by James Bongiorno [Director of Engineering, Electronics Division, S.A.E., Inc.]
IT HAS BEEN several years since I have written for this journal, and at that time I was working on an amplifier of very radical design both electronically and mechanically. Unfortunately, at that time the transistors involved were exasperatingly expensive and it would have been impossible to market an amplifier based on that design at anywhere near a reasonable cost. This situation has changed, of course, and today there are quite a few high-powered amplifiers on the market. It is my feeling that these amps are definitely needed to drive today's low efficiency loudspeaker systems.
Again, unlike the past, designers seem to be a little reluctant to write about their designs or philosophies, and as a result, only a few articles have been published concerning high power amplifier design. I also believe that this reluctance is due in part to reliability problems encountered in the common approach to these designs. There are presently on the market six high power amplifiers from major firms with at least 200 watts per channel, and no less than four of them are using this common type of circuit design. There is one other manufacturer who is using another similar design approach, but with different transistors of an older vintage.
Most high power amplifiers today use what is known as the high voltage triple-diffused output transistor in the output stages with supply voltages reaching ± 100 volts. This approach is, of course, a far cry from the days when we were nervous about using even ± 35 volts, but it is necessary to use the ± 100 volts to achieve the needed power levels. Most engineers seem to feel that the triple-diffused power device is the answer for these designs. I feel this is unfortunate, as they are not really the best choice for output devices.
There are basically three different types of power transistors available: single-diffused homo-taxial, double-diffused epitaxial, and triple-diffused. The single-diffused devices are considered to be very rugged. This is true; however, their extremely slow speed, combined with excessive leakage versus temperature, makes them a relatively poor choice for truly high quality designs, even though one manufacturer is touting them.
Even though triple-diffused devices are nice to work with, thanks to their high voltage capabilities, they have several problems associated with their use. They have several volts saturation loss even at reasonable currents, extremely poor beta holdup at higher currents, and virtually no safe area capability beyond 70 volts. Most difficult of all, they are not available in PNP configuration which means that these devices must be used in a quasi-complementary design. This is unfortunate as the classic cross-over notch cannot be eliminated with this type of output stage design, and one engineer has openly stated in a recent article that this amp does indeed have this problem. Another difficult problem associated with quasi-complementary output stages is their high frequency instability, which is more commonly described as common mode conduction or latch-up in the output stage itself. This problem is due to phase shifts within the output stage itself and is almost impossible to cure.
Full complementary stages do not have this type of problem when properly designed, but they can only be made using double-diffused epi-base transistors. These complementary devices are also available in power Darlingtons, which we use exclusively at S.A.E. These epi-base transistors do not suffer from the problems associated with either single- or triple-diffused devices, as they have excellent beta hold-up out to several tens of amperes, virtually no saturation losses even at 20 amperes of collector current, and excellent safe-area properties. But, best of all, they don't leak at elevated temperatures. When they are used in an emitter-follower configuration, they have ten times more gain-bandwidth product than either of the other types, which must be used in the quasi-complementary format. As the result of lengthy development, these epi-base devices are now available in much higher breakdown voltages than were previously available. This automatically means more ruggedness in the safe-operating area, which has been extended further into the high voltage region. At S.A.E., we use these devices only in series connection in our output stages, rather than in the more commonly used parallel connection, and we have found that it is almost impossible to initiate secondary breakdown, which is the most common cause of all amplifier breakdowns.
Performance
Fig. 1--Safe operating curves at d.c. for triple diffused vs. epi-base;
TC is 25° C.
Fig. 2--Composite soar curves for three devices of each type; TC is 25°
C.
Fig. 3--Load line capabilities of three type 411's in parallel and three
type 2N5630's in series for both resistive and reactive loads.
Fig. 4--Schematic of Mk IIIC and CM power amplifiers.
Figure 1 illustrates the safe-area capability of one epi-base device versus one triple-diffused type 411. Figure 2 shows the difference between three epi-base devices in the series configuration versus three type 411 devices in the parallel configuration. Since the saturation losses for the triple-diffused devices are enormous, they can only be safely used in the shaded area and are practically useless outside these limits. Figure 3 shows the load line capabilities of the two different configurations for both resistive and reactive loads. As can be seen, the epi-based device has considerably better load line capability over the triple diffused. In use, this means that amps incorporating these devices, as all new S.A.E. amps do, are capable of driving any load angle at full rated power from 0 degrees to 90 degrees. Further, we have found in testing that they will deliver a full-power 20 kHz square wave into a 1µF capacitor with no instabilities with less than 0.5 per cent distortion.
Further circuit refinements can be seen in Fig. 4. As shown, all of the circuitry is fully complementary push-pull from input to output using full complementary differential inputs and full complementary series-connected outputs using epibase Darlington output devices. With this particular approach, there are no bias shifts when the amplifier is driven by transient low frequency information, which results in undesirable low frequency modulation noise with single differential-input circuitry. For this latter, it is sort of like a person with one wooden leg or a sprained ankle who favors one side.
Another innovative refinement to the circuitry is the integrated circuit bias regulator which does away with the possibility of runaway while at the same time maintaining accurate tracking of the output stage quiescent current. This means that there is no shifting of the operating points after prolonged operation, which with most designs causes overcompensation, resulting in classic cross-over notch increase.
This, of course, results in long-term listening fatigue. This characteristic can be verified by examining the IM curve, which is flat all the way down to the milliwatt region, and, indeed, it should be if there are no crossover products. THD and IM curves are shown, and for all practical purposes these are the residuals of the test equipment.
Fig. 5-THD at 8 ohms, full power 40 V rms.
Fig. 6-IM distortion at 8 ohms, full power 40 V rms.
Fig. 7-Damping at 8 V rms.
Fig. 8-Crosstalk at 8 ohms, full power 40 V rms.
Fig. 9-Square wave response at A, 20 Hz; B, 2 kHz, and C, 20 kHz. All
at 8 ohm 200 watt output.
Other parameters attained with this design include a damping factor of 250, which we feel is far beyond the minimum necessary, and a power bandwidth to over 150 kHz. We have found the design to be quite stable, and we have driven it into hard overload with frequencies up to 20 kHz and loads from 4 ohms up without oscillation, blown fuses or other sorts of misbehavior. We think the square wave photos speak for themselves.
Construction
Fig. 10--Showing method of heat sinking.
We call our construction method "Unisink," and believe that it is as different as the circuitry, as the entire amplifierless the power supply--is made in a single unit, including the entire heat sink, output stage, and the low level stages.
There are no wires, except for the four power supply wires, as everything is mounted on two p.c. boards, each of which can be popped out in less than a minute. The main virtue of this is, of course, repeatability, but we have found the design to be extremely reliable, with less than 0.8 per cent total failure rate among consumers.
The heat radiator works on the principle of heat expansion, rather than the more common convection principle. This is quite effective and allows us to set the thermal cutout at 70° C., rather than the more usual 85 or 100° C. Cooler operation, of course, means longer life.
But what does it sound like? Well, during the initial listening period, we couldn't believe our ears, but since we're prejudiced, we suggest that you don't believe us either. Instead, we suggest you go listen for yourself.
( Audio magazine, Feb. 1974)
Also see:
Thermal Design of A High-Power Amplifier (Feb. 1974)
Transient IM Distortion in Power Amplifiers (Feb. 1975)
FTC Power Ratings: An Optimistic View (Feb. 1975)
"I See What You Mean!" How the Westrex 45/45 System Was Adopted by the Record Makers (Mar. 1975)
= = = =