Spectrum Analyser

GoldWave general discussions and community help
Post Reply
Evon
Posts: 4
Joined: Fri May 18, 2007 2:14 pm

Spectrum Analyser

Post by Evon »

This is a newbie but how does a Spectrum Analyser help in Audio Recording? I realy am interested to know.

DougDbug
Posts: 2067
Joined: Wed Feb 16, 2005 3:33 pm
Location: Silicon Valley

Post by DougDbug »

Hmmmm.... I don't think it's that useful in "normal" recording. If you are recording in a "bad" room, It could help you identify the resonances. It could also help you identify a noise frequency so you can adjust a notch-filter to reduce the noise.

Kummel
Posts: 141
Joined: Sat Sep 23, 2006 7:10 pm

Post by Kummel »

Image

When I digitize a vynylic or a tape, I always verify on the spectrum analyzer the real end of the legitimate signal. I wait for some brillant sound like trumpet, which makes very regular peaks, and I see from where I have only random noise. That allow me to determine the upper limit of the bandpass filter and from which frequency I can try to recover some briiliance, when the type of audio and the quality allows it. Above, it is the typical kind of trace that is easy to recognize from a random noise.

But often, the use of the denoiser removes enough of background noise to make the bandpass filter useless, and if you encode your music in mp3, the encoding gets rid too of background noise when it throws away some harmonics.
:D

Evon
Posts: 4
Joined: Fri May 18, 2007 2:14 pm

Post by Evon »

Thanks guys. But I am only just a little clearer.

What would be nice is to be able to see the frequencies and the amplitude of each frequency across the entire spectrum being used to represent a piece of music in real time. Is this what it does?

cdeamaze
Posts: 87
Joined: Thu Oct 05, 2006 3:19 pm

Spectrogram

Post by cdeamaze »

For songs, a very powerful tool in spectrum analysis is called spectrogram.

Instead of the usual spectrum, which is amplitude vs freq, we can plot a spectrogram, which is a two-dimensional (freq vs time)description of sheet music. Using "Twinkle, Twinkle Little Star" as an example, the picture below shows the first few notes of the song and the corresponding spectrogram obtained by MATLAB. For those people unfamiliar with MATLAB, MATLAB is another software very popular in science and engineering community. It is published by The MathWorks, Inc.

What can we expect to see in our spectrogram?
Well, from sheet music or MATLAB input data, we see that, assuming tempo=120, the song will last for 4 sec and each quarter note is 0.5 sec. We expect to see 2 quarter notes, each at f4=349 Hz and last for .5 sec, followed by 2 quarter notes each at c5=523Hz, followed by 2 quarter notes each at d5=587Hz, and finally followed by a single half note at c5=523Hz and last for 1 sec.

Sure enough, that's exactly what we see from MATLAB output!

Image
Last edited by cdeamaze on Mon Aug 13, 2007 8:04 am, edited 3 times in total.

cdeamaze
Posts: 87
Joined: Thu Oct 05, 2006 3:19 pm

Spectrogram by GoldWave

Post by cdeamaze »

Listen to the first 7 notes of the selected song. Watch its time waveform as if you were in GoldWave environment. Remember to close these windows to return to forums.

For comparison, the corresponding spectrogram shown below was obtained by GoldWave. It is clear that the two diagrams are almost identical! Notice that, like MATLAB, the vertical axis is Frequency and horizontal axis is Time. However, they are not labeled in GoldWave. Carefully compare the center of horizontal bar in orange color, pay attention to its magnitude(freq) and duration(time), with our expectation.

Checking magnitude first. As expected, the center of horizontal bar in orange color are approximately 349, 523, 587 and 523Hz respectively. Next we check its duration to verify the type of notes we got. Notice that in addition to horizontal bar, GoldWave also introduces a green vertical line, as opposed to a tick in MATLAB plot, for a particular frequency. For the first 3 seconds, it is clear that there are two quarter notes in each second as there is a vertical line in the middle separating the two quarter notes. Since there is no vertical line in the middle for the last second, it is very easy to identify it as a half note!

Again, GoldWave output meets our expectation just like MATLAB output did.

Chris, Great job! :D

Image

DougDbug
Posts: 2067
Joined: Wed Feb 16, 2005 3:33 pm
Location: Silicon Valley

Post by DougDbug »

You threw me off-track with the question about "recording". Normally, you just want to capture (record) the true sound, and you use your ears to determine if something is wrong. Record producers are listening, not looking at the visual displays. (The recording engineer is looking at the displays, but he's usually just trying to maintain good recording levels.)

If something does sound worng, a spectrum analyzer might help to confirm or pinpoint the problem.

Besides the fact that they just look cool... Probably, the most common use for a spectrum analyzer is to set-up a sound system. I have an equalizer with a built-in spectrum analyzer, a pink-noise generator, and a microphone input. With that set-up, you can compensate (to some extent) for frequency response variations (errors) from your speakers, and for variations due to reflections & resonances in the room. (True pink noise looks "flat" on a spectrum analyzer, so you can see if the frequency response of your set-up is flat or not.)

When a new sound system is installed in a theatre, a spectrum analyzer is going to be used to get the sound "balanced". Before a live concert, you might see a similar set-up being used. This is standard procedure for any big-name concert. If you are waiting outside the gates at an outdoor concert venue, and you hear pink noise, that's what they are doing. (If it's indoors, you probably won't hear it.)

Frequency analysis can also be used to analyze the sound of a particular instrument, and the results can be used to make a synthesizer sound like that instrument. However, this can all be done "behind the scenes" without any visual display.

Post Reply