why.

I’m not sure what you mean additional distortion. I tried downloading the same song from different places, and they all looked pretty much the same except the frequency cut offs for the corresponding kbs.

I know I can’t hear the difference between 192kbs and 320 or lossless in a blind test, that’t not what I’m getting at though. I thought a spectrogram would show where the cutoff frequency is for the mp3. Like 320kbps is at 20.5khz and 192kbps is 19khz. Looking at this 192kbps mp3 on a spectrogram, there is no cutoff frequency whatsoever. I’m not sure if it is a constant bit rate or variable, but the song is in the google drive link if you want to see for yourself.

Yes, you can see in the spectrogram where the “cutoff” is (the “low-pass filter” frequency), but the low-pass filter is an option that some encoders allow you to override. By reducing the frequency bandwidth the audio becomes ‘easier’ to encode - it helps the encoder to reduce the file size while minimizing more noticeable artefacts.

For best quality with the LAME encoder, it’s generally best to use the VBR presets (http://wiki.hydrogenaud.io/index.php?title=LAME#Technical_information).

Further to my previous reply, here is a short sample of white noise encoded at 64 kbps CBR with the low-pass filter disabled.

The LAME command for this was:

lame --noreplaygain -q 0 -b 64 --cbr --lowpass -1 <input file> <output file>

and the file properties:

Format                                   : MPEG Audio
Format version                           : Version 1
Format profile                           : Layer 3
Duration                                 : 3s 56ms
Bit rate mode                            : Constant
Bit rate                                 : 64.0 Kbps
Channel(s)                               : 1 channel
Sampling rate                            : 44.1 KHz
Compression mode                         : Lossy
Stream size                              : 23.7 KiB (99%)
Writing library                          : LAME3.99r
Encoding settings                        : -m m -V 4 -q 0 -b 64

Ok, it makes sense now. It’s the first time I’ve seen something like this. Thanks for the help guys.