Whistle making help needed

[wormil]

Thanks to everyone for the advice. I haven't drilled the holes yet on the C whistle, I think I'm a little nervous. But the whistle blows well. Playing around with it as-is, I can hit the higher and lower octave with little effort. After drilling, hopefully I can say the same. After building an Aeolian harp and a few strumsticks, the whistles are the most challenging instrument yet and here I thought it was just a pipe with some holes in it. Maybe if I can figure out this whistle calculator I'll have more success. It kind of baffles me that twcalc and flutomat give entirely different lengths for the same whistle. It makes me wonder if I'm using the calculator correctly.

[Feadoggie]

the whistle blows well. Playing around with it as-is, I can hit the higher and lower octave with little effort

Bravo!

Maybe if I can figure out this whistle calculator I'll have more success. It kind of baffles me that twcalc and flutomat give entirely different lengths for the same whistle. It makes me wonder if I'm using the calculator correctly.

Couple thoughts for you to consider.

You can take the view that making a whistle is rocket science and takes a lot of calculation and physics (which it does if you want to embrace that) or you can just go into it wide-eyed, really and truly believing that it is just as pipe with six holes in it (which it well and truly is at the end of the day). Or you can take any view in between those extremes.

The calculator, as I said earlier, describes a perfect view of the whistle. And furthermore, it describes someone else's whistle, using their own end-correction values (aka: fudge factors). It doesn't make whistles. There are many reasons why a whistle does not match a particular calculator's results. Here's how I view it. Whistle calculators are like the Pirate's Code. They are more like guidelines than rules.

You have to make a pile of trash to make a good whistle. Some folks will make a bigger pile than others. But in all cases, once you make a good whistle, you will be able to make more good whistles. It's a pleasant past time making swarf and sawdust - just don't breath it in.

And there is no reason that you have to use a whistle tone-hole calculator to make a working whistle. Ugg didn't use one when he made the first whistle sitting around the campfire long ago. And we have all been chasing his dream ever since. Trial and error works - just embrace the errors as learning experiences. Emulating classic, working deisgns has merit and a long history. Why else would flute makers have to conjure up the ghosts of Nicholson, Pratten and Rudall to sell a flute? Oh yeah, then there is this method. It may seem laughable at first but it has lots of merit.

But in any case, keep at it. Have fun. You'll get it right.

Feadoggie

[DrPhill]

Hi, I wrote TWJCalc and thouroughly endorse what others, but especially Feadoggie, have been telling you.

First I have to clarify something - I do not understand the maths involved in the calculation. Or more accurately I do not understand the theory behind the maths in the calculation. I 'borrowed' the algorithm from two places: Flutomat (by Pete Kossel IIRC) and TWCalc (Daniel Bingamon). I think there is a very small error in the Flutomat algorithm, but it is really insignifcant.

I wrote TWJCalc to answer the question 'how are the hole positions and sizes related', and since I am visually oriented I wrote a program to display the holes as the parameters changed. From there it got a bit out of hand...

Now you need to understand that the calculations are not accurate predictors - not like a metre rule, or a shop till where there is a correct answer that pops out. The calculations are more like a model of reality. Within limits the model behaves roughly like reality - it makes statements like: "increasing this hole diameter means it needs to move down the whistle, and that the next hole up gets smaller". It does not make statements like "increasing this hole diameter by one millimetre will move the hole 0.75mm towards the end". It is not that accurate. In order to increase the accuracy there are several places that 'fudge factors' are applied, but these are derived from reality (making whistles that work and measuring them) rather than from theory. There is no guarantee that the fudge factors are meaningful in a different whistle geometry.

Now if you are making a range of whistles with closely related geometries it would seem reasonable to propose that you could fine-tune the fudge factors so that the accuracy of the model increases. I believe Hans Bracker has done a lot of work on this some of which made its way into TWJCalc as HBFlutomat (Hans' fudge factors at the time applied to the flutomat algorithm). Hans' whistles are highly regarded, so maybe this approach worked.

I used TWJCalc to guide me in making a (just one ) whistle (bass A). I found the calculations consistently out - ie each hole was off by approximately the same amount. This consistency indicates that the model has some merit if fine-tuned.

