Robot drummer

#1

This is an idea that I can’t get out of my head, so I’m going to get it down here.

Then either it gets torn apart, sinks without trace, or I get all the help I’m going to need to even think about pulling this off :slight_smile:

First, if you google “Robot Drummer” you get things like this:

which is fun to look at but it’s trying too hard to emulate a human drummer. As a result it’s over complicated and the movement is sloppy, leading to horrible timing.

What I want to do is drive a small contact point onto a drum/cymbal/paint bucket/saucepan, hard and with reliably accurate timing (ie. ±1ms from when I want it to happen). More like the rhythm section of a fairground organ.

I want it to play hard, fast music that would ordinarily come out of an electronic drum machine. I imagine adding all sorts of other non-drum instruments to it. The whole thing would be controlled by MIDI and would effectively allow the “acoustic” performance of techno. Kinda like this, but without the human in the loop:

(Note that a lot of what sounds like complexity in this performance is just the “drums” rattling around on the floor!)

There are many problem spaces here.

Power Source - Electric or compressed air? Electric is more portable and cheaper to connect. Compressed air is more versatile, especially when thinking of things that spin or expell air (air-raid sirens, organ pipes etc.)

Mechanical design - What moves? Drumstick on a pivot? Linear impactor “gun”? BB gun? Linear actuators or rotating?

Duty cycle - For some instruments I’d like to play at least 8ths of a beat at 180bpm which is ~40ms between hits. How soon after playing one shot can you play another? This is a function of impactor speed, travel distance, mass and actuator power. Multiple impactors per instrument is fine. After all, drummers have two hands and double bass pedals are a thing.

Control & Calibration - A small computer receives a MIDI signal. An exact number of ms later, the impactor must hit. This will differ due to variations in impactor mass/material and actuator power. Would be great if was auto-tuning and could listen for the impact transient.

Expressiveness - Can I control the volume/tone of a hit?

Moving Target - Hitting things moves them, sometimes by a lot. A human drummer can compensate, a robot drummer cannot. Hey that rhymes. Either the movement is detected and the timing adjusted, or the actuator moves with the instrument, or the instrument is prevented from moving too much.

If I can get kick/snare/hat working I’d call that a success tbh :smiley:

7 Likes
#2

Wow. That guy is insanely good! Trying to replicate that is very ambitious. I can’t help technically, but can throw up some shapes once you get it up and running!

On a slightly different note, did you see the Honda advert I think it was? The one with the car driving in circles and as it drives over large pedals spaced out at intervals it plays the drums and makes a tune.
Could that possibly be condensed down and help you out in any way?

#3

He’s amazing isn’t he?

Mechanical sequencing is a step too far :smiley:
I’d like to be able to offer it up to composers so it has to be plug and play with MIDI

1 Like
#4

Mazda, not Honda.

#5

Your mad , love it , let me think about it for a while.

1 Like
#6

Great idea for a project !!
Initial thoughts are that electric would have to be the way to go as fast modulating air valves are v expensive where as mosfets and coils aren’t.
Small dc solenoids would be fast and you could use them directly to strike the item or via small levers.

Courty

4 Likes
#7

Yeah direct drive with solenoid is probably the way to go. There’s hundreds of the damn things out there though. Need to figure out the requirements…

1 Like
#8

I was thinking moving coil transducers driven by class d amplifier modules which are dirt cheap, and will give millsecond timing, and microgram force Control the transducers should be easy to make as we have a lathe for the pole peace’s.
Moving iron is a bit last century :grinning:
I agree about air it would be a pain .

#9

What about magnets? Could you have a pin that strikes the instrument in a tube with a magnet above it and raises/drops it on demand? Is that fast enough?

#10

You’re basically describing a solenoid, so yeah :slight_smile:

#11

You can always have multiple actuators running at a slower rate to solve this.

2 Likes
#12

Will need a bit more than that I think! (Note to self, check the maths)

#13

my first thought was solenoid.

Plenty of examples on the tubes

#14

In terms of expressiveness. It’s pretty boring hitting the same spot so varying the impact point and contact time would be two fairly straightforward ways of adding interest

#15

Yeah, that but with 10x the power and an actual musician doing the sequencing :slight_smile:

#16

Have you heard of the californian musical road (failure)?

#17

My bad English.
Can be controlled to within a microgram.
So you could generate 1.001 kg

Moving coil is better in every respect but not available off the shelf at a reasonable price. But should be reasonable easy to make.

Moving iron /magnet is available of the shelf but not as good in any respect.

If moving magnet can get the job done go moving magnet .

I would still look into cheap class D amplifier modules to drive it .

#18

Or don’t hit the drum at all just glue a small magnet to it and hit it with a magnetic field.

No contact drumming

#19

So I technically invented a solenoid?

Not bad, huh?

:wink:

1 Like
#20

It may be helpful to think about this project from the other end: how will you program the robot when you have built it? How would you program a MIDI sequence with the necessary timing, subtlety, feel and all the other things that a human drummer brings to his/her instrument? Of course you can program some of this with any decent midi sequencer using a “piano roll” or similar interface, but have a pretty limited set of variables: velocity (i.e. how hard the thing is hit) and aftertouch which may be able to assign to something. But that’s about it as far as MIDI goes, unless you got into layering different percussive hits on the same surface. Eventually you would need a human percussionist to hit the actual instrument or pads that imitate it, and record the midi live for later editing. Some producers work like this, only using the drummer to hit sensors on the drum heads, recording a midi track that will later trigger drum samples - after more editing.

Take a hi hat for example; there are hundreds of variables in playing one, but just consider the pedal. It’s not just closed or open; the pressure on the pedal when depressed affects the sound, or so drummers tell me. And how soon the pedal is released, and by how much, and how quickly. Yet midi only offers closed/open hat usually.

But in the early stages, i’m guessing you will not insist on a really human feel (like the one shown in the video). You just want it to work. I like the idea of midi control to make it accessible to different composers. It might be fun to make it work “live” so the drummer could be in a different location to the kit. Based on my experience with drummers in bands, there are a lot of musicians and producers who would be willing to pay good money for that… :slight_smile: