by Brian G. Wachner [BGW Systems.]
High power, solid-state audio amplifiers are not new to audiophiles. Shaker table amplifiers, which are used in our aerospace programs to test equipment for mechanical stress failure, have been capable of delivering more than ten kilowatts at low audio frequencies for more than a decade. The 1960's saw a dramatic transition from tube equipment to solid-state electronic equipment. But because early power transistors had limited voltage ratings, a step-up out-put transformer was necessary to produce voltage swings large enough to drive speakers (see fig. 1). Even now we still find manufacturers of public address systems using step-up output transformers to provide a 70-volt line.
Fig. 1--Transformer coupled output stage.
As the technology in the semiconductor industry improved, the break-down voltage ratings of power transistors also improved. Capacitor-coupled designs (see fig. 2) came into vogue. Soon afterwards, a few brave audio amplifier manufacturers entered the market with direct-coupled output stages instead of the traditional indirect-coupled output stage.
As a result of the experiences of these early audio amplifier manufacturers, substantially all high performance audio amplifiers now use direct-coupled output stages. But the evolution of the dual supply, direct-coupled output stage has not been trouble-free.
Fig. 2--Capacitor coupled output stage.
The utilization of a direct-coupled output stage requires that the speaker system be connected in series with the output stage transistors and the power supply. If a transistor should fail, then excessive current would flow directly through the speaker system, producing extensive speaker damage. With transformer and capacitor-coupled output stage designs, the power supply voltage simply cannot appear at the output terminals of the amplifier, so no speaker damage can ever result from transistor failure.
Direct-Coupled Output Stage Design
Fig. 3--Direct coupled output stage using split supplies.
At the outset it is important to understand just what happens in a typical direct-coupled output stage when a power transistor develops a short from its collector to its emitter. The diagram (fig. 3) is a simplified direct-coupled output stage and shows the speaker system in direct series with the output transistors and power supply. There are no transformers or capacitors to serve as a "buffer" to prevent excessive current from flowing through the speaker. A collector-to-emitter short of one or both output transistors places the speaker system directly in series with the power supply. This, of course, means "instant death" for prized woofers and tweeters unless excessive power supply current can somehow be prevented from reaching the speaker system before its point of destruction.
Those manufacturers who use fuses generally place the fuse either in the power supply leads to the output transistors or in series with the speaker (fig. 4). The obvious problem with this approach is that a fuse must be large enough to carry peak amplifier output current, but small enough to melt if an output transistor should fail, i.e., develop a collector-to-emitter short. These contradictory requirements are rather unfortunate and mean that a fuse must have a relatively long melting time. If the fuse were called into play, by the time it melted the speaker system may have been destroyed.
Fig. 4--Fused output stage.
Those manufacturers who use relays to protect the speaker system generally place the relay contacts in series with the speaker (fig. 5). When an output transistor fails, the relay opens and removes the speaker system from its potentially destructive current source. But a relay is a mechanical rather than an electronic protective device, and it is extremely slow to react so the speaker system is more likely to be destroyed before the relay has had an opportunity to open.
What is ultimately required to save the speaker system from output transistor failure is a fast reacting electronic circuit capable of diverting excessive current flow in the failure mode. Electronic computers requiring regulated low voltage direct current power use an electronic "crow bar" to prevent similar potentially destructive current surges from reaching expensive integrated circuits. This "crow bar" is a silicon controlled rectifier (SCR) which operates by actually shorting the power supply to prevent excessive current from damaging delicate integrated circuits. The same SCR crow bar circuitry is used in all BGW amplifiers. (fig. 6). No fuses are necessary. There is no chance for human error and no having to run out to the local hi-fi store if a fuse should accidentally blow.
Output Stage Design Considerations
There are several constraints that dictate the circuit parameters which output transistors must be capable of meeting.
They are (1) breakdown voltage specification, (2) current handling capability, (3) gain bandwidth product, and (4) safe operating area. The mechanical and thermal characteristics of output transistors and their thermal resistance also are important factors.
