VoxBox Robo-Drummer Mock Design Review

Group 27

Craig Bost, Nick Dulin, Drake Proffitt TA: Zhen Qin

Objective Our design seeks to enable a human user to convert live “beatboxing” performance into real-time physical performance by a robot. As such, our design is best thought of as musical instrument in its own right: when “played” as intended, our design will respond by instructing a robot drummer to strike the corresponding drum kit component. Correctly “playing” our system amounts to the human user constraining their vocal performance to a subset of sounds which can be identified as comprising traditional “beatboxing” (see Background below for elaboration). This mapping of performative sounds to the correct corresponding robotic performance comprises the core challenge of our design. Our design is distinct from existing designs which similarly process “beatboxing” into equivalent musical forms, as these existing systems typically only convert recordings of “beatbox” performance into drum samples for music production; as such, these existing systems are purely software-based and non-real-time, whereas our design will feature real-time performance, driven by hardware-software embedded systems. One possible social application for our system, which is not met by these other existing systems, is the capacity for persons with mobility restrictions of their extremities to be able to participate in live music performance, in a manner consistent with typical rock music performance. Background: The Sounds

Before refining the system in detail, the step in the process is to define and characterize the real-world stimulus we want our system to respond to: the audio/sound input. For the sake of feasibility, we’ve limited our scope to three core percussion sounds, which will correspond to real-world instruments, per conventional beatboxing, as follows:

Clap = “K” sound (originally “chh” ) Bass = low “Doom” sound snare= “Tss”

Background: Initial Findings & Analysis:

Our initial investigation into the sound profiles was conducted in MATLAB, using the spectrogram library function, so that we could see the energy distribution and time-evolution of each type of sound we’ve defined as a “valid” input stimulus.

MATLAB Spectrogram of Sounds (starting at around 1.2 seconds, from left to right: “k”, “chsh”, “Tss”, “doom”. Repeats twice.) High Level Description Requirements of overall system:

1) Device must be able to receive beatbox audio input from a human user/performer and distinguish between 3 different key sounds. This implies that there are 4 valid inputs in total: clap, snare, bass kick, and no input.

2) Device must be able to perform requirement (1) in real-time (defined in accordance with conventional electronic musical instrument latency standards) and drive a machine based on said inputs.

3) The machine will have 3 drums, one for each input, and will strike the correct drum which corresponds to its input.

4) The drums must generate enough noise so that they can be heard in 20 ft2. Design approach to meeting these requirements:

To meet the operational behavior criteria defined above, we have decomposed the system into 5 subsystems, corresponding to the primary tasks involved, which we will discuss in detail next. For visual reference, an overall high-level block diagram is provided below:

Power Subsystem

Power supply network. Includes a DC power jack and 3.3V linear regulator.

Input Subsystem Audio Amplifier Characterization of the microphone showed that it output a signal amplitude of approximately 1mV from a human voice input. In order for this microphone output to be usable, we need the line level signal amplitude to be 1.5V (63dBV). To accomplish this amplification we will implement the 2-stage inverting amplifiers shown below.

2-stage microphone amplifier. Voltage gain of 63dBV

Microphone amplifier magnitude and phase response.

Audio Filter

Using Texas Instruments’ FilterPro software, we designed a 5-stage active filter to provide lowpass filtering on the amplified microphone input signal, to help with high frequency noise suppression. Shown below are a block diagram depicting the stage connections and the schematic for each stage of the filter, with accompanying component value table.

5-stage active filter (Linear Phase .05 degree), with 12 kHz passband frequency

Filter stage schematic (above), value table (below)

Stage R1 R2 R3 R4 C1 C2

1 15 kΩ 15 kΩ 330 kΩ 330 kΩ 1.2 nF 1nF

2 6.8 kΩ 6.8 kΩ 330k Ω 330k Ω 2.86 nF 1nF

3 2.7 kΩ 2.7 kΩ 330k Ω 330k Ω 6.53 nF 1nF

4 1.5 kΩ 1.5 kΩ 330k Ω 330k Ω 13.84 nF 1nF

5 680 kΩ 680 kΩ 330k Ω 330k Ω 38.16 nF 1nF

Control Subsystem

The control subsystem essentially performs the ADC on the conditioned input signal from the input subsystem, applies the event detection algorithm, and generates the trigger stimulus to the electromechanical conversion subsystem accordingly. Shown below are a schematic of the I/O assignments for the control components, and a flowchart of our preliminary algorithm.

DSP Sub-block Schematic

Software Algorithm

Electromechanical Subsystem This system takes digital input from the control unit and drives the solenoids which are responsible for forcing the drum strike. The control unit uses a Teensy MCU which outputs 3.3V to trigger the solenoid driver. The ideal gate voltage is 10-12V, so to achieve this we will use an op-amp in the non-inverting configuration with a voltage gain of 3. The solenoids will be custom made with a design goal of 2in throw and push-pull action.

Solenoid driver circuit.

Solenoid view (yoke excluded)

Full solenoid view.

Solenoid cross section view.

Mechanical/Output Subsystem The drum system will be small (think pots and pans). To induce a successful drum strike from the solenoid action, we will implement a fulcrum upon which the drum stick can pivot. By balancing the drumstick on the fulcrum, this will reduce the amount of force the solenoid needs to output.

Cost and Schedule

Labor

Labor costs are calculated based off a $65,000 annual salary with a 50 hour average work week. This is equivalent to a $25.00 hourly wage. We estimate on average 15 hours a week per individual group member, over a 16 week period (1 semester).

Parts

Labor $18,000.00

Teensy Microcontroller $33.25

Behringer Microphone $20.00

Microphone Cable $9.00

Microphone Stand $7.58

Microphone Preamplifier

LM741 (op-amp)$ x2 $1.50

Filter Components

2 x LM741 Operational Amplifiers 2 x $0.50 = $1.00

Various ceramic capacitor ( 0.1uF) $0.10

Power Components

Linear Regulator $2.00.

Mechanical Components

Iron Pipe $4.00

Iron Rod $2.00

PVC pipe $0.50

Gorilla Glue $5.00

Pen Case $0.00

Magnet Wire $0.00

Springs $1.00

PLA $5.00

Total = $87.93

Table 3.1.2: Parts & Labor Breakdown

Grand Total Grand Total = Labor + Parts =$18091.93 Schedule

Task Target Completion Date

Filter Prototype 2/24/2018

Preamplifier Prototype 2/24/2018

Actuator Prototype 2/24/2018

Power Circuit Prototype (Battery, Linear Regulator, Circuit Protection)

3/10/2018

Characterize Teensy 3/10/2018

LED debugging circuit (for output testing) 3/10/2018

Interface all electrical prototypes 3/17/2018

PCB designed and ordered 3/17/2018

Mechanical Drummer 3/31/2018

Interface all components 4/7/2018

Sound Detection Algorithm Implemented and refined throughout the semester.

Ethics and Safety No ethics concerns are raised by this product. Electrical safety is the biggest concern, though the product poses no greater risk than your standard electronic device. We will design our product to comply with all electrical device standards present in the United States.