How do you tell when a flute is in tune?

[Tunborough]

When I first started making whistles and tried to get them in tune, I found I couldn't get repeatable numbers when I measured the current tuning. I'd play up the scale and get one set of frequencies, then play down the scale and get a totally different set. I finally resorted to measuring two frequencies for each note: the highest I could play before it broke into the next register (increasing breath), and the lowest I could play before it stopped sounding properly or dropped down a register (decreasing breath). This gave reasonably repeatable numbers. For the most part, (almost) any player should get similar results for these two measurements (although there will be legitimate debates about when a note "stops sounding properly").

Turns out, these two frequencies work well for building a computer model of whistle tuning, which has been implemented in WIDesigner.

Now, I'd like to extend this model to flutes. Paddler has been doing the heavy blowing for this, collecting a pile of tuning information for a sample flute. Trouble is, there's many more variables when playing notes on a flute. As paddler has found, the procedure for whistles doesn't make sense for flutes.

Any suggestions for a predictable, repeatable, procedure to find the playing range for each note on a flute? Maybe guidance on what playing position, embouchure, ..., to use as you go up and down the scale?

If we can get repeatable numbers for the actual tuning of a flute, we have a chance to build a computer model that can predict that tuning.

[Conical bore]

It seems to me the obvious answer here is to use a RTTA analysis program, like Terry McGee links on his web site. I've been using the Android version (TTtuner) while getting used to my newly acquired flute, including a bit of experimentation in cork position. Playing through a full tune using the recording function without thinking about my tuning, is the ultimate reality check on where the pitch actually is in both octaves.

But that only gives you the results for one flute player, because our individual embouchure development and experience will be different. For a "general" computer model, you might need to collect RTTA graphs for a bunch of different players on one test instrument. Maybe a "pass-around" project with a 3D printed flute based on the modeling?

Or wait a few years until we all have 3D printers in our kitchens and can just print a test flute and compare results (I'm not kidding, I think this is where we're going).

[Terry McGee]

I set up the RTTA system, then play for a while to warm up me and the flute. When I think I'm about ready, I run a test playing a few notes coming back to A4 to make sure I have the slide extended to give a good A440.

When I'm ready, I record the notes, checking back to A4 often to make sure I haven't wandered. If I find A4 sharpening, I abandon, pull out a bit and start again, or if only a smidge sharp, I'll roll the flute in a bit until it stops sharpening, then press on with that degree of rolling locked in.

With the exception of the A4 reference, I focus more on the 'number of samples' column than on the pitch column, as sometimes I've found it's recording in the wrong octave (particularly is I've slurred an octave jump), or if the flute is a long way out of tune (e.g. too long body scale or flat foot or both), it's recording in an adjacent note (e.g. low D is recording as low C#).

When I'm playing, I don't seek the ends of the available pitch range. I try to play as I would in a session - strongly but comfortably.

[fatmac]

The only way that I can think of to get repeatable data is to use a mechanical blower, at a given angle on the embouchure lip.

[Terry McGee]

But it would have to be a programmed or at least adjustable artificial blower, fatmac. The late Prof Neville Fletcher built one of these for his flute studies, but found that as we go up the scale and particularly jump octaves, we intuitively change our embouchure in angle, speed, and size of jet.

You can try this out yourself. Play A4 (lower octave A) then fix your delivery in all those parameters and run through the fingering of the two main octaves. Now, relax and run through the octaves as you normally would. I reckon you'll get a better quality and tuning if you let the force be with you.

[fatmac]

Ha! So much for that thought then.

I know I vary when trying to play, & I would expect each person to play differently too.

[paddler]

The real problem here is that the player has a large degree of control over the pitch produced by the flute for each note.
A good player can fairly easily lip a note up or down by a semitone or more, especially on the notes that have left hand
tone holes open. Similarly, when we play in the instrument's higher registers we adjust our embouchure to do so with the
required volume. Since the tone holes on our flutes are different sizes (for ergonomic reasons), different notes speak with
a different voice and tend to be played with slightly different embouchure, especially when held for a while. For example,
the E and e notes often require a bit of coaxing, and cross fingered C natural and c natural are often "special".

The bottom line is that being adaptive with your embouchure is an integral part of flute playing. It determines the tuning
to a significant degree. I like to draw a parallel with fiddle playing. If we asked a fiddle player how the tuning of his/her
instrument was on the notes that are not open strings they would find it difficult to understand the question. It just depends
how you place your fingers to stop the string. Flutes are like this, maybe not to the same degree, but to a much larger degree
than whistles.

I don't think this means that the tuning of the flute itself doesn't matter, of course. Different flutes are harder or easier to
play in tune, and I think this is one of the things we are trying to identify and quantify here. This is different that trying to get
repeatable results though.

