How CADY Is Helping Engineers Build Better Products

Engineering workspace illustrating how CADY is helping engineers build better products, with highlighted schematic connections and a prototype PCB.

CADY helps engineers build better products by adding an AI schematic review between schematic capture and manufacturing.

It takes the netlist and bill of materials (BOM), reads the component datasheets, and checks the circuit against more than 1,000 rules covering connection integrity, voltage mapping, and communication protocols.

I still make every design decision, but the errors that tired eyes miss get flagged before they turn into a printed circuit board (PCB) re-spin.

I saw how much that matters on a healthcare Internet of Things (IoT) device built to help patients recovering from fractured legs regain their mobility.

That board worked on its very first power-up, and the final review I almost skipped is a big part of the reason.

How Does CADY Help Engineers Build Better Products?

CADY helps by giving every schematic an independent, datasheet-aware review before the design is committed to copper.

You upload a netlist and BOM exported from your electronic computer-aided design (ECAD) tool, and CADY returns an interactive report of errors, warnings, best-practice notes, and component insights.

Standard analysis needs only the netlist and BOM, with no component libraries or manual data entry, so starting a run takes about five minutes.

The analysis itself takes from a few minutes to a few hours depending on the design, and reviewing the finished report takes 30 minutes or less, which makes it practical to run on every revision instead of only at the final gate.

If CADY cannot obtain or read a component’s datasheet, the report marks that part as unsupported and lets you upload the datasheet, while the analysis continues on the rest of the circuit.

For field-programmable gate arrays (FPGAs) and microcontrollers (MCUs) with multi-purpose pins, an upload interface lets you assign the configuration you chose for each pin.

The engineer then decides which findings need a design change, which are acceptable, and which are false alarms.

Why Do Experienced Engineers Still Miss Schematic Errors?

Experienced engineers miss schematic errors because familiarity blinds them, not because they lack knowledge.

After days or weeks inside the same design, your brain starts reading what you intended to draw instead of what is actually on the sheet.

A net label with a one-character typo, a pin mapped from the wrong package variant, or an enable pin left floating can all look correct on the hundredth pass.

Peer review helps, but colleagues are busy, and a reviewer who is new to the design rarely has time to open the datasheet behind every symbol.

That is the gap a tireless second reviewer fills, because it checks the last sheet with the same attention as the first.

What Does Datasheet-Based Schematic Verification Catch?

Datasheet-based schematic verification catches errors where two pins are connected correctly in the drawing, but the connection breaks a rule written in the component’s datasheet.

CADY’s published checks cover connection integrity, voltage mapping, and communication protocol errors, plus component insights powered by SiliconExpert.

These are the categories standard rule checks struggle with, because the rule lives in a PDF rather than in the schematic symbol.

The examples below are typical of each category, not findings from a specific board.

Is That 1.8V Signal Really a Logic High?

A 1.8V output driving a 3.3V complementary metal-oxide-semiconductor (CMOS) input is often not a valid logic high.

Texas Instruments specifies a minimum high-level input voltage (VIH) of 2V for its SN74LVC1G04 inverter at a 3V to 3.6V supply, so a 1.8V signal sits below the guaranteed threshold.

The circuit may work on the bench at room temperature and then fail intermittently across temperature or supply variation.

The fix is a level translator, such as TI’s SN74LV1T125, which supports 1.8V to 3.3V up-translation at a 3.3V supply, or a receiver whose input thresholds suit the lower voltage.

This is the kind of mismatch a voltage mapping check exists to flag, because it compares the driver’s output level with the receiver’s input thresholds on each net it analyzes.

Are the I2C Pull-Ups Present and Sized Correctly?

A typical inter-integrated circuit (I2C) bus relies on pull-up resistors, because its serial data (SDA) and serial clock (SCL) lines are open-drain and devices can only pull them low.

The NXP I2C-bus specification connects both lines to the positive supply through a pull-up resistor or a current source.

The value matters too: TI’s I2C bus pull-up resistor calculation note sets the minimum by sink current, Rp(min) = (VCC – VOL(max)) / IOL, and the maximum by rise time, Rp(max) = tr / (0.8473 × Cb).

