Analog Electronics II · 144 h · Topic 1 of 6

Special semiconductors

The bipolar transistor is controlled by current. The components of this topic are controlled by voltage (the FETs), by a pulse that gets latched (the thyristors) or by light (the photoelectric devices). Three different principles and three different fields of application.

Semiconductors FET and MOSFET SCR and triac Optoelectronics

01Three families, three ways of controlling

Field effect

An electric field narrows or widens a channel through which the current flows. The gate draws no current: it behaves like a capacitor. They are used to amplify, to switch and as analog switches.

Controlled triggering

A four-layer structure that, once triggered, keeps conducting on its own until the current drops. They are used to handle mains power: they do not regulate, they switch.

Photoelectric

They convert light into an electrical signal or the other way around. Their great virtue is not only detecting: it is isolating electrically two circuits that cannot share a ground.

02The field-effect transistor

It has three legs —gate (G), drain (D) and source (S)— that play a role similar to base, collector and emitter, but the mechanism is different: the gate voltage modulates the width of the channel through which the current flows, with no current flowing through the gate.

Figure 1. N-channel JFET, animated. As the gate becomes more negative, the depletion region grows and the channel narrows until it closes. At V_GS = V_GS(off) the current is practically zero: the transistor is cut off.

The first thing to fix in your mind: normally on or normally off

JFETs and depletion-mode MOSFETs: with V_GS = 0 they conduct. A negative voltage (N channel) has to be applied to cut them off. They are said to work in depletion mode.
Enhancement-mode MOSFETs: with V_GS = 0 they do not conduct. The threshold voltage V_GS(th) has to be exceeded for the channel to appear. They are the ones used in switching and the ones inside every CMOS integrated circuit.

I_{D} = I_{D S S} {(1 - \frac{V_{G S}}{V_{G S (o f f)}})}^{2}

Shockley equation, valid in the saturation region of the JFET. I_DSS is the current with the gate at zero, and V_GS(off) the voltage that cuts off the channel. The relationship is quadratic, not linear like that of the bipolar.

Worked example · Operating point of a JFET

A 2N5457 with I_DSS = 5 mA and V_GS(off) = −3 V, biased at V_GS = −1 V:

I_D = 5 mA × (1 − (−1)/(−3))² = 5 × (1 − 0.333)² = 5 × 0.444 = 2.22 mA
With V_GS = −1.5 V: I_D = 5 × (0.5)² = 1.25 mA
With V_GS = −2 V: I_D = 5 × (0.333)² = 0.56 mA

Half a volt on the gate cuts the current in half. And there is an important practical consequence: factory spread is enormous. Within a single batch, I_DSS can range from 1 mA to 5 mA. That is why the biasing of a JFET always includes a source resistor with feedback, just as with the bipolar: you never fix the gate at a calculated value and call it done.

	Bipolar transistor	FET
Controlled by	Base current	Gate voltage
Input impedance	Low: kΩ	Extremely high: 10⁹ Ω for the JFET, 10¹² for the MOSFET
Gain	High (β from 100 to 800)	Lower: we speak of transconductance g_m
Noise	Higher	Lower: preferred in first stages
Thermal drift	Runs away: more temperature, more current	Self-limiting: more temperature, less current
In parallel	Needs equalizing resistors	Can be paralleled directly
As a switch	Drops 0.2 to 0.7 V (V_CE(sat))	Behaves like a resistance: R_DS(on) of milliohms
Handling	Rugged	The MOSFET is damaged by static electricity

The power MOSFET

It is today the most widely used switching component. Its advantage over the bipolar is twofold: it does not need a permanent drive current and its voltage drop when conducting is not a fixed voltage but a very small resistance.

Worked example · Losses of a MOSFET versus a bipolar

Driving a continuous 10 A:

Bipolar with V_CE(sat) = 0.5 V → P = 0.5 × 10 = 5 W. And it needs 10/50 = 200 mA of base current continuously.
MOSFET with R_DS(on) = 10 mΩ → P = I²R = 100 × 0.01 = 1 W. The gate draws nothing in steady state.

Five times less dissipation, a smaller heat sink and simpler drive. The drawback appears in fast switching: charging and discharging the gate capacitance (nanofarads) demands drive current peaks, and that is why gate driver circuits exist.

