For example, Nelson Pass2 appears to recommendmultiple-transistor complementary output stages, asmandated by class-A operation, to circumvent theneed for V-I protection of bipolar devices, while

Rod Elliot3 suggests that V-I limiters can be dispensedwith altogether by adopting e-MOSFETs.

These views appear to be rather more widely acceptedthan they should, and constitute a charter for near heroicunreliability in amplifiers so designed. The zener diode-clamping of gate-source voltage for e-MOSFET’s isthought by some4,5 to be all that is required with regard toprotection. While the zener diodes are mandatory, (ideallywith 10V<Vzener<20V to prevent premature clamping),they only serve to protect the e-MOSFET gate oxideinsulation from over-voltage destruction6, and do nothingwhatever to protect the device from accidental shortcircuits and forbidden voltage-current combinations thatmay occur when the amplifier is called upon to drive

reactive loads.The positive temperature coefficient of on-resistance7,

(and therefore negative temperature coefficient of draincurrent), enjoyed by e-MOSFETs eliminates the secondarybreakdown phenomenon which is the bane of bipolartransistors, but does not constitute a licence for wilfulviolation of power dissipation limits in linear, audio-frequency applications. This is in contrast to ultrasonicswitching usage, where e-MOSFET dissipation boundscan be blissfully ignored, and adherence to drain currentand drain-source voltage limits will suffice.

All output stage semiconductors used in complementary,or quasi-complementary, (full or half bridge), linear audiopower amplifiers, without exception, require V-Iprotection for reliable operation. However, such circuitrymust be carefully designed to prevent premature activationduring normal amplifier operation.

Single-slope, linear-foldback limitingMany low to medium-power, (sub-100W), commercialaudio amplifiers incorporate a single slope, linearfoldback, voltage-current protection circuit, (figure 1),attributed to S.G.S. Fairchild Ltd. by Dr A.R. Bailey8. Inpractice the complimentary output transistors, To1 and To2,may each consist of a compound arrangement of at leasttwo transistors in series. The collector-emitter voltage,Vce, across To1 is sensed by R1, and R3, while the outputcurrent, in the guise of a voltage developed across emitterresistor Re, is simultaneously monitored by R3, and R2.The voltages are thus summed algebraically at the base ofthe protection transistor, Tp1, which is driven intoconduction, shunting voltage drive to To1, in the event ofan over-voltage, over-current or simultaneous occurrenceof both conditions in the output device.

The series resistor, Rs, (typically 100R), expedites thisprocess by limiting the current required by Tp1 to shuntvoltage drive to To1. The freewheeling diode, DF, protectsthe output device from excessive base-emitter reversebias9, due to over-rail voltage spikes generated byinductive loads, while DF performs the same function forthe small-signal protection transistor, by preventing itsbase-collector junction from being forward biased10.

If the output approaches the negative supply rail whiledriving a sufficiently low impedance, the current sunk byTo2 generates an appreciable voltage drop across itsemitter resistor. The output is therefore at a significantlyhigher potential than the common input to thecomplementary output stage. Transistor To1 is reversebiased, and Tp1’s base-collector junction, in the absence ofDF, would be forward biased, resulting in current flowfrom emitter to collector.

Diode DF prevents this form of spurious, inverse-activemode limiter activation by decoupling Tp1’s collector asTo1’s base-emitter is reverse biased. The potential at To1’semitter is then equal to the output voltage since, contraryto Duncan11, To1 is non-conducting and no current, except

46 ELECTRONICS WORLD September 2002

The desirability or lack thereof, of over-voltageand over-current protection for powersemiconductors in audio power amplifiersremains a point of contention in the field. MichaelKiwanuka explans.

Transparent V-I protectionin Audio Power Amplifers (Part 1)

V+

TO1

V-

TO2

TP1

DP

DP

TP2

Re

Re

Output

R1

R1

R2

R2

RS100R

RS100R

R3

R3

Df

Df

Figure 1: Single slope,linear foldback protection

circuit applied to acomplementary emitter

follower.

negligible leakage, flows through itsemitter resistor. By symmetry, theexplanation above also applies to thenegative half of the circuit.