For a fast-mode bus at 3.3V with 200pF of bus capacitance, the note’s worked example allows any resistor between about 967Ω and 1.77kΩ.

A 10kΩ pull-up copied from an older design would give a rise time of roughly 1.7µs on that bus, far beyond the 300ns fast-mode limit.

Missing pull-ups, pull-ups tied to the wrong rail, and parallel module pull-ups whose combined resistance falls below the minimum value are all classic communication protocol errors.

Is Every Part Still Recommended for New Designs?

A component can be electrically correct and still carry supply risk if its manufacturer has marked it not recommended for new designs (NRND) or end of life (EOL).

Texas Instruments, for example, defines an NRND part as available for existing designs, while recommending an alternate for new designs.

On an earlier development board, CADY flagged an ESP32-C3 as no longer recommended for new designs and an SS8050-G transistor as end of life, which let me update the BOM before production.

Lifecycle problems rarely show up in a manual schematic review, because nothing about the symbol looks wrong.

They can show up months later as a purchasing problem, when the part becomes hard to source, or the board needs a redesign.

Checking lifecycle status during schematic review costs minutes, while discovering it during a production build can stall the whole build.

How Is Agentic Schematic Review Different From ERC and DRC?

Agentic schematic review differs from an electrical rule check (ERC) and a design rule check (DRC) because it reads component datasheets, while ERC and DRC only check rules defined inside the ECAD tool.

ERC flags connectivity problems based on the pin types assigned in each symbol, such as unconnected inputs or two outputs driving the same net.

DRC checks the board against the constraints you set, such as clearances, trace widths and missing connections.

Basic ERC can catch a floating input when the symbol’s pin types are set correctly, but neither check reads datasheets, so both can miss device-specific requirements, such as how a particular regulator’s enable pin must be driven or whether a sensor’s input/output (I/O) pins tolerate 3.3V.

The comparison below shows where each method fits.

Review methodWhat it checks againstReads datasheets?Setup neededTypical blind spot
Manual peer reviewReviewer experience and the datasheets they openPartly, depending on reviewer timeNone, but hours of reviewer timeFatigue and datasheets nobody opened
ERC (in the ECAD tool)Pin electrical types defined in each symbolNoCorrect pin types in every symbolNets that are legally connected but electrically wrong
DRC (in the ECAD tool)Layout constraints you defineNoConstraint setup per boardSchematic logic and voltage errors
AI schematic analysis (CADY)Component datasheets plus 1,000+ built-in rulesYes, automatically for supported componentsUpload a netlist and BOMLayout, since it analyzes schematics only today

The methods work best together rather than as replacements for one another.

ERC and DRC remain essential gates, and a human reviewer still owns design intent and trade-offs.

AI schematic analysis sits between them and covers the datasheet-level checks that are too slow to do by hand on every net.

How Did an AI Schematic Review Get My Medical IoT Board Right the First Time?

A final AI schematic review before manufacturing flagged issues I had missed. I fixed them, and the first batch of boards powered up and worked without a single modification.

The project was a healthcare IoT device, so reliability was not optional.

Here is how it unfolded, from the design brief to the first power-up.

What Did the Medical IoT Hardware Design Involve?

The design was a healthcare IoT system for a PhD researcher in the United Kingdom, built to help patients recovering from fractured legs regain their mobility.

The device monitors movement and other health-related parameters during rehabilitation.

The hardware combined sensing circuitry, wireless communication, embedded processing, power management, and all the supporting electronics on one board.

In a system like this, one wrong connection can corrupt the data the researcher depends on or take the device offline in the middle of a session.

Weeks of engineering went into it, and by the end the schematic and PCB layout were both complete and looked good.

Why Did I Run One More Review Before Manufacturing?

I ran one more review because the moment a design looks finished is the moment familiarity bias is strongest.

I could have generated the Gerber files and sent them straight to the fab, as many engineers do once the layout is clean.

