by J.M. Woodgate B.Sc.( Eng.), C. Eng., M.I.E.E., MA.E.S., M.Inst. S.C.E.
There are many useful devices, from greenhouse heating controls, through motor speed controllers to disco lights, that need some way of controlling mains power by means of low-level signals, either analogue or digital. Relays can be used for simple on/off control, and recent developments have made switching cur rent ratings of the order of 20A available in small relays of reasonable cost. For many applications, however, solid-state control devices are preferable, on the grounds of physical size or cost, or because proportional control is required.
Thyristors are used for DC, and high -power AC/rectifier applications, while for other AC applications the triac, being a bi-directional device, is better. It can control powers up to at least 10kW (single-phase), and substantially higher powers in multi-phase systems.
High power bipolar and MOS transistors are also available, at reasonably competitive prices, for power -control applications, and are useful with high-frequency supplies, or where controlled turn-off is required: in these cases gate turn-off (GTO) thyristors are also used for powers up to about 100kW.
In this article, however, we shall concentrate on a qualitative explanation of the working of the triac, and a look at design methods in some typical, basic applications.
Construction of the Triac
The triac is a four-layer semi conductor, but, being bi-directional, its structure is at first sight somewhat complex. We can build up an explanation of the structure and operation by first considering the thyristor.
The thyristor can be modeled by an interconnected complementary pair of bipolar transistors, as shown in Figure 1.
From equation A.1 of Appendix 1, the current in response to a positive voltage on the p-n-p emitter (anode) relative to the n -p -n emitter (cathode) is critically dependent on the product of the DC current gains, and becomes unlimited when the product is equal to 1. This occurs at quite low values of the collector currents, in the region where the current gains are proportional to the collector currents. A small current injected at either base can therefore trigger the device into conduction. The n -p -n base, or 'gate', shown in Figure 1, is known as a cathode gate. A low power, four -layer device, BRY39 (page 294 of the 1986 Maplin Catalogue), is available which gives access to both anode and cathode gates. The device is very versatile, offering many small signal switching applications, but these are outside the scope of this article.
For full-wave AC power control, two of these model devices would be required, connected in inverse parallel.
Luckily, it is possible to integrate the two devices on a single die, and an example of such a triac structure is shown diagrammatically in Figure 2. This is a center -gate structure; other structures are manufactured, and the geometry has significant effects on the device characteristics.
Figure 1. 2-transistor equivalent of a Thyristor
Figure 2. Center-gate Triac structure
Figure 3. Triac main voltage/main current characteristics with gate current
as parameter
Electrical Characteristics of Triacs
If the anode to cathode voltage applied to the model device of Figure 1 is increased sufficiently, the increased leakage currents will themselves raise the current gains to the critical values, and the device is then said to have 'broken down'. This is not a normal mode of operation however, and conduction is usually started by applying to the gate terminal a positive voltage (in the case of the model device) relative to the cathode.
The characteristics of the semi conductor materials used in a triac are different from those of a bipolar transistor (for example, the same region has to work both as a collector and as a base), and the gate input characteristics are rather different from those of a 'conventional' base-emitter junction.
Part of this difference can be represented by the resistors RI and R2 in Figure 1. These differences in characteristics raise the gate voltage at, say, 10mA gate current from the 600mV of a typical silicon transistor junction to about 900mV.
The presence of R1 means that conduction in the main circuit cannot be stopped by connecting the gate to cathode or MT1; thus the gate current can be applied in the form of a short pulse.
Conduction is normally stopped by allowing or forcing the main circuit current to fall below the minimum value necessary to hold the current gains at or above their critical values. This current is known as the holding current 1H, and, in AC applications, the current normally falls below this value as a matter of course once every half -cycle.
If the gate current is increased slowly from zero, with a resistive load in the main circuit (see Figure 3), the voltage across the main terminals falls, eventually to a low value, represented in the model by the saturation voltage of the p -n -p transistor plus the base-emitter voltage of the n-p-n transistor, or vice versa for an opposite supply polarity.
