Power & RF Design: Move Real Energy and Real Signals
This is the advanced path. Where the earlier modules were about getting a circuit to work, this one is about making it work efficiently and cleanly: converting voltage without wasting it as heat, driving motors with precision, getting power down a cable without reflections, and radiating (or receiving) it with an antenna. Each module pairs the theory with a live calculator so you can feel the design trade‑offs before you commit to a part.
Two ideas tie this whole path together. The first is energy storage: an inductor and a capacitor let you move energy in time, which is how you change voltage levels, smooth a supply, or feed a motor. The second is wavelength: once your signals get fast enough that the cable length matters, "a wire" becomes a transmission line and "an end of a cable" becomes an antenna. Master those two and power electronics and RF stop being separate subjects and become the same physics viewed at different scales. Work through the four modules; the calculators let you check each design against real numbers.
By the end you should be able to: size the inductor and capacitor of a buck converter, estimate a motor's speed and stall current from PWM duty, compute VSWR and mismatch loss for a load, and cut a dipole to frequency while estimating free‑space path loss.
1. Switching power supplies & converters
A linear regulator drops the difference between input and output across a transistor that behaves like a variable resistor, and that excess power becomes heat: Pwaste = (Vin − Vout)·I. Regulate 12 V down to 3.3 V at 1 A and you burn nearly 9 W as heat, so linear efficiency is often 50–70%. A switching converter avoids this by never sitting in the "partly on" region: a fast transistor (usually a MOSFET) is driven either fully on (tiny resistance, little loss) or fully off (no current, no loss), and energy is shuttled by an inductor (which stores it as a magnetic field) and a capacitor (which stores it as an electric field). The control variable is the duty cycle D — the fraction of each switching period the main switch is on. Pulse the input at a high frequency, filter the pulses with L and C, and you get a clean DC output, typically 80–95% efficient.
There are four basic topologies, distinguished by where the switch, diode, and inductor sit. The buck (step‑down) gives Vout = D·Vin, so the output is always below the input. The boost (step‑up) gives Vout = Vin/(1 − D), producing a higher output than the input. The buck‑boost inverts the polarity, |Vout| = D·Vin/(1 − D). The flyback is a buck‑boost that stores energy in a transformer instead of a plain inductor, which gives you galvanic isolation (no direct electrical connection between input and output) and easy multiple outputs — it is the workhorse of phone chargers and PC power supplies. In every case the inductor is the energy "bucket" and the output capacitor is the reservoir that keeps the voltage steady between switching pulses.
Designing one comes down to a few linked choices. Pick a switching frequency fsw: higher lets you use smaller L and C (cheaper, tinier) but raises switching and gate‑drive losses. Choose the inductor so its ripple current ΔIL is a sensible fraction (often 20–40%) of the load; for a buck, L = (Vin − Vout)·D / (fsw·ΔIL). Choose the output capacitor for acceptable voltage ripple (C ≈ ΔIL/(8·fsw·ΔV)). Then account for the real losses: switch conduction (Rds(on)·I²), the freewheeling diode (or a synchronous MOSFET), core losses in the inductor (hysteresis plus eddy currents), and ESR losses in L and C. Everything not delivered to the load becomes heat, so thermal management — derating parts, adding airflow or a heat sink, leaving output headroom — is part of the electrical design, not an afterthought. Good designs also add protection: over‑current, over‑voltage, over‑temperature shutdown, and a soft‑start so the output ramps up gently rather than inrushing.
Buck converter designer
Enter the rails, switching frequency, inductor and load, and get the duty cycle, inductor ripple current, and the output capacitance needed for your ripple target.
| Topology | Output relation | Step | Polarity | Isolated | Typical use |
|---|---|---|---|---|---|
| Buck | Vout = D·Vin | Down | Same | No | Logic rails, 5 V/3.3 V |
| Boost | Vout = Vin/(1−D) | Up | Same | No | Battery → higher rail |
| Buck‑boost | |Vout| = D·Vin/(1−D) | Either | Inverted | No | Wide‑input, fixed‑out |
| Flyback | set by turns ratio | Either | Either | Yes | Chargers, isolated supplies |
Go deeper — where the efficiency really goes
Losses split into conduction (I²R in the switch's Rds(on), the diode, the inductor's DCR, and the capacitor's ESR) and switching (energy lost each time the switch transitions, roughly ½·V·I·tsw per edge, so it scales with frequency). Add core loss, which depends on the flux swing, frequency, and the magnetic material (often modelled with Steinmetz's equation). The advanced answer to switching loss is soft switching: in a resonant (e.g. LLC) topology you time the switch to turn on when its voltage is already zero (ZVS) or its current is zero (ZCS), so there is almost no overlap of high voltage and high current. That is why modern high‑efficiency adapters and data‑centre supplies push toward resonant or phase‑shifted topologies rather than plain hard‑switched ones.
Picking an inductor whose saturation current is below the peak (load + ½·ΔIL) — once it saturates the inductance collapses and current runs away. Using a slow rectifier diode at high fsw (reverse‑reccovery spike). Forgetting the input capacitor, because the switch pulls a pulsed current from the supply rail. And skipping the thermal check — a "90% efficient" 100 W supply still dumps 11 W into the air.
2. Motor drives & generators
A DC motor is secretly a generator. As the armature spins in the magnetic field it generates a back‑EMF that opposes the supply, E = ke·ω, proportional to speed. The current is therefore I = (Vapplied − E)/R, and the torque is proportional to that current, T = kt·I. Two consequences follow immediately: at startup (ω = 0, so E = 0) the current is the large value V/R — that is your starting torque and your overheating risk if the motor is held stalled; and at steady speed the back‑EMF rises until the current (hence torque) drops to just what the load needs. To drive a motor both directions and control its speed, you use an H‑bridge — four switches in an "H" around the motor. Driving one diagonal pair sends current one way (forward), the other diagonal sends it the other way (reverse), and turning off both lets the motor freewheel (coast) or, with a dedicated path, brake.
Speed control in practice is PWM: the bridge is switched at a high frequency and the average voltage applied to the motor is Vavg = D·Vsupply, where D is the duty cycle. The motor's own inductance smooths the current, so it sees roughly the average voltage and the speed follows D — while the ripple current stays small. At light load the motor accelerates until its back‑EMF nearly equals Vavg and the current (torque) settles low. Precision drives go further: they measure speed or position with an encoder and close a feedback loop (PID) so speed holds under load. Brushless DC (BLDC) motors drop the mechanical commutator and brushes; a controller senses rotor position (Hall sensors, or "sensorless" from the back‑EMF) and energises the correct phase pair in sequence — a simple six‑step pattern, or, for smoother and more efficient running, field‑oriented control (FOC), which drives the motor in a rotating d/q reference frame. Stepper motors move in fixed angular steps and are used where precise open‑loop positioning matters.
On the generation side, a motor driven mechanically is a generator, and in industrial and grid equipment the standard is three‑phase power: three sinusoids 120° apart that deliver constant instantaneous power and use copper efficiently. Two facts are worth memorising: in a wye (star) connection the line voltage is √3 times the phase voltage (Vline = √3·Vphase), and swapping any two lines reverses the rotation of a three‑phase motor. In an induction motor the rotor turns slightly slower than the rotating field — the slip, s = (ns − nr)/ns, where the synchronous speed is ns = 120·f/P (P = number of poles). That small slip is what induces the rotor current that produces torque; with zero slip there would be no changing flux and no torque at all. Power factor — the phase relationship between voltage and current — determines how much of the apparent power is actually doing work, which is why motors and drives are rated on it.
Motor PWM estimator
Enter the supply, the motor resistance and back‑EMF constant, and the PWM duty to estimate the average voltage, no‑load speed, and the (dangerous) stall current.