If we just think in terms of repeatability, then all we need is for the player to use some kind of reference. For example,
if I look at the needle of my tuner while playing a note I can blow each note fairly precisely in tune, repeatably. If I take away
the tuner and try to play a tune while recording with RTTA, then the only reference I have is my ear, for the relative tuning of
notes. I can use my ear to tell me whether I'm playing in tune and I can use my embouchure control to try to make the tune
sound good. When I do this, RTTA is measuring my ear and embouchure control as much as it is measuring the tuning of the
flute itself. If I don't use my ear either, and I just blow all notes the same way without even listening, then, well, it is not really
flute playing is it?

This line of argument might make the whole approach sound hopeless, but I don't think it is. If WIDesigner can be calibrated to
give accurate results for a particular player, playing in a natural way on a particular instrument (which I think we are already getting
evidence that it can), then it should be able to accurately predict the impact of changes to the flutes design, for that player playing
that same way. If it can do this, then it will be VERY valuable to flute makers, because it will allow them to do extremely rapid virtual
prototyping of refinements.

Different players will respond in subtly different ways, of course, but one of the skills I think many flute maker's possess is the ability
to blow their own instrument's full range of notes with some consistency, while fairly accurately estimating the experience of their target
group of players. This is a critical skill for a flute maker, and I think one of the reasons so many flute makers are also pretty good players.
A tool like WIDesigner is never going to replace the maker, but it can be a valuable addition to a maker's tool box.

[Terry McGee]

Ha! So much for that thought then.

I know I vary when trying to play, & I would expect each person to play differently too.

No, don't write the thought off, because it is possible to program an artificial blower to meet the requirements of changing embouchure as you go up the scale or change register, and of course you learn a lot from having to do so.

But it does probably mean that the plan to use the vacuum cleaner on blow through a squashed drinking straw will need some embellishment....

[fatmac]

But it does probably mean that the plan to use the vacuum cleaner on blow through a squashed drinking straw will need some embellishment....

[Tunborough]

I don't think this means that the tuning of the flute itself doesn't matter, of course. Different flutes are harder or easier to
play in tune, and I think this is one of the things we are trying to identify and quantify here.

For Native American Flutes and reeds, WIDesigner predicts only a single playing frequency for each note, because of the nature of those instruments and the models we have. This answers questions like, "Is [a typical player] likely to play this note in tune?" For whistles, the more ambitious WIDesigner predictions can answer two sorts of questions: "Is it possible to play the notes of this whistle in tune?" and "Is it easy or comfortable to play this whistle in tune (without a lot of breath adjustments from note to note)?"

For flutes, we could simplify the model, and predict only a single playing frequency. That is definitely an option. However, the predictions would be player-dependent, and would reduce the questions we could answer. I'm holding out hope we can predict playing ranges. (I also confess a fondness for the cool tuning graphs this allows us to draw.)

Different players will respond in subtly different ways, of course, but one of the skills I think many flute maker's possess is the ability
to blow their own instrument's full range of notes with some consistency, while fairly accurately estimating the experience of their target
group of players. This is a critical skill for a flute maker, and I think one of the reasons so many flute makers are also pretty good players.

I think you're speaking here of blowing one playing pitch for each note consistently (more consistently than I was doing on my whistle, for example). We could try calibrating the flute model using only nominal playing pitches, another option. We might get decent results, but would leave suspect the quality of predictions for minimum and maximum frequency.

Do you think we could extend this to ask the flute maker to blow the top and bottom pitches that their target player could comfortably get on each note? Not the absolute limits of the player's ability, but what they could produce with the expected embouchure and mouth position for that note? These might be the the limits on what would possibly show up on RTTA. Would that be a reasonable request, and would it produce consistent results? If so, we have a good shot at modelling and predicting these results.

(WIDesigner could support both of the last two options, calibrating either with nominal playing pitches, or min/max pitches. If I have time over Christmas, I'll try implementing calibration from nominal playing pitch.)

[paddler]

For flutes, we could simplify the model, and predict only a single playing frequency. That is definitely an option. However, the predictions would be player-dependent, and would reduce the questions we could answer.

For flutes, I think the predictions will always be player-dependent to a significant degree, but I do not think this diminishes the value of the tool in any way.

I think you're speaking here of blowing one playing pitch for each note consistently (more consistently than I was doing on my whistle, for example). We could try calibrating the flute model using only nominal playing pitches, another option. We might get decent results, but would leave suspect the quality of predictions for minimum and maximum frequency.

I think that is worth trying. I'm note sure how much value the min and max frequency predictions have for a flute anyway, given the vast amount of influence the player has over these.

Do you think we could extend this to ask the flute maker to blow the top and bottom pitches that their target player could comfortably get on each note? Not the absolute limits of the player's ability, but what they could produce with the expected embouchure and mouth position for that note? These might be the the limits on what would possibly show up on RTTA. Would that be a reasonable request, and would it produce consistent results?