This is the normal on -state of the device, and operation at lower values of gate current than is necessary to achieve it should normally be avoided, because the power dissipation in the device is considerably increased by this. Thus it is wise to ensure that, subject to the limits of peak and average gate dissipation, the worst-case gate current available from the trigger circuit comfortably exceeds the gate current required to trigger a least-sensitive device.
Capacitance within the device, between the main terminals and the gate, can cause the triac to trigger if the main-terminal voltage rise rate (dV/dt) is sufficiently fast. This, too, is normally an undesired effect, but can be avoided by correct design. Some types of triac are very resistant to this sort of false triggering, having maximum dV/dt values of several hundred volts per micro second, but are limited in their triggering modes (see below). Where the load is inductive, and the triac is required to go from the conducting to the non-conducting state, the maximum permissible dV/dt is dependent on the current flowing in the main circuit and its rate of fall (- dl/dt), a high current or rate of fall reducing the permissible dV/dt, which is then known as 'commutation dV/dt'.
If we look at the main circuit voltage/current characteristics in Figure 4, we can see significant hysteresis: the main circuit current has to rise to IL, the latching current, before the gate loses control, whereas the current can fall to the holding current, before conduction substantially ceases.
The off-state leakage current I_n is sufficient to disqualify the triac as a circuit isolator, and a mechanical isolator switch should always be provided for maintenance and/or service operations! The quadrants of the V/I graph are conventionally numbered I to N. Operation in quadrant I is satisfactory with the gate either positive or negative with respect to MT 1; and in quadrant III with a negative gate, but for a positive gate in quadrant III the gate sensitivity is, for most types of triac, much reduced, and this triggering mode may not be recommended by the device manufacturer. The latching current may also be different in the two quadrants, and decreases some what with increasing gate -current pulse width (see below). The datasheet values, rather than measured values, should be accepted for design work.
When the gate current is in pulse form, as is the case in many practical circuits, the rather large capacitance between the gate and MT I effectively makes triggering dependent on the total charge at the gate, i.e. for a square pulse, the product of gate current and application time. The precise nature of this dependence is not linear however.
Typical characteristics are shown in Figure 5. Very short pulses of less than lAs duration should be avoided, because only pan of the die area may be triggered. This will cause a hot -spot to form, and may destroy the device. The manufacturer may specify a pulse width, which should be regarded as a recommended minimum value. Once the device is properly triggered, complete conduction is established typically in about 1µ,s. There is, however, a delay time of several microseconds before conduction begins, as shown in the timing diagram of Figure 6.
Figure 4. Dynamic main current/main voltage characteristics.
Figure 5. Peak gate trigger current as a function of trigger pulse duration.
Figure 6. Timing diagram for the triggering process.
A group of characteristics which is not usually mentioned in data sheets describes the effects of the anode voltage and the main circuit current on the gate voltage. The gate drive circuit has to have a high impedance for these to be observed. For example, the continuity of the load in the main circuit can be confirmed, while the triac is in the blocking state, by monitoring the gate voltage. This is particularly of value where the load is a projector lamp, whose failure might spoil an audio-visual presentation, and the use of this technique is advocated in a draft international standard 1.
The gate voltage in this condition is due to the capacitance between the gate and MT2. This is usually large enough to be practically useful only for triacs of 10A rating or greater, smaller triacs requiring an unreasonably high impedance at the gate. The effects of main circuit current on the gate characteristics depend on the triac geometry. When the gate voltage is of the same polarity as the MT2 voltage, the existence of main -circuit current simply reduces the gate current produced by a given voltage, see Figure 7a.