A small-value capacitor issometimes connected across the base-collector junction of each protectiontransistor1, with a view to eliminatingbenign parasitic oscillation12 that mayoccur sporadically in the networkduring the limiting process. Thesecapacitors appear in parallel at A.C,and are entirely unsatisfactory, as theycreate an ill-defined and thereforeundesirable feed-forward path aroundthe output stage, shunting it out of theglobal feedback loop at high audiofrequencies, precisely where theamplifier is most vulnerable withrespect to non-linearity. Suchvulnerability is due to a necessarilydiminished feedback factor at highaudio frequencies in the interest ofNyquist stability. Connecting thecapacitor, (of the order of 1n0), acrossthe base-emitter junction of eachprotection transistor is the preferredsolution.

The resistor values for thearrangement in Figure 1 are obtainedby drawing the desired protectionlocus onto a linear scale graph of theoutput transistor’s safe operating area,(S.O.A). One of the three resistors (usually R3), is assignedan arbitrary value (typically 100R•R3•1k), and the tworemaining resistors calculated from simultaneousequations developed from two convenient points on theprotection locus.

This arrangement requires that the linear protection locusintersects the S.O.A’s Vce axis at a value greater than thesum of the moduli of the amplifiers voltage supplies,otherwise Tp1 turns on under normal loading when theoutput swings negative, even with the output open-circuit.Similarly Tp2 would be activated under normal outputloading when the output swings positive. This effectivelyshort-circuits the small signal circuit preceding the outputstage directly to the output, causing gross and very audibledistortion. Failure to adhere to the above condition appearsto have caused some designers to erroneously abandonelectronic S.O.A protection of any form altogether1,13.

This requirement however, constitutes a significant

limitation with regard to efficient utilisation of thecomparatively large S.O.A in the low-Vce region of thegraph, especially at high supply-rail voltages where, in thecase of bipolar transistors, secondary-breakdown severelycurtails flexibility in optimal placement of the protectionlocus. This is graphically illustrated in figure 2, for anamplifier with ±40V supply rails, using Motorola’sexcellent14 200W, MJL3281A-MJL1302A complementarypower transistors.

Only the positive half, Fig. 3., of the circuit in figure 1needs be used to calculate the required component values.Ideal devices are assumed, with infinite input impedance,zero saturation voltage, and zero ohmic resistance, theerror thus accrued is negligible. Let Vbe=0V6, R3=220R,Re=0R22. Taking two arbitrary points, A and B on thelocus, such that,

0 6 0 2 80. = ≤ < =( ) V V Vce cc

PROJECTS

47September 2002 ELECTRONICS WORLD

Figure 2: MJL3281A safe operating area with single slope, linear foldback protection locus

40V

TO

TP

0V6 Re

IC = 6A85

R3

1V5070R22

R1

R2

RS100R

39V093

38V493

40V

220R

Output

Figure 3: Output conditions at point A on the protectionlocus in figure 2.

40V

TO

TP

0V6 Re

IC = 4A

I1

I2

I3R3

0V880R22

R1

R2

RS100R

33V72

3V12

4V

220R

Output

Figure 4: Output conditions at point B on the protection locus in

where for point A, Ic=6.85A; Vce=0V, and for point B,Ic=4A; Vce=36V, it follows from figure 3:

(1)

With reference to figure 4:(2)

(3)

Solving (1) and (3) simultaneously gives R1²12K4 and R2²143R0. To afford an acceptable degree of precision, it isrecommended where necessary, that these values be madeup from series, or parallel combinations of 1% resistors.

When the output swings to –40V, then 80V appearsacross R3 in series with R2//R3, to a good firstapproximation. Therefore the voltage present at the base ofthe protection transistor, Tp, is given by:

It follows therefore that subject to instantaneouscollector current, ic, being less than the maximumpermissible collector current, IC(MAX), at Vce²2|Vcc|,spurious activation of Tp cannot occur. A generalexpression which allows the rapid verification of thecompliance of any amplifier using single slope, linearfoldback limiting may developed:

(4)

Equation 4 is valid subject to the following condition:

(5)

This condition is invariably fulfilled during normaloperation, as no practical loudspeaker system woulddemand that the output transistor sustain Vce²2|Vcc|, whileproviding any appreciable current.

