Solar MPPT Charge Controller Test Bench with Logged I-V Curves

Solar MPPT Charge Controller Test Bench is a laboratory instrument our EEE batch built to do two jobs that commercial products keep hidden: trace a photovoltaic panel's I-V and P-V curves, and measure how well a maximum-power-point-tracking (MPPT) algorithm actually tracks that curve as sunlight, temperature and load change.

The bench is a buck-converter-based charge controller around an Arduino Uno, with two INA219 sensors measuring panel voltage, current and power on both sides of the converter. The firmware implements perturb-and-observe (P&O) and incremental-conductance (INC) MPPT, logs every sample over USB serial, and a Python script converts the log into I-V curves, P-V curves, tracking-efficiency figures and duty-cycle traces. A resistive load bank and a small battery let us sweep operating points safely.

India's rooftop-solar capacity has grown rapidly under the MNRE push, and Karnataka ranks near the top in installed capacity — yet most college labs demonstrate MPPT only in slides. The few teaching kits that exist cost tens of thousands of rupees and are black boxes. Our bench costs roughly ₹5,500 in components, and every design choice — switching frequency, inductor size, sampling rate — is documented and defendable in the viva.

The build follows the VTU 2022 scheme's toolset: the converter is simulated first in Multisim/PSpice, the firmware is written in embedded C on the Arduino Uno (a tool the VTU profile names for EEE), and the curves are fitted and compared in MATLAB/Scilab and Python.

A solar charge controller that claims MPPT gives the user no evidence. In a college laboratory this creates two problems. First, students learn maximum power point tracking only as equations, because there is no affordable instrument that shows a panel's actual I-V curve and the controller's tracking behaviour on it. Second, even where a college owns a commercial controller, its firmware is proprietary: duty-cycle decisions, sampling rates and stability under fast irradiance changes cannot be inspected, so no measurement-based comparison of P&O versus incremental conductance is possible at teaching level.

There is a need for an open, logged, low-cost MPPT test bench — a buck converter with panel-side and load-side power sensing, selectable tracking algorithms, and a host-side analysis pipeline — that a four-member B.Tech team can build for under ₹6,000, calibrate against standard instruments, and use to produce quantitative I-V curves and tracking-efficiency results that an examiner can verify step by step.

Abstract

This project builds an open, instrumented test bench for solar charge controllers. It traces a PV panel's I-V and P-V characteristics under real sunlight, runs two classic MPPT algorithms (perturb-and-observe and incremental conductance) on a buck converter, and quantifies tracking efficiency, converter efficiency and transient response from logged data. The deliverables are the working bench, firmware with selectable algorithms, a Python analysis pipeline, and a calibration report that compares our measurements with a reference multimeter and states the uncertainty.

Introduction

Grid-connected rooftop solar in India is now measured in tens of gigawatts, and Karnataka consistently ranks among the top states. Charge controllers with MPPT are standard in these installations, but in a college lab the technology is usually a sealed box: students see terminals in and terminals out, never the tracking decision itself. Our bench opens that box. It is deliberately built from parts any Indian electronics market (Bengaluru's SP Road, for us) stocks: a 50 W panel, an IRF3205 buck stage, INA219 sensors, and an Arduino Uno.

Existing system and gap

• Commercial MPPT controllers (₹3,000–₹15,000) report only LED states; you cannot see the duty-cycle trajectory or measure tracking efficiency.
• PV-characterisation setups in industry use electronic loads and source meters that cost several lakhs.
• Typical student demonstrations use a lamp and a small panel with no logging, so "MPPT works" is claimed but never shown with data.
• Teaching papers describe algorithms, but few give a reproducible low-cost rig with calibration data attached.

Proposed system

• Instrumented buck converter: 50 W panel, IRF3205 MOSFET with IR2104 gate drive, Schottky rectifier, 100 µH inductor, battery + resistor load bank as the sink.
• Dual INA219 sensing at panel side and load side, giving simultaneous input and output power and hence converter efficiency.
• Firmware with two MPPT algorithms selectable at run time, fixed 31 kHz PWM, and logging of V, I, P and duty every 10 ms tick.
• Host analysis in Python and MATLAB/Scilab: I-V/P-V sweeps from near short-circuit to open-circuit, daily tracking-efficiency figures, and comparison of P&O vs INC oscillation and convergence.
• A traceable calibration: shunt values, sensor offsets and multimeter cross-checks recorded so every result carries a stated uncertainty.

Feasibility

• Technical: every building block is second-year EEE material (switching converters, sampling, C programming) with well-documented datasheets; simulating the power stage in Multisim before soldering removes most risk.
• Economic: the bill of materials is about ₹5,500 (panel ₹700, Arduino ₹600, two INA219 ₹300, MOSFET/driver/inductor ₹350, load resistors ₹800, battery ₹1,200, enclosure and wiring ₹1,500, sundries) — roughly one-tenth of a commercial teaching rig.
• Operational: the bench is benchtop-sized, powers itself from the panel, needs no mains wiring, and demonstrates on the department corridor balcony on any bright day — important because viva dates cannot be scheduled around the weather.

