Research on trombone slide dynamics

[Kdr152004]

Greetings trombone community,

For the technically-oriented, scientists-engineer-musicians out there , I'd like to share my latest research, recently published as a pre-print on TechRxiv (link below):

Multimodal Sensing the Human-in-the-Loop Haptic Dynamics of the Trombone Handslide

https://www.techrxiv.org/doi/full/10.36 ... 6884162/v1

Would certainly value any constructive feedback, especially from any trombonists experienced with sensing, dynamic systems, human-machine interaction, haptics etc.

Regards,
-K

[Sesquitone]

Greetings trombone community,

For the technically-oriented, scientists-engineer-musicians out there , I'd like to share my latest research, recently published as a pre-print on TechRxiv (link below):

Multimodal Sensing the Human-in-the-Loop Haptic Dynamics of the Trombone Handslide

https://www.techrxiv.org/doi/full/10.36 ... 6884162/v1

Would certainly value any constructive feedback, especially from any trombonists experienced with sensing, dynamic systems, human-machine interaction, haptics etc.

Regards,
-K

Nice work! Thanks for this. Of course, the sensors themselves add considerable inertia. Is there any way of "subtracting out" this effect?

[Kbiggs]

A very interesting study. I don’t understand the physics, but the narrative explanation helps.

Re: sesquitone’s comment, would you be able to make similar measurements if you were to film a trombonist in a 3D CGI environment, like Gollum in LOTR or superheroes in the Marvel comics series? For example, you might dress a trombonist in a black bodysuit, place a bunch of color-coded dots on them and the instrument, and then film them playing scales, etudes, excerpts, etc. What kind of measurements are available?

[mgladdish]

A very interesting study. I don’t understand the physics, but the narrative explanation helps.

Re: sesquitone’s comment, would you be able to make similar measurements if you were to film a trombonist in a 3D CGI environment, like Gollum in LOTR or superheroes in the Marvel comics series? For example, you might dress a trombonist in a black bodysuit, place a bunch of color-coded dots on them and the instrument, and then film them playing scales, etudes, excerpts, etc. What kind of measurements are available?

I'm struggling to understand it myself.

But I _think_ the reason for the force sensor is to be able to measure the effect of friction/resistance. Tracing the movement of the slide tells you where it is and how fast it got there, so you can infer *minimum* forces from that, but you can't infer the additional force needed to overcome resistance to get it there.

Why you would want to know such a thing, and what it can tell you about playing the trombone is a whole other question. If I've understood the conclusion, and I'm really not sure I have, then it's essentially suggesting that we take feedback from the resistance/inertia of the slide into account when moving between positions.

[Sesquitone]

Greetings trombone community,
Would certainly value any constructive feedback, especially from any trombonists experienced with sensing, dynamic systems, human-machine interaction, haptics etc.

Regarding presentation, in Figure 3, the positions for the C major and Eb major scales do not correspond to those for F major and Bb major. The C major trace needs to be lowered by five positions (beginning at 6th position), the Eb major trace needs to be lowered by two positions (beginning at 3rd position). Some light horizontal lines showing "theoretically correct" positions might be helpful. In Figure 4, arrows indicating forward time and different colours for ascending and descending scales would clarify the trajectories a lot. Again, horizontal lines (with position numbers) would help. The (roughly) semi-circular trajectories between each note are very interesting. Is there some simple explanation for this?

I'm still concerned about the overwhelming inertial mass of the sensors compared with that of the hand-slide itself. An experienced player does not need to hear the notes for these diagnostics. So the metal outer hand-slide could be eliminated entirely, replaced by a short section of double tubing with cross braces (e.g. in carbon fibre) to which the sensors are attached.

You mention "neural processing latencies, typically on the order of 30 to 300 milliseconds". This is in addition to the so-called "Libet time"—the time it takes (for humans) to be aware of external (and internal) stimuli, typically 500 milliseconds. In other words, "the world" is actually "happening" around us (and inside us—including our own decisions) a full half second before we are even aware of what's going on. Let alone, responding to it. It means, among other things, that a jazz player (for example) might have ripped off half-a-dozen notes before they are aware of which notes they have decided to play! It also means that when you are driving your car at highway speed, your car (and everything in it, including yourself) is about four or five car-lengths ahead of where you believe it (and yourself) to be!

