The world is continuous
Think about the last sound you heard. The air pressure on your eardrum did not jump from one fixed level to another — it slid, smoothly, through *every* value in between. Sound, light, temperature, the radio wave hitting your phone: all of them are continuous. There is no smallest step, no list of allowed values. Between any two readings there is always another reading, and another inside that, forever.
Digital design lives in the opposite world. A digital signal is allowed only two values — a voltage near the supply rail (call it 1) or near ground (call it 0) — and it is *expected* to snap cleanly between them. That cleanliness is a gift: it lets you stack billions of gates and never worry that a tiny smudge of voltage noise will flip a bit. Analog gives up that gift. An analog signal is the smooth, in-between thing itself, and every value it can take is meaningful.
Why we still need analog
If digital is so clean, why not make the whole chip digital? Because the world refuses. A microphone, a camera pixel, a temperature sensor, an antenna — each of these *starts* as a continuous physical quantity. Something has to receive that smooth signal, strengthen it, clean it up, and hand it to the digital core in a form the bits can use. That something is analog, and there is no way around it: the first and the last millimetre of every signal's journey through a chip is analog, because that is where the chip meets reality.
Analog also does jobs digital simply cannot. It generates the steady reference voltages and the clean clock the digital part needs to even run. It manages power — turning a battery's drooping voltage into the rock-steady supply the logic demands. And whenever a faint signal must be coaxed up out of the noise before anyone can read it, that is an amplifier — the op-amp and its relatives — doing analog work.
A different mindset
Coming from digital, the hardest adjustment is psychological. In digital you ask *is this a 1 or a 0?* — a yes/no question with a right answer. In analog the questions never close that neatly: *how much* gain, *how close* to the noise floor, *how far* from oscillating, *how much* does this drift when the chip warms up. Every answer is a number on a sliding scale, and almost every number is a trade-off against another one.
The single most important number in that mindset is transconductance (written gm): how strongly a transistor's input voltage steers its output current. It is the raw material of gain. But you never get gm for free — you buy it with bias current, and more bias current costs power and chip area. That pattern repeats everywhere in analog: gain trades against bandwidth, speed trades against power, low noise trades against both. You are never *solving* a problem so much as *balancing* one.
gm = 2*Id / Vov ; transconductance of a MOSFET in saturation
; Id = bias (drain) current
; Vov = overdrive voltage = Vgs - Vth
; intuition: want more gm? spend more bias current Id.Why analog resists abstraction
Digital design works because of a beautiful lie: that a gate is a perfect, instant, noise-proof 1-or-0 device. The lie holds well enough that you can forget the physics underneath and just compose logic. Analog cannot tell that lie. There is no clean 0/1 to hide behind, so the physics is never abstracted away — it always has a vote.
Three physical facts in particular refuse to be ignored. Noise: electrons jiggle with heat, so every wire carries a faint, irreducible hiss; the smallest signal you can use is set by that floor, not by your cleverness. Temperature: a transistor's behavior shifts as the chip heats and cools, so a circuit that is perfect at room temperature may sag at 85 degrees. Mismatch: two transistors drawn to be identical never quite are — atoms land slightly differently during fabrication — and analog circuits often lean on two devices being *truly* equal. A digital designer can mostly ignore all three. An analog designer designs *around* them, every single day.
The mixed-signal boundary
A modern chip is not purely one or the other — it is mixed-signal, with an analog edge wrapped around a digital core, and a sharp border between them. Crossing that border in one direction is the job of an analog-to-digital converter (ADC): it watches a smooth, continuous voltage and produces a stream of numbers. Crossing back the other way is the digital-to-analog converter (DAC), which turns numbers into a smooth voltage again.
That crossing is never free. To turn a continuous signal into numbers you must do two violent things: chop time into discrete instants (sampling) and round each measurement to one of a finite set of levels (quantization). The rounding throws away a little of the signal — it adds a noise of its own — and that loss puts a hard ceiling on quality. The standard yardstick is the signal-to-noise ratio (SNR): how far the real signal stands above the noise floor.
SNR = 6.02*N + 1.76 ; ideal SNR in dB of an N-bit converter ; each extra bit of resolution buys ~6 dB more SNR ; 8-bit -> ~50 dB ; 16-bit -> ~98 dB
Where analog lives on a chip
Once you know the shape of analog, you can spot it on almost any chip. It clusters in a handful of recognizable places, always at the edge where the chip touches something physical — power, time, the antenna, the sensor, or the digital core itself.
- Sensor front-ends: the low-noise amplifiers that take a microphone, camera pixel, or biosignal — a tiny, fragile voltage — and lift it cleanly up to where it can be measured.
- SerDes and PHY: the high-speed I/O that drives bits down a cable or backplane and recovers them at the far end, where multi-gigahertz edges are pure analog even though they carry digital data.
- Clocking (PLLs): the phase-locked loops that multiply a slow, clean reference into the fast, precise clock the digital core runs on — a feedback loop that polices its own timing.
- Power management: the regulators and references that turn a noisy, drooping battery voltage into the steady supplies every block depends on, including the temperature-stable reference at the heart of it.
- Data converters: the ADCs and DACs sitting exactly on the mixed-signal boundary, shuttling signals between the continuous world and the bits.
All of it is built from the same humble parts you already know — the MOSFET, wired up in CMOS, plus resistors and capacitors. What changes is how you *think* about them: not as switches snapping between 1 and 0, but as continuous, gently-controlled valves for current. The rest of this track teaches that way of thinking, one stage at a time, starting with the simplest magic of all — making a single transistor amplify.