The MOSFET and static electricity

The gate is insulated by an oxide layer a few nanometers thick that is punctured by a few tens of volts. The human body builds up thousands of volts walking on a carpet. A MOSFET can be damaged without it showing: it works on the bench and fails a week later. Work with an antistatic wrist strap, with the soldering iron tip grounded, and keep components in their conductive foam. Those with internal protection Zener diodes tolerate it better, but that is no excuse.

Other uses of the FET worth knowing

Analog switch (CD4066): connects or disconnects a signal without a relay or contacts.
Voltage-controlled resistor: in the ohmic region, the FET behaves like a variable resistor. It is the basis of automatic gain control.
High-impedance input stage: it is what gives an op amp such as the TL081 input currents of picoamperes, and what makes it possible to measure the output of a pH sensor or of an electrode.
All CMOS logic: every gate in the 40XX series is a pair of complementary MOSFETs.

03Thyristors: the SCR

The silicon-controlled rectifier is a diode that does not conduct until it is given permission. Inside it has four P-N-P-N layers and it is best understood as two transistors with feedback: a pulse on the gate turns one on, that one turns on the other, and the second keeps the first one on. That is why, once triggered, the gate loses control.

Figure 2. The four layers and their equivalent. The feedback between the two transistors is what produces the latching: that is why the trigger can be a very brief pulse and the SCR keeps conducting anyway.

The four figures that define an SCR

V_DRM — repetitive peak voltage it withstands without triggering on its own. For 220 V mains, 400 or 600 V devices are used.
I_T(RMS) — RMS current it can conduct, with its heat sink.
I_GT — gate current needed to trigger it. Typically between 5 and 50 mA.
I_H — holding current: below it the SCR switches off by itself. It is the key to its whole behavior.

How an SCR is turned off

It is not turned off with the gate. It turns off when the anode current falls below I_H. In AC that happens by itself, at every zero crossing, a hundred times per second: that is why thyristors are natural for the mains. In DC, on the other hand, you have to force it with a commutation circuit —a capacitor that reverses the anode voltage for an instant— and that complicates the design quite a bit. It is one of the reasons why in DC a MOSFET or an IGBT is preferred today.

Phase control

It is the central application. If the thyristor is triggered later in each half-cycle, it conducts for less time and the load receives less power. The angle from the zero crossing to the trigger point is called the firing angle α.

Figure 3. Phase control, animated. The gate pulse moves earlier and the conducting part of each half-cycle grows: the power in the load goes from almost zero to almost everything. It is what the knob of a dimmer does.

V_{e f} = V \sqrt{\frac{π - α}{π} + \frac{\sin 2 α}{2 π}}

RMS voltage across a resistive load with full-wave phase control, with the angle α in radians.

Angle α	RMS voltage on 220 V	Power delivered	Appearance on a lamp
0°	220 V	100 %	Full brightness
45°	209.7 V	90.9 %	Barely noticeable
90°	155.6 V	50 %	Half brightness
135°	66.3 V	9.1 %	Filament barely glowing red
180°	0 V	0 %	Off

The scale is not linear, and you can feel it in the knob

Between 0° and 45° almost nothing changes, and between 90° and 135° it changes a great deal. That is why dimmers seem to have “all the adjustment at the end of the travel.” The useful zone is between 40° and 140°, and a good knob uses a logarithmic-taper potentiometer to spread the range better.

04Diac and triac

The SCR conducts in one direction only: on AC, it lets half a cycle through. The triac solves that: it is two SCRs in antiparallel inside the same package, with a single gate. Its terminals are no longer called anode and cathode but MT1 and MT2, because either can be the positive one.

Triac

Conducts in both directions: it controls the full cycle.
It is triggered by a positive or negative pulse, but not with the same sensitivity in all quadrants: quadrants I and III are the good ones; IV is the worst and should be avoided.
It turns off by itself at every zero crossing, just like the SCR.
Typical: BT136 (4 A), BT139 (16 A), TIC206.

Diac

A two-terminal component that does not conduct until a certain voltage (about 32 V) and then fires abruptly, its voltage dropping.
It is symmetrical: it does the same in both directions.
Its function is to generate a clean, uniform pulse for the triac gate. Without it, the triggering depends on the spread of the triac and the dimmer flickers and shows hysteresis.
Typical: DB3.

Figure 4. Classic dimmer. Three control components —potentiometer, capacitor and diac— control a load of hundreds of watts. It is the most widespread thyristor circuit there is.

This circuit is at mains potential

The entire control circuit, including the knob, is connected to the mains. It has no transformer or insulation. Never measure it with a grounded oscilloscope without an isolation transformer: it produces a dead short through the instrument's ground. For the labs, work with an isolation transformer and, better still, with reduced voltage.

