ETI-466 High Power Amplifier
Technical Review

Table of Contents

  1. Introduction

  2. How it Works

  3. Variations

  4. Notes

  5. Set-up Procedures

Ref. also:

Appendix 1: SOAR Calculation Results

Appendix 2: ETI-466 Schematic

Appendix 3; The original ETI article, divided into three as Part 1 and Part 2 and Part 3.

Appendix 4; A key article used throughout this text, written by David Eather, "A Practical Approach to Amplifier Output Design", Silicon Chip, February 1991, pp.14~18 Part 1 and Part 2 of the same article, from Silicon Chip, April 1991, pp.64~67

Click here for the full size image. From the original published article, this ETI466 is dominated by what can only be described as an "interesting" heatsink arrangement!
The ETI466 amplifier module as it featured on the cover of the Feb. 1980 magazine. Constructional details included those for the metalwork for that quite spectacular heatsink!

1. Introduction


Project ETI-466, by Barry Wilkinson, appeared in the Australian edition of Electronics Today International, (ETI) for February 1980. It was introduced as successor to both the ETI-480 and later ETI-470. Much more sophisticated, complex and expensive to construct than these earlier projects, it is rated at 200W into 8 Ohms, and a little over 300W into 4 Ohms. It thus represents around the power limit of conventionally connected, reasonably common BJT outputs. The ETI-466 is a solid, reliable amplifier, and a good foundation for up-rating, variation or plain sustained dependable operation.

As one application example Figure D and Figure F show views of a pair of "stock" ETI-466's mounted side-by-side on a common, fan cooled heatsink, intended for semi-pro use. Another pair, Figure PR, running slightly lower than normal B+ rails, delivered 150W/ch. into nominal 8 Ohms loads, on separate convection cooled heatsinks in a friend's disco set-up for a number of years.

Fig. D
Fig. D
Click here for larger image.

Fig. F
Fig. F
Click here for larger image.

Fig. PR
Fig. PR
Click here for larger image.

Despite at least one letter to ETI following the original article [ref. 1]complaining of stability problems, these are easily avoided by good construction practice, and problems encountered early on with particular kits related to particular parts supplied in those kits. While minor modifications will improve reliability, there are no vital major changes to the original ETI-466 circuit necessary for operation as originally published.

RCS Radio can still supply ETI-466 PCB's!

It remains do-able to construct these amp boards, as PCB's may still be had from RCS Radio in Sydney, Australia. You'd need to contact RCS directly via their website if you wanted to obtain PCB's - in December 2001 they were priced at around A$34 or so each - about US$17.00.

2. How it Works


2.1 Input Differential Amplifier

Q1, Q2 and Q5, Q6 form complimentary differential pairs at the input of the amplifier. The differential pair is a topology that appears at the input of almost all conventional amplifiers. It confers a number of advantages, (over a single ended input), including an inherent distortion cancellation mechanism independent of transistor parameters such as hfe. It is critical however that the design makes the collector currents of the pair equal, and like the ETI-480 this is not quite the case with this amplifier. There are benefits in terms of substantially less (arguably desirable second harmonic) distortion to be gained from maintaining strict balance between the Q1, Q2 and Q5, Q6 currents in a differential amplifier [ref. 2].

The input differential amplifiers are "transconductance" amplifiers, meaning that they take a voltage difference input and produce a current drive output. Input differential amplifiers are normally operated at low gain, ensuring bandwidth much wider than the voltage amplifier stage (VAS) following, thereby minimising the amount of lag compensation required for stability. Each of the differential pairs has its own constant current source, Q3 and Q4, operating at (5.1 - 0.7)V/2.2kOhm = 2mA. 22 Ohm emitter degeneration resistors R8, R9 and R15, R16 provide local negative feedback which improves linearity, reducing distortion at higher audio frequencies and improving slew rate [ref. 3].

The gain of each input differential amplifier is given by R4/2(re +RE) [ref. 4]. re is approximately 26/Ie where Ie is in milli-amps, here 1mA, i.e. re is 26 Ohms. RE is R8, R9, R15 or R16 = 22 Ohms. The differential amplifier gain is therefore approximately 4.7kOhms/(2 x [26 + 22]) = 49. The input impedance of each of the following VAS elements, (Q7 and Q9), is approximately Zin = hfe(re + RE) = 25 x [(26/Ic(Q7, Q9)mA) + 390 Ohms] = 25 x[26/7 + 390] [ref.5], or approximately 9.85kOhms as a minimum. This impedance lowers Q1, (and Q6's) collector load impedance to 4.7kOhm//9.85kOhm = 3.2kOhm, and the differential amplifier stage gain to 3.2kOhm/(2 x [26 +22]) = 33. The two differential amplifier stages (Q1, Q2 and Q5, Q6) act in tandem, so the gain figures given above must be doubled for an overall result of approximately 66 minimum.

Addendum - Errors!

Good friend and veteran amplifier designer Hugh Dean, 30th June 2000: You calculate the gain of the input differential amp in the ETI-466 at 49. Does this mean voltage gain? How do you specify gain of a transconductance amp? As soon as the output is strapped across a b-e junction, you are generating a varying current, so really the diff.-amp gain should be expressed as mA/V, transconductance.

Hugh R Dean, 2 July 2000: I read some time back in a Peter Blomley article where he aims for a transconductance in a two stage input stage of 1A/V - a transconductance of around 1000mA/V must surely rank as unbelievably high by valve standards. Quite achievable with transistors, he says. I think this figure is as close to a gain as you can get.

Hugh Dean

Paul Cambie, 1 July 2000: The substantial majority of that overall voltage gain is contributed by the VAS, but not all. The diff.-amp contributes too. The diff. amp's contribution is as transconductance, and the VAS's contribution is as transresistance, (not outright voltage gain as such), and the product of the two is the overall open loop voltage gain of the amp, assuming an exclusively follower based output stage.

Hugh R Dean, 2 July 2000: Probably better then to merely describe the voltage amplifier as a two stage affair. The VAS can't work without a varying current input anyway, so it's a bit like saying you need a transmission and an engine to make a car!

Paul Cambie; I see that I've expressed the gain of the diff.-amp as though it were a voltage amp, (ref. p.99, Art of Electronics, 2nd Ed.), and not at all as a transconductance amp. I can only imagine that this was because I took one horrified look at the (correct) formula for the diff. amp relationship between Iout and Vin that Doug Self gives, and had a brain stoppage!

Doug Self gives a formula for the relationship between Iout and Vin for the diff.-amp, (p.63, Audio Amplifier Design Handbook, 1st Ed.), referencing "Grey and Meyer, Analysis and Design of Analog Integrated Circuits, Wiley, 1984, p.194.

