

Perceptron
KiCad design files for an analog perceptron with 16 binary inputs and 16 adjustable weights.
- Size
- 258.7 × 120.1 mm
- Copper layers
- 2
- Thickness
- 1.6 mm
- Parts
- 120
- Nets
- 122
- Pads
- 441
- Tracks, vias
- 1,071, 138
- Pours
- 1
- Routed
- Yes
- KiCad
- KiCad 10.0 (file 20260206)
- Created
- 28 May 2026
- Updated
- 28 May 2026
Overview
Perceptron is an open-source printed circuit board by wagnermoritz, from the wagnermoritz/analog_perceptron repository on GitHub. It is a 2-layer board of 258.7 × 120.1 mm with 120 parts on 122 nets, routed with 1,071 tracks and 138 vias.
It is built around the ADG1414BRU. Its parts include 48 resistors, 28 capacitors, 16 diodes, 6 ICs and 3 connectors. It belongs with the AVR (ATmega, ATtiny) boards of this library.
Main components
1README
README.mdAnalog Perceptron PCB
KiCad design files for an analog perceptron with 16 binary inputs and 16 adjustable weights.
The project implements a small hardware perceptron where each input is either 0 or 1, each weight is set manually with a potentiometer, and the weighted sum is displayed on a center-zero analog amp meter.
It is intended as a hands-on demonstration of how a simple neural network classifier can be implemented using analog electronics.
How It Works
Inputs
The inputs x_i are binary values, either 0 or 1.
They are generated by an Arduino Nano, which is used only to define four different input patterns. These inputs are shown visually on the LED matrix.
In the included setup, the LED matrix displays two J patterns and two T patterns.
Weights
Each weight w_i is set using a potentiometer.
The potentiometers act as voltage dividers between 0.5V and +0.5V.
This allows each weight to be positive, negative, or approximately zero.
Multiplication
The multiplication w_i * x_i is implemented using ADG1414 analog switches.
For each input:
input = 1 -> switch closed -> weight is passed through
input = 0 -> switch open -> weight is disconnectedSo the analog switch performs multiplication by either 1 or 0.
Summing and Output
The selected weight voltages are summed using two inverting op-amp stages.
The result is displayed using a 50µA center-zero amp meter:
positive current -> classified as J
negative current -> classified as TThe goal is to manually adjust the potentiometers so that the meter deflects positively for J patterns and negatively for T patterns.
Build Notes
The build uses a eurorack power supply to deliver +5V, +/-12V, and 0V.
In the final build, the following substitutions were used:
510Ωresistors with green LEDsOPA2172op-amps instead ofOPA2197
Parts
120| Qty | Value | Footprint | References |
|---|---|---|---|
| 20 | 100k | Resistor_SMD:R_0603_1608Metric | R25, R26, R27, R28, R33, R34, R35, R36, R37, R38, R39, R40, R41, R42, R43, R44 +4 |
| 18 | 100n | Capacitor_SMD:C_0603_1608Metric | C1, C3, C4, C5, C10, C11, C13, C14, C16, C17, C18, C19, C20, C21, C24, C26 +2 |
| 17 | 10k | Potentiometer_THT:Potentiometer_Alpha_RD901F-40-00D_Single_Vertical_CircularHoles | RV2, RV3, RV4, RV5, RV6, RV7, RV8, RV10, RV11, RV12, RV13, RV14, RV15, RV16, RV17, RV18 +1 |
| 16 | 560R | Resistor_SMD:R_0603_1608Metric | R5, R6, R7, R8, R9, R10, R11, R12, R13, R14, R15, R16, R17, R18, R19, R20 |
| 16 | LED | LED_THT:LED_D5.0mm | D1, D2, D3, D4, D5, D6, D7, D8, D9, D10, D11, D12, D13, D14, D15, D16 |
| 8 | 100R | Resistor_SMD:R_0603_1608Metric | R21, R22, R23, R24, R29, R30, R31, R32 |
| 8 | 10u | Capacitor_SMD:CP_Elec_5x3 | C2, C9, C12, C15, C22, C23, C25, C27 |
| 4 | ADG1414BRU | Package_SO:TSSOP-24_4.4x7.8mm_P0.65mm | U1, U2, U4, U5 |
| 2 | 10k | Resistor_SMD:R_0603_1608Metric | R1, R4 |
| 2 | 10n | Capacitor_SMD:C_0603_1608Metric | C6, C28 |
| 2 | OPA2197xD | Package_SO:SOIC-8_3.9x4.9mm_P1.27mm | U3, U6 |
| 1 | 01x02_Pin | Connector_PinHeader_2.54mm:PinHeader_1x02_P2.54mm_Vertical | J2 |
| 1 | 01x08_Pin | Connector_PinHeader_2.54mm:PinHeader_1x08_P2.54mm_Vertical | J3 |
| 1 | 02x08_Odd_Even | Connector_IDC:IDC-Header_2x08_P2.54mm_Vertical | J1 |
| 1 | 100k | Potentiometer_SMD:Potentiometer_ACP_CA6-VSMD_Vertical_Hole | RV1 |
| 1 | 1k | Resistor_SMD:R_0603_1608Metric | R2 |
| 1 | 9k | Resistor_SMD:R_0603_1608Metric | R3 |
| 1 | Arduino_Nano_v2.x | Module:Arduino_Nano | A1 |
Similar boards
- 2 layers182.9 × 172.7 mm

- 2 layers281.7 × 95.6 mm

- 2 layers96.5 × 55 mm

- 2 layers163 × 137.5 mm

- 2 layers30 × 95 mm

IotThermalCamera
BerkeleyHCI/PolymorphicBlocks
A Python-based hardware description language for PCBs for higher-level (block-diagram / system-architecture level) PCB design through libraries of directly reusable subcircuit generators.
120 parts94BSD-3-Clause - 2 layers227.7 × 112.8 mm

Wren universal
oliviahanley/wren-keyboard
A split ergonomic keyboard with options for numpad, nav cluster, and more.
120 parts93MIT - 2 layers84 × 69 mm

- 2 layers185.4 × 105.4 mm

Questions
How big is the Perceptron PCB?
It measures 258.7 × 120.1 mm and has 2 copper layers, 1.6 mm thick.
What parts does the Perceptron use?
120 parts, built around the ADG1414BRU. The full list of its parts is on this page, with links to each part's symbol and footprint where PCBKit's library has them.
Where can I download the Perceptron KiCad files?
From the wagnermoritz/analog_perceptron repository on GitHub, at the commit this page was made from: the board is KiCAD_files/perceptron.kicad_pcb and its schematic KiCAD_files/perceptron.kicad_sch. PCBKit opens the board in the browser, with nothing to install.
Can I use the Perceptron design in my own project?
Its repository states no licence, so its author keeps all rights: ask wagnermoritz before reusing the design.