Figure 5 shows a common variation11,14,15, on the

i IC C MAX V Vce cc

<≈( ) 2

20 62 3

2 3 1 2 1 3

V R R

R R R R R RVcc

+ +( ) <

20 62 3

2 3 1

V R R

R R RVcc / /

/ /

( )( ) + <

VR R

R R RVbe ≈

+≈80

0 552 3

2 3 1

( / / )( / / )

0 6 36 28 0 28 2202 1. . .= +( )R R

0 6 40 3 72 4 3 722 1 3. . .R R R= −( ) + −( )

I I I2 1 3= +

1 507220 220

2

2 1 1

. R

R R R+ +( )

PROJECTS

48 ELECTRONICS WORLD September 2002

V+

TO1

V-

TO2

TP1

TP2

Re

Re

Output

R1RS100R

RS100R

R3

R3

R1

Figure 5:Compromised

single slope, linearfoldback scheme

resulting in grosslyinefficient S.O.A

utilisation.

Figure 6: Linear protectionlocus clearly shows

inflexibility of scheme infigure 5.

single slope, linear foldback limiter of figure 1, withresistor R2 excised, so that from equation 4:

Since , then:

The optimal protection locus for this network Fig. 6., isplotted so that calculated resistor values comply with theabove condition. This scheme is atrociously inefficient, asfor a nominal Vce²0V and Re=0R22, resistors R1 and R3 arein parallel, and collector current, Ic, is perforceprematurely limited to,

A value of Re=0R1 gives a modest improvement, with,

The protection locus is realised by deriving output stageconditions Fig. 7. for a single, arbitrary point, B on thelocus, subject to, 0<Vce<2|Vcc|. With Vcc=40V,Re=0R22, R3=220R and noting that R1, R3 constitute asimple voltage divider:

This unwarranted dependence on the value of Re isunacceptable, as in some applications such as output stagescomprised of paralleled, complementary e-MOSFET pairs,(0R1<Re•1R0), may be required to ensure equable currentsharing.

Driving reactive loads.A clear appreciation of the nature of the amplifierís load isrequired to establish the bounds within which the V-Ilimiter must remain inactive. Figure 8 shows an idealcomplementary emitter follower, (in ElectronicsWorkbench’s excellent Multisim professionalsimulator16), used to drive a standard (8ohms 0°) test loadto ±40V supply rails.

The plots obtained in Fig. 9. show that the voltage vce,across To1 is precisely 180° out of phase with the current,ic,, its required to source; the voltage across the device is aminimum when its collector current is at a maximum, andvice versa. Instantaneous power dissipation is merely theproduct of instantaneous device voltage and current. Peaktransistor dissipation, Pd(max)²50W, occurs twice in To1’sconducting half-cycle, at half the peak load voltage,(Vout/2²Vcc/2) and half the peak load current, ic(peak)/2.

As the (8ohms 0°) load line lies well below the linearprotection locus in figure 10, (reproduced from figure 2),

it is clear that a single pair of MJL3281A-MJL1302Apower transistors, operating from ±40V rails willcomfortably drive an 8ohms dummy load to clippingwithout V-I limiting. This however, will certainly not bethe case with loudspeaker loads, which are invariablyreactive17,18. An amplifier with ‘high-fidelity’ aspirations,intended to drive full-range, multiple-transducerloudspeaker systems, including electrostatics, should atleast be capable of driving a (4ohms ±60°) impedancewithout V-I limiting.

A (4ohms –60°) impedance was devised by driving a2ohms0 resistor in series with a 45µ 9441 capacitor at1Khz with the ideal complementary emitter follower infigure 8. The traces thus obtained, Fig. 11., were used toplot the (4ohms ±60°) load line in figure 10. Peaktransistor dissipation, Pd(max)²352.9W, occurs atvce²45.97V, and ic²7.68A.