| Type | Commutation | Control | Strength | Typical use |
|---|---|---|---|---|
| Brushed DC | Mechanical brushes | Voltage / PWM | Cheap, simple, high starting torque | Toys, appliances, simple drives |
| BLDC | Electronic (Hall / sensorless) | Six‑step or FOC | Efficient, long life, high power density | Drone props, EVs, tools |
| Stepper | Electronic, fixed steps | Step/pulse | Precise open‑loop positioning | 3D printers, CNC, plotters |
| Induction (AC) | Self‑starting (slip) | V/Hz or vector | Rugged, high power, self‑starting | Factories, pumps, compressors |
Go deeper — from six‑step to field‑oriented control
Six‑step commutation energises two of the three phases at a time in a fixed sequence — simple, but torque "pulsates" as the current jumps between steps. FOC (also called vector control) transforms the three phase currents into two orthogonal quantities in a frame that rotates with the rotor: d‑axis (aligned with the field, producing no torque) and q‑axis (90° ahead, producing torque). By holding the d‑current near zero and commanding the q‑current, the controller produces smooth, constant torque at any speed and can even drive negative q for regenerative braking. It needs a good estimate of rotor position and a fast current loop, which is why FOC drives use current‑sense resistors and run their control in a DSP at tens of kHz. The payoff is the silent, efficient, precisely‑torqued motion you get in a modern electric vehicle or servo.
Holding a motor stalled (locked rotor) draws V/R continuously and cooks the windings. Undersizing the FETs/freewheel diodes for the inrush and stall current. Leaving no freewheeling path for the inductive kick when the switches turn off — the resulting voltage spike can destroy the drive. Choosing a PWM frequency that is too low (audible whine, big current ripple) or too high (excess switching loss and EMI). And shooting through — ever driving Q1 with Q3, or Q2 with Q4 — which shorts the supply.
3. Impedance matching & transmission lines
At DC a cable is just a conductor. But when the signal's wavelength becomes comparable to the cable's length (roughly when the line is longer than about λ/10), the wire stops behaving like a short and starts behaving like a transmission line with a characteristic impedance Z₀ — 50 Ω for most RF test gear and transmitters, 75 Ω for video and cable TV. If the load at the far end doesn't equal Z₀, part of the signal reflects back toward the source, bounces again, and sets up a standing wave. The strength of that reflection is the reflection coefficient Γ = (ZL − Z₀)/(ZL + Z₀); its magnitude runs from 0 (perfect match, no reflection) to 1 (total reflection, as at an open or short). Engineers usually quote the standing‑wave ratio, VSWR = (1 + |Γ|)/(1 − |Γ|): 1:1 is perfect, 2:1 is workable, and above about 3:1 you have a real problem.
To get power from a source to a load (an amplifier to an antenna, a probe to a circuit) you insert a matching network — a small arrangement of lumped inductors and capacitors, a λ/4 transformer, or an open/short stub — that transforms the load impedance to Z₀ at the operating frequency. The classic L‑network (one series and one shunt reactance) can match any resistance to Z₀ with the right choice of which element is series vs shunt and inductor vs capacitor. The catch is that matching is inherently narrowband: it is exact at one frequency and degrades as you move away, which is why wide‑band radios need a tuner that re‑matches continuously. The Smith chart is the traditional hand tool for this: it plots impedance and reflection coefficient on one circle, and moving around it corresponds exactly to adding a series or shunt reactance, so you can "walk" from the load to Z₀ by reading off component values.
Why care? Because mismatch means reflected power and mismatch loss (less power delivered), voltage peaks along the line (which can break down insulation or overstress an amplifier's output stage), and — in a transmitter — the reflected power is dissipated as heat in the power amplifier and can destroy it. Practical rules: keep RF runs short relative to wavelength or properly terminated; use the correct 50 Ω or 75 Ω all the way through (don't mix); verify VSWR with a meter before applying full power; and remember a quarter‑wavelength line inverts impedance (Zin = Z₀²/ZL), which is the basis of the λ/4 transformer (choose its impedance as √(Z₀·RL) to match RL to Z₀) and of many antenna feed systems. Also remember that "matching an antenna" is two jobs at once: you want low VSWR and you want the antenna to actually radiate efficiently (high radiation resistance, low loss).
