Breadboards, Prototyping, and the Bench
A breadboard is a tool for finding out whether an idea is wrong — not for finding out whether a circuit works.
🎯 By the end you'll understand
What a solderless breadboard is actually made of — contact resistance, row-to-row capacitance, and jumper inductance — with numbers, and how each one distorts a specific class of circuit. Why switching regulators and anything above a few megahertz produce results on a breadboard that do not transfer, and why 2.4 GHz does not work at all. How to decide, before you build, what a given prototype can and cannot prove, and which platform to use for each question. The bench supply's current limit as the cheapest bug-catcher you will ever own, ESD as a latent-damage mechanism rather than a dramatic one, and why the ESP32-C3-DevKitM-1 is the honest prototyping platform for Beacon and a dishonest one for measuring its sleep current. And how to build Beacon's Stage 1 milestone: LDO, LED, button, and the first current measurement.
KEY INSIGHTA prototype is an experiment, and every experiment has a domain of validity
Three days of chasing an unstable regulator. The layout was fine, the parts were right, the maths checked out. The instability lived in twenty centimetres of jumper wire and disappeared the moment the same circuit was soldered.
Breadboards are not bad. They are an instrument with a specification, like every other instrument in Lesson 1.7, and the mistake is not using one — it is using one without knowing what it adds. A solderless breadboard puts a few picofarads between every adjacent row, tens to hundreds of milliohms in every contact, and a hundred nanohenries or more in every jumper. For an LED and a resistor those numbers are invisible. For a 1.5 MHz switching regulator they are the dominant elements of the circuit, and the thing you build is not an approximation of your design — it is a different circuit that happens to share a schematic. The engineering skill this lesson installs is deciding, before you push the first wire in, which of your questions this apparatus can answer.
Ask of every prototype: what question am I asking, and does this platform's error budget leave the answer intact? A breadboard reliably answers questions about DC behaviour, logic, firmware, protocol, and rough current levels. It cannot answer questions about switching-regulator stability, fast-edge signal integrity, RF, or microamp-scale leakage, and no amount of care makes it able to. When the question is one it cannot answer, the right move is not to be more careful with the wiring — it is to change the platform, or to accept that this experiment is only testing whether an idea is wrong, not whether a circuit works.
CORE CONCEPTWhat a solderless breadboard is actually made of
Underneath the plastic, each row is a phosphor-bronze spring clip. A component lead or a jumper pushed in is held by friction against that spring. Nothing is soldered, nothing is specified, and no manufacturer publishes an electrical characteristics table — which by the standards of Lesson 1.6 tells you exactly how much to trust any number you are about to read.
| Parasitic | Typical magnitude | Where it comes from | What it ruins |
|---|---|---|---|
| Contact resistance, per contact | tens of mΩ when new and clean; hundreds of mΩ when worn, oxidised, or holding a thin wire | A friction joint between two metals, degrading with every insertion | High-current paths, precision references, anything where a millivolt matters |
| Row-to-row capacitance | a few picofarads between adjacent rows; more along the power rails | Parallel metal strips a couple of millimetres apart | Fast edges, high-impedance nodes, oscillators, 400 kHz I2C with weak pull-ups |
| Jumper loop inductance | roughly 1 nH per millimetre of loop; a 10 cm jumper and its return is 100–200 nH | A wire far from its return path — the loop is the inductor | Switching converters, decoupling, anything with fast di/dt |
| Surface and bulk leakage | tens to hundreds of nanoamps, occasionally microamps, rising sharply with humidity and flux residue | Conduction across the plastic between adjacent rows | Microamp-scale sleep-current measurement |
| Current handling | commonly quoted around 1 A per contact, limited by heating at the spring | The contact area is small and the spring is thin | Motor drivers, heaters, anything above a few hundred milliamps |
All of those figures are illustrative. Independent measurements published by hobbyists and engineers vary by an order of magnitude between boards and between contacts on the same board, and that variability is the specification: a breadboard is a component whose value you do not know and cannot control.