In scope

• One 12 V / 50 W polycrystalline panel on a tiltable mount, tested outdoors during the project weeks.
• Buck topology only (panel to 12 V battery bus), hard-switched at 31 kHz.
• Two MPPT algorithms: P&O and incremental conductance, in fixed-point arithmetic on the Uno.
• Load bank of stepped power resistors plus a 12 V 7 Ah battery as a realistic sink.
• Logging of V, I, P and duty cycle at up to 50 Hz over USB serial; CSV export.
• Calibration against one reference 4.5-digit multimeter with a stated uncertainty budget.

Out of scope

• Grid-tie operation, inverters and net-metering hardware.
• Full battery-charge management (a simple current limit only; no complete BMS).
• A calibrated pyranometer — irradiance is an LDR-based estimate, clearly labelled as indicative.
• Partial-shading reconfiguration and distributed MPPT.
• Cloud dashboards and IoT apps; the bench is deliberately offline for viva-day reliability.

We follow the V-model, which mirrors the two-phase VTU structure: every design phase has a matching verification phase, which is exactly what Phase-I (BEE685) planning promises and Phase-II (BEE786) must demonstrate.

Test plan: (1) unit tests — sensor accuracy versus the reference multimeter at five points; (2) integration tests — converter efficiency at 25/50/75/100 per cent load; (3) functional tests — MPPT convergence after a shadow step-change; (4) endurance test — a full-day run to expose thermal drift. Every case is recorded with input, expected, actual and status in the Phase-II report, and each objective above maps to at least one test case.

The bench is a three-layer instrumentation system: a power layer (panel, converter, load), a measurement-and-control layer (INA219 sensors, Arduino Uno, gate driver), and an analysis layer (Python/MATLAB on a laptop).

flowchart TD
  PV["PV panel 12 V 50 W on tiltable mount"] --> S1["Input sensor INA219 No. 1"]
  S1 --> BUCK["Buck power stage: IRF3205 MOSFET + 1N5822 diode + 100 uH inductor"]
  BUCK --> S2["Output sensor INA219 No. 2"]
  S2 --> LOAD["Sink: 12 V 7 Ah battery + stepped resistor load bank"]
  UNO["Arduino Uno firmware: P&O and INC MPPT, 31 kHz PWM, safety cut-offs"] -->|"gate pulses via IR2104"| BUCK
  S1 -->|I2C bus| UNO
  S2 -->|I2C bus| UNO
  UNO -->|"USB serial at 115200 baud"| PC["Laptop: Python logger writes CSV"]
  PC --> AN["Analysis: I-V and P-V curves, tracking efficiency, MATLAB/Scilab fits"]

Power layer. The panel feeds a nonsynchronous buck stage. The IRF3205 is driven by an IR2104 so the high-side switch gets a proper gate swing; the Schottky diode and 100 µH inductor were chosen from simulation for 31 kHz operation with under 5 per cent current ripple. The load bank (parallel power resistors, 0.5–5 Ω steps) drags the operating point across the I-V curve; the battery provides a realistic near-constant-voltage sink for MPPT runs.

Measurement and control layer. Two INA219 breakout boards sit at the input and the output. They report bus voltage, shunt current and computed power over I2C — exactly the data needed for converter efficiency. The Uno runs a fixed 10 ms control tick: read sensors, run the selected MPPT algorithm, update the PWM duty, stream one CSV line. Safety cut-offs force the PWM to zero on over-current, bus over-voltage, or panel under-voltage at dusk.

Analysis layer. The Python script timestamps, filters and stores each run, then produces the traced I-V/P-V curves with the single-diode model fitted in MATLAB/Scilab, tracking-efficiency figures per algorithm, and duty-cycle traces that show P&O's characteristic three-point oscillation against INC's steadier behaviour near the MPP.

flowchart LR
  A["Start sweep"] --> B["Set duty toward heavy load"]
  B --> C["Step load / duty upward"]
  C --> D["Log V, I, P after 200 ms settle"]
  D --> E{"Near open-circuit?"}
  E -->|No| C
  E -->|Yes| F["Fit single-diode model and save curve"]

The second diagram is the I-V tracing routine — the same sweep the viva demo runs live.

• Add a boost converter stage so the bench can track panels whose MPP sits below the battery bus, and compare topologies on the same instrumentation.
• Replace the LDR estimate with a calibrated silicon pyranometer for proper performance-ratio calculations comparable with published figures.
• Port the firmware to an ESP32 for Wi-Fi dashboards while keeping the offline CSV path as the viva-day fallback.
• Build a shading jig and run partial-shading experiments comparing global MPPT techniques on the traced curves.
• Log panel back-surface temperature with a DS18B20 and derive the temperature coefficient from our own data.
• Turn the design into a repeatable lab experiment: a PCB, a written 3-hour procedure, and a batch of kits for the department.