"The differential pair has the great advantage that its transfer characteristic is highly predictable. The output current is related to the differential input voltage Vin by:

Iout = Ie.tanh(-Vin/2Vt)"

Ie is the diff.-amp tail current, and Vt is the thermal voltage, 26mV at 250C

Not being even vaguely familiar with the "tanh" function, I found that it's a "hyperbolic" function;

tanh(x) =3D (ex - e-x)/(ex + e-x)

Plotted graphically it looks roughly like an "S" shaped curve in its side, with the origin right in the centre of the "S" curve. Overall then, the diff.-amp contributes gm = Ie.tanh(-Vin/0.056)/Vin and the VAS contributes R = hfe x Rc, where Rc is "the effective value of VAS collector impedance", to give an overall amplifier open-loop voltage gain of Av = gm x hfe x Rc. Diff.-amp typical gm appears to be roughly in the 20mA/V region, hfe of the VAS might be say 50 and a "typical real value for Rc" might be about 20kOhm at low frequencies. This suggests an overall voltage open loop gain for a "typical" amplifier of 20,000.

Hugh R Dean, 2 July 2000: It seems a little high to me; if we have bypassed the lower divider network in a conventional solid-state input/VAS stage and come up with a gain around 3,000~7,000 depending on degeneration of the differential pair and the VAS.

2.2 DC Output Offset


For a DC coupled amplifier, with advantages inherent in avoiding AC coupling, the input differential amplifier provides a minimal DC offset. In the ETI-466 the base current of one differential pair offsets that of the other pair in terms of setting up a DC offset, (from zero volts), at the input, (bases of Q1 and Q6), and hence the output. So quite low DC offset is inherent with this input stage, and close matching in value of R19 and (R2 + R3) is not required to minimise it.

Steps to ensure low DC offset at the output remain necessary. Q1, Q2 and Q5, Q6 should be gain [ref. 6] matched using a transistor hfe tester, as closely as practicable, to within say 10%. The gain of each should be high, (in excess of about 200 preferably). Mismatched gains give an offset at the output of as much as 50mV, but a gain difference between each member of the pairs transistors of say no more than 10% gives a DC offset at the output of just a few mV. It is not necessary to have the gains of the PNP pair matched to that of the NPN pair, but they should be similar. Widely differing gain values between the two pairs will worsen output DC offset.

Ic(Q1) and Ic(Q2) should ideally have the same nominal value, as should Ic(Q5) and Ic(Q6) likewise. Replacing R7 and R17 with short circuit links would be one modification that should achieve this, although I have not verified this in practice. Having R6 = R7 and R17= R18 gives apparent symmetry, (visually when looking at the schematic), but that's all. The constant current sources, in the differential amplifier tails, fix the total current through the Q1, Q2 and Q5, Q6 pairs. The feedback action between the two emitters would cause the c-e currents in each pair to remain the same, with R7 and R17 shorted. Collector resistors R6 and R18 are necessary to provide a load across which to take an output and set up the current through the VAS, (Q7 and Q9), but R7 and R17 are simply not required at all.. Replacing them with short circuit links will allow the circuit to find its own equilibrium such that the collector currents of each pair equalise at about 1mA each.

In an amplifier with a single differential input, DC offset at the output occurs when there is an imbalance between the base current flowing in, (say), Q1 and then R3 then R2, and that flowing in Q2, (say), and then R19. Here, in the example case of the Q1, Q2 pair, currents in the reverse direction occur in the Q5, Q6 pair. The two largely offset each other, and dependence on strict balance between R19 and (R2 + R3) does not occur.

Output DC offset does occur to some degree in practice in this circuit. As a general rule this should be no more than +/-20mV maximum. Anything above this should be corrected. Minor changes can be made by changing one of the transistors in one of the transistor pairs, and using one with slightly greater or lesser gain. Switching, (say), Q1 and Q2, one for the other, can produce the desired adjustment; while the gains of each transistor in the differential pairs should remain closely matched some small difference will remain, and in practice this can be used to advantage here.

Major DC offsets at the output, 50mV or more, should be considered as "faults", and fixed. Grossly mismatched gains in the differential pairs can produce this result, for example.

2.3 Voltage Amplifier Stage (VAS)


2.3.1 General Operation

The Voltage Amplifier Stage (VAS), Q7 and Q9 and associated components, provides almost all the voltage gain of the amplifier. It must be able to provide the full output voltage swing of the complete amplifier. It is important that the voltage gain of the Q7~Q9 circuit is high. The two devices are driven by out of phase inputs, from their respective differential amplifier drive circuits, so they act in complimentary fashion. "...the collector current of one decreases while the other increases", as described in the "How it Works" section of the original article, "...This configuration provides quite an amount of gain".

The VAS is a "transresistance" amplifier, converting current drive at the bases of Q7 and Q9 to voltage drive at their collectors. The transresistance amplifier thus has the reverse current-voltage relationship to the input differential transconductance amplifier considered earlier. The output from the differential pairs is current (not voltage) drive to the VAS (ref. section 2.1). In fact the voltage appearing at the bases of Q7 and Q9, as observed with a CRO, can be quite misleading. It is a very low amplitude, (a few milli-volts and substantially less than the voltage amplitude of the signal being input to the whole amplifier in the first place), distorted triangle shape, (for a sinewave test signal input).

2.3.2 Stability

As frequency increases the gain of an amplifier falls and phase shift imparted on the amplified signal increases. Each amplifying stage adds phase shift due to junction capacitance; (this applies to both voltage amplifiers and current amplifiers in the output stage). When this phase shift at high frequencies accumulates to 1800 then the amplifier feedback, which is 1800 negative feedback anyway, becomes 1800 + 1800 = 3600 positive feedback.

This isn't a problem provided that, by a healthy margin, gain has already fallen below unity. If the gain at some high frequency at which amplifier phase shift has reached 1800 is still above unity, then the amplifier becomes an oscillator, self sustaining even the slightest circuit disturbance, (e.g. the effect of switching it on), leading promptly to self destruction in all likelihood. While there are a number of components in the circuit involved with ensuring stability, including Zobel network C15, R47 and output stage high frequency response roll-off capacitors C13 and C14, there are two principle networks specifically for the purpose. These are C12, R26, and C8 in combination with the gain setting resistors, R19 and R5.

