Tag Archives: spectral signature

How to tap test a guitar – part 2

In a previous page – How to tap test a guitar Part 1 – I talked about the first steps  in producing a tonal signature by tap testing, and how to use Audacity software to record and analyse the tap.

In How to tap test a guitar Part 2 I want to complete the description of how I analyse the spectral signatures I end up with, using Excel to produce charts that allow direct comparison between taps. Just looking at Audacity Fast Fourier Transform charts like the one below is all very well, but they’re so detailed and spread across the whole audible sound spectrum that it’s hard to make sense of them, even though it’s obvious there’s pattern and structure there as you’d expect from something musical.

Guitar makers are mainly focused on the 80 to 1,000Hz range because that carries information we can actually use to take control of the sound. Players wanting to compare guitars, though, may be interested in the whole range from 80 to around 3,000Hz because that’s the whole response of the instrument we can hear.

Just a reminder: remember we’re measuring response in deciBels (dB) where a difference of around 3dB is a DOUBLING of sound level. 10dB is a difference of 10 TIMES, and 20dB is a difference of 100 TIMES.

Just to quickly run over what I’ve covered already:

  • use a small padded hammer, or if you want to go all organic, your knuckle to tap the guitar top, with the instrument held in the usual position, pointing at the computer microphone (the built-in microphone on my MacBook seems okay, and I’ve compared it with a reasonable quality Apogee USB microphone)
  • tap as consistently as you can, and in the same spot on or close to the bridge (unless you’re interested in how the response varies across the top)
  • record the tap series – around 20 of them – using Audacity (try to keep the recorded pulses from “clipping” – that is, hitting or overlapping the upper and lower track borders)
  • select the tap series, go to the ANALYSE menu and select PLOT SPECTRUM
  • marvel at the wonders of modern technology (not so long ago that analysis would have taken hours) and scratch your head while thinking “What does that mean??!”

You will have ended up with something like this from one of my guitars:


One nice thing is that this window allows you to scroll the cursor along the tonal signature and it will tell you not only what frequency you’re on, but also the exact frequency of the nearest peak. Each of these peaks relates to one aspect of the guitar’s structure, which is why we guitar-makers like it. For example, the first large peak at just under 100Hz is the Coupled Helmholtz response that comes from a combination of the top, back, and air body “breathing”. The frequency of it is determined by the size of the soundbox, the vibration of the top and back, and the size of the soundhole. So you can see what I mean, here’s a tap from a soprano ukulele:


Because the ukulele has a much smaller body, the Coupled Helmholtz peak is higher at 129Hz. (Does the smaller soundhole of the uke also help shift the peak upwards? Weirdly, no – it’s pulling the other way. Ah, physics…don’t you love it?)

It would be perfectly reasonable to stop at this point and admire all the different tap signatures you collect one by one just using the Audacity window, but I think you’ll quickly see the limitations of that. You can save the tap tracks on Audacity if you want, and come back to analyse the recording as often as you like. Or, you can take screen shots and keep the signatures as JPEG files. What you can’t easily do is overlap different taps so you can compare directly, as you might for example if you want to win an argument that says Matons are better than Martins, for example (a silly example, because you can’t prove any such thing – but you get my drift I hope).

So the next step is to export all this information in a form that will allow you to plot it as a graph using Excel. At this point I’m going to assume you know how to use Excel charts reasonably well, but I will perhaps give more detailed instructions later on if anybody wants – you can contact me through the Contact me page on this site.

The format I use for exporting is .txt because it’s easy to import into Excel, which will be the step after this.

So, as you already guessed while I’ve been droning on, now hit the Export… button. Save the data file somewhere handy, making sure it’s in .txt form.

Create a new Excel workbook. I usually start out with two tabs, one labelled DATA and the other GRAPH. In the DATA worksheet, select Data menu from the toolbar, then click on Get External Data/Import Text File… . Follow the string of Excel-ish dialogue boxes through to its inevitable conclusion, and you should find a very large set of data magically appearing in the worksheet.