The arithmetic that decides everything
Two calculations turn those parasitics from trivia into decisions.
Inductive spikes. V = L · di/dt. A buck converter's switch node changes current by hundreds of milliamps in nanoseconds. Take 500 mA in 10 ns through the 150 nH of a jumper pair: V = 150 × 10⁻⁹ × 0.5 / 10 × 10⁻⁹ = 7.5 V. On a 3.7 V board. That spike is not in your design, it is not in the vendor's reference layout, and it will ring, radiate, couple into everything nearby, and quite possibly exceed the absolute maximum ratings of the regulator you are evaluating. The same converter on a 4-layer board with a tight loop sees a few nanohenries and a few hundred millivolts.
Capacitive slowing. A few picofarads per adjacent row sounds negligible, until you count the rows a signal passes through and add the probe from Lesson 1.7. An I2C bus routed across a breadboard easily accumulates 30–60 pF of extra capacitance, on top of the devices' own. With a 4.7 kΩ pull-up, 60 pF adds 2.2 × 4.7 kΩ × 60 pF ≈ 620 ns of rise time. At 400 kHz the bit period is 2.5 µs, so the rise consumes a quarter of it — the bus works, marginally, and behaves differently on the PCB where that capacitance is gone. Which direction the surprise goes depends on which side of the margin you were on, and that is the worst kind of surprise.
HOW IT WORKSWhich circuits lie on a breadboard, and in which direction
It matters enormously whether a breadboard makes a circuit look worse or better than it will be. A prototype that looks worse costs you time; a prototype that looks better costs you a board spin.
Looks worse than reality — you will waste time
- Switching regulators. Loop inductance produces ringing, overshoot, and EMI that the PCB will not have. You will chase instability that does not exist, add snubbers you do not need, and possibly conclude the part is unsuitable. This is the single most common misuse of a breadboard in modern embedded work, and the vendors know it — which is why every switching regulator ships with an evaluation module whose layout is the point.
- Fast digital edges. Ringing, crosstalk between adjacent rows, and reflections from unterminated stubs make clean signals look ugly. A 3 ns edge into 150 nH and a few picofarads rings visibly.
- High-current paths. Contact resistance in series with a load produces droop that reads as a weak supply or a bad regulator.
Looks better than reality — you will ship a bug
- Marginal edge timing. Breadboard capacitance damps ringing. A circuit that needs a series termination or a slower edge can look perfectly clean on a breadboard and ring badly on a tight PCB.
- Decoupling. The long, inductive supply wiring behaves as a filter, so a design with inadequate local decoupling may work on the breadboard and fail on the board where the supply is stiff and the transients are fast.
- Bus loading margins. An I2C bus with pull-ups that are too weak still works on a breadboard because everything is slow anyway; the same weakness on a PCB, with a hotter sensor and a longer bus, shows up as intermittent NACKs.
Does not work at all
- Anything at 2.4 GHz. At Beacon's radio frequency the free-space wavelength is about 125 mm, so a 30 mm jumper is a quarter-wave stub. The breadboard's rows are transmission lines with random impedances and resonances. Antenna matching, RF layout, and radiated-emission behaviour cannot be prototyped this way, and this is precisely why Beacon uses a pre-certified module with an integrated antenna: the RF design is done for you inside a part whose behaviour is characterised.
- Microamp-scale current measurement. Surface leakage across the plastic is comparable to Beacon's entire 8 µA sleep budget, and it varies with humidity and with how much flux and finger grease is on the board. You cannot measure 8 µA on a breadboard and believe it.
Solderless breadboard — fastest to change, worst fidelity. For firmware, logic, protocol, DC behaviour, and finding out whether an idea is wrong. Dev kit plus breakout modules — the same silicon and, crucially for a wireless product, the same certified radio module as the product, with jumpers between boards. Excellent for firmware and radio behaviour, poor for power measurement because the kit's own regulator and USB interface dominate. Soldered protoboard or stripboard — slower to change, much better contacts and shorter leads, still no ground plane. Good for analog, moderate-current, and anything you need to run unattended for a week. Two-layer PCB with a ground pour — a real board, cheap and quick from any fabricator, and it removes almost every parasitic in the table above. This is the right answer far more often than people expect: at prototype quantities a small two-layer board costs less than a decent dinner and arrives in days. The real board — Beacon's 4-layer, 35 × 25 mm Rev A. The only platform on which the RF, the power integrity, and the sleep current are the product's own.