Conventional stabilisation technique is " always to put in one dominant lag to attenuate the loop gain at 20dB/decade (6dB/octave), starting from a corner frequency which is sufficiently low to ensure that the loop gain is reduced to unity before the other lags inevitably present at high frequencies have produced too much further phase lag." [ref. 7] In other words, the conventional method for ensuring amplifier stability is the addition of a "first order" filter, rolling off the amplifier response earlier than it otherwise would, at -6dB per octave. This is referred to as inserting a " pole" in the amplifier's frequency response. Since its value is deliberately structured to precede all naturally occurring poles for a given amplifier as frequency increases, it is referred to as the "dominant" pole.

The C12, R26 filter is placed between the VAS transistor collectors and one power supply rail, being AC ground. Connecting to the actual zero volt line may appear more desirable, but presumably during the design it was found that using the negative rail, (as AC ground), instead was more practical and equally effective.

C8 provides phase advance or "lead" compensation by being in the feedback path. The C12, R26 network is a phase "lag" compensation method, in the forward gain path. C8, 330pF, is connected in parallel with feedback resistor R19, increasing feedback at high frequencies and thereby reducing gain. " is often advantageous to add a capacitor of quite small value across the feedback resistor ... sufficient to cause a little phase advance around the unity-loop-gain frequency and a reduction in the rate of attenuation of loop-gain at frequencies above this." [ref. 8]

2.3.3 Constant Current Sources

The input differential pairs Q1, Q2 and Q5, Q6 are operated with constant current sources Q3 and Q4. These increase the effective load impedance at which Q1, Q2 and Q5, Q6 are operated, without requiring a correspondingly large supply voltage to provide the same collector currents. Use of a constant current source to control the current for the input differential pair results in improved power supply ripple rejection [ref. 9], and better rejection of common mode input signals [ref. 10] than a simple tail resistor.

R20 and R25 set the reverse current through Zener chain ZD1~ZD3 at around (68V - [-68V]) - (5.1V + 62V + 5.1V), all divided by (1kOhm + 1kOhm), which gives about 32mA, less the differential amplifier currents totalling about 4mA, and less about 7mA flowing through Q7, Q9, leaving around 20mA [ref. 11]. ZD1 and ZD3 in turn fix the voltage across the series combinations of R13 and Q4 base-emitter junction, and R11 and Q3 base-emitter junction. The differential amplifier tail currents are thereby fixed at (VZD1 - VbeQ3,4)/R11,R13 = (5.1 - 0.6)V/2.2kOhm = 2mA.

Q7 and Q9 are considered as operating as constant current sources in the "How it Works" section of the original article, operating at around 7mA. There is very little voltage variation seen at the collectors of transconductance amplifiers Q1 and Q6. The bases of Q7 and Q9 are therefore held almost constant with the voltage drop across R6 and R18 impressed across the base-emitter junction of either Q7 or Q9, plus R21 or R24 respectively. The voltage drop across R6 and R18 is half the 2mA Q3, Q4 tail current, multiplied by R6, R18, i.e. 1mA x 4.7kOhm = 4.7V. The Q7, Q9 current is therefore (4.7 - 0.6)V/390 Ohm = 10.5mA, by calculation.

2.3.4 Vbe Multiplier

Resistors R22 and R23 together with potentiometer RV1 control the voltage across Q8, the degree of forward DC bias on Q10 and Q11, and hence the quiescent current of the output stage. Temperature stability is attained by mounting Q8 on the heatsink. The Q8 collector-emitter current increases with temperature rise caused by heat dissipation in the output transistors. Increased Q8 collector-emitter current causes the voltage dropped between Q8 collector and emitter to therefore decrease. Reducing the voltage difference between these two points means reducing the forward DC or quiescent bias on Q10 and Q11, and hence the output devices. Thus the increased quiescent current produced in the output transistors by the temperature increase that caused the change in the first place, is countered [ref. 12] .

2.4 Output Stage


2.4.1 General Operation

The ETI-466 uses one of the most common high power complimentary output transistor pairs of all; the MJ15003 and MJ15004. Vcer is 140V, Ic is 20A, Pdiss is 250W and they are TO 3 package devices [ref. 12A]. The circuit topology used provides voltage gain with local negative feedback applied, as well as current gain, and is a variation on the conventional Complimentary Feedback Pair (CFP) [ref. 13] configuration, (also called a "Sziklai Pair"). The CFP confers significant output stage linearity advantages over a purely emitter follower topology, and lower quiescent dissipation and better thermal stability.

2.4.2 Design Calculations

Design calculations for 200W and 300W output into 8 Ohms and 4 Ohms respectively, may be summarised as follows [ref. 14].

2.4.3 Peak Voltage & Current Delivered to the Load

Vmax load = (2 x P x Z)0.5 [ref. 15]

For 200W/8 Ohms, Vmax load = (2 x 200 x 8)0.5 = 56.6Vpk

For 300W/4 Ohms, Vmax load = (2 x 300 x 4)0.5 = 49.0Vpk

Imax load = (2 x P/Z)0.5

For 200W/8 Ohms, Imax load = (2 x 200/8)0.5 = 7.07A

For 300W/4 Ohms, Imax load = (2 x 300/4)0.5 = 12.25A

2.4.4 Selection of Emitter Resistors

"As a guide you would normally try for about 0.6 Volts across the emitter resistors at Imax load ... These resistors help provide thermal stability of the output stage bias current and, in designs with output transistors in parallel, they help to ensure equal current sharing. The higher the resistance the better the thermal stability and current sharing but the more power they waste. The final value is a compromise." [ref. 16]

This suggests R37~R40 in the 200W version equalling 0.6V/(7.07A x 0.5) = 0.17 Ohms [ref. 17]. For the 300W version 0.6V/(12.25A x 0.5) = 0.098 Ohms. In the design preferred and readily obtainable value 0.1 Ohms is therefore used. The Figure shows views of "suitable", (an IRH resistor at left), and "unsuitable", (no-name, at right), 5W 0.1 Ohm emitter resistors.

2.4.5 Required Supply Voltage



V(B+/-) = Voverhead + Vripple+ Vmax load




(Imax load x R37) + Vce sat(Q12)

Vce sat(Q12) being about 1V at 5A Ic, from Motorola's device data sheet.



(7.07A x 0.5 x 0.1 Ohms) + 1V

again allowing for current sharing between two devices




for the 200W case, and



(12.25A x 0.5 x 0.1 Ohm) + 1V





for the 300W amp.



6,300 x Imax load/C

where where Vripple is Vpk-pk ripple on the power supply, and C is the filter capacitor size in F on each supply rail. For the original design the capacitance on each rail (C1//C3 and C2//C4) is 5,000F.



