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: Comparing spectral frequency displays (Read 6744 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Comparing spectral frequency displays

Here are two versions of the same track. Does it necessarily mean that the one with the higher spectral frequency is the one with the better sound quality?

Which one, in your opinion, by looking at the frequencies, is the better quality?

I'm not sure which one to delete and I can't hear any difference, but I'd like to keep the better quality one on my hard drive.

Here are the two spectral frequencies:

Comparing spectral frequency displays

Reply #1 –
One cannot discern audio quality by looking at graphs.  One must listen to the audio and not bother to look at it if any true audio quality statements are made (ie blind ABX tests).  See another thread here for all the reasons why.

Comparing spectral frequency displays

Reply #2 –
Quote from: Damo1888 on
Here are two versions of the same track. Does it necessarily mean that the one with the higher spectral frequency is the one with the better sound quality?
No.

Quote
Which one, in your opinion, by looking at the frequencies, is the better quality?
It's impossible for anyone to tell you that with any degree of certainty.

Quote
I'm not sure which one to delete and I can't hear any difference, but I'd like to keep the better quality one on my hard drive.
If they both sound identical to you, then delete either.

Cheers, Slipstreem. 

Comparing spectral frequency displays

Reply #3 –
Have you tried ABX'ing the two tracks?  Most people here in HA will tell you to do so.  This is the way to go.  If in case you can't hear the difference after the hassle of ABX'ing, here is my two cents worth.

1. Bitrate: Are these tracks of same bitrate? CBR or VBR? More often than not, if one is VBR then it may be of the higher quality (it depends on the encoder used) versus almost equivalent bitrate CBR.  If both are of same bitrate CBR, proceed to the next step.

2. Encoder: Did you encode these tracks yourself?  (Most likely, not) At least you know what encoder was used if you did.  Try to use Encspot on these two tracks.  If one of the tracks has been encoded using improper techniques or a lousy MP3 encoder, whether that track has a higher frequency cut-off, most likely it would be of lower quality.

2. Spectral graphs: Somehow, they have their merits.  For one thing, you could see at what high freq cut-off the track was encoded, but this does not guarantee you that a higher cut-off is of better quality.  (Of course, you ABX'ed them already, and you can't hear the difference, so the higher cut-off track would be of no difference to you too).  This is helpful if a track was not encoded with LAME, and you can't tell what frequency cut-off was used when using Encspot.  If however, both are encoded with the same encoder at a higher bitrate (like 320 CBR) and one has a cut-off of 16kHz and the other 19kHz, then you should keep the 19kHz MP3.

After going through careful deliberation and you still can't decide, don't delete... yet.  But if the difference is obvious, for example, an extreme case of "128 CBR dual stereo encoded using old Xing with high frequencies intact" versus "VBR quality of 77 joint-stereo using LAME (which means -V2 or --preset standard)", then you ought to delete the CBR MP3.

Cheers!

EDIT:  I might add that I was a spectral graph fanatic before, when I just got my Cool Edit Pro software a long long time ago.  But after getting here at HA, things have changed a lot to me.
"Listen to me...
Never take unsolicited advice..."

Comparing spectral frequency displays

Reply #4 –
I'll do the ABX test.

Just to clarify, they were both ripped using LAME, one using 3.96, and the other using 3.97. Also, both are 320 CBR.

I ripped one from my CD and the other was bought off junodownload.

I can't quite understand why there's a difference in the two graphs if both were ripped using LAME at 320 CBR.

Comparing spectral frequency displays

Reply #5 –
I just checked that graphs, one was cut-off at 19kHz and the other at 20kHz.

If both were encoded using LAME with default settings (maybe one was at 320 CBR and the other at VBR -V0), most people would agree that the 320 CBR would be of the of marginally higher quality that -V0).  This won't need an ABX result since it has been already established here in HA.

Quote from: Damo1888 on
I'll do the ABX test.

Just to clarify, they were both ripped using LAME, one using 3.96, and the other using 3.97. Also, both are 320 CBR.

I ripped one from my CD and the other was bought off junodownload.

I can't quite understand why there's a difference in the two graphs if both were ripped using LAME at 320 CBR.


Maybe the MP3 of higher frequency was encoded with the -k switch.  This tried to encode with all frequencies intact.  Most here would disagree that it would be useless, because the additional bits for the higher frequencies (which would you would hardly hear) could have been used for the more audible parts of the song (and the sfb21 issue).

EDIT:  try this for an experiment, encode using "--cbr 320" and "--cbr 320 -k". One will have a higher frequency cut-off, this would reconcile your doubts.

Cheers!
"Listen to me...
Never take unsolicited advice..."

Comparing spectral frequency displays

Reply #6 –
They produce different graphs as each one is a different version of Lame.  That and some other parameters could have been entered during the encoding process of one or both files.  I can produce different graphs using 320kbps CBR and Lame 3.98 by adding some additional commands here and there.  One graph just let's Lame uses its default lowpass frequency cutoff while another one forces Lame to encode the whole frequency spectrum.  One looks better on paper than the other yet the full frequency one will actually sound worse.

Comparing spectral frequency displays

Reply #7 –
Consider if it were a recording of a solitary sine wave.  The spectral graph would be a thin, narrow, line against a mostly black background.  If I then showed you a spectral graph littered with colour covering every part of the spectrum, do you think it would be reasonable to say that the latter is a recording of the sine wave that sounds better than the former?!

Having more colours in a picture means absolutely nothing about the sound!  In the same way, the noise level in a artist's room doesn't correlate with the quality of the painting!  Vision... hearing... different senses!! 

Comparing spectral frequency displays

Reply #8 –
The first graph looks like Lame VBR V0, there is a lowpass right after 19000Hz. The second looks very much like Lame 320CBR, lowpass close to 19900Hz. Although they present differences, you won't be able to hear them at all. If there is the so called "killer sample" you will be able to hear in both. But after saying thing, I agree with others, graphics can't tell you quality. A VBR 5 can be as good as those.
 

Comparing spectral frequency displays

Reply #9 –
Quote from: Damo1888 on
I'm not sure which one to delete and I can't hear any difference, but I'd like to keep the better quality one on my hard drive.

I'd run the files in a program such as Encspot or MrQuestionman to determine which encoder being used. That would give you a hint of which one is better. That is, if you don't have the option to ABX them.
//From the barren lands of the Northsmen