Triac Control Meets an Inductive Load
Looking for more precise power control， Ed used a triac and a few
supporting components to control high-power equipment. In this
article， he describes everything you need to know about the
circuits. The results might surprise you.
Back in the not-so-good old days，controlling high-power equipment
required big relays and motor-starting contactors； an armature‘s
snap preceded the hum of large machinery and the whine of motors.
Semiconductor power control has removed mechanical motion，
arcs-and-sparks， and all the drama from high-current switching.
While it's entirely possible to control a resistance heating unit
with a simple relay， I decided to recycle the triac and some
supporting components，as well as the hulking power transformer，
from a microwave oven in order to get more precise power control.
My February column described the transformer‘s properties
（"Transformers，" Circuit Cellar 211， 2008）。 In this column， I‘ll
examine how a triac behaves when confronted with a highly inductive
load， which isn't as obvious as the datasheets lead you to believe.
If you‘ve ever wondered what makes a triac tick （or how to make one
tick！）， read on.
TRIAC BASICS
At the simplest level， a triac is a three-terminal bidirectional
power switch. No current flows between its two high-power （"Main"）
terminals MT1 and MT2 until a specified minimum current flows
between its Gate terminal and MT1. When the Gate current exceeds
the triac's triggering threshold， the voltage between MT2 and MT1
abruptly drops to about 1 V and current may then flow between the
two Main terminals in either direction.
The triac remains turned on until the current between MT2 and MT1
decreases below the holding current threshold， typically a small
fraction of an ampere， whereupon conduction abruptly ceases. The
triac remains off until the next Gate-triggering pulse. Many
low-power triac applicationsprovide fractional-cycle control by
triggering at a specific delay after the start of each half-cycle
of the AC power line； that‘s how fan speed controls and
incandescent lamp dimmers work.
In contrast， microwave ovens and high-power heating applications
use duty-cycle modulation by turning the triac on for a specific
number of complete AC power-line cycles， then holding it off for
another group of cycles. The average output power depends on the
ratio of on cycles to total cycles，with the load's thermal mass
averaging the pulses. This may cause a buzz or produce audible
thumps as the load switches at a fairly low frequency；you‘ve
certainly noticed that in your own microwave oven.
Unlike NPN or PNP junction transistors，a triac is inherently an
ACoperated device. The datasheets refer to the four possible
combinations of Gate and MT2 polarity with respect to MT1 as
"operating quadrants"： Q-1 with both positive， Q-2 with MT2
positive and Gate negative， Q-3 with both negative， and Q-4 with
MT2 negative and Gate positive.
The Gate threshold current in Q-1 and Q-3 tends to be much lower
than in Q-2 and Q-4， to the extent that datasheets discourage Q-4
operation because of its poor sensitivity. The best triac
triggering circuits operate in Q-1 and Q-3 where the Gate polarity
matches the MT2 polarity. Operation in Q-2 and Q-3 with negative
Gate current trades off Q-2 sensitivity for the simplicity of a
unipolar Gate bias power supply.
A triac driving a resistive load such as an oven heater behaves
largely as you'd expect from the datasheet，while a reactive load
like a motor or a capacitive power supply poses some interesting
problems. If you‘ve ever tried to control a ceiling fan with an
ordinary light dimmer， you have some idea of how nasty things can
get.
Figure 1 shows a Spice simulation of the voltage and current on the
primary side of my rewired transformer after a single Gate trigger
pulse at the start of a power-line voltage cycle. The triac starts
conduction normally，but it turns off at the current's second zero
crossing. The initial current waveform is nearly all positive， not
symmetric around the X axis as a good sine wave should be. A good
Spice model is the first step in figuring out what‘s happening.
CIRCUIT MODELINGThe circuit in Figure 2 divides neatly into three sections： the
resistive heating load on the far right，the transformer and triac
in the middle， and a stack of triggering controls on the left.
While this schematic can serve as the basis for an actual
circuit，some components represent pure simulation fictions.
点击查看Figure 2I chose a 25-mΩ load resistor to set a load current of 200 A and an
output power of 1 kW，based on the nominal 5-V secondary voltage and
the original microwave oven‘s power rating. The properties of a
resistance- soldering joint depend on the workpiece's condition，
but this should be in the right ballpark.
The standard Spice transformer model consists of two inductors and
a coupling coefficient， a simplification ignoring many of the
real-world details that make transformers so puzzling. I used a
more accurate π-model incorporating actual measurements while
excluding the core‘s nonlinearities and hysteresis that you saw in
my February column.
I found a good explanation of the transformer π-model in ON
Semiconductor application note AN-1679-D，which includes the
equations converting measurement into Spice component parameters.
The application note covers switching converters， but its
transformer models work equally well in other circuits.
Finding the values in Table 1 required a few minutes. I measured
the transformer's voltage ratio with a relatively low primary
voltage to avoid core saturation. I then applied a 2-A DC current
to the windings and measured the voltage drops with a
Kelvin-connected DVM to get better accuracy than my simple
ohmmeter‘s display. Finally， although my inductance meter uses a
1-kHz test signal，the inductances should be close to their 60-Hz
values.