SEE ITPrototype fidelity checker — decide what the experiment can prove
Choose what you want to learn and how you plan to build it. The checker estimates how much of the result will transfer to Beacon's 4-layer PCB, names the artefacts you will see that are not real, and names the real behaviours the platform will hide from you. The scoring model is a weighted sensitivity estimate, not a simulation — its job is to make the trade explicit before you spend a day on the wrong apparatus.
Things to try. Start with the LED on a breadboard: over eighty percent, and the missing part is contact resistance you do not care about — run it, and stop thinking. Move to the 400 kHz I2C bus and the score collapses, because row-to-row capacitance is the dominant term and it is exactly the parameter the experiment is about; then set the discipline to short wires and watch it partially recover, which tells you that this experiment is worth running if you build it carefully. Now pick the buck converter on a breadboard: single-digit percent, dominated by loop inductance and return-path integrity, which is the checker's way of saying that three days spent here will teach you about jumper wire. Switch the platform to a two-layer PCB and the same experiment becomes worth doing. Then pick the 8 µA sleep measurement: the breadboard fails not on inductance or capacitance but on leakage, a completely different mechanism, and adding a multimeter in series makes it worse rather than better. Finally pick the antenna match on anything other than the real board and read the verdict, which is the only one in the widget that says the experiment cannot be repaired.
IN PRACTICEThe bench: a current limit, a wrist strap, and some discipline
The bench supply's current limit is the cheapest debugging tool you own
A bench power supply has two knobs, and most beginners use one of them. Set the voltage, then set the current limit to roughly one and a half to two times what the circuit should draw. Now a short circuit, a reversed part, a solder bridge, or a firmware bug that turns on two MOSFETs at once does not destroy anything: the supply drops out of constant-voltage into constant-current mode, the voltage sags, an indicator lights, and you look at the board instead of at smoke. For Beacon's breadboard, a limit of about 100 mA is generous for a 40 mA burst and tight enough to catch a wiring error before it matters.
Two refinements worth adopting. First, set the limit before connecting the board, by shorting the output leads together and turning the current knob until the display reads what you want; then remove the short. Second, when a new board is powered for the first time, bring the voltage up slowly from zero and watch the current — a board that draws 300 mA at 1 V has a short, and you will know before anything has enough energy to be damaged. This is the first step of Lesson 6.1's bring-up procedure, and it is the reason experienced people are relaxed about powering an unknown board.
ESD: not dramatic, and that is the problem
Electrostatic discharge is modelled for testing purposes as a charged human body — about 100 pF through about 1.5 kΩ in the human body model — and walking across a synthetic carpet in dry air can charge a person to several kilovolts. A CMOS gate oxide is tens of nanometres thick and breaks down at a few tens of volts across it, so the arithmetic is not close. The reason ESD is under-respected is that the dramatic failure is rare: modern parts have on-chip protection structures that survive most events. What they do not always survive intact is the latent damage — a partially punctured oxide or a degraded junction that works today, works through your testing, and fails in the field weeks later, at which point the failure looks like anything but ESD.
The proportionate response for a 25-unit pilot is not a full static-controlled workstation. It is: a grounded wrist strap when handling bare boards and loose parts, parts left in their antistatic packaging until they are placed, a mat if you have one, and awareness that dry air and synthetic clothing raise the risk sharply. Beacon's own defence is at the design level — the canon's PRTR5V0U2X TVS on USB VBUS and the 22 Ω series resistors on the D± pair — because the field is where the product will meet static, not the bench.
