Back to Automation & Control

Applied Sensor Systems

From Photodiodes to a Controller

Design the current-to-voltage, ADC, multiplexing, wiring and calibration chain for many optical channels.

  • By MySunDirection
  • 17 min read
Share this explanation
One-minute answer

You have several photodiodes in an optical head and need a controller to turn their small currents into stable, comparable direction data.

Do not connect a bare photodiode to an arbitrary controller pin and call the ADC count a direction. Define the optical current range, choose photovoltaic or biased operation, convert current to a bounded voltage, filter and protect it, allow the ADC to settle, calibrate every channel, and reserve recognizable states for faults.

Select or check first
Choose the channel architecture from signal magnitude and bandwidth, required simultaneity, controller pin and timing budget, cable length, calibration ownership, diagnostic coverage and service boundary.
Controlling assumption
The detector identity, spectrum, optical geometry, minimum and maximum photocurrent, temperature range, update rate, cable environment and controller input limits are known.
Top risk
Excess gain can saturate in bright light, insufficient gain can bury direction information in noise, and multiplexing without settling can make one channel contaminate the next.
Proof
Inject darkness and repeatable optical levels into every assembled channel, record offset, noise, gain, settling and saturation, then prove open, shorted, clipped and implausible states produce a safe response.
Start with the basics Optional foundation before “From Photodiodes to a Controller”

Automation-and-control basics

An automatic system measures a condition, decides what it means and commands an actuator while checking whether the expected result actually happened.

Sensor
A sensor converts a physical condition into a signal. Its range, uncertainty, response time, environment and failure behaviour matter as much as its nominal output.
Controller
A controller reads inputs, applies logic or a control law and issues outputs. Software cannot correct missing measurements, inadequate hardware authority or unsafe architecture.
Actuator
An actuator turns a command into physical action, such as motor torque, valve position or switching. Its real limits depend on load, supply, drive and duty.
Feedback
Feedback compares measured behaviour with the desired state so error can be corrected. Open-loop command assumes the action occurred; closed-loop control checks it.
Fault and safe state
A fault is an abnormal condition that may invalidate control. The safe response must be defined from the hazard and may require independent hardware, not only software.

How to use this article: Trace one complete chain—physical condition, sensor, signal conditioning, controller, drive, actuator and feedback—then ask how each link can fail.

On this page
  1. Follow the signal from photons to qualified data
  2. Choose gain, bandwidth and bias as one decision
  3. Ten detectors do not require ten controller ADC pins
  4. Reserve evidence for healthy, weak and failed channels

A photodiode is a light-sensitive junction that produces a small current related to the light it receives. The controller ultimately needs a voltage or a digital code, but the useful engineering object is the complete measurement chain: detector, bias and conversion, protection, bandwidth limiting, sampling, calibration, diagnostics and the software rule that decides whether data is trustworthy. Every stage can alter range, noise, timing or failure behaviour.

Follow the signal from photons to qualified data

A controller code is meaningful only when the preceding analog path and its limits are known.
Diagram description

The diagram begins with an illuminated photodiode and passes through a resistor or transimpedance conversion stage, protection and filtering, an ADC or multiplexer, and controller logic. Labels identify range, bandwidth, settling, calibration and diagnostic decisions at their relevant boundaries.

System map · photocurrent, voltage, samples and qualified direction data

One measurement chain, six owned interfaces

  1. 01

    Photodiode

    Converts incident light into current with wavelength, temperature, area, capacitance and dark-current limits.

    Output: Photocurrent plus leakage and noise

    Failure clue: Open lead, contamination, damage or no useful illumination

  2. 02

    Current-to-voltage stage

    Uses a resistor or active transimpedance stage to create a bounded voltage and define gain.

    Output: Conditioned analog voltage

    Failure clue: Saturation, instability, bias error or insufficient signal

  3. 03

    Protection and filter

    Limits damaging transients and unwanted bandwidth without erasing the required response.

    Output: Band-limited protected signal

    Failure clue: Clamping in the normal range or excessive delay

  4. 04

    ADC and multiplexer

    Selects and acquires channels with adequate resolution, reference quality and settling time.