As an example of output stage design, consider the parameters for an amplifier that is capable of delivering 200 watts into an 8 ohm load, and 250 watts into a 4 ohm load.
The equations that relate power output to voltage, current, and load resistance are shown below. Solving these equations for the above example output stage, the following table can be constructed:
By inspection, we see that the peak voltage occurs for the 8 ohm load but that the peak current occurs for the 4 ohm load. Our power supply must be capable of handling the peak current and peak voltage requirements required by the various output stage load resistances, which in this example are two.
Although our peak voltage requirement is technically only 56.5 volts (and occurs for the 8 ohm load), power supply regulation and the desire for a few volts "headroom" suggest that the power supply voltage requirement should be somewhat greater, say about 70 volts. The output transistor breakdown voltage is, at a minimum, twice the power supply voltage. Hence, we require output transistors with breakdown voltages of at least 140 volts. Ohm's Law tells us that the output transistors must also be capable of handling peak currents of at least 11.1 amps, (the peak current for the 4 ohm load). Thus, our output transistors must have a breakdown voltage of at least 140 volts and be capable of handling peak currents of at least 11.1 amps.
Fig. 5--Relay protected output stage.
Today there are four basic types of power transistors: single diffused, triple diffused, epi base, and multiple emitter site double diffused. A comparison of the specifications of these various types in tabular form would look like the information presented in the following table:
These devices are all NPN types, since the availability of high voltage, high power PNP types is limited to epi base' transistors. Practically all large amplifiers use NPN output devices, so we will confine ourselves to quasi-complementary designs. The devices which have the largest safe operating area are the single diffused types. These are the most rugged, but also the slowest. The 2N6259 clearly can handle over three times as much current at 75 volts as the triple diffused DTS410.
Fig. 6--SCR crowbar protected output stage.
Now let's finish our analysis of the output stage and determine the safe operating points which the transistors must be capable of handling. We have made the following assumptions: The power supplies are fixed at plus and minus 70 volts, the maximum power requirement is 250 watts into 4 ohms and 200 watts into 8 ohms. The problem is to determine how many of each type of device in parallel are necessary to produce a unit that will not exceed the manufacturer's safe operating area at a maximum operating case temperature of 70 degrees centigrade, since hotter temperatures would prove hazardous to the transistor and cause burns if accidentally touched.
When the output of the amplifier is crossing through zero volts and we are driving a resistive load, we have 70 volts across the output transistors and no appreciable collector current. But when the output of the amplifier is at 56.5 volts (peak voltage for the 8 ohm load), we will have approximately 13.5 volts (70.0 56.5 volts) across the power transistors. The current will then be, according to Ohm's Law, 7.1 amps. (which is equal to the peak current for the 8 ohm load). Now we must examine the voltage and current conditions impressed on the output transistors for each point in the cycle of a sine wave. Using 5 volts increments, we would generate a table that looks like the following:
We can determine the minimum number of devices of each type that must be paralleled to handle the power requirements of our output stage design. For the 2N6259, by inspection we see that at 25 volts output, the current into 4 ohms is 6.3 amps. (line 6 of the table). The safe operating area capability is only 4.1 amps., so 2 devices must be placed in parallel for safe operation. Checking each line of the table to make sure that we are within the manufacturer's safe operating area requirements using two transistors in parallel, we find that we are. Continuing this exercise, we will obtain the following results:
Device type | Number Required
2N6259 2
RCA 410 3
DTS 410 4
2N5634 3
The results are striking and point out that the penalty we pay for extra bandwidth is quite extreme. If we were to use triple diffused devices such as the DTS410, a very popular transistor, we would require twice as many transistors as compared to the single diffused example. Using these fast devices, we are able to produce full power at higher frequencies (above 20 kHz), but this capability is not required for audio use. Conservative design practice does not allow any possibility for output devices to be overstressed, regardless of how long the overstress lasts.
Fig. 7--Photomicrograph of 2N6259 chip.
Fig. 8--Showing inside view of a RCA 410.