Inductive loads: the snubber

With a motor or a transformer, the current is out of phase with the voltage: when the current crosses zero and the triac turns off, the voltage has already risen, and that dV/dt can trigger it again by itself. The circuit loses control and the load stays on. It is corrected with a snubber: 100 Ω in series with 100 nF, in parallel with the triac. Modern “snubberless” triacs tolerate much more dV/dt, but for a strongly inductive load the snubber is still the right choice.

Phase control and integral-cycle control

Phase control chops the sine wave in the middle and therefore generates interference: there is a very fast current step a hundred times per second, which propagates through the wiring and can be heard on an AM radio. That is why every dimmer has an inductor and a filter capacitor at the input.

When the load is thermal and slow —an oven, a heating element— the other method is preferable: integral-cycle control (or zero-crossing control). For example, 7 cycles out of every 10 are let through, always triggering at the zero crossing. It generates no interference and it is what solid-state relays with zero-crossing optotriacs do. It is no good for a lamp, because it would flicker.

05Photoelectric components

Component	What it does	Speed	Where it is used
LED	Converts current into light. Forward drop depends on the color: 1.8 V red, 2.1 V green, 3.2 V blue and white.	ns	Indicators, lighting, emitter of every optical sensor. Always with a current-limiting resistor.
LDR	Its resistance drops with light: from 1 MΩ in the dark to 1 kΩ when illuminated.	Slow: tens of ms	Outdoor lighting, brightness control. Cheap and simple, but slow and not very repeatable.
Photodiode	Generates a current proportional to the light. It is used reverse-biased.	Very fast: ns	Fiber optics, remote-control receivers, precise light measurement.
Phototransistor	A photodiode with the gain of a transistor: much more current.	Medium: µs	Presence sensors, encoders, optocouplers.
Photovoltaic cell	Generates voltage: 0.5 V per cell in full light.	—	Solar power.
Optocoupler	An LED and a phototransistor facing each other inside one package: they transmit a signal with no electrical contact.	µs	Isolating the control from the power side, breaking ground loops, protecting a microcontroller.
Optotriac	An LED and a trigger triac: it drives a power triac with full isolation.	—	Solid-state relays, controlling mains loads from a micro. MOC3021 and MOC3041.

Figure 5. Optocoupler, animated. The information crosses as light: the two halves of the circuit share no conductor. That barrier withstands thousands of volts and is what allows a 5 V microcontroller to drive a mains load without risk.

Why the optocoupler is so important

Galvanic isolation: 2500 to 5000 V between input and output. A fault on the power side does not reach the control circuit.
Breaks ground loops: two distant devices with grounds at different potentials cannot be joined by a signal cable. An opto solves it.
Adapts levels: 5 V on one side, 24 V or 220 V on the other, without converters.
Its key specification is the CTR (current transfer ratio): how much output current you get for each milliampere of input. And you should know that it declines with the years: a serious design leaves margin.

06In the lab

Lab 1 · JFET curve

With a 2N5457, a variable supply on the gate (from 0 to −3 V) and 12 V on the drain through 1 kΩ, measure I_D every 0.25 V of V_GS. Plot it. Compare with the Shockley equation using the measured I_DSS (V_GS = 0) and the measured V_GS(off) (I_D ≈ 0). Repeat with another transistor from the same batch: the difference between the two is the most important lesson of the lab.

Lab 2 · The MOSFET as a switch

With an IRFZ44 driving a 12 V lamp, measure the voltage between drain and source with the gate at 12 V and calculate R_DS(on) = V_DS/I_D. Then repeat with the gate at 4 V: the transistor does not saturate, the drop goes up and the package heats up. It is the demonstration of why an ordinary MOSFET is not driven directly from a 3.3 V microcontroller and logic level devices are needed.

Lab 3 · SCR latching (at low voltage)

With 12 V DC, a small SCR (C106) and a lamp: a pushbutton between the gate and the positive supply turns the lamp on. Release the pushbutton: it stays on. A second pushbutton that shorts the anode to the cathode, or that cuts the supply for an instant, turns it off. This is how latching and holding current are understood without touching the mains.

Lab 4 · Phase control, with an isolation transformer

Build the dimmer with a small incandescent lamp powered through a step-down isolation transformer. With the oscilloscope across the load, turn the knob and watch how the conducted part of each half-cycle grows. Measure the RMS voltage with a True RMS multimeter and compare it with the formula: a calibrated average-responding meter will read wrong, and that difference is in itself part of the exercise (AC measurements).

