Short answer
Electronics repair and soldering data covers work done under magnification: lifting components, applying flux, heating joints, reflowing solder and inspecting the result. Repair centers and refurbishers may hold records and images, while step-level video is usually commissioned. It supports task and defect understanding; fine-motor control still needs robot or teleoperation data.
Why this task is different
Most workplace tasks in this cluster happen at arm's length with ordinary cameras. Board-level repair happens at a few centimeters from the work, often under a stereo microscope or with a magnifying lamp, using hand tools that move by fractions of a millimeter. A technician steadies a wrist on the bench, holds a soldering iron in one hand and tweezers or a hot-air nozzle in the other, and watches a screen or eyepiece rather than their hands.
That changes what recording means. A head-mounted camera sees the technician's hands but not what the technician sees through the optics. A bench camera sees the tools but not the joint. Useful capture may need several views: a microscope feed, an overhead view and a side view. This page assumes the buyer wants these joined to repair records and defect images.
Neighboring pages cover related ground: PCB and solder joint inspection is about assessing joints, hand tool use demonstrations covers generic tool handling and e-waste dismantling covers disassembly at volume. The cluster hub is maintenance, repair and tool-use data, and the section overview is the humanoid training data guide.
Task anatomy
- Fault confirmation. Power on, observe symptoms, measure rails with a meter; see measuring instrument use.
- Localization. Narrow to a component or area by probing, thermal imaging or visual inspection.
- Preparation. Secure the board, apply flux, set temperatures and choose tips.
- Removal. Desolder through-hole parts or lift surface-mount parts with hot air, taking care with neighboring components.
- Replacement. Align the new part, tack, solder, and clean residue.
- Verification. Visual check under magnification, functional test and sometimes re-measurement.
Repair centers and refurbishers differ in how they work and what they record; see repair shop networks as suppliers. Failure modes include lifted pads, bridged pins, cold joints, heat damage to neighbors, component misorientation and static damage. A repair where the first attempt failed and was corrected is more informative than a clean one, and rarely appears in polished training material.
What each part of the data can support
| Data | Can support | Cannot show |
|---|---|---|
| Repair records and symptoms | Fault vocabulary, parts frequency, outcomes | Motion or technique |
| Board photos before and after | Defect and repair recognition | Process |
| Magnified video of rework | Step segmentation, tool recognition, joint appearance changes | Tip temperature, contact force, robot action |
| Overhead and side video | Posture, tool paths in coarse form | Fine contact |
| Test logs and measurements | Success criteria and diagnosis | Hand skill |
Public references, with limits
DeepPCB offers 1,500 image pairs, each a template and a test image, annotated with six common PCB defect types (Tang et al., 2019). It is an inspection dataset of printed circuit boards, not a repair or soldering dataset, and it contains no process footage. Use it as a format reference for template-and-test image pairing and defect annotation.
IPC-A-610 provides visual acceptance criteria for electronic assemblies, covering solder joints, placement, cleanliness, coating and marking, in three product classes: general, dedicated service and high performance, with Class 3 the strictest (IPC, 2024). Buyers sometimes use such criteria to define pass or fail labels for repaired joints. Which class applies to a given product is a question for the supplier, and the standard's text must be obtained from IPC; this page does not restate its detailed criteria.
Choosing the repair level
"Electronics repair" covers very different work. Module swaps, such as replacing a battery, screen or connector board, use screwdrivers, spudgers and adhesive handling at normal desk scale. Component-level repair uses soldering and hot air at millimeter scale. Microsoldering, such as repairing pads and tiny traces, goes smaller still. Each level has different cameras, tools and error costs. A buyer should state the level, since a dataset of module swaps says little about component rework, and refurbishers who process large volumes usually do the former far more than the latter.
Limits: what this data cannot support
It cannot train fine-motor control. Soldering and rework depend on micro-scale force, heat transfer and timing that video of a human does not record. A humanoid hand with current actuators may not match the precision of a technician's hands, so many teams would test whether the task suits a specialized tool or end effector. Control training needs robot or teleoperation data, plus instrumented tools such as temperature-logging irons; see instrumented tools.
It does not capture parameters unless recorded. Tip temperature, solder alloy, flux type and dwell time are usually invisible in video and must be logged or narrated.
It reflects one shop's methods. Technique and standards differ by employer, product class and training.
Existing footage is rare. Repair centers seldom record their bench work, so step-level video is typically a commissioned collection.
Illustrative capture specification
The following is an illustrative specification with synthetic values, not an existing dataset.
# Illustrative only: synthetic values, not a real or available dataset
task: smd_capacitor_replacement
board_family: board_type_17
views:
microscope_feed: {resolution: 1920x1080, fps: 30, scale_bar: yes}
overhead: {fps: 60}
side_on_tool_tip: {fps: 60}
logged_parameters: [iron_tip_temp_setpoint, hot_air_flow, solder_alloy, flux_type, dwell_time]
labels:
steps: [flux, heat, lift, clean_pad, place, solder, inspect]
outcome: pass / rework / fail
acceptance_ref: "supplier's stated class and criteria"
sync: shared clock across cameras and logger
excluded: robot actions and forces
Quality and acceptance criteria
Examples a buyer can set: every step visible in at least one view; parameters logged for every attempt; timestamps synchronized across views to a tolerance you specify; outcome labeled by a reviewer applying a stated acceptance criterion; failed attempts retained. Synchronization and calibration matters because actions cannot be reconstructed from unsynchronized views.
Permissions and provenance
Boards from customers can carry personal data in storage, device identifiers and proprietary firmware. Repair records identify customers and devices. Recordings show technicians' faces and voices. Ask what data was on devices when recorded, what customers agreed to and whether manufacturers' authorized-repair agreements restrict use of procedures, schematics or service content. See worker consent and identifiable people. This is not legal advice.
Questions to ask suppliers
- Which device categories and board types do your technicians repair most?
- Are repair records structured, with fault, action and result fields?
- Do you hold before-and-after photos, and at what magnification?
- Would technicians repeat tasks under a microscope camera with recorded parameters?
- Which acceptance criteria, such as IPC classes, apply to your work?
- Are failed rework attempts documented?
- What customer data or firmware could appear in recordings?
- Do authorized-repair or manufacturer agreements limit sharing procedures?
- How would you handle static-safe practice during recording?
Discuss your requirements
If you need electronics repair records or magnified repair recordings, describe the device types, views and permitted uses to SourceX. SourceX looks for US businesses that hold suitable material, checks the data and the supplier's licensing permissions, and coordinates the license and delivery. If no existing asset fits, SourceX can scope a commissioned collection of new recordings. A request does not guarantee a match. See also the physical-world collection path and the catalog page on first-person video of manual work.
Describe the workplace demonstration data your robotics project needs
Sources
- Online PCB Defect Detector On A New PCB Defect Dataset — DeepPCB image pairs and defect types, as an inspection format reference.
- IPC-A-610 Acceptability of Electronic Assemblies — visual acceptance criteria and product classes.