At BGW Systems, two types of devices are used. The BGW professional line uses large single diffused devices such as the 2N6259 for maximum reliability and controlled bandwidth. Professional experience has shown that it is sufficient to have half as much power available for the mid and high-range transducers as for the low frequency transducers. Our hi-fi amplifiers use the double diffused multiple emitter site devices such as the RCA410. These amplifiers will produce full rated power at 20 kHz. This type of transistor has all the advantages of the triple diffused part, but also has a significantly larger safe operating area.
The size of the transistor die, the actual semiconductor chip inside the familiar TO-3 transistor case, varies with the type of process used to diffuse the device. For example, the 2N6259 chip measures approximately 0.250 x 0.250 inches.
This is an extremely large geometry device, and a photomicrograph of the dies is shown below (fig. 7). The double diffused and triple diffused devices are much smaller. The RCA410 is approximately 0.135 x 0.135 inches, or about 30 percent as large as the 2N6259 device. This large size difference also accounts for the much larger power handling ability between the two devices. The 2N6259 can dissipate 250 watts as opposed to 125 watts for the RCA410. The RCA410 chip is soldered to a copper slug which is placed on the steel header. A compression clip with little protrusions corresponding to each emitter is then placed on top of the die. Spring tension clips make contact to the emitter and base contacts, and are soldered in place. A photomicrograph of two assembled RCA410's and a 410 die and clip are shown below (fig. 9).
Fig. 9--Heat sink with large, radiating area.
The material chosen for the transistor package or case has been shown to affect the life of the transistor. It has recently been demonstrated that steel packages will outlast aluminum by many years under repeated thermal cycling. In these tests the transistors are thermally cycled from 40 degrees centigrade to 130 degrees centigrade with 16 watts of dissipation until failure occurs. The results indicate that the aluminum package will fail after less than 5,000 cycles, while the steel package typically is good for more than 100,000 cycles.
Fig. 10--Forced air sink used with high power amplifiers.
TABLE I
TABLE II
Heat Dissipation
A major problem that confronts all manufacturers of solid-state audio amplifier equipment is how to dissipate the heat produced by the output transistors. The term used to measure the ability of a radiator to dissipate heat is called "thermal resistance." There are three components of thermal resistance in any output state: There is the resistance from the junction to the transistor case, the resistance from the case to its heat sink, and finally, the thermal resistance from the heat sink to the air. Thermal resistance numbers are expressed in terms of degrees centigrade per watt. Calculating the temperature at any transistor junction only requires that we add the three thermal resistance coefficients of the three components mentioned above, and then multiply by the power dissipated (expressed in watts).
Of the three components of thermal resistance mentioned, it is only the thermal resistance from the heat sink to the air (or ambient) that can most easily be lowered. An amplifier capable of delivering several hundred watts requires considerable heat sinking for safe, conservative operation. Each of two heat sinks used in the BGW Model 500 has over 560 square inches of heat radiating area (Fig. 9).
In industrial amplifier design, forced air cooling is more practicable. A forced air quadrant heat sink (Fig. 10) is used in the BGW Model 4X250 which delivers 1,000 watts of output power.
1. "Second Breakdown and Safe-Area Ratings of Power Transistors," by C. R. Turner; Reprinted from EEE Magazine, July 1967, Volume 15, Number 7.
2. "Thermal-Cycling Rating System for Silicon Power Transistors by W. D. Williams; RCA App. Note AN4783.
3. "Evaluation of Hermeticity of Aluminum TO-3 Packages under Thermal Cycling Conditions," by D. Baugher, RCA Reliability Report St-6071.
4. "High-Speed, High Voltage, High Current, Power Transistors," RCA Tech. series, PM-81.
(Audio magazine, Feb. 1973)
Also see:
A 700 Watt Amplifier design (by Bob Carver) (Feb. 1972)
Buying Watts and Other Things (Feb. 1973)
Transient IM Distortion in Power Amplifiers (Feb. 1975)
FTC Power Ratings: An Optimistic View (Feb. 1975)
= = = =