[Uri Shaked] is really into Latin music. When his interest crescendoed, he bought a trumpet in order to make some energetic tunes of his own. His enthusiasm flagged a bit when he realized just how hard it is to get reliably trumpet-like sounds out of the thing, but he wasn’t about to give up altogether. Geekcon 2018 was approaching, so he thought, why not make a robot that can play the trumpet for me?
He scoured the internet and found that someone else had taken pains 20 years ago to imitate embouchure with a pair of latex lips (think rubber glove fingers filled with water). Another soul had written about measuring air flow with regard to brass instruments. Armed with this info, [Uri] and partners [Ariella] and [Avi] spent a few hours messing around with air pumps, latex, and water and came up with a proof of concept that sounds like—and [Uri]’s description is spot-on—a broken robotic didgeridoo. It worked, but the sound was choppy.
Fast forward to Geekcon. In a flash of brilliance, [Avi] thought to add capacitance to the equation. He suggested that they use a plastic box as a buffer for air, and it worked. [Ariella] 3D printed some fingers to actuate the valves, but the team ultimately ended up with wooden fingers driven by servos. The robo-trumpet setup lasted just long enough to get a video, and then a servo promptly burned out. Wah wahhhh. Purse your lips and check it out after the break.
If [Uri] ever gets fed up with the thing, he could always turn it into a game controller a la Trumpet Hero.
Cute little project. Sadly though the valves on a brass instrument don’t change the instrument’s notes, they just alter the length of the piping to be resonant at the right frequencies. It’s the player’s lips and airflow doing all the work. That’s why the classic style of bugles and early brass instruments don’t have valves at all.
lol, He needs the most important input.
The valves don’t control the frequency. They merely control the timbre of the sound. The frequency is controlled or input from the “lips”. He needs to design some “lips” that can be stretched and relaxed to create the desired range of notes.
A trumpet (or cornet) has three valves and as everyone here would realize that is a binary total of eight permutations Only. The same valve combination is used for the same note in two (or three if your good) octaves.
I hope he can get this working, it will be a feat of “lip” engineering! The trumpet is inherently a difficult instrument to play even for a human. It’s no were as easy and a reed instrument or flute.
I’d say the valves select the length of pipes and therefore the resonant frequencies of the trumpet. The lips need control to add to this to select which resonant frequency to play. I agree the lips need more control to even get to an octave range.
In the article a linked video looks like toyota already mastered this https://www.youtube.com/watch?v=T1etIeZFm14
That Video is fake.
Read the second comment –
Quote: “thats extremely offensive to all of us who play trumpet.There is just no way.”
The robot has no lips or any means to expel air. It is probably using a common speaker.
A trumpet is not a resonator or oscillator (where Q is greater than or equal to 1). It is a filter (where Q is far less than 1).
What you are describing relates to tuned length oscillation like a reed instrument where the reed forms part of a resonant chamber and resonates at the “fundamental frequency” as dictated “by” the resonant chamber.
A trumpet on the other hand takes the sound you make with your lips which is rich in harmonics and filters it to a sound timbre which is also rich in harmonics. The buttons adjust the filter parameters only and do not dictate the sound or fundamental frequency that is made with your lips. (however with experience you can “feel” the chambers influence and work “with” it). This is why it is hard to play trumpet when compared to a flute or reed instrument. You actually have to make the correct sound in the first place. This instrument doesn’t do that for you.
You can play the exact same fundamental frequency with any valve combination but it will sound like shit with most combinations because the harmonics are wrong. Just like you can play several keys on a keyboard (piano) but only the cords sound ok and the rest sounds like crap.
If the tuned length dictated the fundamental frequency then this wouldn’t be possible –
https://www.youtube.com/watch?v=93o_cUhhBfI
Doing the same with a human voice box sounds like this (which is pretty mind blowing) –
https://www.youtube.com/watch?v=vC9Qh709gas
I was waiting for her to break out in FSK!
aparenty the robot does have “lips” and “lungs” https://youtu.be/6fctULDctuA
Wow. Mind blown.
Thanks for sharing!
If humans had two mouths for one brain, we could do overtone BINAURAL singing, and that would be really really crazy.
I agree with you walter, I’m the example…. LoL
I think the point the first video is making is that trombones (with a trigger) have flexibility in notes produced without the using slide. The trumpet in the video sticks pretty rigidly to the tuning provided by the valves.
From my experience the bigger the brass instrument the more they player can play fast and loose with valves and their tuning. A tuba player can basically sing down their instrument if required. Trumpets not so much.
Having just got my trumpet out I have failed to get anything beyond about a semitone beyond what the valves want. If my trumpet is purely a filter, mate does it filter in harmonix well.
As a sceptical trumpet player myself I wanted that to be fake but after some minutes of googling I’m leaning towards it being real. Still not fully convinced you’d get a good seal between the the robot’s mouth and mouthpiece as the lip imitating mechanism seems to be hidden in the mouth but I did find this paper by a Phd student who tried to create an artificial mouth for comparing to real brass players.
https://www.era.lib.ed.ac.uk/bitstream/handle/1842/1966/Bromage+S+thesis+07.pdf?sequence=1
instead of robot lips, should be a variable air input, like pneumatics MXP… sensor with PID control in a secondary smaller chamber, or could be extended to multiple mini air chambers interconnected with solenoids to control it, each chamber just with air/pressure for a upcoming note …
You can’t dispense with the lips–pushing air through the trumpet without “buzzing” lips will only produce a hissing sound.
Controlling the air pressure would be useful to change the loudness.
Varying air pressure won’t change the frequency. The frequency is controlled by the tightness of the lips. Though more air pressure is normally required for the higher frequencies.
To mimic a human I would make lips that can be stretched tight or relaxed loose. I would probably also cheat a little and use an electro-mechanical device like a solenoid to induce the base frequency.