6,300 x 7.07A/5,000F





for the 200W case, and;



6,300 x 12.25A/5,000F





for the 300W version.

Vmax load



from section 2.4.3





And thus V(B+/-) = 1.4V + 8.9V + 56.6V = 66.9V for the 200W amplifier, and = 1.6V + 15.4V + 49V = 66V for the 300W amplifier.

Now the Vcer of the output transistors limits the allowable DC supplies to marginally more than 140Vdc/2 = +/-70Vdc. Both the 200W case and the 300W case are thus shown to be achievable using +/-70Vdc rails in practice, allowing for power supply voltage sag to some degree at full load.

2.4.6 Transistor Load Lines

"...we need to make an estimate of the maximum phase shift caused by the inductive portion of the speaker load. 450 seems to be the accepted standard..."[ref. 18].

Tables 1 to 3 in Appendix 1 show the results of calculations needed to plot graphically the load lines for the output transistors for the 200W amplifier, with 450, 600 and 900 maximum phase shift [ref. 19] Figure 1 in Appendix 1, shows these results plotted graphically. Tables 4 to 6 in Appendix 1 show the corresponding data for the 300W version and Figure 2 in Appendix 1, shows the 450 results plotted against the SOAR boundary for not only two output transistor pairs, but three pairs as well.

The tables are constructed as follows; ["w" is used instead of the Greek lower-case omega (= 2.pi.f) and "T" is used instead of the Greek lower-case theta, representing (phase) angle in degrees].

The same calculations are now done for the driver transistors [ref. 20]. The load impedance presented to the drivers is Rload x hfe min [ref. 21] hfe min is the minimum gain of the output transistors; a figure of 25 @ 5A for the MJ15003 and MJ15004. Imax for the drivers is Imax load/hfe [ref. 22]. Tables 7 through 12 in Appendix 1, show the data for the 200W and 300W versions respectively, at 450, 600 and 900 maximum phase shift, and Figure 3 in Appendix 1, shows the 450 results plotted graphically.

The calculated data and graph plots are revealing. Figure 1 demonstrates the suitability of the ETI-466 for driving an 8 Ohm load to 200W, making the reasonable allowance of 450 as the maximum phase angle that need be tolerated. Figure 2 shows how inadequate two output pairs are for driving a 4 Ohm load at the same phase angle. At least three, (rather than the specified two), pairs of outputs are needed for 4 Ohm loads.

Appearing just four months after the publication of the ETI-466, the Electronics Australia Playmaster 300 [ref. 23] used as standard three MJ15003/ MJ15004 pairs for the same power output rating. Even then designers J. Clarke and L. Simpson noted that "typical 4 Ohm reactive loads are likely to cause the protection circuitry to operate when driven hard and thus reduce the maximum power delivered. This compromise is absolutely necessary. While a six transistor output stage of this sort may appear very rugged, it cannot be allowed to drive 4 Ohm loads without protection circuitry. The only alternatives are unreliable operation or more transistors in the output stage." [ref. 24].

Figure 3 in Appendix 1 reveals additional and serious inadequacies insofar as the BD139, BD140 drivers are concerned, for both the 200W and 300W cases, at all but the most minimal phase angles. Up-rating of these devices is surely warranted. Figure 4 shows the SOAR curve for the MJE340 driver used in the Playmaster 300 referred to earlier, and it is readily discerned that the MJE340/ MJE350 [ref. 24] pair would make much more suitable drivers. The BD139/ BD140 pin-out is the same as that of the MJE340/MJE350 pair, and the TO-126/SOT-32 and TO-225AA/SOT-32 packages of the two, (respectively), are also essentially the same. The change in drivers from the original design therefore appears eminently suitable.

2.4.7 Output Stage Voltage Gain

Local feedback is applied to the output stage by the networks R35, R29, R30 and R36, R31, R32 giving the output stage a voltage gain of about five [ref. 25].The local voltage gain of the output is demonstrated by applying op-amp gain calculation methods to Q10 and Q12//Q13 [ref. 26] jointly acting as a non-inverting voltage amplifier. (The same applies to Q11 and Q14//Q15). R35, (220 Ohm), is the feedback resistor connected between output and inverting input, and the relationship of this to R29//R30 sets the gain;

Av = (R29//R30 + R35)/(R29//R30)

= (50 + 220)/50 = 5.4 (+14.7dB)

2.4.8 Zobel Network

The ETI-466 includes a shunt Zobel network, C15, 0.1uF, and R47, 4.7 Ohm 1W, for stability into inductive loads, and a series output inductor L1 [ref. 27]. (but with only a high value parallel damping resistor) [ref. 28]. for stability into capacitive loads and long speaker cables. The Zobel network resistor approximates to the expected load and can be a wirewound type without reducing its effectiveness. Use of a 5W resistor rather than the 1W original would be more likely to prevent its burn-up in the event of high frequency instability, and therefore is considered a useful modification to the original design. As stated in the original article, it is important that the capacitor in the Zobel network, C15, is a non-inductive type; "Elna brand 250V or 630V greencap's are suitable or Philips type poly capacitors are suitable All amplifiers should include the output inductor, as load capacitance contributes additional lag phase-shift to the overall negative feedback loop. Note particularly that the Zobel network comes before the output inductor; i.e. between the amplifier and the output inductor, not between the inductor and the load [ref. 29] In practice, a 7uH to 14uH air-cored [ref. 30] inductor is often used in amplifiers generally, with a 10 Ohm damping resistor in parallel to avoid resonance problems between the inductor and load capacitance. A wirewound type is okay but unnecessary, as the power requirement for this resistor is very low. A 1/2W or 1W type is more than sufficient, as the power dissipated even under worst case conditions is of the order of mW.

