People new to hardware tend to picture the printed circuit board as the design. It is not. The board is a manufacturing document. The design is the schematic — the diagram that says which parts exist, what they are, and how they connect — and by the time the schematic is signed off, roughly 70 to 80 percent of the product's eventual bill of materials cost, its power consumption, its certification burden, and its supply chain risk are already locked in. Layout can make a good schematic work badly. It cannot make a bad schematic good.
Understanding what happens in this phase is worth the time, because it is the last point where changes cost hours instead of weeks.
Block diagram before any part number
Good schematic work starts one level above the schematic. Before a single component is placed, the engineer draws boxes: the compute block, the power tree, the sensors, the radio, the user interface, the connectors, the protection. Arrows between them carry a label — a bus name, a rail, an interrupt, a current. That drawing fits on one page and answers the questions that matter most:
- What is the compute block and does it have enough peripherals, pins, memory, and headroom? Choosing that part sets the toolchain, the price floor, and the availability risk for the next five years, which is why picking a microcontroller deserves more than an afternoon.
- How many voltage rails exist and where does each load sit?
- Which buses connect what, and are there address collisions or voltage-level mismatches waiting to happen?
- Which functions are worth buying as a certified module and which are worth designing from parts?
Reviewing that diagram with the founder or product manager is the highest-leverage meeting in an electronics project: vague requirements like "it should last a while on a charge" become numbers here, and numbers change the architecture.
Part selection is the real work
Drawing a resistor takes seconds. Deciding which resistor takes judgment, and on the parts that matter it can take a day. For each significant component the engineer is weighing:
- Availability today and stocked quantity, not just a listing on a distributor page. A part with 400 units in stock worldwide is a part your product cannot be built from.
- Lifecycle status. Manufacturers publish product change notices and longevity commitments. A part already flagged "not recommended for new designs" will force a redesign inside two years — the exact scenario described in what to do when a component goes end-of-life.
- A genuine second source. For every critical part, is there a drop-in alternative from a different manufacturer with the same footprint and pinout? Where there is not — often true of the microcontroller and the radio — that is a risk you accept knowingly rather than discover during a shortage. Planning around this is covered in designing a product around component shortages.
- Package and assembly implications. A 0201 passive, a fine-pitch BGA, or a bottom-terminated part with no visible fillet raises assembly cost and inspection difficulty.
- Certified modules versus discrete design. A pre-certified radio module costs several dollars more per unit and saves $15,000 to $40,000 in certification. At low volume that trade is obvious; at high volume it flips.
The output of this thinking is not just a diagram — it is a first real bill of materials with manufacturer part numbers, and the first honest cost estimate the project has had.
What a finished schematic package contains
A schematic phase that is genuinely complete delivers more than a PDF of pages.
| Deliverable | Why it matters |
|---|---|
| Hierarchical schematic sheets | One function per sheet — power, MCU, sensors, connectors — so a reviewer can follow a signal instead of hunting a spaghetti page |
| Full BOM with manufacturer part numbers | Feeds quoting, availability checks, and cost modeling |
| Power tree diagram | Shows every rail, its source, and its budgeted current |
| Pin assignment and net list | Lets firmware start before boards exist |
| ERC report, clean or annotated | Proves the electrical rule check was run and every waiver was deliberate |
| Layout guidance notes | Tells the layout engineer what is critical: high-current loops, differential pairs, keep-outs, thermal pads |
Net naming, hierarchy, and reference designators
These sound like housekeeping. They are not. A net called VBUS_5V0 on four sheets is a connection; a net called VBUS on one sheet and V5 on another is two nets, and the board gets built with the connection missing. Sensible naming, sheet hierarchy, and designators grouped by function are what make a schematic reviewable by a second engineer, debuggable by a technician six months later, and translatable into an assembly document a contract manufacturer can follow.
Decoupling by inspection, not by habit
Most schematics sprinkle 100 nanofarad capacitors near every IC because the reference design did. Better practice is to read each datasheet, note what the part actually asks for, and understand where return current for its fastest edges will flow. This is also where you annotate layout intent: this capacitor within two millimeters of pin 14, this ground a single-point connection. Layout engineers cannot infer intent nobody wrote down, and the results are catalogued in first PCB design mistakes.
Simulation and the parts worth checking early
You do not simulate a whole product. You simulate the three or four spots where getting it wrong costs a board spin: a switching regulator's loop stability, an analog front end's gain and noise, a filter's corner frequency, a level-shifting or protection network's behavior at the extremes. An hour in a circuit simulator on a sensor amplifier is cheaper than a four-week respin, and simulating circuits before you build hardware covers where it pays and where it is theater.
The schematic review
Before layout begins, a second engineer who did not draw the schematic should walk it page by page against a checklist. A serious review asks:
- Does every IC's power and ground pin actually connect? Multi-pin power nets on large packages are a classic miss.
- Is every rail within the absolute maximum of every part on it, at maximum input voltage?
- Are all unused inputs tied off? Floating CMOS inputs oscillate and burn current.
- Do pull-ups exist on every bus that needs them, at the right value, exactly once? Three sets of I2C pull-ups on one bus is a real and common failure.
- Are I2C addresses unique, and are chip-select and interrupt lines mapped to pins the firmware can actually use?
- Is there a programming and debug header? More first boards than you would believe arrive with no way to load firmware.
- Are there test points on every rail and every critical signal? They cost nothing and save days of bring-up.
- Does anything need a sequencing or reset supervisor, and is it there?
- Is polarity, ESD, and overcurrent protection present at every externally accessible connector?
- Does the design leave headroom — spare pins, spare flash, a spare gate — for the features that will inevitably be requested later?
A good review also brings in manufacturability early, since part choices, package types, and panelization constraints made here are hard to undo later. That is the argument for pulling DFM considerations for electronics into the schematic conversation rather than the layout one.
How long it takes and what it costs
For a modestly complex product — a microcontroller, a radio, a handful of sensors, a battery, and a charger — block diagram through reviewed schematic typically runs three to seven weeks and a meaningful fraction of the electronics budget. That feels slow to founders who expected to see a board. It is the cheapest phase in the project, and the one where an extra week routinely removes a month later on. The wider picture of how the phases stack up is in how long it takes to develop an electronic product.
Projects House treats the schematic as the deliverable it is: a reviewed architecture, a costed BOM with second sources, and written layout intent — not a drawing produced on the way to a board. If you are starting an electronics project or want a second set of eyes on a schematic before layout, get in touch through our contact form.