In other words Fig 12., because current leads voltage ina capacitive impedance, the NPN transistor, To1 in figure8, is required to source ²7.68A when the output swings

RR

K1

40 39 439 4 39 78 220

46≈ +( )− +( )

≈.. .

I V R Ac V V be ece ≈

< ≈( ) 06 0

I V R Ac V V be ece ≈

< ≈( ) 02 7

20 63

3 1

V R

R RVcc

+( ) <

R2 → ∞

20 63

3 1 13

2

V R

R R RR

R

Vcc

+ +

<

PROJECTS

49September 2002 ELECTRONICS WORLD

40V

TO

TP

0V6 Re

IC = 1A

R3

0V220R22

R1RS100R

ñ39V4

-40V

-39V78

220R

40V

TO1

TO2

Ic

Pd

Vout

Vce

Load

1Ω

40V / 1kHz / 0º

0V

8Ω

C

A

B

40V

XX

Y

1V / V /0V

Figure 7: Outputconditions at point B onthe protection locus infigure 6.

Figure 8: Ideal emitter follower used todetermine instantaneous collector current, Ic,collector-emitter voltage, Vce, and devicedissipation, Pd.

Figure 9:Instantaneous Vce,Ic, and Pd in sourcingoutput transistor,driving 100W into(8ohms 60°).

away from the negativesupply rail to ²–5.97V.Similarly, the PNP device,To2, must sink ²7.68A whenthe output swings to +5.97Vfrom +Vcc. Note that thecrossover discontinuity inthe output voltagecharacteristic (figure 12),now precedes zero crossingby 60°, at |Vout|²35V.

For a (4ohms +60°)inductive impedance, inwhich current lags voltage,the output conditions arereversed, with the loaddemanding 7.68A from To1when the output swings fromthe positive supply to–5.97V. Regardless of thenature of the load however,device voltage, vce, and load

voltage, vout, are always 180° out of phase, and being avoltage follower, the input voltage is always in phase withvout.

The linear foldback protection locus of figure 10 onlypermits 3.1A at Vce=45.97V, therefore a minimum ofthree, (ideally four), output pairs are required to drive anotional (4ohms ±60°) loudspeaker system from ±40Vsupply rails without intrusive limiter activation. On thisbasis and using other established techniques12,19, includingD.C. offset, and thermal overload protection, a reliable,low distortion, 100W into (4ohms ±60°) class-B amplifiermay be constructed.

As the cost of power transistors is significant, there is acompelling financial incentive to minimise the number ofdevices used by utilising the S.O.A as efficiently aspossible. To this end it has been suggested11 that ideallythe protection locus should closely match the bounds ofthe S.O.A. This is unnecessary, as reactive load driveprimarily requires that current delivery in the|Vcc|•Vce<2|Vcc| region be maximised without violatingD.C safe operating limits. In general an optimally located,non-linear protection locus with no more than onebreakpoint should suffice.

Single slope, single breakpoint non-linear foldback limiting.Introducing a zero-gradient segment Fig. 13, at someoptimal point in the protection locus permits theenhancement of current delivery at the low-Vce end of theS.O.A., without significantly compromising availablecurrent at higher device voltages. The single slope, linearfoldback ‘protocol’, (equation 4), is made redundant, asthe protection locus does not cross the Vce-axis at anypoint. This scheme is briefly mentioned in reference [20],

PROJECTS

50 ELECTRONICS WORLD September 2002

Figure 10: Reactive loadgives rise to an ellipticalline, resulting in morethan seven times greaterpeak device dissipationthan for the (8ohms– 660°)

Figure 11: Instantaneous Vce, Ic, and Pd in sourcing output transistor,driving 150W into 4ohms– 660°).

Figure 12: Transistor To2 delivers 7.68A to the(4ohms– 660°) load when output swings away from -Vccto -5.97V. Note that the crossover discontinuity markedX precedes zero voltage crossing by 60°.