Reconstructing the quills

See the post The Quills: the forgotten American folk woodwind for general information about the American instruments known as the quills.

African players make their own instruments, and I imagine that Henry Thomas did as well. Recent players of the cane fife in America do so as well.  So, what do you need to know in order to make a historically accurate instrument?

First off, you need to understand the difference between the cane fife (quills) and the panpipe form of the instrument, and decide which one you want to pursue.  There are modern players and makers of the cane fife form of the instrument, so for the moment I’m going to concentrate on the panpipe version made famous by Henry Thomas.

There are no surviving instruments identified as American (pan pipe) quills.  I was told by Smithsonian conservator and banjo scholar Scott Odell that the collection in the Smithsonian once contained a set of quills that had been donated in the late 1800s along with a mountain dulcimer. The dulcimer survives, but the current whereabouts of the quills is unknown, and may have been lost.

Fortunately, we have a great deal of information in the form of Thomas’s audio recordings, and we can study these to determine fairly precise measurements that will allow us to reproduce the instrument that Thomas played.

I started with the classic Bulldoze Blues, which stands out for its especially well played quills part. Here is a short except from Bull Doze Blues in case you are not familiar with the song.

An analysis of the Quills featured in Bull Doze Blues

I first listened to the piece to determine the number of pipes, and then applied audio analysis tools to a digital recording of the song in order to determine the exact pitch of the pipes–some of them sound for too short a period of time (for me, at least) to get an accurate feel for the pitch of the pipe by ear.

This worked fairly well, and I was able to determine that 8 different notes were sounded, so that the instrument contained at least 8 pipes.  I had less luck with the audio analysis. I used the Spectrum tools in Sound Forge, but some of the notes sounded for too short a period of time still to get an accurate rendering.  This was obvious after I had made some bamboo prototypes (thanks to some generous neighbors who let me raid their bamboo patch).

Although the pipes were tuned within .25 Hz. using a frequency generator and an oscilloscope, the finished instrument did not play well with the original instrument – clearly, something was wrong.

I decided to make larger samples for analysis by condensing all notes of a particular pitch into a single file. This proved to be an straightforward if time consuming task, again using Sound Forge.  I simply worked my way through the master file, and clipped each quills note, and appended it to a file that held other notes of the same pipe.

Here is what I was able to determine:

Pipe number Total duration of sound Approximate pitch of highest point of graph Approximate pitch of note estimated by ear Approximate Amplitude of note
Pipe 1 2.4 676 Hz 680 Hz -37 dB
Pipe 2 2.9 731 Hz 731 Hz -36 dB
Pipe 3 14.6 854 Hz 851 Hz -34 dB
Pipe 4 8.3 961 Hz 961 Hz -36 dB
Pipe 5 17.9 1061 Hz 1055 Hz -36 dB
Pipe 6 10.5 1247 Hz 1256 Hz -33 dB
Pipe 7 0.6 1421 Hz 1442 Hz -40 dB
Pipe 8 3.6 1666 Hz 1696 Hz -32 dB

The approximate pitch is just that–approximate. When playing the quills, it is possible to play the pipe so that it’s pitch varies a considerable amount.  It is also possible that the record it’s self was pressed slightly off center, causing different notes to playback at slightly different frequencies.

In any case, the record is what we have to work with, and it is a starting point for a more accurate analysis.  This time, I selected the entire body of each one of these audio files, and subjected them to Sound Forge’s Spectrum Analysis tools. The results are shown below. You can click on each graph to view a full-size image.

Pipe number Spectrum analysis from Sound Forge 7, full sample is selected. Approximate pitch of highest point of graph Approximate pitch of note estimated by ear
Pipe 1 The note is split into two peaks at 676 and 687.  When analyzed by ear, the pitch seems to fall around 680, which makes sense, since the ear tends to average out rapid variation in pitch.

 

676 680
Pipe 2 731 731
Pipe 3 854 851
Pipe 4 961 961
Pipe 5 There is some variation and “flutter-tounging on this note, which gives a range of perceived pitches.  I’ve chosen 1055 as the strongest note heard, but you might feel otherwise.  The pipes play sharp when the player blows harder, and that is clearly what is happening in some of the samples used to make this track.

 

1061 1055
Pipe 6 1247 1256
Pipe 7 Another short sample with two nodes.  Although the peak is on the lower node, the note sounds higher, as shown to the right.

 

1421 1442
Pipe 8 1666 1696

As you would expect, the pipes with the longest samples provide the most accurate results.

The recording of Bulldoze blues plays back a bit sharp of the key of  Ab. It is said that Henry Thomas played with a capo high on the neck of his guitar, so this is very possible. It is also possible that the recording machine ran a bit slow, which would have given a faster sounding performance and a higher pitch. This was done on occasion by 78 rpm recording engineers to add more punch to a recording by making the player just a bit faster than they were in real life. If this was the case, and the recording was as much as a semitone sharp, the key of G would be a reasonable guess for the original key of the song.