A transient appears at the onset of main circuit current, due to the gate to MT2 capacitance. If the gate polarity is opposite to that of MT2, (I- and III+ modes), there is spectacular disturbance of the gate voltage caused by transients in the main terminal voltage waveform, see Figures 7b and 7c. This characteristic can be traced by applying an alternating voltage through a resistive load to the main circuit, and an out-of-phase current to the gate. Figure 7b shows the IgNg characteristic of a TIC226D, which has a high dV/dt rating but is not characterized in the III+ mode. Figure 7c shows the behavior of another type, which is characterized for all modes, but has a lower dV/dt rating. Figure 7d shows that the gate voltage of the TIC226D is affected by the MT2 voltage, even if the external trigger gate current is insufficient to trigger the device. The occurrence of fast pulses of reverse polarity gate voltage due to capacitive effects, and of negative resistance regions in the gate characteristics, can give rise to considerable r.f. interference. This suggests that operation in the reverse-polarity modes (I- and III+) should be avoided, where possible, for all triacs, even if they are characterized for this service.
Thermal Characteristics
There are limits to both the peak power and the average power that can be dissipated at the gate, and care is necessary to design the gate drive such that fast and reliable triggering is achieved without exceeding these limits.
This should always be verified by taking measurements.
Figure 7
(a). Gate dynamic characteristics with gate and main currents in phase.
(b). Gate dynamic characteristics of TIC226D with gate and main currents in phase opposition.
(c). Gate dynamic characteristics of BT139 with gate and main currents in phase opposition.
(d). Gate voltage waveforms in un-triggered and triggered conditions.
Figure 8. Electrical analogue of thermal circuit
Figure 9. Simple Triac switch
Figure 10. Typical Diac characteristic
Figure 11. Diac relaxation oscillator
The main power loss in the device is due, naturally enough, to the main-circuit voltage drop during conduction, and the manufacturer normally provides data on this, for both half -wave and full -wave operation, as well as for various triggering points along the waveform. When the power loss has been determined, an electrical analog circuit, see Figure 8, can be used to calculate the heat -sink requirements (see Appendix 2).
The performance of the chosen heat -sink design, in its final environment or housing, should always be measured, allowing for worst -case conditions of load current, ventilation and ambient temperature.
Gate Drive Techniques
For a simple switching circuit, it is possible to supply gate triggering cur rent from a resistor connected between gate and MT2, as shown in Figure 9. The device is switched on by closing the (low -power) switch. As the load current falls through zero each half-cycle, control is regained by the gate, so the load current will cease at the end of the half-cycle during which the switch is opened.
This circuit has the advantage of simplicity and is widely recommended in American textbooks.
Figure 12. Practical Triac switch for a motor -driven appliance (for
component values see Appendix 3).
But with a British 240V mains supply, it may be difficult to ensure reliable triggering of least -sensitive devices without exceeding the permitted peak gate dissipation. It is preferable to apply gate current in pulse form, and one common way of doing this is to use a device known as a diac, or Silicon Bilateral Switch (SBS).
This is the solid-state equivalent of a neon -tube, that is to say as far as its electrical behavior is concerned, and has a V/I characteristic as shown in Figure 10, with prominent negative resistance regions. Its action can be likened to a bi-directional zener diode.
Connected as a relaxation oscillator, shown in Figure 11, a series of current pulses are produced in the resistor R2, representing the triac gate. For measurement of the gate -current pulse -width, an actual triac should be connected in place of R2, because the impedance at the gate is very non-linear and cannot be satisfactorily represented by a fixed resistor.
Figure 12 shows the resulting triac switch circuit. Normally the triac fires on the first gate pulse, but if it does not, trigger pulses continue to be applied until firing occurs. This is particularly significant with inductive loads, since the current may not rise above IL, the latching current, until some milliseconds after the triac has first been fired.
Operation of this circuit with highly inductive loads may in any case be unsatisfactory for other reasons. With partially inductive loads, the snubber network R3, C2, is necessary. C2 reduces the rise rate of the MT2 voltage (dV/dt) at current cut-off, while R3 controls the current dumped from C2 into the triac as it begins to conduct.
With resistive loads, it is necessary to add interference suppression components. The interference r.f. is generated by the rapid collapse of the voltage across the triac as it fires. The detailed design of this circuit is dealt with in Appendix 3.
