3

I have two spectrograms of different copies of the same song. I just wanted to know how to tell which one has a greater quality than the other. Could someone help me?

I think that the first one has a better quality because the lines move to a higher frequency. But am I correct? What do those lines actually mean?

Copy 1

Copy 2

Thanks!

2

The upper one is definitey the better one.

Spectrograms (like the ones you produced with Spek) are used to display and inspect tendencies over time. For example if some noisy signal is present all the time, it will show up as a straight horizontal line somewhere:

enter image description here

You can determine some general properties, like that 20 Khz cut, but an average frequency spectrum plot for the whole song is really more ideal for that, like this one (har-bal is a handy tool here):

Frequency Plot

That said, there are some other things, like lossy compression (mp3 etc) that shows up pretty significantly (I guess that was what you were comparing?), and being able to see that over time, confirms that is not a momentary tendency of some sort:

Comparison of compression spectrograms

Based on this you could generalize that quality signals are continuous and smooth on the Y axis, i.e. no sudden cut offs. On the X-axis you should see some variation and no straight horizontal lines going through the whole spectrogram. Ofcourse this depends very much on the material (if someone thinks holding down that C key all the way .. it will show up as a line).

Here is the difference plot from your two plots. It shows some action all over the frequencies, which could be due to some variable bitrate (or in general lower sample rate):

enter image description here

0
  • Color is for the volume (Min - black, Max - Red)
  • Y axis for the frequency (0Hz - 22Khz)
  • X axis is the Time (0:00 - 4:14)

Every vertical lines that compose your spectrum represents the sprectrum analysis of your sound at the moment x.

The first audio source has a better quality. as the high frequencies range (more frequencies in a tighter representation space) have more volume and nuances than on the second one.

In fact your frequency cut is so neat (at precisely 20 Khz which is a well known frequency) on the second spectogram, that it looks a bit like your second spectrogram represents your first one with lower sampling rate.

Am I right?

0

A spectrogram will be determined by it's own analysis/spectrum settings and resolution (FFT Window), so you could likely represent the same audio signal in many different ways. You should be aware of what those settings are and the Legend for the image, as well as the Frequency curve (in your case it is Linear, so about 3/4 of the image actually displays high frequencies above 4Khz). You could check more of the spectrum if you can change the Curve to Mel or Logarithmic. Your spectrum also consists of frequencies from -120dB to -20dB.

There are two sections of the track (the Chorus perhaps?) that show up to 20Khz as at least-60dB (light green), looking like a rectangular block. This could indicate limiting or high compression in that area. It also appears that the track is not Mastered and is at a low level (a tiny portion or frequencies only reach -20dB).

I would not rely on a spectrogram to determine the quality of the track. You could also guess that the first track is higher quality since it has more frequencies over 20Khz, but those frequencies are inaudible so why would it matter?

You will have to listen to determine the quality, as well as cross check the waveform for other artifacts like clipping, etc.

0

You are barking up completely the wrong tree. This is sound. We don't hear sound with our eyes, we listen with our ears. You cannot tell whether a sound is "better quality" than another sound by simply looking at a spectrogram. You have to listen. You learn to listen by training your ears over many years of practise.

Your question is just as absurd as asking if you can tell which picture is better by listening to the paint.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.