Mismatch & VSWR calculator
Enter the line impedance and the (complex) load to get the reflection coefficient, VSWR, return loss and the mismatch loss.
| Metric | Meaning | Perfect | "Good" |
|---|---|---|---|
| Reflection coeff. |Γ| | Fraction of amplitude reflected | 0 | < 0.3 |
| VSWR | Standing‑wave ratio on the line | 1:1 | < 2:1 |
| Return loss (dB) | −20·log₁₀|Γ|, higher is better | ∞ dB | > 14 dB |
| Mismatch loss (dB) | Power lost to reflection | 0 dB | < 0.3 dB |
Go deeper — the Smith chart in one minute
The Smith chart plots the complex reflection coefficient Γ = u + jv in the unit circle, and it is simultaneously a map of normalized impedance z = 1 + Γ/(1 − Γ) (divide by Z₀). Circles of constant resistance and arcs of constant reactance tile the chart, and the outer rim is the unit‑|Γ| circle (pure reactance / total reflection). The magic property: adding a series reactance moves you along a constant‑resistance circle, while adding a shunt (parallel) admittance moves you along a constant‑conductance circle. So to match a load you plot it, then hop by series/shunt elements until you land on the centre (Γ = 0, z = 1, a perfect match) — and the angular distance of each hop tells you the reactance value. It is the fastest way to design an L‑match or a double‑stub match by hand, and it makes "what does this 50 Ω line + this load actually look like?" a visual question.
Matching at the wrong frequency (a network tuned to 100 MHz is useless at 10 MHz). Assuming a long unterminated cable "just passes the signal" — its length changes the impedance the source sees. Confusing VSWR with delivered power: a 2:1 VSWR is only ~0.2 dB mismatch loss, yet the reflected power still stresses the PA. And assuming 50 Ω is always best — 75 Ω has lower resistive loss for some long‑run uses. Always de‑rate a mismatched transmitter instead of pushing full power into a bad load.
4. RF fundamentals & antennas
RF energy travels as a self‑sustaining wave of coupled electric and magnetic fields, moving at the speed of light in free space, c ≈ 3×10⁸ m/s. The single most useful relationship is λ = c/f: a 100 MHz signal has a 3 m wavelength, a 433 MHz signal about 0.69 m, and a 2.4 GHz Wi‑Fi signal about 12.5 cm. This is why antenna size tracks frequency — a half‑wave dipole is roughly λ/2 long, so it is 1.5 m at 100 MHz but only ~6 cm at 2.4 GHz. RF power is almost always quoted in dBm, decibels relative to 1 mW: P(dBm) = 10·log₁₀(Pw/1 mW). So 0 dBm = 1 mW, 10 dBm = 10 mW, 30 dBm = 1 W. Decibels make huge gain and loss numbers compact and, crucially, additive — you can sum them up a whole link with ordinary addition.
An antenna converts guided waves on a feedline into radiated space waves (and back again). The simplest is the half‑wave dipole — two conductors, total length ≈ λ/2, fed at the centre — with a gentle figure‑8 pattern in its plane and a gain of about 2.15 dBi. Gain measures how much the antenna concentrates power in a direction compared with an ideal omnidirectional (isotropic) radiator: dBi is dB over isotropic, and dBd is dB over a dipole, with dBi = dBd + 2.15. You buy directivity by making the antenna physically larger or adding elements. A Yagi‑Uda antenna keeps a driven element and adds one or more passive reflector and director rods spaced a fraction of a wavelength apart, focusing the beam and adding gain — the classic directional TV and amateur‑radio antenna. Polarization is the orientation of the radiated electric field: a vertical dipole radiates vertically polarized waves, a horizontal one horizontally, and the transmitting and receiving polarizations should match or you lose signal (cross‑polarization loss can be large).
Whether a receiver actually hears a transmitter is set by the link budget: received power = transmitted power + transmit gain − path loss + receive gain − other losses. The dominant term is path loss, and in open space the Friis transmission equation gives Pr = Pt·Gt·Gr·(λ/4πd)², or in the more convenient dB form Pr(dBm) = Pt(dBm) + Gt(dBi) + Gr(dBi) − FSPL(dB), where FSPL(dB) = 20·log₁₀(4πd/λ). The physics to internalize is the inverse‑square law: doubling the distance cuts the received power to a quarter, i.e. −6 dB, and it compounds — every further doubling costs another 6 dB. Real links add margin for obstacles, rain, multipath fading, connector and cable loss, and the receiver's own noise figure; a good RF engineer designs with explicit margin and verifies with a spectrum analyzer or the receiver's sensitivity specification. And remember the lesson that runs through the whole path: "more transmit power" is rarely the whole answer — antenna gain, a clean impedance match, and low cable loss usually matter just as much, or more.