A unijunction transistor may also be employed as a trigger -pulse generator, but this will require a DC supply. Note that to avoid operation (or non-operation!) in the III+ mode, any single -polarity gate drive should consist of negative -going pulses.
In the triac switch circuit, it is necessary for the triac to fire as early in each half -cycle as possible, so as to minimize the loss of load power. Conversely, by varying the delay between the start of the half -cycle and the time at which the triac fires, the load power can be controlled. This can be done by varying the phase relationship of the gate voltage to the main terminal voltage, and this method is therefore called 'phase control'. Simple household lamp -dimmers use this technique, which has the considerable disadvantage of generating a great deal of r.f. interference, due to the sudden fall in the voltage across the triac as it fires. Suppression components L1, C4 are essential, and a typical circuit is shown in Figure 13.
This is one of those circuits whose operation is more complex than appears at first sight. If the components R2, R3 and C2 were omitted, the dimmer control RI would suffer from considerable hysteresis. The lamp comes on suddenly, and quite brightly, as the control is advanced (reducing the resistance).
When the control is turned the other way, the lamp becomes much dimmer, until it finally goes out at a control position noticeably different from that at which it came on. This effect is due to the loss of charge from C1 into the triac gate when the diac breakover voltage is only just exceeded.
Suppose that the resistor R1 is set to a value such that the diac breakover voltage of, say, 34V is just not exceeded.
The capacitor C1 charges to +34V on one half-cycle, then discharges to zero and charges to -34V on the next half-cycle, giving a change of capacitor voltage of 68V. If the resistor RI is then slightly reduced, so that the diac con ducts, this quickly reduces the voltage across the capacitor to, say, 28V. Thus, on the following half -cycle, the change in capacitor voltage is only 62V, so that the breakover voltage is reached at an earlier 'epoch' (i.e. time during the half-cycle), and, when the steady-state is reached, breakover is occurring considerably before the end of the half-cycle, and the lamp is quite bright.
Increasing R1 then smoothly reduces the conduction angle (firing time approaches the start of the half cycle), and, hence, the brightness of the lamp is reduced. This effect can be overcome with the additional components R2, R3 and C2. Gate current is drawn from C2, which is recharged via R2 from the much larger C1, with hardly any effect on the voltage across C1 and consequently on the firing epoch. R3 serves to limit the discharge current from C2, which is desirable on reliability grounds anyway, and reduces the recharging demand made of C1. With these additional components, 'backlash' is practically eliminated. This circuit also works well as a speed-controller for series-wound commutator motors, such as are found in power tools. Speeds can be reduced by around 10 times without an unacceptable loss of torque. A snubber network, R4 and C3, is included to avoid false triggering due to excessive dV/dt from the inductive load and the commutator noise spikes, is necessary. However, the circuit is not ideal for phase-control of loads having significant inductance, because if the firing epoch is not precisely the same for both polarities of supply voltage, then the load current will contain a DC component due to the unbalance, and this may cause undesirable effects due to the saturation of the magnetic circuits of the load.
When resistive loads of greater than about 1kW dissipation have to be switched or controlled, the r.f. interference generated by the circuits described above becomes a serious problem.
For such applications as stage and disco-lighting, there are few alternative techniques, and relatively costly filters are used to eliminate the interference. But for heaters with a large thermal inertia, burst -firing with synchronous switching can be used.
Figure 13. Lamp-dimmer or speed-controller for an AC series commutator
motor
Synchronous Switching
Synchronous switching is a technique for minimizing the amplitude of the voltage transient across the triac as it fires, and it works best when the latching current of the triac is very much less than the full load current. Gate drive is applied in pulse form, the leading edge of the pulse occurring at, or preferably before, the zero -crossing of the supply voltage, so that the triac remains conducting as the load current falls below the holding current and goes through zero. The pulse must, of course, last long enough to maintain the device in the 'on' condition for the load current to rise in the next half -cycle, to a value exceeding the latching current.
There is then no voltage transient, and consequently no r.f. interference.