Homework： Run my numbers from Table 1 to verify the Spice component
values in Figure 2. Bonus： Measure your own transformer and derive
a model.
Based on those calculations， Lm represents the real transformer‘s
magnetizing inductance from the magnetic field that actually
couples energy between the windings. LI1 and LI2 model the primary
and secondary leakage inductances caused by stray fields that do
not link the windings. I included the winding resistances as series
resistance within the Lm and LI2 Spice models.
L1 and L2 form an ideal transformer with tight coupling（K1 =
0.9999） and a turns ratio matching the actual transformer's voltage
ratio. I set L1， the primary winding， to 1 H so that its 60-Hz
reactance greatly exceeds that of Lm， the 18-mH inductor
representing the real transformer.
The ideal transformer losslessly converts voltage and current,
while the components surrounding it model the real transformer‘s
coupling （about 0.93） and resistive losses. You could， if you wish，
build a π-model Spice component with all these effects in one
place， but keeping them visible may be a better idea.
The inductor models exclude core effects， which seemed a reasonable
trade-off for this column. Most cores do not operate as far into
saturation as this transformer， making the results you see here
more typical of normal designs.
Homework： SwitcherCAD includes an inductor with core effects.
Derive the critical H and B values from the measurements in my
February column，replace Lm with that model， and see what happens.
The components around the triac in the middle of Figure 2 model
some parts salvaged from the microwave oven. A snubber package
includes C1 and R5， while C2 and R6 represent a capacitor soldered
between the Gate and MT1. R7 models a simple resistive clamp， with
1 kΩ having little effect.
Although I managed to find a datasheet for the microwave oven's
triac， it does not have a Spice model and， in fact， predates the
era of cheap Spice simulation. I used a model for a similar triac
from Teccor Electronics，hoping that any differences will be minor
compared to other circuit effects.
The standard method of triac triggering uses a small optically
isolated triac between the main triac‘s MT1 and Gate terminals. A
trigger pulse across the optical barrier fires the small triac and
dumps current into the larger triac's Gate terminal. The potential
across the small triac drops and it turns off when the main triac
fires.
That method has four key advantages：one isolated device delivers
inphase（Q-1 and Q-3） trigger pulses without an additional power
supply. Other methods require more components or more design time，
both in short supply for commercial products. Those limits don‘t
apply here， so I used a more complex trigger.
Transistors Q1 and Q2 apply about 120 mA of current （negative and
positive，respectively） to the Gate， while Q3 and Q4 clamp the Gate
terminal to MT1. The current sources at the transistor base
terminals represent optocouplers：2 mA corresponds to an LED driven
with 20 mA and a current transfer ratio of 10% from its isolated
transistor. Your couplers will certainly differ from those values！
Each transistor has only one optocoupler in the real circuit， but I
find it easier to use multiple Spice current sources for starting，
periodic， and ending current pulses rather than fiddle with one
source‘s configuration values. The real firmware will drive four
microcontroller output pins， each controlling an optoisolator's
LED， at the proper times.
It‘s important to remember that a Spice simulation can show why a
circuit might not work， but it cannot prove that a circuit will
work. When a circuit design violates electrical rules and component
operating conditions，Spice can show you what's wrong. However，
Spice models do not include all the subtle real-world effects that
can prevent proper operation， even when you think you know what
you‘re doing and the model works perfectly.
As you've seen， this model excludes transformer core saturation，
many parasitic effects， and the precise values of many components.
The simulated waveforms should serve as a guide，not a justification
to tweak component values to five significant figures until "it
works."With that in mind，watch what happens！
LIVE SWITCHING
I‘m sure that engineers most often associate the
adjective"zero-crossing"
with "triac." Triggering a triac when the voltage across it is zero
ensures that the current is also zero， at least for a resistive
load. Eliminating abrupt voltage and current transitions helps
reduces high-frequency energy coupled into delicate circuits or
radiated into free space.
Although both the current and voltage start out at zero in Figure
1， the current waveform quickly lags behind the voltage. That‘s
what you expect for an inductive load： the voltage and current
zerocrossing points simply don't coincide. As a result，
zero-voltage switching isn‘t desirable for highly inductive loads.
The top plot in Figure 3 shows the triac triggering at 90
electrical degrees，the maximum voltage point， after the voltage
zero-crossing. The resulting current waveform is a nearly perfect
sine wave with only a small initial DC offset and no huge peak. In
round numbers，the current is 27 Apeak and 19 Arms.
The apparent power drawn by the circuit is the product of the
voltage and current values：
Calculating the real power， however，requires the phase angle， which
you can measure from the plot by comparing the peaks of the current
and voltage waveforms. I measured the times of the second positive
peaks as 20.7 ms and 23.9 ms, making the
phase angle:
Figure 3b shows events on the secondary side of the transformer.