2.5 Protection


2.5.1 Fuses DC Supplies

Protection for the amplifier, (against shorted output leads), is provided by fuses in the positive and negative supply rails to both amplifiers." These do not protect the output devices but are intended to minimise collateral damage when the output devices have already failed [ref. 31] and so should be of the slow blow type. David Eather [ref. 32] however recommends a quick blow fuse for this position, and suggests Imax load/3.18 as a suitable empirically determined start point for value selection. This would suggest 7.07A/3.18 = 2.22A, say 2.5A for the 200W/8 Ohm case, and 12.25/3.18 = 3.85A, say 4A for the 100W/4 Ohm case. "This size fuse should allow the amplifier to produce a continuous sinewave output and allows a bit of clipping during music. Gross levels of clipping should blow the fuse." The fuses specified in the original design are 5A and although unspecified appear to be quick blow types. Speaker Lines

Speaker fuses are not included in the original design, are not dependable as a means of DC offset protection in the event of amplifier failure, invite wrong value replacement, and should not be considered as a possible useful addition to the ETI-466, (or any other amplifier), design. Mains

The original suggested design for the ETI-466 power supply did not include a mains fuse. Adding one is a sensible modification. The mains fuse used with the amplifier module or modules should be a slow blow type. Its value is calculated by approximately determining the maximum current to be drawn from the mains, and then making suitable allowance for power drawn by other sections of the complete amplifier. Maximum power to be delivered to the amplifier, (in turn to be delivered as power to the speaker or lost as heat), is VB+ x Imax load. VB+ is Vsec of the mains transformer x 20.5. Thus, 240Vac x Ifuse rating = VB+/20.5 x Imax load

Ifuse rating = VB+ x Imax load x 0.71/ 240 ([ref. 32] also)

This suggests a 1.5A~2A fuse for one ETI-466 200W module, 3A~4A for a stereo version, and 2.5A~3A for a 300W version, 5A~7A for stereo.

2.5.2 Addition of Catching Diodes

Reversed biased diodes, e.g. 1N5404's, between each supply rail and the output, limit the voltage appearing at the output due to energy discharge back into the amplifier and the supply caused by inductive load effects, to one diode drop above each rail DC voltage. They are an easy and worthwhile addition to the original circuit design.

2.5.3 Protection Circuit

Q16 and Q17 are protection circuits sensitive to both current through the output transistors and voltage dropped across them. When caused to conduct, Q16 and Q17 shunt the drive current to the Q12, Q13 and Q14, Q15 output pairs. The current limit is set by the condition in which sufficient collector-emitter current flows in Q13, and hence emitter resistor R38, to produce approximately 0.65V drop across R38. This then is sufficient to forward bias protection transistor Q16 and cause it to start shunting drive away from the output transistors. 0.65V across 0.1 Ohm corresponds to 6.5A, and since Q13 is but one of two output transistors and, (assuming ideal equal current sharing between the two), this corresponds to a peak output current of 2 x 6.5A = 13A. 13A correlates well with the design peak current into 4 Ohms of 12.25A, (ref. section 2.4.3 earlier).

The R44, R43 network applies a measure of the voltage across Q13 collector-emitter to the base of the protection transistor Q16 at the same time, as the current sense voltage across R38 is applied. Thus Q16 is caused to operate by a simultaneous combination of output Vce and Ic. In the voltage-sensing network R38 acts in series with R43. However since R38, 0.1 Ohm, is very much less than R43, 39 Ohm, it is safe to ignore it in calculating voltage protection threshold levels. The voltage limit is set then by the condition in which the voltage between the speaker output and one supply rail produces approximately 0.65V at the base of Q16, after voltage division by R44 and R43.

(39 Ohm/5.6kOhm + 39 Ohm) x Vmax = 0.65V

Vmax = 0.65 x (5600 + 39)/39 = 94V

Intermediate values between the solely voltage limit and the solely current limit may now be calculated, such that the two together produce the 0.65V tun on threshold level at the base of protection transistor Q16.

[R43/(R43 + R44)].Vout + R38 x Iout/2 = 0.65V

(39 Ohm/5.6kOhm +39 Ohm)Vout + (0.1 Ohm/2)Iout = 0.65V

Resolving this gives the formula 0.007Vout + 0.05Iout = 0.65V, which may be made to look a little nicer for simplicity by multiplying all terms by 1/0.05 = 20, giving;

0.14Vout + Iout = 13

A set of points to plot graphically may now be created by nominating some values for Iout and plugging them into the formula and calculating what Vout must be in each case. This gives results as seen in the following table;

























These results can now be plotted graphically as shown in Figure 6 in Appendix 1 to see how the locus of the protection circuit relates to the SOAR curve for two MJ15003/ MJ15004's. The protection circuit clearly follows within the SOAR curve boundaries, preventing the output transistors being stressed beyond any of the SOAR limits. Figures 7 and 8 show more clearly the relationship of the protection locus to the V-I operating areas of 8 Ohm and 4 Ohm loading. Figure 7, Appendix 1, both the protection circuit locus and the 450 V-I performance characteristic locus of 200W/8 Ohm operation appear reasonable. It is to be expected that in normal operation the output stage is acceptably rugged, and the protection circuit would rarely if ever be called upon to operate. Figure 8 reveals more significant concern with the 300W/4 Ohm mode of operation. The 450 V-I performance locus falls significantly outside the SOAR boundary for just two MJ15003/ MJ15004 pairs, and the protection circuit trigger locus falls well within the 450 V-I performance curve plotted. This suggests that in 300W/4 Ohm operation, while the output stage will in fact survive loads producing a typical practical degree of V-I phase shift, frequent intervention of the protection circuitry is to be expected.

Frequent operation of the protection circuitry carries its own risks, including severe voltage spikes appearing at the output [ref. 33] (suggesting at least that the addition of catching diodes, ref. section 2.5.2, might be even more relevant), and oscillation [ref. 34] during protection circuit operation.

2.6 Gain


2.6.1 Calculation Method

The voltage gain of the amplifier as a whole is given by applying op-amp gain calculation methods and treating the amplifier as a non-inverting op-amp. Input is applied to the non-inverting terminal, (bases of Q1 and Q6), and feedback is applied to the inverting terminal, (bases of Q2 and Q5). R19 is the feedback resistor connected between output and inverting input, and the relationship of this to R5 sets the gain.

2.6.2 As 200W into 8 Ohms

Here R19 = 10kOhm and in this case;

Av = (R19 + R5)/R5 = (10,000 + 220)/220 = 46.5 (33.3dB)

For an output of 40Vrms, being the voltage required to produce 200W into 8 Ohms, this suggests a required amplifier input voltage for full output of;

Vin (rms)= 40/46.5 = 860mV

2.6.3 As 300W into 4 Ohms

For the 300W version, R19 is changed to 6.8kOhm in the original design, lowering the gain.

Av = (6800 + 220)/220 = 31.9 (30dB)

For an output of 34.6Vrms, being the voltage required to produce 300W into 4 Ohms, this suggests a required amplifier input voltage for full output of;

Vin (rms) = 34.6/31.9 = 1.085V

2.6.4 Other Components

C8 in parallel with R19 is present to help provide overall stability. C7, 100F, in series with R5 sets low frequency roll off. Its impedance increases as frequency decreases, progressively increasing the effective value of R5 in the gain formula shown previously. At DC "R5" therefore is an open circuit, there is 100% negative feedback via R19, and the amplifier therefore has unity gain. The -3dB point occurs when voltage gain has fallen to 1/20.5 = 0.707 times that calculated above for the amplifier passband.