We could do this, but I think this is kind of what a maker does in gathering the single pitch per note information above. They blow a note a few different ways and then settle in on what they think the comfortable middle of the range for that note is on this instrument, drawing on their experience about how various players blow and how they themselves blow. When they say that a particular note is flat or sharp by a certain number of cents, I think they are really just giving an indication of which direction they would have to lip the note to get it in tune, and how much effort they would have to put in (the effort, in this case, being measured in cents). This is always going to be an approximation that is player dependent, but its what we work with as flute makers today, and I don't think a computer model is going to change this aspect of flute design.

[david_h]

Isn't "being in tune" just one aspect of how a flute plays?

I haven't played many different flutes. I borrowed a Boehm flute for a while. What struck me about is was that playing up and down the scale , within each octave, I could get a series of in tune notes of roughly the same tone and volume without changing my embouchure much. With a simple system flute that doesn't seem happen because of the uneven venting. On one flute if I play G-A-B trying not to change my embouchure then the A is slightly sharp and has a 'lighter' tone than the notes either side. However, by blowing more downwards I can get a reasonable match in tone and volume and at the correct pitch. If the note was in-tune when blown the same as its neighbours wouldn't it go flat if I tried to match the tone?

My reading suggests that these sorts of thing this is just a fact of life with a D flute. A low-cost PVC picollo - with relatively larger and more evenly sized holes - is much more even.

Is a flute with within reach holes tuned so that it is in tune without embouchure adjustments what we want if the tone and volume are uneven across the scale?

[Terry McGee]

Isn't "being in tune" just one aspect of how a flute plays?

Is a flute with within-reach holes tuned so that it is in tune without embouchure adjustments what we want if the tone and volume are uneven across the scale?

I think that question invites a range of answers, indeed as many answers as there are players.

There are many (most?) who prize accurate pitch with the minimum of player intervention. Playing fast, ornamented and rhythmically-nuanced music on the flute in good time and with good tone provide all the challenges they need.

But there are also those hardy souls who will put up with some pretty weird tuning (and somehow work out how to deal with it) in order to get a flute that sounds in some way better to them. Some of these are what I call "genetically pre-disposed to flute" - they could get a tune out of a broomstick. Others get there by dogged persistence.

And there are the disappointed ones sandwiched in the middle who hoped they could get a leg up by buying a "legendary" flute, but find they can't sustain the greater demands of tuning without sacrificing the other desiderata.

Getting back to Tunborough's model-developing work, and Paddler's observations, here's an interesting thought. We may be able to use the model (once bedded-down) to explore some of what I've talked about above. If we model a flute that has the "pretty weird tuning but attractive sound", we might be able to learn more about how players who like them deal with the tuning, and determine what are the technical reasons behind any benefits they perceive. That might give us clues as to how to reduce the tuning demands while optimising the benefits by other means.

Thank you for your work, Tunborough. I reckon we've come about as far as we can organically and intuitively fiddling the 19th century flute. Time to start meddling inside the space-time continuum. What can possibly go wrong?

[Conical bore]

There are many (most?) who prize accurate pitch with the minimum of player intervention. Playing fast, ornamented and rhythmically-nuanced music on the flute in good time and with good tone provide all the challenges they need.

(Raises hand!) That would be me, as an advancing beginner. Or maybe an almost-intermediate player depending on the phase of the moon. I don't want to fight with the flute. Learning flute is hard enough as it is, and when I use my Android TTtuner app that shows I'm getting real close to good intonation for playing with others in a session at full tempo, I get a warm fuzzy feeling about both the flute and my skills. I know those are interdependent, but I like a flute that makes it easy to play in good intonation.

I also have a fiddler Significant Other who will let me know right away if I'm not in tune with her fiddle. It's a good reality check; avoiding that particular flavor of raised eyebrow.

But there are also those hardy souls who will put up with some pretty weird tuning (and somehow work out how to deal with it) in order to get a flute that sounds in some way better to them. Some of these are what I call "genetically pre-disposed to flute" - they could get a tune out of a broomstick. Others get there by dogged persistence.

Now you're talking about the more advanced players, those who can lip a difficult flute into tune because they already own a bunch of flutes (or have bought and sold a bunch) and they're chasing a certain tone.

I admire and respect those players. I don't think I'll live long enough to get to that point (including being killed by my S.O. for spending that much money on different flutes like I used to do on guitars). But I understand it. However, if we want this tiny niche area to grow, or at least survive, I think it's important to focus on the flutes that intonate well without too much work. That's what the beginners and intermediate players need.

[Tunborough]

We could try calibrating the flute model using only nominal playing pitches, another option. We might get decent results, but would leave suspect the quality of predictions for minimum and maximum frequency.

I think that is worth trying. I'm note sure how much value the min and max frequency predictions have for a flute anyway, given the vast amount of influence the player has over these.

I'll give it a shot.

Isn't "being in tune" just one aspect of how a flute plays?

Is a flute with within-reach holes tuned so that it is in tune without embouchure adjustments what we want if the tone and volume are uneven across the scale?

In the WIDesigner model, "in tune" means "in tune with uniform adjustments to breath speed through the whole gamut." Even changes in breath speed translate to even changes in volume. Changes in tone, however, are another matter.