An Informal Analysis...

[Congratulations]

Hi cpelsor

Have you thought about doing Fourier analysis on the various waveforms and using that to provide a more direct comparison.

David

I might also suggest running them all through a Hoozenfarkle algorithm, and then graphing the derivative of the Sporkeniter Sum.

It's a fairly simple process, and proves very effective for comparing their relative Tittlebum values.

(I'm sorry)

[Denny]

Hi cpelsor

Have you thought about doing Fourier analysis on the various waveforms and using that to provide a more direct comparison.

David

I might also suggest running them all through a Hoozenfarkle algorithm, and then graphing the derivative of the Sporkeniter Sum.

It's a fairly simple process, and proves very effective for comparing their relative Tittlebum values.

(I'm sorry)

...and we believe you...

[fearfaoin]

Have you thought about doing Fourier analysis on the various waveforms and using that to provide a more direct comparison.

I would think Fourier Analysis was used to generate those frequency charts in the first place...

[cpelsor]

Have you thought about doing Fourier analysis on the various waveforms and using that to provide a more direct comparison.

I would think Fourier Analysis was used to generate those frequency charts in the first place...

Audition uses Fast Fourier Transforms (FFT) to graph the results. From there, you can use different types of FFT's to get different levels of accuracy. I used Blackmann-Harris. From the Audition docs:

"...[T]he Blackmann-Harris window has a broader frequency band, which isn't as precise, but the sidelobes are very low, making it easier to pick out the major frequency components."

So there ya have it!

[arlan]

Have you thought about doing Fourier analysis on the various waveforms and using that to provide a more direct comparison.

I would think Fourier Analysis was used to generate those frequency charts in the first place...

Audition uses Fast Fourier Transforms (FFT) to graph the results. From there, you can use different types of FFT's to get different levels of accuracy. I used Blackmann-Harris. From the Audition docs:

"...[T]he Blackmann-Harris window has a broader frequency band, which isn't as precise, but the sidelobes are very low, making it easier to pick out the major frequency components."

So there ya have it!

Congratulations!

Regarding your last observation - about sound stability over time - I think a time-frequency transform, like Wavelet, could be useful.

[Dale]

For what it's worth, all the spectral analyses ended up looking EXACTLY the way I had them pictured in my head.

[mikey_r]

Hi, I'm new here on the forum. I would normally lurk for a while to get to know the place better, but this thread sparked my interest. I had a slightly different interpretation of the power spectra.

First, the width of the first peak (around the fundamental, which should be 587.33 Hz for the D) tells you how close to a pure tone the whistle gets. The wider it is, the more "off-pitch" components there are in the tone. The wider and flatter it is, the mellower (or sloppier) the sound will be. Vibrato will cause the same broadening of the peak. The narrower it is, the "purer" the tone - i.e. the closer it gets to being a pure harmonic series. All the whistles were about the same in this respect.

Second, you'll notice that the spectrum between the harmonic peaks is essentially flat (or close to it). This represents the noise floor of the spectrum, and is essentially the breath noise. If it's higher relative to the peaks, the instrument has a "breathier" sound. Look at the spectrum for the Shaw, and you'll see that the ratio of noise floor to the harmonic series is higher than the others, corresponding to a breathier sound. Is this borne out by your experience?

Third, all of the instruments produced what looks like a superposition of sawtooth waves. They contain both even and odd harmonics, in the familiar sawtooth roll-off. However, there is more than one fundamental - there are multiple modes of vibration excited in the whistle, the most important being the root and the octave. The Burke whistle seemed to have both these modes in almost equal strength, which would suggest that (a) it overblows into the second octave really easily, and (b) it has a rich sound. However, it's hard to conclude anything from this because this effect will really be dependent on how hard you were blowing when you did the test - it looks like you were right at the point where the sound was going to break from the lower to the higher octave.

So, basically, there were two main differences that jumped out among the whistles tested. One was how breathy the sound was, and the other was how close to the overblow threshold they were.

I'm sorry if this post alienates everybody - I couldn't resist. I do acoustics and audio processing for a living I'll try to be much less serious from now on.

- Mike (newbie whistler)

[Congratulations]

Hi, I'm new here on the forum. I would normally lurk for a while to get to know the place better, but this thread sparked my interest. I had a slightly different interpretation of the power spectra.

First, the width of the first peak (around the fundamental, which should be 587.33 Hz for the D) tells you how close to a pure tone the whistle gets. The wider it is, the more "off-pitch" components there are in the tone. The wider and flatter it is, the mellower (or sloppier) the sound will be. Vibrato will cause the same broadening of the peak. The narrower it is, the "purer" the tone - i.e. the closer it gets to being a pure harmonic series. All the whistles were about the same in this respect.

Second, you'll notice that the spectrum between the harmonic peaks is essentially flat (or close to it). This represents the noise floor of the spectrum, and is essentially the breath noise. If it's higher relative to the peaks, the instrument has a "breathier" sound. Look at the spectrum for the Shaw, and you'll see that the ratio of noise floor to the harmonic series is higher than the others, corresponding to a breathier sound. Is this borne out by your experience?