2.7 Power Supply


2.7.1 General Operation

The power supply originally specified uses a full wave rectifier and a centre tap to derive 68Vdc,using a 47-0-47V, 300VA, (Ferguson designated PF4363), transformer. This could be expected to give DC voltages after rectification of slightly less than 47Vac x 20.5 = 66.5Vdc. At 300VA the transformer is inadequate to support the 300W/4 Ohm rating and the DC voltage derived from 47Vac appears only marginally adequate to give 200W/8 Ohm (ref. the calculation in section 2.4.5 previously). Eminently more suitable and readily available is a 50-0-50 Vac toroidal transformer, giving rails of approximately 70Vdc. 300VA is fine for 200W/8 Ohm applications but 500VA or 625VA types are required for 300W/4 Ohm and professional applications. As a rule of thumb, allow one and a half to two times the desired power output from the amplifier or amplifiers connected to the power supply, as a suitable VA rating for the mains transformer. For professional or semi-professional use choose the higher of the two figures. For example, a 300VA transformer would probably be suitable for a single 200W ETI-466 module where only an 8 Ohm load was envisaged. At the other extreme, two modules operating in bridge mode into 8 Ohm to produce 600W would certainly suggest a 1kVA~1.2kVA transformer. Opt for toroidal transformers at every opportunity, as they surpass E-I types in all respects in this application, including availability and cost.

2.7.2 Output Transistor Vcer

A limiting factor for the power supply voltage, and hence the power output into 8 Ohm and 4 Ohm loads, (and therefore about which the rest of the amplifier has been designed), is the Vcer of the MJ15003/MJ15004 pair, of 140V. When biased OFF, each output transistor Q12~Q115 must sustain the difference between the nominally fixed supply rail voltage at its emitter, and full supply voltage from the opposite rail when its opposite output transistor is fully ON. The supply is thus limited to about 140V(Vcer)/2 = 70Vdc which in turn sets nominal power output capabilities of 200W into 8 Ohms and about 300W into 4 Ohms.

2.7.3 Filter Capacitors

"The filter capacitors should be a minimum of 100F~200F per Watt of output power for a Class B amplifier with a full wave rectifier" [ref. 35] This suggests 20,000F~40,000F per rail for the 200W version, and 30,000F~60,000F per rail for the 300W version. This is considerably more than the two 2,500F, (i.e. 5,000F), per rail provided in the original design, and an area of the amplifier suitable for modification accordingly. While home hi-fi probably only warrants the lower of the two totals suggested, the reverse is certainly true for professional or semi-professional applications. Serious increase in value of the filter capacitors in this amplifier is certainly warranted.

Increasing filter capacitance substantially means higher peak forward currents in the rectifier diodes. In practice, replacing each of the four on-board 2,500F filter capacitors with 4,700F, 6,800F or even 10,000F PC mount electrolytics has not resulted in failure of 1N5404 rectifiers. Additional 1N5404's might be added in parallel with the existing four, (perhaps on the underside of the PCB for ease of mounting), to improve reliability however. Improving the power supply still further would mean constructing a discrete off-board power supply using a much heavier duty bridge rectifier and more filter capacitance still. I have not attempted this, nor found it to be warranted, but you might.

3. Variations


3.1 Bridge Mode Operation

3.1.1 General

ETI-466 modules do not appear particularly suitable for bridge mode operation and I have not attempted this. (It's referred to as amplifier "strapping" in the US in some literature, rather than bridging). As originally designed, reliable operation into 4 Ohm loads appears doubtful, with frequent operation of the safe area limiter protection circuit. Modification by the addition of a third output pair and improved drivers would be necessary before considering bridge mode operation with an 8 Ohm standard load impedance. In bridge mode each amplifier effectively sees only half the load. Hence 8 Ohm is the minimum sensible load, meaning that each module "sees" 4 Ohms. Use of a 4 Ohm load in bridge mode means that each module would "see" 2 Ohms, which is courting failure absolutely, as each amplifier module is certainly not designed for anything like the output current necessary for such a load. As originally designed the safe area limiter protection circuit would be unlikely to allow 2 Ohms operation in any event. Further, it should also be clearly remembered that in bridge mode neither speaker output terminal is at ground potential; shorting either speaker connection to chassis or ground will produce a short circuit across the output of one module and associated failure.

3.1.2 Output Power

Output power for two amplifiers in bridge mode into 8 Ohms is theoretically four times the output power of one module into 8 Ohms. This assumes the power supply has sufficient VA rating for its output voltage to be unaffected by the substantial additional load, and no additional losses due to the significantly higher currents in each output stage. This does not occur in practice. Realistically, the paired modules should produce into 8 Ohms twice what each will deliver into 4 Ohms, which makes simple sense as each module effectively "sees" half the load when connected in bridge mode. Modified, (up-rated), ETI-466's therefore might be expected to deliver 600W into 8 Ohms, depending on the power supply which needs to be capable of 900VA~1200VA or more.

3.1.3 Drive Methods From the Output of One Module

I'm not aware of proven practical driving of one module from the output of the other, as can be reliably done in the case of the ETI-480, for example. While similar inter-connection might well be possible, the relative ease of constructing a suitable inverter to provide drive for the second module makes experiments in this area unwarranted. From a Separate Inverter Circuit

Various amplifier bridge operation inverter/driver adapter circuits have been published [ref. 36] These provide a unity gain inverting amplifier stage that has its input the same signal as drives the first amplifier module, and its output driving the second power amplifier module.

3.2 Up-rating


3.2.1 General

I have only limited experience with up-rating the ETI-466. Others have indeed constructed higher power versions. Five MJ15024/MJ15025 pairs, twin 500VA or 650VA toroidal power supply transformers at 70-0-70Vac producing 100Vdc are reported to produce 400W into 8 Ohms, 650W into 4 Ohms. Four MJ15024/MJ15025 pairs, and at least a 650VA toroidal power supply transformer at 55-0-55Vac producing 75Vdc,are reported to produce 260W into 8 Ohms, 420W into 4 Ohms. As an example, Figure E shows a single ETI-466 board, up-rated by adding two additional pairs of output transistors. A protection and switch-on delay circuit has also been added and can be seen at right.

Fig. E
Fig. D
Click here for larger image.