Because R8， the simulated resistanceheating load， is a pure
resistance and IL2， the secondary leakage inductance，is so small，
the voltage and current waveforms are almost precisely in phase.
The secondarys apparent and real power are， therefore， essentially
identical. Their RMS values give：
855 W = 4.6 V * 185 A [8]
The combined circuit losses outside the core account for 105 W.
Compared to the 76 W of actual core loss I measured in February，
its obvious that this model excludes some significant real-world
effects！
Notice how the primary and secondary voltages are in phase even
though the primary current lags by 65。 The real component of that
current is， by definition， in-phase with the primary voltage，
because it represents the real power in the load as well as heat
dissipated in primary-side losses.
GATE CONTROL
Figure 3c shows that， contrary to the tidy explanations found in
datasheets， voltages and currents at the Main terminals do have a
significant effect on the Gate. For example，the Gate voltage and
current move in opposite directions when the MT2 current passes
through zero at 11.1 ms. Despite the .120 mA Quadrant-3 trigger
current applied through R1， the Gate voltage becomes more negative
and the Gate current more positive.
No triac datasheet specifies the parasitic capacitance between the
triacs MT2 and Gate terminals. Instead，youll see an upper limit on
the rate of change of MT2 voltage， typically a few volts per
microsecond. Voltage transitions faster than that will drive enough
current through the parasitic capacitance to falsely trigger the
triac.
At the moment of switching near 11.1 ms， the voltage on MT2 begins
falling from nearly zero toward .150 V. That change pulls current
through the parasitic capacitor and spikes the Gate current from
.140 mA to .86 mA. If the driver transistor werent supplying
sufficient current （as in Figure 1）， that current must come from
MT1.
Think about it： negative voltage on MT2 combined with 50 mA of
current flowing from MT1， past the Gate terminal，to MT2. Thats
exactly what happens in Quadrant-3 triggering： negative MT2 voltage
and MT1 current. The fact that the MT1 current doesnt flow through
the external Gate circuit has no effect； the semiconductor layers
dont care where the current originates.
Actively clamping the Gate to MT1 can prevent this type of false
triggering，as shown after transistors Q3 and Q4 turn on at 25 ms.
The Gate current spikes from .70 to 24 mA as the triac turns off at
28 ms， but that current flows from the driver transistors through
the parasitic capacitance to MT2， rather than from MT1 through the
Gate layers. Without any Gate-to- MT1 trigger current， the triac
shuts off as you‘d expect.
The usual false-triggering solution is a simple snubber similar to
C1 and R5 that resonates with the circuit's inductance to limit the
dV/dt across the Main terminals. Unit cost and control complexity
generally prevents active gate clamping， but sometimes you must
combine both snubbing and clamping to get enough control. Keep that
trick up your sleeve！
CONTACT RELEASE
While a Spice model provides the easiest way to understand triac
triggering behavior and also reduces the risk of venting the
triac‘s magic smoke，there's not enough room here for all of the
variations. You should download the model from the Circuit Cellar
FTP site and do some exploration of your own.
Purists will quibble that my power calculations ignore the
distorted waveforms and fractional cycles. I‘m approximating the
long-term value by assuming the power is applied continuously at
the measured values. Homework for purists only： find the
differences！
Ed Nisley is an EE and author in Poughkeepsie，NY. Contact him at
ed.nisley@ ieee.org with "Circuit Cellar" in the subject to avoid
spam filters.
PROJECT FILESTo download code, go to
ftp://ftp.circuitcellar.com/pub/Circuit_Cellar/2008/213.
RESOURCESFairchild App Note AN-3008 on snubberdesign: Fairchild
Semiconductor, "Application Note AN-3008: RC Snubber Networks for
Thyristor Power Control and Transient Suppression," 2002,
www.fairchildsemi.com/an/AN/AN-3008.pdf.
Littlefuse, "Teccor Triac App Notes,"
www.teccor.com/cgi-bin/r.cgi/en/know_content.html?ContentID=86&WhichApp=6.
Thyristors: Teccor Electronics, Inc.,"Thyristor Design Guide,"
www.teccor.com/data/en/Application_Notes/01_trigg.pdf.
Transformer modeling and measurement:C. Basso, "How to Deal
withLeakage Elements in Flyback Converters,"ON Semiconductor,
AN1679/D,2005,
www.onsemi.com/pub/Collateral/AN1679-D.PDF.
Triac basics: J. B. Calvert, "ElectronicsIndex: Thyristors,"
http://mysite.du.edu/~etuttle/electron/elect5.htm.
Wikipedia, "Magic smoke,"
http://en.wikipedia.org/wiki/Magic_smoke.
SOURCESLinear Technology simulation tools
Linear Technology Corp.
www.linear.com/designtools/software/
Teccor Triac Spice models
Littlefuse
www.littelfuse.com/data/en/PSPICE_Models/libtriac.zip
;){kind=link}