Third, all of the instruments produced what looks like a superposition of sawtooth waves. They contain both even and odd harmonics, in the familiar sawtooth roll-off. However, there is more than one fundamental - there are multiple modes of vibration excited in the whistle, the most important being the root and the octave. The Burke whistle seemed to have both these modes in almost equal strength, which would suggest that (a) it overblows into the second octave really easily, and (b) it has a rich sound. However, it's hard to conclude anything from this because this effect will really be dependent on how hard you were blowing when you did the test - it looks like you were right at the point where the sound was going to break from the lower to the higher octave.

So, basically, there were two main differences that jumped out among the whistles tested. One was how breathy the sound was, and the other was how close to the overblow threshold they were.

I'm sorry if this post alienates everybody - I couldn't resist. I do acoustics and audio processing for a living I'll try to be much less serious from now on.

- Mike (newbie whistler)

...wow.

Thank you, that was seriously very interesting.

On a side note, what type of job do you have that requires processing of audio like this? I'm intrigued.

[Denny]

I'm sorry if this post alienates everybody - I couldn't resist. I do acoustics and audio processing for a living I'll try to be much less serious from now on.

- Mike (newbie whistler)

Mike,
Please feel free to be serious at any time, or not!

Thanks,
Denny

[Wanderer]

Mike

That was extremely interesting. And many of your speculations were right on though there would be some debate about burkes having a "rich" sound..most folks think of them as "pure". The Shaw is definitely a breathier whistle. It amazes me that you could see that just by looking at the graph. But really, woudln't it be easier to know if a Shaw is breathier by listening to it?

[mikey_r]

...wow.

Thank you, that was seriously very interesting.

On a side note, what type of job do you have that requires processing of audio like this? I'm intrigued.

I'm an electrical engineer by training, and have worked in acoustic signal processing my entire professional life - worked on everything from speech analysis tools for acoustic phonetics research, sonar signal processing for anti-submarine warfare, speech coding for digital wireless telephony (if you have a CDMA phone, you're using my work), and most recently have been heading a research department that looks at a number of different areas in acoustics, audio and video processing at an industrial research laboratory.

Analyzing audio signals with tools like the ones used in the original post, as well as a host of others, is part of the methodology used in a wide variety of fields including audio equipment design, automatic speech recognition, text-to-speech synthesis, sonography - the list is nearly endless. I even worked on a project in cooperation with the Woods Hole Oceanographic Institute that used these techniques to track whale migrations!

Now that I've totally labeled myself as a hopeless geek, I'll find a quiet corner and listen to people who know a whole lot more about playing the tin whistle than I do.

[mikey_r]

Mike

That was extremely interesting. And many of your speculations were right on though there would be some debate about burkes having a "rich" sound..most folks think of them as "pure". The Shaw is definitely a breathier whistle. It amazes me that you could see that just by looking at the graph. But really, woudln't it be easier to know if a Shaw is breathier by listening to it?

It sure would. I don't happen to have one, though. And if someone gives me a power spectrum to look at, well, what's a guy supposed to do?

The main thing about the Burke was that it looked like the lower D and the octave were about equally strong, meaning that it was either just at the jumping point, or that the lower octave has a strong component in the higher octave - kind of like am extremely well-tuned mandolin. I could see where this could make a tone appear more "pure" , since the "signal to noise ratio" (if we define the "signal" to be the harmonic components of the sound, and the "noise" to be the more or less flat spectral components between the harmonics) is higher for this whistle than for the others. I guess words like "pure" and "rich" and "fat" and "thin" are subjective, but we meant the same thing - a strong tone, rich in harmonics, with less breathiness.

- Mike

[cpelsor]

Woah! Mike! Perfect!!!!

My (secret) hope had been that if I did this, and was wildly wrong in the analysis, that someone with more knowledge would step up, and get it straight! Thanks!!!

[mikey_r]

Woah! Mike! Perfect!!!!

My (secret) hope had been that if I did this, and was wildly wrong in the analysis, that someone with more knowledge would step up, and get it straight! Thanks!!!

I'm really glad you did it. Of course, the best way to select whistles (or any other musical instrument) is to try a bunch of them and pick the ones that feel and sound the best to you. But it isn't easy for me to try or even to hear a bunch of different whistles, particularly the more expensive ones, before deciding what to buy. So people like me (and I guess a lot of others) rely on word of mouth - reviews on websites like this one, etc. And then we comparison shop by sequential buying.

If subjective reviews were accompanied by more objective data - kind of the way they do for high-end audio equipment - it would be very helpful. In fact, I would be very interested in expanding your study to include other whistles, and combining it with subjective reviews. It sounds like the analysis piece was done using readily available software and didn't take a lot of time. I, for one, would find it very helpful to include a picture of the spectrum of the bell tone with the review, maybe with slightly higher resolution in the frequency domain.

In the meantime, I'm going to mail order some whistles - I have a few, but none in D! (They were acquired very haphazardly over the course of 26 years or so). I was thinking of a tweaked Sweetone and a two piece Clare, and maybe a traditional Clarke (untweaked), all in D. I have a couple of very old Generation whistles and an old Clarke in C, and I wanted to experience a range of sounds in D (since that seems to be the most useful key).

In any case, thanks for doing the analysis - I enjoyed it a lot.

[cpelsor]

Mike:

The software I used was Adobe Audition (formerly Cool Edit Pro):

http://www.adobe.com/products/audition/main.html

One feature I just discovered is that I can dump all of the spectral analyses out as a text file, which means I can do much more with it visually (like comparisons, combo graphs, etc.) It actually is surprisingly little work, just takes a bit of time to collect the sound samples.