Wire discipline, and why it is not aesthetics
Neat wiring on a breadboard is usually presented as a matter of pride. It is actually a matter of loop area. A jumper that goes straight from A to B, lying flat against the board, has a small loop with its return; a rainbow jumper that arcs twenty centimetres through the air has a large one, and Lesson 1.4's inductance arithmetic applies directly. Four rules that are all the same rule:
- Use solid-core wire cut to length and pushed flat, not pre-made flexible jumpers, for anything carrying current or fast edges.
- Run power and its return adjacent to each other, not around opposite sides of the board. The loop between them is what matters, not the length of either.
- Put a 100 nF capacitor across the supply pins of every active device, on the closest rows available, and a 10 µF bulk capacitor where power enters the board.
- Keep the supply for a noisy load — a motor, an LED string, a radio — on a separate path from the supply for anything sensitive, joined at one point near the source. This is the star-ground idea, and it is the breadboard version of Stage 5's plane discipline.
Beacon prototypes on the ESP32-C3-DevKitM-1, which carries the same ESP32-C3-MINI-1 module the product uses. That is what makes it the honest platform: the silicon is the silicon, the certified radio module with its PCB antenna is the same part, and firmware written against it is firmware that will run on Rev A. Where it is dishonest is power. The kit has its own LDO, a USB-serial interface, an on-board LED, and pull resistors, and together these draw orders of magnitude more than the module's 5 µA deep-sleep current. Any current measured at the kit's USB connector is a measurement of the kit. To measure the module you must break out its own supply pin and measure there — which is exactly why Beacon's Stage 1 milestone builds an LDO on a breadboard rather than using the kit's, and why the canon's 8 µA figure comes from the real board and not from this one.
BEACONBuilding Beacon's first breadboard, and taking the first current measurement
Stage 1's milestone is small on purpose: an LDO, an LED, and a button on a breadboard, plus a first current measurement compared against a back-of-envelope budget. Every part of that is a decision this stage has already taught, and every one of them is deliberately inside what a breadboard can prove.
The regulator: an XC6220 LDO, which is not the product's regulator
The canon is explicit that Beacon ships with a TPS62840 buck and that the XC6220 LDO is used on the breadboard in Stages 1 to 3. That is not a compromise, it is the correct application of this lesson: a buck converter cannot be meaningfully prototyped on a breadboard, and an LDO can — it has no switching node, no inductor, and no loop whose area matters. The LDO's cost is its quiescent current, around 25 µA typical, which is three times Beacon's entire sleep budget and would make the six-month requirement impossible. But Stages 1 to 3 are not measuring sleep current; they are running firmware. Using the wrong regulator on purpose, for the stages where its wrongness does not affect the question, is exactly the judgement this lesson is about. Lesson 4.2 makes the swap, on a real board, where the buck can be evaluated honestly.
The LED and its ballast: Lesson 1.2's arithmetic, on real parts
A plain LED with a series resistor, not the WS2812B the product uses — the addressable part needs a data protocol and has its own controller drawing hundreds of microamps, neither of which teaches anything at this stage. Size the resistor from the supply, the forward voltage, and the target current, then check the resistor's dissipation against its package rating with the 50 % rule. On a breadboard the contact resistance adds perhaps a few hundred milliohms in series with a 130 Ω ballast, which changes the current by well under a percent. The experiment is comfortably inside the platform's error budget, which is why it is the first thing you build.
The button: a pull-up, a series resistor, and a first look at bounce
A mechanical switch does not close once; its contacts bounce for something on the order of a millisecond to several milliseconds, producing a burst of edges. On a breadboard you can see this on a scope and measure it — and the measurement mostly transfers, because bounce is a mechanical property of the switch, not of the wiring. What does not transfer is the exact RC of any hardware debounce you add, because the breadboard's stray capacitance is part of it. So measure the bounce here, and implement the debounce in firmware (Lesson 3.3), where the platform cannot lie to you. Note also that the canon puts Beacon's button on GPIO9, a strapping pin, and holding it at power-up confuses boot mode — a defect worth reproducing deliberately on the breadboard so that you meet it here rather than in the field.