The technique can also be applied to phase control, where detection of the zero -crossing of the supply voltage is used as a reference for timing the trigger pulses more accurately, and more controllably, than can be achieved with RC phase -shift networks.
Synchronous switching circuits can be designed using discrete transistors, but integrated circuits are now available which offer improved performance, and usually a number of additional features.
An example is the TDA 1024. This is an advanced phase -control device especially suitable for very low -differential temperature control or speed -control of series commutator motors. It can be powered either directly from the mains supply via a suitable voltage dropper, rectifier and stabilizer circuit, or from a local DC supply, and will control the conduction angle of the associated triac in accordance with a DC voltage, which may be obtained from a tacho-generator or a temperature sensor, for feedback control, or from a potentiometer, for open-loop control.
Synchronous switching also pro vides a solution to the problem, mentioned above, of firing -point asymmetry causing DC saturation in phase -con trolled inductive loads. In this case, it is particularly important to use a synchronous switching circuit, or zero -crossing detector, which does not itself suffer from asymmetry. Most of the current integrated circuits satisfy this requirement.
Burst Firing
Burst firing, with synchronous switching, is a technique for controlling load power without generating r.f.i.
Instead of varying, as in phase control, the fraction of each half -cycle for which load current is allowed to flow, the current is allowed to flow for an exact number of half -cycles, followed by an interval, also an exact number of half-cycles in duration, when no current is permitted.
Isolated Driving Circuits
It is essential for safety reasons that low-level drive signals derived from microprocessors, remote sensors etc., should not be applied directly to the triac, if it is, as usual, connected directly to the mains supply. All low-level circuits must be isolated from the mains. If they are earthed, the isolated coupling (opto isolator, pulse transformer etc.) must withstand 2000V for 60 seconds, but if not, then circuit isolation to withstand 3000V for the same period is necessary. It should be noted that many opto-couplers and low-cost pulse trans-formers will not meet these requirements! Of the two devices mentioned, the opto-coupler is perhaps simpler to incorporate in a design, and there is little difference in cost. Furthermore, for AC applications, an optically -triggered triac is much easier to use than a coupler with a bipolar output device. Such a device is the Triac Isolator (Maplin Stock Code: QQ10L), and this can directly replace the switch SI in Figure 12 and in similar circuits. The input of this device can be driven directly from TTL logic, and easily interfaced with CMOS. There is also a version, Order Code RA56L, which includes a zero -crossing detector.
References:
1. IEC Publication 574-3A: Audio visual, video and television equipment and systems. Part 3A: Connectors for automatic slide -projectors with built-in triacs, for audiovisual applications. Inter national Electrotechnical Commission, Geneva. (To be published.)
2. D.R. Armstrong, 'Zero -crossing detector circuits', Mullard Technical Communications No.132 Page 63-68.
October 1976.
Appendix 1
Appendix 2
Figure A. 2. 1. Triac dissipation verses main current, with conduction angle as parameter.
1. From a graph similar to Figure A.2.1, which will be found in the triac data sheet, find the power, PT, dissipated by the triac at the given value of load current, IT, and conduction angle a.
2. From the data sheet, substitute values in the electrical equivalent circuit, Figure 8.
R,_a = thermal resistance of triac junction to ambient.
Ri_mb = thermal resistance of triac junction to mounting-base.
Rmb_h = thermal resistance between the mounting -base and the heat -sink.
Allow 1°C/W for a mica washer, 1.5°C/VV for contact resistance with out thermal jointing compound, 1°C/VV for contact resistance with silicone grease, and 0.5°C/W for contact resistance with oxide -loaded thermal jointing compound. These values apply for TO-220, TO-3 and similar encapsulations.
We have to find Rha of the necessary he
3. Substitute I, =
- Tamb), where T, permitted junction
= thermal resistance at-sink.
PT, also V, = max max is the maximum temperature of the triac, from the data sheet, and Tamb is the maximum value of ambient temperature (35°C for the normal household environment). Preferably, reduce the value of V, by 10°C, for greater reliability.