Instead, I exported the netlist and BOM and ran the whole project through CADY.

Because the layout was already done, any schematic change could also mean rework in layout, which is why I now run the review before layout as well.

Even so, a late review is far better than discovering the problem on a powered board.

What Did the CADY Report Flag?

The report flagged several areas that deserved a second look, including issues I had not noticed and recommendations that made the design more robust.

None of the findings felt random, because each one pointed to a specific connection or component and the reason behind it.

I checked every item against the relevant datasheet, implemented the changes that made engineering sense, and verified the design again before finalizing it.

CADY did not design the board for me. It helped me verify the work I had already done.

What Happened When the Boards Came Back?

The boards worked on the first try, with no rework between fabrication and a fully functional prototype.

The files went to PCBWay for fabrication, and then came the waiting, refreshing the tracking page and replaying every design decision in my head.

When the boards arrived, my client assembled the hardware, loaded the firmware, and powered up the system.

There was no redesign, no second PCB order, no cut traces and no blue jumper wires pretending to be design improvements.

For a hardware engineer, a board that does exactly what it was designed to do on the first power-up is one of the most satisfying moments in the job.

Why Does First-Pass PCB Success Matter So Much for Medical Devices?

First-pass PCB success matters for medical devices because every re-spin pushes back the testing, studies, and approvals that depend on stable hardware.

For medical electrical equipment, IEC 60601-1 sets general requirements for basic safety and essential performance.

Any hardware change made after verification work has started raises the question of which of those tests need to be repeated.

Research devices like my client’s add another constraint, because study schedules and funding timelines rarely leave room for a second fabrication cycle.

In my experience, one missed issue means at least another two weeks of fabrication, assembly, and shipping, plus the engineering time to diagnose it.

That is why the cheapest place to reduce PCB re-spins is the schematic, before a single board has been ordered.

Where Does AI Schematic Review Fit in a Hardware Design Workflow?

AI schematic review fits after schematic capture and ERC, before layout begins, with a second run just before Gerber release.

It does not replace peer review, simulation, or layout checks.

It adds a datasheet-level pass that is fast enough to repeat whenever the schematic changes.

When Should You Run an Agentic Schematic Review?

Run an agentic schematic review at schematic freeze, before layout starts, because a fix at that stage only needs schematic edits, while a fix after layout may also need placement or routing changes.

Run it again after every engineering change order (ECO) and once more before you release manufacturing files, since late changes are where new errors creep in.

On my healthcare project, the review happened after layout, which worked out, but meant fixes could also affect the layout.

CADY currently analyzes electrical schematics only, with a layout analysis module planned, so DRC and design for manufacturing (DFM) checks still cover the layout side.

How Should You Handle the Findings in the Report?

Treat every finding as a prompt to verify, not an instruction to change the design.

Open the datasheet section behind each flag, confirm whether the condition applies to your use case, and either fix it or record why it is acceptable.

CADY states an overall accuracy of about 90%, with a false negative rate of 2 to 3%.

Verify each flagged finding before acting on it, and keep in mind that a small share of genuine errors may still go undetected.

Marking items as reviewed and leaving comments in the report gives the rest of the team a record of what was checked and why.

How Can You Prevent PCB Design Errors Before Production?

You prevent PCB design errors before production by checking every net against the datasheet of each component it touches, then confirming the BOM matches the netlist.

The checklist below covers the items I verify on every board before release, whether by hand or through an automated tool.

  • Every supply pin connects to the correct rail, within the datasheet’s recommended operating conditions.
  • Every driver’s output levels meet the receiver’s VIH and low-level input voltage (VIL) thresholds, with a level translator where they do not.
  • Every open-drain line, including I2C, interrupts and resets, has a pull-up sized for its bus capacitance.
  • Every integrated circuit (IC) power pin has a decoupling capacitor, placed as close to the power pins as possible.
  • Enable, reset, and boot-mode pins are tied to a defined state.
  • Unused pins are terminated the way the datasheet specifies.
  • Every BOM line has a manufacturer part number that matches its netlist reference, and any NRND or EOL part has been reviewed for availability, with an alternative selected where appropriate.