Using the key of Ab major pentatonic, what notes do the pitches determined above give us?

Pipe number Note name and reference pitch, A=440 Difference Approximate pitch of note by ear Approximate Amplitude of note
Eb (6th octave) 622.25Not represented in this octave.
Pipe 1 E (6th octave)  659.26.Not a member of the scale, but rather a “blues” note that always slides into F. 20.74 Hz 680 Hz -37 dB
Pipe 2 F (6th octave)  698.45 32.55 Hz 731 Hz -36 dB
G (6th  octave) 783.99
Pipe 3 Ab (6th octave)  830.60 20.4 Hz 851 Hz -34 dB
Pipe 4 Bb (6th octave)  932.32 28.68 Hz 961 Hz -36 dB
Pipe 5 C (7th octave)  1046.50 8.5 Hz 1055 Hz -36 dB
Db (7th octave)  1108.73
Pipe 6 Eb (7th octave)  1244.50 11.5 Hz 1256 Hz -33 dB
Pipe 7 F (7th octave)  1396.91 45.09 Hz 1442 Hz -40 dB
G  7th octave)  1567.982.
Pipe 8 Ab (7th octave)  1661.21 34.79 Hz 1696 Hz -32 dB

Why is the first pipe so “out of tune”?

It is likely that the whole set of pipes is tuned a bit sharp (or the cutting lathe for the recording was running slower than 78 rpm). Even considering that ,one oddity that stands out – the first pipe plays a E, and the 6th pipe plays an Eb.  This causes some problems when determining the scale and mode of the instrument, although I’m certain it did not bother Thomas at all–the sharpened E is always used to slide into the F pipe, and is never played on it’s own. In other words, it might be a “blue” note, used for effect.

It may be that cane was not available in a proper size for a longer pipe to play Eb, or it may just be that this is the effect that Thomas wanted.  In any case, it works very well in the context of the song.  The Eb in the second octave is played as a true scale element, and is used much more often (10 seconds total, compared to 2 total for the first pipe).

The easy way: a simple set of pipes in G Pentatonic

For ease of playing with other instruments such as the guitar, it makes sense to assume that the recording is about a semitone fast, and that the first note is un-intentionally sharp. If you take the values shown above and drop them by about a semitone, and flatten the first pipe to match it’s octave, you end up with a pentatonic scale in the key of G, starting on the note D.  This is a very reasonable pitch to use for a set if all you want to do is play quills.

Given those pitches, we still need to know two things before we can make a pipe – the length of the pipe and it’s internal diameter that would provide the pitches shown in the chart. Since there is no easy way of calculating this, I did some empirical research on the physics of stopped pipes which can be found at the following page: The Acoustics of Pan Pipes.

Adjusted pitch Estimated internal length from graph Internal diameter Bore
G 53.8 4.55
E 63.12 5.74
D 70.85 6.44
B 84.26 7.66
A 94.58 8.60
G 106.16 9.65
E 126.24 11.8
D 141.7 12.88

The numbers are far more precise than are needed for an instrument made out of a irregular material like bamboo. The bore can be a good bit larger or smaller (experiment with what gives you a good, strong sound). You should start each pipe longer than you think you need by at least a couple of mm.  You can always make a pipe shorter, but never a longer.  One way is by melting wax into the pipe to raise its sound.  If you do that, don’t leave the instrument in a hot car.

The hard way: what size pipes match the recording?

On the other hand, you might just want an instrument just like Thomas used. Knowing what we know about the acoustics of stopped pipes, what would this set of pipes have looked like?

The first step here is to figure out what size pipes (length and width) would provide the pitches shown in the chart. Since there is no easy way of calculating this, I did some research on the physics of stopped pipes which can be found at the following page: The Acoustics of Pan Pipes.

Based on the information on that page, the following values look like good starting points for making a reproduction of Thomas’s instrument to match the pitch of the recordings.

Pitch from recording Estimated internal length from graph Internal diameter Bore
1696 49.07 4.46
1442 57.72 5.62
1256 66.26 6.30
1055 78.89 7.31
961 86.60 7.93
851 97.80 8.82
731 113.85 10.11
684 121.68 10.73

What material were the quills made of?

Now that we know how many pipes were in the instrument and roughly what the length and diameter of the pipes would have been, the next step is to find the appropriate materials. Many historical sources that mention the quills mention that they were made of cane. In the American south, this can be only one plant–Arundinaria Gigantea, also known as Southern Cane, Switch Cane, and Canebrake Bamboo.  It is the only native bamboo found in North America, and is common in southern states.  For more information, see the page: Cane.