The first current measurement, and what it is and is not worth
Put a meter in series with the LDO's input and measure the board's current with the LED on, then with everything idle. The active figure is meaningful: tens of milliamps is well above the breadboard's leakage and contact effects, and the meter's burden voltage on a milliamp range is a few millivolts. The idle figure is not meaningful at the microamp scale, for three independent reasons — the LDO's own 25 µA swamps the module, the breadboard's surface leakage is comparable to the number you are looking for, and the meter's burden voltage on a microamp range may prevent the board from running at all. Write the active measurement in the log with its conditions (Lesson 1.7); write the idle measurement as "greater than 25 µA, dominated by the LDO, not usable as a sleep-current figure". That second sentence is the whole lesson, and putting it in the log rather than a number is what keeps a wrong figure from propagating into Stage 4.
Compare against the back-of-envelope budget
Lesson 1.6's widget assembled Beacon's budget from datasheet rows: about 8 µA of sleep and about 30 µA of averaged radio activity, for roughly 38 µA. The breadboard cannot confirm the 8 µA. What it can confirm is the active-window figure's order of magnitude — that the module plus sensors during a wake draw tens of milliamps, not hundreds — and that is worth checking early, because the active term is the dominant one in the budget and an order-of-magnitude error there would invalidate the whole design. Confirming the big term on a platform that can measure it, and explicitly deferring the small term to a platform that can, is a better use of a week than trying to measure everything badly at once.
At the end of this stage you have: a requirements table with numbers (Lesson 1.9), a breadboard that demonstrates the LDO, LED, and button working under firmware, a measured active current that agrees with the budget's dominant term to within an order of magnitude, and a written statement of which numbers the breadboard could not produce and why. That last artefact is the one that distinguishes an engineer from someone with a working breadboard, and it is what Stage 2 builds a schematic on top of.
WEAKNESSWhere this bites
This is the expensive one, because the conclusion feels earned. You spent two days, you have waveforms, the regulator really did oscillate. What you measured was a circuit containing your regulator and about 300 nH of wire that is not in the design, and every switching regulator on the market misbehaves with 300 nH in its switching loop. The reflex to build instead: before blaming a part, ask what the platform contributed, and whether the vendor's evaluation module — which exists precisely because layout is part of the design — behaves the same way. If the evaluation board is clean and your breadboard is not, the difference is the apparatus. Two-layer prototype PCBs are cheap and fast enough that "build a small board" is a reasonable response to this question rather than an escalation.
Engineers scrutinise bad results and accept good ones, and a breadboard's parasitics are perfectly capable of producing a good result that will not survive the PCB. Extra capacitance damps ringing, so a signal that needs a series termination looks fine. Inductive supply wiring filters transients, so inadequate decoupling passes. Slow edges everywhere mean marginal setup times are never tested. The habit worth building: when a prototype works on the first try, ask which parasitic might be helping, and design the PCB as though it were not there. Beacon's I2C pull-up values are chosen against the specification's rise-time limit for 400 kHz and the expected PCB bus capacitance, not against what happened to work on the breadboard.
The tempting plan is: get the firmware working on a dev kit, get the sensors talking on a breadboard, then combine everything on the PCB. The parts that get deferred are always the ones the cheap platforms cannot test — power integrity, sleep current, RF, thermal behaviour — and they are also the parts most likely to require a board spin. The mitigation is not to prototype everything perfectly; it is to identify the questions your platforms cannot answer and schedule them explicitly. Beacon's plan does this: the buck is deferred to Stage 4 with a note that it needs a real board, the antenna is deferred to Stage 5 with a note that it needs the module at a board edge with a defined keep-out, and the sleep current is deferred to a measurement with the right instrument on Rev A. Each of those is a known unknown with a date, rather than a surprise.
KEY INSIGHTHow this connects
- Passive Components for Real (1.4) — a breadboard is an exercise in unintended parasitics: contact resistance, row capacitance, and loop inductance are the same physics as ESR, ESL, and stray C, arriving uninvited.