Lab 5 · Isolation with an optocoupler

With a 4N25, drive from a 5 V pushbutton an LED powered by a separate 12 V supply, with no connection between the two grounds. Measure with the multimeter between the two grounds: there is voltage and the circuit works just the same. Then measure the CTR: output current divided by input current.

07Common mistakes

Symptom	Usual cause
The MOSFET gets very hot when switching	Insufficient gate voltage: it works in the linear region instead of saturated. Or it switches slowly and the losses are in the transition.
The MOSFET always conducts, even with the gate at zero	Floating gate (the 10 kΩ gate-to-source pull-down resistor is missing) or a transistor damaged by static.
Two FETs in parallel and one gets hotter	Threshold difference between individual devices. In switching it is equalized with individual gate resistors.
The JFET amplifies differently on each circuit board	I_DSS spread. Source feedback is missing from the biasing.
The SCR never turns off	It is powered from DC: the current never drops below I_H. Forced commutation is needed, or use another component.
The triac triggers by itself with a motor	dV/dt at turn-off from the inductive load. The snubber (100 Ω + 100 nF) is missing.
The dimmer does not turn on until a certain point and then jumps	Hysteresis of the trigger R-C: the capacitor remains charged from the previous half-cycle. It is corrected with the double R-C network that commercial dimmers have.
The AM radio buzzes when the dimmer is at half brightness	Interference inherent to phase control. The input filter is missing, or you have to switch to integral-cycle control if the load allows it.
The optocoupler stops working over the years	The LED's CTR falls over time. Design with margin and not at the current limit.
The phototransistor responds to the room light	An optical filter or modulation is missing. Remote-control receivers use a 38 kHz carrier precisely for this reason.

08Self-assessment

What is the essential difference between controlling a bipolar and controlling a FET?

The bipolar is controlled with base current and draws drive power continuously; the FET is controlled with gate voltage and in steady state it draws no current, because the gate behaves like a capacitor.

A JFET with I_DSS = 8 mA and V_GS(off) = −4 V biased at V_GS = −2 V: how much current flows?

I_D = 8 × (1 − (−2)/(−4))² = 8 × (0.5)² = 2 mA. Half of the cutoff voltage gives a quarter of the current: the relationship is quadratic.

Why does a power MOSFET dissipate less than a bipolar at the same current?

Because its drop is not a fixed saturation voltage but a very small resistance: P = I²·R_DS(on). With 10 A and 10 mΩ that is 1 W, versus the 5 W of a bipolar with 0.5 V of V_CE(sat).

How is an SCR turned off and why does that make it ideal for the mains?

It turns off when the anode current falls below the holding current I_H. In AC that happens by itself at every zero crossing, a hundred times per second, so the control comes down to deciding when to trigger it in the next half-cycle.

With α = 90° on 220 V, what RMS voltage and what power does a resistive load receive?

V_rms = 220 × √0.5 = 155.6 V, and since power goes with the square, 50 %. Half the angle is not half the power: the relationship is not linear.

What is the diac for in a dimmer?

To generate a well-defined trigger pulse: it does not conduct until about 32 V and then discharges the capacitor onto the gate all at once. Without a diac, the triggering depends on the sensitivity of each triac and the circuit ends up erratic and with hysteresis.

Why can an inductive load make a triac lose control?

Because the current lags the voltage: when the current reaches zero and the triac turns off, the voltage across it has already risen abruptly. That dV/dt triggers it again. It is corrected with an R-C snubber in parallel.

When is integral-cycle control preferable to phase control?

When the load is thermal and slow —ovens, heating elements—. By always triggering at the zero crossing, no interference is generated. It is no good for lighting, because the flicker would be visible.

What advantage does an optocoupler give that an ordinary transistor does not?

Galvanic isolation: input and output share no conductor, and between them there are thousands of volts of isolation. It allows a 5 V circuit to drive mains power, and it breaks the ground loops between distant devices.

Why does a remote-control receiver use a 38 kHz carrier?

To tell the signal apart from ambient light. The transmitter turns the infrared LED on and off at 38 kHz and the receiver has a filter tuned to that frequency: sunlight or the light of a lamp, which is constant or 100 Hz, is left out.

← Previous topic · Digital Electronics II

Microprocessor and microcontroller architecture

Next topic · site publication →

Linear integrated circuits · the 555