If you live in the south, cane might be a real possibility, but if not bamboo is a adequate second choice, and that is where I focused my attention. Another good source is South-American “trade” panpipes, which are often sold in import stores.  The cane in these instruments is thin-walled, and the pipes are usually finished enough to play well. A good friend and neighbor, Chris (who has helped me with my work and research for this web page) has created an excellent set using this method:

Girl scout camping trip summer 2008 012.JPG (2243002 bytes) Girl scout camping trip summer 2008 016.JPG (4121026 bytes) Girl scout camping trip summer 2008 020.JPG (2926328 bytes)

Practical Acoustics of pan Pipes

Chris2largeAs of this writing, there is no good source of practical information available regarding the construction of pan pipes, so I did a couple of tests to establish some general principles for pan-pipe scaling.  The question I was trying to answer was:

Given a pitch, what is the ideal internal length and diameter of the pipe?

Since cane is a natural material, it will not conform to the precise dimensions defined here, however these numbers should provide a reasonable starting point for experimentation.

Length

Acousticians say that the sounding length of a stopped tube (such as a panpipe pipe) is 1/4 the wavelength of the sound produced. In actual practice, the open end of the pipe “loads” the resonating chamber, and the formula I have come up with, (based on simple tests using Lucite tubes of various lengths) is this:

Sounding length = 2.4123*Wavelength

The wavelength of a pitch can be determined as follows:

Wavelength = speed of sound / frequency

Given a sea level speed of sound of 345 meters per second, this produces the values shown in the following graph.  The red line is an idealized curve of all notes between F#3 (261.63) and A#6 (1864.66).  The dark blue indicates values physically determined from cut Lucite tubes, and the green line are the intervals from Henry Thomas’s Bull Doze Blues, which is a early blues recording featuring the rare American panpipe, the Quills – which is what got me thinking about this in the first place.

The following is the table used to calculate the values for the idealized curve shown in the graph above.

 Note  Frequency (Hz) Calculated wave length (mm) Pipe length (mm) Pipe width (mm)
F#3/Gb3 185 1864.9 449.87 40.90
G3 196 1760.2 424.62 38.60
G#3/Ab3 207.65 1661.4 400.80 36.44
A3 220 1568.2 378.30 34.39
A#3/Bb3 233.08 1480.2 357.07 32.46
B3 246.94 1397.1 337.03 30.64
C4 261.63 1318.7 318.11 28.92
C#4/Db4 277.18 1244.7 300.26 27.30
D4 293.66 1174.8 283.41 25.76
D#4/Eb4 311.13 1108.9 267.50 24.32
E4 329.63 1046.6 252.48 22.95
F4 349.23 987.9 238.31 21.66
F#4/Gb4 369.99 932.5 224.94 20.45
G4 392 880.1 212.31 19.30
G#4/Ab4 415.3 830.7 200.40 18.22
A4 440 784.1 189.15 17.20
A#4/Bb4 466.16 740.1 178.54 16.23
B4 493.88 698.6 168.51 15.32
C5 523.25 659.3 159.06 14.46
C#5/Db5 554.37 622.3 150.13 13.65
D5 587.33 587.4 141.70 12.88
D#5/Eb5 622.25 554.4 133.75 12.16
E5 659.26 523.3 126.24 11.48
F5 698.46 493.9 119.16 10.83
F#5/Gb5 739.99 466.2 112.47 10.22
G5 783.99 440.1 106.16 9.65
G#5/Ab5 830.61 415.4 100.20 9.11
A5 880 392.0 94.58 8.60
A#5/Bb5 932.33 370.0 89.27 8.12
B5 987.77 349.3 84.26 7.66
C6 1046.5 329.7 79.53 7.23
C#6/Db6 1108.73 311.2 75.06 6.82
D6 1174.66 293.7 70.85 6.44
D#6/Eb6 1244.51 277.2 66.87 6.08
E6 1318.51 261.7 63.12 5.74
F6 1396.91 247.0 59.58 5.42
F#6/Gb6 1479.98 233.1 56.23 5.11
G6 1567.98 220.0 53.08 4.83
G#6/Ab6 1661.22 207.7 50.10 4.55
A6 1760 196.0 47.29 4.30
A#6/Bb6 1864.66 185.0 44.63 4.06

Diameter

I’ve found that a ratio of between 10 to 1 and 15 to 1 for length to bore works well.  I have been using 11 or 12 to 1 in my experiments, and 11 to 1 is used in the example above. Good sounding pipes can diverge from this value and still sound well, which is a good thing because bamboo or cane is rarely exactly the size you need.

More information on panpipes and the quills can be found in the following posts:

The Quills: the forgotten American folk woodwind

Reconstructing the quills, a lost American panpipe