RF & antenna calculator
Enter a frequency and a distance to get the wavelength, a half‑wave dipole length, and the free‑space path loss — the three numbers behind almost any RF link.
| Frequency | Wavelength λ | Half‑wave dipole (≈λ/2) | Common use |
|---|---|---|---|
| 100 MHz | 3.0 m | ≈ 1.5 m | AM band, some aviation |
| 433 MHz | 0.69 m | ≈ 0.33 m | Key fobs, remote control |
| 1 GHz | 0.30 m | ≈ 0.15 m | Wi‑Fi (lower), radar |
| 2.4 GHz | 12.5 cm | ≈ 6.2 cm | Wi‑Fi, Bluetooth, microwave |
| 5 GHz | 6.0 cm | ≈ 3.0 cm | Wi‑Fi (upper), point‑to‑point |
Go deeper — the Friis equation and noise figure
The Friis equation assumes a clear path and matched, lossless antennas. Real links stack on top of it: cable and connector loss (significant at microwave frequencies — a few dB of coax can dominate), implementation loss (the gap between an antenna's theoretical and real gain), and atmospheric/obstacle loss. On the receive side the limiting factor is often the receiver's noise figure NF (how much extra noise the receiver adds on top of thermal noise, kT·B). The minimum detectable signal (sensitivity) is roughly −174 dBm/Hz + 10·log₁₀(B) + NF + a required SNR; everything before the first amplifier (the LNA) matters most, which is why low‑loss front‑ends and a good match beat brute transmit power. A disciplined link budget lists every dB — gain, loss, NF, margin — in one column and checks that the total clears the sensitivity with room to spare for the worst day.
| Type | Pattern | Typical gain | Polarization | Use |
|---|---|---|---|---|
| Half‑wave dipole | Figure‑8 | 2.15 dBi | Linear (along element) | Reference, portable, base |
| Monopole (¼ wave) | Omnidirectional | ≈ 3 dBi | Vertical | Handsets, base masts (needs ground) |
| Yagi‑Uda | Directional | 5–20 dBi | Linear | TV, amateur radio, point links |
| Patch (microstrip) | Directional (broad) | 6–12 dBi | Linear | Wi‑Fi, radar, flat arrays |
| Parabolic dish | Narrow beam | 10–40+ dBi | Linear | Satellite, long point‑to‑point |
| Helix | Directional | 6–15 dBi | Circular | Satellite, where polarization fade matters |
Mismatched polarization (vertical transmitter, horizontal receiver — easily 20+ dB of loss). Cutting a dipole to exactly λ/2 and ignoring the end effect (it resonates slightly low, so cut a few percent shorter). Assuming a higher‑gain antenna is always better (a tight beam misses off‑axis sources). Forgetting coax loss at microwave frequencies. Mixing up dBi and dBd (a constant 2.15 dB error). And expecting a quarter‑wave monopole to work without a ground plane — it is literally half a dipole and needs the ground to complete the other half.
Power and RF are the same discipline at different scales. Energy stored and released by L and C is how you make efficient supplies and drive motors; once the wavelength shrinks to the size of your hardware, that same L and C become transmission lines and antennas, and impedance matching becomes as important as the circuit itself. Duty cycle, back‑EMF, VSWR, and wavelength are just four names for the idea that timing and matching decide whether energy and signal arrive where you want them.
Now try it
Chain the modules: design a 12 V → 5 V buck (D = 0.42) to power a brushed motor, use the PWM estimator to predict its speed and stall current, then imagine the drive's control line running over a 50 Ω cable to a sensor node — check the VSWR — and finish with a 433 MHz link budget from a 10 cm dipole to a receiver 500 m away. That full chain — conversion, drive, transmission, radiation, reception — is the skeleton of countless real products, and completing it is a good measure that the advanced path has clicked.
Step back to the Learn Electronics hub, or open the calculator directory for the resonant‑frequency, decibel, and unit‑conversion tools that support these calculations.
Figures here use standard teaching values; for any real build confirm exact part ratings — inductor saturation current and DCR, MOSFET Rds(on) and safe‑operating area, motor ke and winding resistance, cable loss per metre, and antenna gain/polarization — on the specific datasheets you use.