3.2.2 Component Considerations Output Transistors

Motorola MJ15024 and MJ15025's [ref. 36A] make suitable outputs when higher supply voltages are being considered, to increase output power. Vceo is 250V, Ic = 16A and Pdiss = 250W. Note that although Vceo is much higher, (suggesting supply rails to 250/2 = 125V), Ic is lower and Pdiss is the same as the MJ15003/MJ15004 pair originally specified. Consequently a significant increase in the number of devices that must be paralleled together occurs. Drivers

Motorola MJE340/MJE350 pairs, (Vce = 300V, Ic = 0.5A, Pdiss = 20W, Hfe = 30~240), should make much more suitable drivers than the originally specified BD139/BD140* pair, (Vce = 80V, Ic = 1A~1.5A, Pdiss = 8W, Hfe = 40~250), in the first place, (ref. Figure 4 in Appendix 1), before up-rating is considered. Adding a second pair of drivers in parallel with the first is theoretically possible (using say 22 Ohm resistors in the emitter leg of each device to promote good current sharing), but difficult to achieve in practice.

*[For BD139/BD140 component data see [ref. 20] links to Fig. A, Fig. B and Fig. C].

A comparison of various alternate drivers is shown in Figure 5 in Appendix 1. The MJE15030/ MJ15031 [ref. 37] complimentary pair, while appearing to be much more rugged from initial consideration of key specifications only, (Vceo = 150V, Ic = 8A, Pdiss = 36W, Hfe = 40min), in fact demonstrates little real improvement over the MJE340/MJE350 pair. The MJE5730, (also MJE5731 and MJE5731A)/ TIP47~50 [ref. 38] pair however is clearly significantly better and appear a suitable choice for up-rating the ETI-466.

4. Notes


(1) In ETI December 1981 p.69, a letter from A. Stewart, Gumsdale, Queensland, identified a long serious history of difficulty with an ETI-466 suffering from high frequency instability and detonating the output stage repeatedly. Roger Harrison, E.T.I.'s then editor, responded with various points, also noting publication of information responding to similar troubles with the ETI-477 MOSFET amplifier, in ETI August 1981, p.11.

Suitable and unsuitable emitter resistor types.
  • Output transistor emitter resistors must, as always for any amplifier, be low inductance types; "Noble" and "IRH" brand 5W wirewound resistors work reliably, but some "no-name" brands do not. Typical examples of the former and later are shown at left.


(2) "Exact DC balance of the input differential pair is essential in a power amplifier. It still seems almost unknown that minor deviations from equal Ic in the pair seriously upset the second-harmonic cancellation....." Douglas Self, Audio Amplifier Design Handbook, Newnes, 1996, p.65.

Doug Self's "Audio Power Amplifier Design Handbook" is now available in 2nd Edition.


(3) Douglas Self, Audio Power Amplifier Design Handbook, pp.67~68.


(4) Gdiff = Rc/2(re + RE) p.99, "The Art of Electronics", 2nd. Ed., Cambridge University Press, 1989, P. Horowitz and W. Hill.


(5) For Q7 = BD140, Q9 = BD139 min. gain = 25, referring to manufacturer's device data. Ic(Q7,9) = 7mA as given in the "How it Works" section of the original article, and RE = R21,24 = 390 Ohms.


(6) Vbe matching of the input pair is not warranted, as Vbe mismatch is only minimally responsible for output stage DC offset. At most it gives rise to about 5mV of output DC offset (Douglas Self, Audio Power Amplifier Design Handbook, p.74). For the input pair it is gain that is important for matching. The reverse is the case for matching output transistors for good output stage current sharing.


(7) Audio Power Amplifier Design - 4, P.J. Baxandall, Wireless World, July 1978, p.76. Reprints of some Wireless World articles are indeed available, but not this far back, as far as I know.


(8) ibid. p.77.


(9) The Art of Linear Electronics, John Lindsley Hood, Butterworth Heinemann, 1993, p.90.


(10) Douglas Self, Audio Power Amplifier Design Handbook, p.60.


(11) Power dissipation in ZD1 and ZD3 is 20mA x 5.1V = 100mW, and 300mW devices are specified for these positions. R12 shunts some of the 20mA Zener chain current away from ZD2; 62V/22kOhm = 3mA, leaving 20 - 3 = 17mA as the standing current through ZD2. This gives power dissipation in ZD2 of 62V x 17mA = 1.05W. A 5W device is specified, and this runs hot in practice, and needs to be mounted well clear of the PCB. Much of the heat dissipation of this Zener is through the component legs to the surrounding air, and this needs to be given as much opportunity to occur as possible.


(12) Critical however is the tracking of this thermal feedback path. Inevitably time delay occurs between a change in audio drive signal such as to cause a change in temperature in the output transistors and hence a change in their bias point, and thermal sensing by Q8 with consequent countering of the quiescent bias change.


(12A) See MJ15003 and MJ15004 data.


(13) There seems to be only one popular configuration [of the CFP], though versions with gain are possible" Douglas Self, Audio Amplifier Design Handbook, Newnes, 1996, p.99. Ref. also p.121;"The CFP topology has good quiescent stability and low LSN [Large Signal Non-linearity]" . . . its worst drawback is that reverse-biasing the output bases for fast switch-off is impossible without additional HT rails". Prof. W. L. Leach in "Build a Double Barrelled Amplifier", Audio, April 1980, p.40, provides an opposite view; "Because this [CFP] connection forms a negative feedback path from the collectors of the output transistors back into the emitters of the driver transistors, a large reduction of static distortions in the output stage can be realised. It is felt, however, that the [emitter follower output topology] results in a more stable and better sounding amplifier. This is because the output transistors in the [CFP] are operated in their slowest configuration. The driver transistors are forced to supply a higher and higher share of the load current as the frequency is increased, which in turn causes the high-frequency output impedance of the amplifier to increase, resulting in a reduced high-frequency damping factor."


(14) The design procedure that follows is drawn directly from an exceptionally interesting, practical and readable article by David Eather, "A Practical Approach to Amplifier Output Design", Silicon Chip, February 1991, pp.14~18 and April 1991, pp.64~67.


(15) Prms = Vrms2/Rload, and where Vrms = Vpk/20.5, then by inserting this and rearranging, Vpk = (2 x Prms x Rload)0.5.


(16) David Eather, "A Practical Approach to Amplifier Output Design", Silicon Chip, February 1991, pp.14~18 and April 1991, pp.64~67.

Silicon Chip - premier (only?) Australian hobbyist electronics magazine!


(17) Two output devices share the current, so the calculation assumes equal current sharing and half each.