Sorry for the essay - I have added little really to the discussion, except to reassure you maybe that you are striving for an accuracy which I do not believe the whistle calculators will yield. You may be interested to try some of the different algorithms available, or even use the scripting calculator to fine tune your fudge factors (things like 'end correction').

HTH

[MeMyselfandI]

Oh yeah, then there is this method. It may seem laughable at first but it has lots of merit.
Feadoggie

Haha! I've never thought of that, it's a pretty good idea! There's only one problem with it, and that's that your pipe has to have the right ratio of length/diameter for it to work properly, but other than that, it's genius! I'll have to try that.

[wormil]

Success! (mostly)

The C whistle plays okay though there is still a little fine tuning with needle files necessary as several holes are just a hair flat. The whistle is much breathier after drilling the holes and the upper register on the E,F,G & A holes is difficult and unstable. It's not very loud but that isn't a problem as I just play it around the house. It is definitely playable as-is so long as you don't need many high notes. I'm hoping a bit of tweaking can improve things.

Some lessons learned: Drilling the first hole sharpens the base note just a tad. Next time I'll leave a couple extra millimeters to the length to compensate, I can always shave it down. The thickness of the pipe, 2mm, is definitely working against me as the more holes I drill the previous holes tend to go sharp then flat. I'm going to experiment with countersinking the holes just a little and see if that helps. The new pipe I bought is slightly (.25mm) thinner than the cpvc I've been working with. With this whistle I trended toward small holes because I was afraid I might have to drill them out larger like last time but the calculator put me right in the ball park so next time I'm going to trend toward larger holes which should help the volume a bit.

Thanks for the advice, all of it has been helpful actually. Once I get it tweaked a bit I'll post a sound file.

[highwood]

The calculator, as I said earlier, describes a perfect view of the whistle.

In fact Pete Kossel's Flutomat and TWCalc use calculations based on first order approximate solutions for the equations and therefore only describe an 'approximate' perfect view of the whistle.

[wormil]

Goofing around in the shop tonight I made a 6" whistle just to see if it would work. I wasn't aiming for any particular key and just picked the length by eye. It's a very shrill D around 1170Hz if my tuner app is correct. I then drilled 2 holes using the percentages given in the rubberband video and they were actually close. Drill bit sizes were arbitrary and got an E & F note from them (well close to those notes, it's off a good bit). The wood plug actually slopes upward toward the blade, an accident that worked out very well.

I plan on giving it to my neighbor's little girl, revenge for the times they sugared my kids up on candy and sent them home.

[Mr Ed]

Welcome to the wonderful world of whistle making, Wormil!

You have to make a pile of trash to make a good whistle. Some folks will make a bigger pile than others. But in all cases, once you make a good whistle, you will be able to make more good whistles. It's a pleasant past time making swarf and sawdust - just don't breath it in.

Ain't that the truth! I had to step away from it for a while, it was becoming too frustrating. Most of my ideas didn't pan out. Glad to see this reminder that there is a learning curve and everyone will have a pile of not so great ones. Now that the frustration has worn off a little, I grabbed the whistles I made and have 3 definite keepers, two Ds and a C. That's out of 8 or 9 attempts. Now I want to make that elusive Bb to rival the Generation. See what you have to look forward to, Wormil?! Again, welcome to the insanity!

[Tunborough]

The calculator, as I said earlier, describes a perfect view of the whistle.

In fact Pete Kossel's Flutomat and TWCalc use calculations based on first order approximate solutions for the equations and therefore only describe an 'approximate' perfect view of the whistle.

I think it is possible to do better. Lately, I've been spending far too little time working on whistles, and far too much time working on whistle models. I've got one that works from exact solutions to equations with only two little fudge factors, and it is gratifyingly close to reality for the whistles I've tried it on. Sometime soon, I hope to start drilling holes in two whistles based on it.

The math for the model comes from recent research in Montreal and Paris, along with my own experiments. I can post references if anyone is interested.

Unfortunately, the model is currently implemented as a collection of routines written in Python, so it's not ready for public release. I'm working on a Java model with a couple of collaborators, but its only in the early stages.

[Tor]

[new model]
The math for the model comes from recent research in Montreal and Paris, along with my own experiments. I can post references if anyone is interested.

