How to read a spectral view?

Can anyone be so kind and explain what to look for in an audio file, in spectral view. As I understand it, a spectral view is most telling in regards to validating a 24/96 needledrop. I did both a 16/44 and 24/96 of the latest Grizzly Bear and the spectral s look identical, I can't tell the difference. How can I tell which is which?

 

The spectral view gives you frequency on the 'y' axis ant time on the 'x' axis. The graphic shows intensity of the frequencies at any given moment. The color indicates the intensity with blue/purple being lowest and yellow/white being highest.

Most of us use this view to declick needledrops as clicks and pops have a distinctive visual signature. You use the view to see the click and your ears to figure out whether it is worth fixing.

As for your resolution question, 16/44 files should not have any content above 20khz. Hi-Rez files allow for content way above that. You can see this most easily on sibilants and cymbal hits. Both of these produce 'clouds' of high frequency content upwards of 40khz. A 16/44 file will flatline at 20. If you are seeing the flatline on both files, that suggests that the original recording was done at 16/44 or that the rip that produced the hi-Rez file somehow involved a down-rez in the process.

 

I think you have this backward. You mean graphical. The spectral gives you the pretty colors that represent the intensities of the spectrum.

As for your resolution question, 16/44 files should not have any content above 20khz. [/QUOTE]

Not true. Mine always have content over 20,000 KHz. Remember, the cutoff frequency is half of the Nyquist, which is 22,000 KHz.

Now, if you rip an audio CD, yeah, you won't see any content above 20KHz.

Here is a screenshot of a needledrop I made. The sample is a 44.1/16-bit file. As you can plainly see, there is considerable content above 20,000 KHz:

 

Here's a better example of one I did just last month of a Madonna 45. Again, plenty of content above 20,00 KHz.

 

The nyquist frequency (or cutoff frequency) will be half of the sample rate you use. So for 44.1kHz samplerate the cutoff frequency is 22.05kHz. For 96kHz samplerate it will be a cutoff frequency of 48kHz.

 

When you rip a CD, though, some do cut off at exactly 20,000 KHz.

 

Ok. Nitpicking aside, it should be dead obvious which file is the hi-rez one when you look at the cymbal hits or vocal sibilants.

BTW Grant, what software is that? I thought you were an Audition guy - but I don't recognize the screenshot.

 

Thanks everybody, I just put in a long day at work, and now I get to relax and try and make sense of all this.

 

In reality, if the original has information up to - for example - 30kHz and you are downsampling to or recording at 44.1 you should be able to get information up to 22.05kHz if you use a steep filter. You'll get some aliasing, though near the filter's cutoff frequency. (An example of an analog recording with information up to about 31kHz is the Getz/Gilberto album.

Another case would be if the original material was recorded from an old player, which does not play back anything above 16 or 17kHz.

 

Actually, Dave's statement was absolutely correct. The spectral does give you frequency on the Y axis and time on the X.

I think you misunderstood his use of the word "graphic" in his next sentence:

I don't think he meant graphical view (as opposed to spectral view) but simply "the graphic" as in "the image". Dave?

 

That's Audition 3 as you can see the little squares in the top left and right corners for creating fades.

 

Speaking of spectral view, one thing I've noticed lately is that switching to a lower FFT setting (64, 128, 256) can actually reveal some clicks and other noises that don't show up at all in the higher resolution settings.

 

It is Audition 3. I darkened it. I like to play with the GUI colors and move things around to my liking. I don't like the default look or settings.

 

It's the choice of words he used.

 

True. I have mine set high. I forget what I was doing at the time I set it.

 

That is what I meant. Thanks! But it's no issue really. So long as the OP gets the answer they were looking for!

 

Anyone have shots of how MP3s should look so i can check my MP3s to make sure they're not transcodes?

of course the problem I have is that a lot of them are concerts so its hard to tell anyways.

 

an mp3 by definition is a lossy transcode.

 

The higher the bit rate, the closer that mp3 will look like it came from a CD. If a good mp3 encoder was used, it can also make it harder to tell. I'd say anything from 256 kbps and up will be harder to tell it's an mp3 by just looking at a spectrograph or listening to the file. Anything under about 256 kbps will be obvious because the file will cut off at 18,000 KHz or lower, with little or no residual sound above the fundamental cut off region. And, don't forget that wvwn a full-band recording that has had limiting applied to it can look like an mp3 in a spectral graph.

The only sure way to know if something is a converted mp3 if it is of a higher bit rate is to havwe the source, or know what the source sounded like in comparison.

As with anything else in audio, things aren't always simple. However, if an mp3 does look like it uses the full spectrum, you can still make an educated guess because an mp3 will look less dense in the upper frequencies above, say 8,000 KHz, if the music is known to be loud or densely mixes.

 

What exactly are you trying to check? For example, if a 320 mp3 was transcoded from a 128 mp3?

 

I find that a good pair of spectracles makes 'em easier to read.

A friendly spectre can be a great help.

 

wave
320 mp3
v0 mp3
v0 transcoded to 320

 

I like Wavelab Spectral View for spotting transcodes...