(18) David Eather, "A Practical Approach to Amplifier Output Design", Silicon Chip, February 1991, p.15. However designing for 900 is fundamental to Douglas Self; reactive load impedances can "...double the Vce seen by the output devices. It is therefore necessary to select a device that can withstand at least twice the sum of the HT rail voltages, and allow for a further safety margin on top of this." Audio Power Amplifier Design Handbook, Newnes, 1996, p.275. In practice this limits power output quite severely. The message is clear however; protection circuitry is mandatory if the full Vcer capability of the MJ15003/ MJ15004 pair is to be utilised, or output transistor series connection topology needs to be implemented to avoid secondary breakdown limitations.


(19) 450, 600 and 900 phase angles with respect to load impedances of 8 Ohms and 4 Ohms correspond to complex impedances as follows;


Phase Angle

Complex Impedance

8 Ohms


5.65 + j5.65 Ohms

8 Ohms


4 + j6.9 Ohms

8 Ohms


0 + j8 Ohms

4 Ohms


2.8 + j2.8 Ohms

4 Ohms


2 + j3.5 Ohms

4 Ohms


0 + j4 Ohms


(20) Device data for the BD139 and BD140 driver transistors.

Fig. A
Fig. A
Click here for larger image.

Fig. B
Fig. B
Click here for larger image.

Fig. C
Fig. C
Click here for larger image.


(21) 8 Ohms x 25 = 200 Ohms, 4 Ohms x 25 = 100 Ohms.


(22) 7.07A/25 = 0.28A, 12.25A/25 = 0.49A.


(23) Electronics Australia May 1980, pp.39~41, June 1980, pp.54~61, July 1980, pp.52~57. Note also corrections in Electronics Australia July 1980, p.100, August 1980, p.141, and derivative article "High Power Amplifiers", October 1979.


(24) Electronics Australia, June 1980, p.56.


ON Semiconductor markets Motorola devices, making available a huge range of dtasheets on-line.

(24A) See On Semi for Motorola MJE340 and MJE350 device data.


(25) "The almost universal choice in semiconductor power amplifiers is for a unity-gain output stage, and specifically a voltage follower. Output stages with gain are not unknown [example cited] but they have significantly failed to win popularity. Most people feel that controlling distortion while handling large currents is quite hard enough without trying to generate gain at the same time." Douglas Self, Audio Power Amplifier Design Handbook, Newnes, 1996, p.91.


(26) Both Q10 and Q12//Q13, (in this example), are inverting, so when considered as a pair the amplifier is non-inverting. Hence the base of Q10 is the non-inverting terminal, (of the pair jointly considered as an op-amp), and the emitter of Q10 is the inverting terminal.


(27) A value for L1 isn't exactly specified in the original article; "The inductor L1 is made by winding a layer of 26 SWG enameled wire, (or the nearest equivalent gauge), along the body of a 1W resistor."


(28) "...The value of the resistor may be anything over 100 Ohms>."


(29) Strangely, amplifier designs are regularly seen, certainly from the same era as the ETI-466, either with this order reversed or an additional RC network placed after the output inductor. Literature explaining or supporting this topology is notably absent.


(30) The inductor is preferably air cored to avoid possible distortion due to saturation of a ferrite core. This however means much greater power loss due to the coil resistance or a physically very much larger component or both. "The resistance of an air-cored 7uH coil made from 20 turns of 1.5mm diameter wire, (this is quite a substantial component 3cm in diameter and 6cm long), is enough to cause a measurable power loss into a 4 Ohm load..." Douglas Self, Audio Power Amplifier Design Handbook, Newnes, 1996, p.159.


(31) Ibid., p.276.


(32) David Eather, "A Practical Approach to Amplifier Output Design", Silicon Chip, April 1991, p.67.


(33) T. Holman, "New Factors in Power Amplifier Design", JAES, Vol. 29, No. 7/8, 1981 July/August, pp.517~522. See especially Fig.'s 6~8.


(34) Douglas Self, Audio Power Amplifier Design Handbook, p.281.


(35) David Eather, "A Practical Approach to Amplifier Output Design", Silicon Chip, February 1991, p.15.


(36) As examples;


(36A) See On Semi for MJE15030, MJE15031, MJE5730, MJE5731 and TIP47 data.


(37) Ibid.


(38) Ibid..


5. Amplifier Set-Up Procedures


The following set-up procedure for newly constructed boards is closely derived from original text, ref. p.48, E.T.I. Feb. 1980.

  1. Remove the two fuses. In this condition, power remains supplied to the front end, but not to the drivers or output stage.

  2. Solder a small link across C11. This shunts collector-emitter of the Vbe multiplier, Q8, and shorts the bases of all the driver transistors together. The output stage devices are thuis all help hard "off".

  3. Solder a wire between this link and the output pad. All DC and AC drive and forward bias to the output stage is now shorted out.

  4. With no load connected and no input signal, switch the power on.

  5. Check the supply rail voltages. These should be about 68Vdc~72Vdc each, (plus and minus).

  6. Check the voltage on the cathode of ZD1, (which should be about +37V) and that on the anode of ZD3, (about -37V), with respect to 0V. 37V is about half the 72V~74V sum of the 5.1V + 62V + 5.1V Zener chain, ZD1, ZD2 and ZD3, setting up the constant current sources for the input stage.

  7. If these two voltages differ with respect to each other by a Volt or so, check other voltages around the input stage to determine the reason.

  8. Check the DC voltage on the output, (with respect to 0V). It should be within 20mV of zero.

  9. Inject a sinewave signal into the input at a level of about 20mVrms. Don't use a higher input level. The output should be 1Vrms. The amplifier voltage gain is approximately 1V/20mV = 50 = 10kOhms[R19]/220 Ohms[R5].

  10. Switch off the main power and allow the filter capacitors to discharge. Remove the input signal.

  11. Solder a 10 Ohm 1/2W resistor across each fuse holder. Rotate the trim-pot RV1 such that it is set at maximum resistance. This turns the Vbe multiplier Q8 on hard to bias the output stage hard "off" initially. Remove the short across C11 and the link from there to the output pad. This reconnects the driver and output stage.

  12. Switch on. If the 10 Ohm resistors immediately vaporise you either have a short or some fault in the output stage. find it, and fix it!

  13. If all is well check the DC output voltage. It should be near zero.

  14. Measure the voltage drop across one of the 10 Ohm resistors placed across the fuse-holders and adjust RV1 to give a reading of 1.0V. This gives a quiescent current in the output stage of 1V/10 Ohm = 100mA, being 50mA through each of the two output transistors.

  15. Switch off, allow the filter capacitors to discharge and remove the two 10 Ohm resistors. Replace the fuses.

  16. Apply normal power, speaker load and input signal and listen to the output for abnormalities.