- Diodes, Transistors, and MOSFETs as Switches (1.5) — the LED and button circuit built here is the first hardware the switching lesson's parts appear in, and the breadboard's contact resistance is negligible against a 130 Ω ballast for exactly the reasons that lesson's arithmetic gives.
- Measurement (1.7) — every instrument you attach to a prototype adds its own parasitics to the platform's, and the widget above scores them together because that is how they arrive.
- Beacon's Requirements (1.9) — next lesson turns the whole stage into numbered requirements, including which of them can be verified on which platform.
- Decoupling, Grounding, and Return Paths (2.5) — the star-ground discipline described here as a breadboard habit becomes a layout rule with a physical justification.
- Regulators: LDO vs Buck vs Boost (4.2) — where the deliberately-wrong breadboard LDO is replaced by the product's buck, on a platform that can evaluate it.
- Board Bring-Up: Smoke Test to First Blink (6.1) — the current-limited slow power-up described here is the first step of a formal bring-up procedure.
The habit to carry forward is a single question asked before every experiment: what would have to be true for this result to transfer? Ask it about breadboards, about simulations, about test firmware, and about the 25-unit pilot in Stage 6. It is the same question every time, and answering it honestly is most of what separates a fast project from a slow one.
GLOSSARYKey terms introduced
- Parasitic
- An electrical property present because of physical construction rather than by design — contact resistance, stray capacitance, loop inductance, surface leakage. Unavoidable, and either negligible or dominant depending on the circuit.
- Loop inductance
- The inductance of the closed path a current takes, set by the area enclosed between the outgoing conductor and its return, not by the length of either alone. Roughly 1 nH per millimetre of loop for typical wiring.
- Prototype fidelity
- The degree to which a result obtained on a prototyping platform predicts the behaviour of the production hardware. A property of the pairing between question and platform, not of the platform alone.
- Surface leakage
- Current conducted across an insulating surface between adjacent conductors, worsened by humidity, flux residue, and contamination. Tens to hundreds of nanoamps on a breadboard — significant only when the signal of interest is comparably small.
- Current limit (constant-current mode)
- A bench supply's second regulation mode, entered when the load demands more than the set current. Converts short circuits and wiring errors from destructive events into an indicator light.
- Latent ESD damage
- Partial damage from an electrostatic discharge that leaves the part functional but degraded, producing a field failure weeks or months later that is almost impossible to attribute to its cause.
- Star ground
- Joining the returns of noisy and sensitive circuits at a single point near the supply, so that a noisy load's return current does not flow through a sensitive circuit's reference. The breadboard ancestor of ground-plane discipline.
- Evaluation module
- A vendor-built board carrying a part with a reference layout. Exists because for switching and RF parts the layout is part of the specification, and it is the correct comparison point when a part appears to misbehave.
ACTIVE RECALLCheck your understanding
Answer in your own words first, then reveal.
1A colleague reports that the TPS62840 "oscillates badly" on their breadboard and proposes choosing a different regulator. Estimate the magnitude of the breadboard's contribution and say what you would ask them to do next.›
Estimate first. A breadboard's jumper pairs contribute on the order of 100–200 nH in the switching loop. A buck's high-side switch turns on and off with current edges of hundreds of milliamps in a few nanoseconds; taking 500 mA in 10 ns through 150 nH gives V = L·di/dt = 7.5 V of spike on a 3.7 V board. That is not a subtle perturbation — it exceeds the supply, will ring at whatever the loop's resonance is, will couple into the feedback node, and may exceed the part's absolute maximum ratings. Any switching regulator would misbehave under those conditions, so the observation carries no information about this particular part. What to do next, in order of cost: (a) look at the vendor's evaluation module and its layout, and if possible run the part on one — that is the controlled experiment, and it is what the module exists for; (b) if the design is otherwise ready, spin a small two-layer PCB with a tight switching loop, which at prototype quantities costs less than the time already spent; (c) if neither is possible, at minimum rebuild it on soldered protoboard with the input capacitor's loop as short as physically possible and re-measure. What I would not do is change the part, because we have no evidence about the part yet.