    Output: Time-stamped channel codes

    Failure clue: Channel memory, missing codes, aliasing or reference drift

  5. 05

    Calibration and diagnostics

    Applies offset and gain corrections, checks range and rejects stale or inconsistent data.

    Output: Qualified measurements

    Failure clue: Plausible-looking but invalid values

  6. 06

    Control decision

    Combines channels into a directional error only while confidence and operating conditions remain acceptable.

    Output: Bounded motion request or safe hold

    Failure clue: Unsafe authority from invalid sensing

  • detector → conversion: current and capacitance
  • conversion → filter: voltage and bandwidth
  • filter → sampling: source impedance and settling
  • sampling → qualification: codes, channel identity and time
  • qualification → control: value plus validity

Choose gain, bandwidth and bias as one decision

First-order transimpedance relationship

Vout ≈ Vref − Iphoto × Rf

This relationship is a starting point, not a complete circuit proof. Output swing, input common-mode range, photodiode capacitance, amplifier bias current and noise, feedback capacitance, stability, temperature and the required bandwidth must also be checked.

Vout
amplifier output voltage, volts
Vref
chosen reference or virtual-ground voltage, volts
Iphoto
photodiode current over the stated illumination range, amperes
Rf
feedback resistance that sets nominal volts per ampere, ohms
Example

If the brightest credible input would drive the calculated output beyond the amplifier or ADC range, reducing gain, adding a switched range or changing the optical geometry is required; software cannot recover a clipped signal.

Visual case comparison

Passive resistor or active transimpedance stage?

Held constant
  • The same photodiode, optical head and controller reference are compared.
  • The required update rate and minimum useful directional difference are stated.
  • Both candidates must remain inside component voltage, current and temperature limits.
01

Simple resistor conversion

Condition: Photocurrent is large enough, bandwidth is modest and the resulting source impedance still satisfies the selected ADC acquisition requirement.

Result: It can minimize parts, but detector voltage changes with current and the resistor, detector capacitance and ADC input can create range and settling trade-offs.

02

Transimpedance amplifier

Condition: Small current, controlled detector bias, lower effective source impedance or a defined gain-bandwidth response is needed.

Result: It can create a more controlled interface, but amplifier input limits, stability, feedback capacitance, noise, output swing and power must be designed and tested.

Decision: Select the simplest architecture that meets the measured range, timing, noise and fault requirements; do not select a circuit merely because it is common in a diagram.

Ten detectors do not require ten controller ADC pins

Count timing, wiring and diagnostic ownership—not only pins.
Diagram description

Three columns compare direct controller ADC inputs, an external multiplexer or multichannel ADC, and a local sensor processor that returns digital data. The comparison highlights channel count, simultaneity, cable length, calibration and service implications.

Decision studio

Select the channel architecture from the evidence boundary

Where should analog conversion, calibration and diagnostics live?

01

Direct controller ADC channels

Use when
A small channel count, short wiring and adequate controller ADC performance make the signal path simple.
Controlling criteria
Available pins, ADC reference and resolution, input source impedance, acquisition time, scan schedule and connector capacity.
Trade-off
Few components, but the controller board owns analog layout, calibration and every channel wire.
Rules it out
Insufficient channels, poor analog performance, long noisy cabling or a need for near-simultaneous samples.
Verify
Measure each channel after the worst preceding channel and at the fastest intended scan rate.
02

External multiplexer or multichannel ADC

Use when
More channels or better conversion are needed near the controller while analog cable lengths remain controlled.
Controlling criteria
Input range, leakage, on-resistance, settling, crosstalk, reference, resolution, sample rate and interface timing.
Trade-off
Saves controller pins and can improve conversion, but adds sequencing, configuration and a shared failure point.
Rules it out
Required simultaneity, isolation or cable environment cannot be met by scanned analog channels.
Verify
Test channel-to-channel steps, settling delay, full-scale clipping and missing-device behaviour.
03

Local intelligent optical head

Use when
The detector wiring should remain short and the enclosure can host conversion, calibration and diagnostics close to the sensors.
Controlling criteria
Outdoor environment, thermal design, local power, digital bus length and addressing, update latency, sealing and service replacement.
Trade-off
Reduces long low-level analog wiring, but moves electronics and firmware into a harsher location.
Rules it out
The head cannot be powered, cooled, sealed, updated or serviced reliably.
Verify
Prove communication loss, corrupted data, replacement calibration and safe controller behaviour with the real cable and environment.