4. From the equivalent circuit:
I, = V + 11,_a V, Rj _ th Pmb-h + Rh -a
This is just Ohm's Law. For many devices and applications, R,_a can be neglected (thereby adding a safety factor).
Then Rh -a = (V/I)) -R -mb Rmb-h = (Ti max - Tamb) - _ Rinb_h PT
Appendix 3
Referring to Figure 12, the triac, T1, is chosen to have a Vgo rating greater than the peak voltage of the supply. For 240V mains, a V50 of 400V must be regarded as a minimum. The IT rating obviously depends on the required load current. It is wise to choose a generously - rated device, because this does not usually increase the cost greatly, and often gives a lower thermal resistance as a bonus. For example, a TIC226D, rated at 400V/8A r.m.s., could be chosen for this circuit, and is quite inexpensive.
We next look at the snubber network, R3 and C2. In the absence of these components, the series inductive component LL of the load impedance will cause a voltage spike to appear across the triac when it switches off. This spike may exceed the dV/dt rating of the triac and cause it to remain conducting! To avoid this, C2 is added to slow down the transient and reduce its amplitude, and R3 is included to ensure that the tuned circuit L1/C2 is, at least, critically damped.
Unfortunately, the value of LL is often not known, and it is not easily measured because, being dependent on an iron-cored component, it varies with the applied voltage and frequency. Under these conditions, it is usual to choose an initial value for C2, such as 100nF, and to examine the resultant voltage transient with an oscilloscope. Any necessary adjustment, to achieve a desired value of dV/dt, can then be easily made. If LL is known, then:
C2 > 2VS2/{LL(dV/dt)2 max}
Similarly, the value of R3 depends on that of LL: R3 > 2 L1/C2 and, in practice, is adjusted so that the voltage transient is observed to be well damped.
C2 should be rated for 250V AC working, and a self-healing `X -type' capacitor should be used.
R3 has to pass a current transient greater than 3A if power is applied to the circuit at a voltage maximum, and should be chosen accordingly. Some low-cost metal -film resistors are unsuitable.
D1, the diac, can be chosen from the few types (e.g. ST2 (QLO8J)) available.
The symmetry of the breakover voltage is important in minimizing the DC component in the load current. It is worth noting that the prices of rather similar devices of this type vary considerably.
C1 is chosen to store enough charge to supply the gate current required for reliable triggering. Most samples of triac require typically one -tenth of the triggering current given in the data sheet, so an experimental approach is unwise here.
C1 should not be made too large, as this will delay the triggering and cause loss of load power and increased r.f.i. A value of 47nF is satisfactory in this circuit. Again, a 250V AC rated component of a -type' is preferred.
R1 should have a resistance much lower than the impedance of C1, so that the triggering delay is minimized. But if R1 is made too low, a short-circuit in, or across, C1 may cause such a severe overload in RI that it burns the printed-circuit board and adjacent components.
A simple fault may thus destroy the circuit. Luckily, a value of 181d1 will avoid this problem, without delaying the trigger point unacceptably.
R2 is required to adjust the trigger current pulse duration to exceed the minimum recommended value, which is 20tis for the TIC226D. A value of 4711 is satisfactory.
The switch SI can be any form of low -current switch, such as a light -touch push or a reed switch, but it must be insulated for use in direct connection with the mains supply.
We can determine the heat -sink requirements using the final result from Appendix 2. From the device data-sheet,
'I', max = 110°C, PT = 2W at IT = 2.08A, and 13.,_flib = 1.8°C/W.
Allowing for a mica washer, without thermal jointing compound, Rmb-h = 2.5°C/W, Tamb = 35°C and the derated T, max = 100°C.
Then Rha =
(Ti max - Tamb) PT 65 2}
- 4.3 = 28.2°C/W
- R,_ nth - Rmb_ h
By comparison, 11,_a = 62.5°C/W, from the datasheet. This is not negligible compared with the heat-sink requirement, and would allow the use of a heat-sink having a thermal resistance of 50`C/VV rating. The final design should be checked by measurement.