So "feedback" on which positions to play (when there are alternates—especially attachment alternates) is heavily dependent on Libet time and learning to adjust for that.

Drive carefully!



.

[Kdr152004]

I appreciate the interesting points and discussion. See my responses below:

@Sesquitone: Indeed, while the sensors do contribute appreciably to the inertial force, canceling-out this effect is possible, albeit non-trivial and beyond the scope of the present work. As noted in the second paragraph of the Discussion, "....future work could explore test conditions involving well-defined input forces, or develop friction models in conjunction with state observers to deconvolute applied and frictional forces." In other words, since the measured force was a convolution of the inertial and non-inertial terms, state observers could be applied to separate the terms, with the subsequent inertial force scaled based on the known mass of the outer handslide, with and without sensors - thereby providing a method to 'cancel out' the added mass. This task traditionally requires an accurate dynamic friction model which is experimentally challenging to identify.

@Kbiggs: Interesting point.... what your describing is essentially a multi-body simulation of the slide, which could certainly be a future direction to explore. The paper from Ekdahl, 'A simple model of the mechanics of the trombone playing' offers insight into this direction, using an LED kinematic tracking system to measure the motion of the slide and performer.

@mgladdish: Allow me to clarify. By measuring the slide motion and the translational force, we can infer both the resistive and applied forces associated with the motion, however the current sensor configuration does not allow these contributions to be fully separated. Nonetheless, this type of dynamic measurement is valuable because it mimics how the performer senses force/motion, and reacts accordingly to track a reference trajectory, thereby providing insight into the human-machine interaction.

@Sesquitone: Figure 3 presents the relative slide position (0 corresponds to the first note of each scale), based on the step-by-step protocol described in the Methods section, to ensure that the sensor zeroing practices were standardized across test conditions. I will consider options to clarify the graphs further. The significance of the semi-elliptical trajectories is discussed further in the Discussion section. I was not familiar with the 'Libet time' concept but allow me to read further on that !

Best!

-K

[timothy42b]

Fascinating.

Thanks for sharing.

Have you tried doing this blindfolded? (and did you touch the bell? <smiley>)

Motion capture technology has revolutionized the understanding of many sports, and would seem easily adapted to this. That would avoid the mass on the end of the slide.

[Sesquitone]

I appreciate the interesting points and discussion. See my responses below:

I have mocked up (hopefully, not mucked up) sketches of what I suggested earlier. As a purely kinematic diagnostic, the position vs velocity plot would seem to be very useful—especially for beginning students under the guidance of a well-trained instructor. Ignoring the stated problem with reaching 6th position, it's clear that the player shown here is well in tune when retracting the slide in an ascending phrase but tends to play a little sharp when extending the slide in a descending phrase. Obviously, this kind of diagnostic could be pedagogically helpful (in the hands of a discerning instructor).

Another point of slide technique that arises from time to time—and involves both multi-mode kinematics and dynamics of the player's anatomy as well as the slide itself—is the "stiff-wrist/active-wrist" controversy. Your diagnostics would seem to be ideally suited to study this question. [And, maybe, suggest some definitive answers.]

PS If I were making the position/time plots, I would have the slide position increasing vertically upward—as I tend to visualise slide positions as if I were looking down on a horizontal slide (from the mouthpiece end). [I wonder how other players visualise positions (if at all)?]


.

[Kdr152004]

Interesting points @Sesquitone, I hadn't fully considered the pedagogical implications . Indeed, it may help to normalize the position data from the player's perspective -- by contrast, I have normalized it w.r.t. the sensor's starting position. This highlights the dualistic nature of the data, it can be analyzed from the perspective of the human performer, or from a sensor/instrument perspective, or from a control systems perspective that models the interaction.