What changes the choice: Re-open the choice when detector count, required simultaneity, signal range, cable route, enclosure architecture, update rate, environmental class or maintenance boundary changes.

Reserve evidence for healthy, weak and failed channels

A number inside the ADC range is not automatically a valid optical measurement.
Diagram description

The chart shows a low-level noise region, a bounded calibrated region and upper saturation. Separate annotations identify open-circuit, short-circuit, clipped, stale and implausible channel behaviour that software should distinguish where the hardware permits.

Commission the assembled chain

  1. Identify every channelRecord detector part, physical position, connector and pin, analog stage, ADC channel, software name and calibration revision so a swapped wire cannot remain invisible.
  2. Measure darkness and offsetWith a controlled dark condition, record offset, variation, warm-up and temperature behaviour without assuming every nonzero count is light.
  3. Apply repeatable optical levelsUse a stable geometry and compare multiple levels across the useful range. Record gain, linearity and channel matching rather than calibrating only one point.
  4. Challenge scan settlingAlternate a bright and dark channel, use the intended scan order and delay, and check whether the result depends on the previously selected input.
  5. Prove fault statesWhere safe, simulate open, shorted, saturated, disconnected-bus and stale-data conditions and confirm the controller rejects or bounds the resulting motion.
  6. Save acceptance evidenceKeep raw readings, conditions, calibration coefficients, software and hardware revisions, tolerances and the retest trigger with the product record.
Why a two-wire digital bus is an architecture, not a promise

A shared digital bus can reduce controller pins, but it introduces voltage-level, addressing, pull-up, capacitance, cable, grounding, timing, error-recovery and device-availability constraints. The exact limits belong to the selected controller, converter and physical bus documentation. An address conflict or missing device must be diagnosable, and a frozen old value must not be accepted indefinitely.

Decide whether coefficients live in the head, controller or production database and what happens when a head is replaced. Serviceability improves when identity and calibration travel together under revision control, while the controller still checks freshness and plausibility before granting motion authority.

One-minute questions

Questions before committing the circuit

Can I read a photodiode directly with an ADC?

Only when the detector, conversion method, voltage range, source impedance, acquisition time, noise and protection meet that ADC's documented requirements. A bare junction does not by itself define a trustworthy voltage interface.

Is a larger feedback resistor always better?

No. It raises nominal volts per ampere but also consumes output range and interacts with capacitance, bandwidth, noise and stability. Select it from minimum and maximum current plus the required dynamic response.

Must all direction channels be sampled simultaneously?

Not always. Sequential sampling can work when illumination changes slowly relative to the complete scan and settling time. Simultaneity becomes important when the scene, obstruction or controlled system can change materially between channels.

Can software correct every channel mismatch?

Calibration can correct bounded, repeatable offset and gain differences. It cannot recover clipping, an unstable amplifier, excessive noise, optical contamination, an ambiguous fault or data acquired before the analog path settles.

What should the controller do with invalid optical data?

Use the system hazard analysis: stop, hold, return to a safe position or fall back to a separately validated calculation may be appropriate. The response must be explicit, bounded and tested rather than silently treating invalid data as zero error.

Sources and further reading

  1. Compensate Transimpedance Amplifiers IntuitivelyTexas Instruments · 2026-10-14Primary application report used for photodiode capacitance, transimpedance response and stability boundaries.
  2. Analog-to-Digital Converter (ADC) Acquisition TimeMicrochip Technology · 2026-10-14Manufacturer guidance used for source impedance, sampling-capacitor settling and device-specific acquisition-time checks.
  3. MCP3561/2/4 Family Data SheetMicrochip Technology · 2026-10-14Representative manufacturer documentation used to verify that multichannel ADC architecture, input impedance and filtering are part-specific decisions.
Educational image