A checklist like this is only as good as the time you have to apply it to every net, which is where most manual reviews fall short.

What Mistakes Lead to Avoidable PCB Re-Spins?

Process shortcuts let preventable errors reach manufacturing, even when the engineering itself is sound.

Skipping the final review because the schedule is tight is one of them, and it is exactly the call I almost made.

Reviewing only the sheets that changed after an ECO is another, because a change on one sheet can break a net that continues on another.

Copying a reference design without checking it against your actual parts, rails, and operating conditions carries its assumptions straight into your board.

Finally, treating a clean ERC as proof that the design is correct ignores everything ERC cannot see, such as voltage thresholds and datasheet requirements.

Want Your Next Board to Work on the First Power-Up?

Hardware engineering will never be perfect, but the best time to find a design issue is before your PCB is manufactured, and an AI schematic review is the most practical way I have found to do that on every board.

As an agentic schematic analysis platform, CADY reads your netlist, BOM, and component datasheets and checks them against more than 1,000 design rules to flag connection, voltage, and protocol errors before your board goes to fabrication.

Start a free trial of CADY or talk to the CADY team to run it on your own design.

Frequently Asked Questions

Can an AI schematic review reduce PCB re-spins?

Yes. An AI schematic review reduces PCB re-spins caused by schematic errors by checking the circuit’s connections against component datasheets before fabrication. It flags issues such as logic-level mismatches, missing pull-ups, and incorrect power connections that manual reviews often miss. It cannot prevent re-spins caused by layout, assembly or firmware problems, so DRC and DFM checks are still needed.

When should I run an agentic schematic review?

Run an agentic schematic review at schematic freeze, before PCB layout starts, because fixes at that stage only require schematic edits. Run it again after every engineering change order and once more before releasing Gerber files. Running it only at the end still catches errors, but fixes found after layout may require component, placement or routing changes.

Does AI schematic review replace a manual design review?

No. AI schematic review supports a manual design review rather than replacing it. The tool applies the same datasheet-based rules on every run, while engineers judge design intent, trade-offs and whether each flagged item actually applies. The strongest process combines ERC, an AI schematic review and a human peer review before the design moves to layout.

What should I check before sending a PCB to manufacturing?

Before sending a PCB to manufacturing, confirm that every supply pin connects to the correct rail, logic levels match between drivers and receivers, open-drain lines have correctly sized pull-ups, and every power pin is decoupled. Also check that enable and reset pins have defined states and that any NRND or end-of-life parts on the BOM have been reviewed for availability.

What does CADY check in a schematic?

CADY checks a schematic’s netlist and BOM against component datasheets and more than 1,000 custom-built rules. Its published checks cover connection integrity, voltage mapping and communication protocol errors, plus component insights powered by SiliconExpert. Components it cannot support are flagged in the report. It works with tools such as Altium, OrCAD, Xpedition and KiCad, and currently analyzes schematics only.

About the author

Joseph Ogbonna

Joseph Ogbonna is a PCB design expert specializing in schematic capture, board layout, and design reviews for electronics products. Beyond hands-on design work, Joseph mentors aspiring hardware engineers, helping them build practical skills in circuit and PCB design.
Schematic analysis

Check your next schematic before layout

Upload your Netlist and BOM, run the analysis, and review the findings in an interactive report. We never store your Netlist or BOM files, nor train models on them.

Back to
Share this White Papers & Case Study
Read Next

How CADY’s agentic technology detected two critical errors in the final stage of a schematic design, preventing the need for a re-spin

A leading French telecom company adopted CADY’s automated electrical analysis solution to reduce PCB re-spins, cut design costs, and accelerate time to market, after first validating its accuracy through controlled testing with intentional design errors.

CADY’s agentic technology detected a potential capacitor breach in WINT’s schematic design, leading to an ECO.

This case study shows how CADY’s analysis of WINT’s main board identified a critical issue, prompting an immediate ECO to replace a capacitor before the product left the assembly line.