I've recently been thinking about such things myself and would love to dive into any interesting material, so I for one would be grateful to see any references you may have.

Unfortunately, the model is currently implemented as a collection of routines written in Python, so it's not ready for public release. I'm working on a Java model with a couple of collaborators, but its only in the early stages.

Python routines would be interesting too, even more than Java!

-Tor

[DrPhill]

The calculator, as I said earlier, describes a perfect view of the whistle.

In fact Pete Kossel's Flutomat and TWCalc use calculations based on first order approximate solutions for the equations and therefore only describe an 'approximate' perfect view of the whistle.

I think it is possible to do better. Lately, I've been spending far too little time working on whistles, and far too much time working on whistle models. I've got one that works from exact solutions to equations with only two little fudge factors, and it is gratifyingly close to reality for the whistles I've tried it on. Sometime soon, I hope to start drilling holes in two whistles based on it.

The math for the model comes from recent research in Montreal and Paris, along with my own experiments. I can post references if anyone is interested.

Unfortunately, the model is currently implemented as a collection of routines written in Python, so it's not ready for public release. I'm working on a Java model with a couple of collaborators, but its only in the early stages.

Or chat with me off-forum. I could get an implementation into TWJCalc within a couple of days. The infrastructure is designed to allow new calculators to slide in seamlessly.......

[wormil]

Welcome to the wonderful world of whistle making, Wormil!

Well I was definitely encouraged after my modest success with a C whistle. The calculator got me very close. I started by drilling each hole two bit sizes small then worked my way up. On most of the holes, I was able to use a bit size smaller then waller out the hole a bit by rocking the drill forward and back until it was right. Only two holes were significantly different from the calculator. I have a failed whistle that I can cut down to make a G, I might try that soon although my hardware came in for my stick dulcimer and I need to spend a couple days on that.

[CarvedTones]

Or chat with me off-forum. I could get an implementation into TWJCalc within a couple of days. The infrastructure is designed to allow new calculators to slide in seamlessly.......

+1 on that idea. Since I found TWJCalc, I really prefer it. Hides some of the complexity, shows me roughly what the whistle will look like and doesn't eliminate the differences between the calculators but at least allows me to flip between them and see how close they are and where the middle ground is. I like having my confusion all in one place.

[Tunborough]

I've recently been thinking about such things myself and would love to dive into any interesting material, so I for one would be grateful to see any references you may have.

Antoine Lefebvre, Computational Acoustic Methods for the Design of Woodwind Instruments, Ph.D. Thesis, McGill University, 2011.

Antoine Lefebvre, Gary P. Scavone, "Refinements to the Model of a Single Woodwind Instrument Tonehole", Proc. 20th ISMA, 2010.

Roman Auvray, Benoit Fabre, Pierre-Yves Lagree, "Regime change and oscillation thresholds in recorder-like instruments",
J. Acoust. Soc. Am. 131 (2), February 2012.

Or chat with me off-forum. I could get an implementation into TWJCalc within a couple of days. The infrastructure is designed to allow new calculators to slide in seamlessly.......

As I recall, TWJCalc and Flutomat provide dimensions for given frequencies. This model, on the other hand, calculates playing frequencies for given dimensions, so going the other way requires an additional layer. And rather than a frequency for each note, it gives a range of frequencies, between where it jumps to the next register up, and where it drops down a register or (in the first register) doesn't play a very stable note.

[CarvedTones]

Or chat with me off-forum. I could get an implementation into TWJCalc within a couple of days. The infrastructure is designed to allow new calculators to slide in seamlessly.......

As I recall, TWJCalc and Flutomat provide dimensions for given frequencies. This model, on the other hand, calculates playing frequencies for given dimensions, so going the other way requires an additional layer. And rather than a frequency for each note, it gives a range of frequencies, between where it jumps to the next register up, and where it drops down a register or (in the first register) doesn't play a very stable note.

It might be an interesting option to run it on results of one of the other calculations. The "Tunburough Test" of a generated whistle plan.

BTW, I recently made a high D out of what is really too wide a bore (14mm) before I found TWJCalc. TWCalc and flutomat both were off the mark. Comparing to what it actually ended up as when tuned, Hans' tweaked flutomat calculation is much more accurate.