Here’s where I assume you know how to create a new chart and plot the data. I normally plot the horizontal Frequency axis from 0 to 500 or 1,000Hz to suit my purposes, and use a linear rather than logarithmic scale. If you want to see all the data range (up to around 3,000Hz usually) you might want to use a logarithmic scale instead.

And on that probably very annoying note this page reaches its end, gets into its jimjams, and stumbles off to bed with a hot water bottle, because here in Canberra tonight it’s FREEZING!

As part of my series on building a bamboo guitar, I’ll go into the tap tone analysis of the instrument in detail.

Getting comfortable with charts – the difference between noise and music

I know there’ll be some of you who turn off when faced with graphs and charts. Because they’re so useful, though, I’d like you to persevere with them. You can get a lot from them once you can relate them to something real – that’s what I’d like to try and do here.

We can all tell the difference between noise and music (although where the boundaries lie can be a matter of opinion and circumstance). First an example of noise – the sound of surf on a beach:

And next the musical sound of a singing wineglass:

Now here are the sound spectra charts of each one for comparison, beach first:

Beach noise

Beach noise

Singing wineglass

Singing wineglass

The horizontal axis of the graph shows frequency (pitch, in everyday terms), and the vertical axis shows loudness at each frequency.

The difference between the two sounds is very clear to the ear, and you can see it is also very clear from their sound spectra. Noise blasts out fairly evenly across a wide part of the sound spectrum, where a musical note has a series of well-defined peaks, often evenly spaced as in the wineglass example.

The value of the charts is that you can actually read off the frequencies of the main frequency peaks, and so compare different sounds more analytically than listening allows.

Here is another example: the noise of an electric kitchen mixer compared to the sound of a tubular gong of the sort you’ll find in wind chimes.

Here’s how the two look when you analyse them, the kitchen mixer in blue and the chime in red (I’ll leave you to imagine the sounds):


The chart shows the sound level put out by each. One is a maddening noise, the other is a pleasant ring (unless small children get hold of it), and the chart makes the difference as clear visually as it is to your ears.

The chime produces its loudest tones at well-defined frequencies, where the mixer blasts it out right across the spectrum, with one peak at around 400Hz which could well be related to the speed of the motor. We could all hum a note to match the gong, and maybe with the mixer too, probably somewhere between G and G#, which is at the 400Hz peak.

One interesting aspect of the gong signature is that each peak is double. That’s because the gong has a split along it at the bottom, giving it two fundamental frequencies instead of just one.

Charts like this can help a lot in understanding music. The singing wineglass is one of the purest sounds you’re likely to hear (this is a different wineglass than before with a different fundamental frequency, but the same type of spectrum):


This shows that the singing glass puts out sound strongly at some very well-defined frequencies: 709Hz, 1664Hz,1816Hz, and 2496Hz. These correspond reasonably closely to the scale notes F5, G#6, A6, and D#7 (the number refers to which octave each note belongs to). Notice that even a “pure” sound contains more than one musical note.

Thoughtlessly, the makers of the glass didn’t take the time to tune the glass to the standard musical pitches.

By contrast, I tuned the 2nd string of my Jumbo 6 guitar to the note B3 at 246.9Hz:

Here’s the analysis of the note:


The peak frequencies in order are:

  1. 246.1Hz    (B3 is 246.9Hz)
  2. 492.2Hz    (an octave above, B4 is 493.9Hz – double 246.6Hz)
  3. 738.3Hz    (F#5 is 739.9Hz)
  4. 990.2Hz    (B5 is 987.8Hz)
  5. 1236.3Hz  (D#6 is 1318.5Hz)
  6. 1482.4Hz  (F#6 is1480.0Hz)
  7. 1728.5Hz  (somewhere between G# and A6)
  8. 1980.4Hz  (B6 is 1975.5Hz)
  9. 2226.6Hz  (C#7 is2217.5Hz)
  10. 2478.5Hz  (D#7 is 2489.0Hz)
  11. 2724.6Hz  (somewhere between E and F7)

Those with a music theory background will recognise B, C#, D#, E, and F# as notes in the B Major scale. This is what’s known as a harmonic series – more about these in another blog entry.