2Beacon's I2C bus works perfectly on the breadboard at 400 kHz with 10 kΩ pull-ups. Should you use 10 kΩ on the PCB? Explain both directions the surprise could go.›
No — the breadboard result does not license that choice, in either direction. The naive worry is that the breadboard adds capacitance, so the PCB will have less and be faster, meaning 10 kΩ is conservative and safe. That reasoning is right about the direction of the capacitance and wrong about the conclusion, because it treats a single working prototype as a margin measurement. The real analysis starts from the I2C specification, which sets a maximum rise time for fast-mode operation, and from an estimate of the PCB's bus capacitance — device pin capacitance, trace capacitance, and the connector or test points. Compute the required pull-up from the rise-time limit and that capacitance, then check the low-level requirement: the pull-up must be large enough that the device sinking the line can pull it below its VOL at the specified sink current. Those two constraints bracket the value, and the breadboard is not one of the inputs. The direction that actually bites is temperature and units: a bus that is marginal by design works on the bench and fails on the warm units in the field, and Lesson 1.7's answer to that failure mode was to put a scope on the edges rather than a logic analyzer. Note also that if the breadboard's extra capacitance had made a good design look marginal, the temptation would have been to reduce the pull-ups unnecessarily and pay for it in current — a smaller error, but the same mistake.
3Why does Beacon deliberately prototype with an XC6220 LDO whose 25 µA quiescent current makes R1 impossible, rather than prototyping with the buck it will actually ship?›
Because Stages 1 to 3 are not asking a power question. They are asking firmware questions: do the drivers work, does the state machine sleep and wake correctly, does the interrupt arrive within R4's budget, does the BLE payload assemble. All of those are answerable on a breadboard, and none of them depend on which regulator supplies 3.3 V. The LDO is chosen because it is the regulator that a breadboard can support honestly — no switching node, no inductor, no loop area that matters — so it contributes no artefacts to the questions being asked. The buck cannot be prototyped this way at all, as the first question in this set established. The 25 µA is not a problem to be tolerated; it is a parameter that is simply irrelevant to these stages, and it is documented as such so that nobody accidentally measures sleep current on this platform and believes it. The general principle: choose the prototype platform and the prototype parts to suit the question, and write down which questions you have therefore deferred. Beacon defers the buck to Lesson 4.2, on a real board, where it can be evaluated on its merits.
4You need to know Beacon's active-window charge before committing to a battery size, and Rev A does not exist yet. Design the best experiment you can with a dev kit, a breadboard, a bench supply, and a scope, and state exactly what it does and does not prove.›
Build: the ESP32-C3-DevKitM-1 for the module, with the sensors on a breadboard over I2C. Bypass the kit's own USB power entirely — power the module's supply pin from the bench supply through a 1 Ω sense resistor on the low side, with the kit's LDO and USB interface out of the path. Set the supply's current limit to about 100 mA. Instrument the firmware with a marker GPIO taken high at the first line of the wake handler and low immediately before the sleep call. Trigger the scope on the marker's rising edge, DC coupling, a 10× probe with a short ground, timebase set to capture the whole 45 ms window, and check that the resulting sample rate (memory depth divided by capture time) still gives several samples across the fastest feature you care about. Then use the scope's integration function to get the area under the current waveform between the marker's edges — that is charge, in coulombs, and it is the number the battery actually sees. What it proves: the active-window duration and its charge, to good accuracy, because tens of milliamps is far above every breadboard parasitic and the 1 Ω shunt's 40 mV burden is harmless. Since the active term dominates Beacon's budget, this validates the largest number in it. What it does not prove: the sleep current, which is below the breadboard's leakage and the kit's own consumption; the buck's efficiency, since there is no buck; RF range or antenna performance, though the module's own radio and antenna are the product's; and anything about the PCB's power integrity. Record all four exclusions in the log next to the result, because the charge figure will be quoted for the next two stages and whoever quotes it needs to know what it excludes.