Ever since the author first time designed and built his first class A amplifier more than 15 years ago, his amplifier has always been a class A design. This amplifier is not an exception. It is the successor of a non-feedback design based on a folded cascode differential gain stage followed by a buffered compound emitter follower. The reason why this amplifier has not been published, was the problem with the stability of the compound stage.
So here is a brand new amplifier, the tested version is a 2x25 W RMS dual mono amplifier, working in class A, of course, with distinguished data.
Power output of 25 W does not sound like a lot, but as many owners of class A amplifiers know; a class A amplifier sounds like a much larger amplifier, when compared to class B (and A/B) amplifiers. This is due to among other factors the large (oversized) power supply, which make the amplifier capable of delivering power to the most peculiar loudspeaker loads. And the amplifier draws over 100 watts from the mains plug…
If the loudspeaker demands more than 25 W (some inefficient loudspeakers do), it is possible to increase the output power up to 2x50 W RMS without any large problems. A further increase in the output power is more difficult, but not impossible.
The high power consumption generates heat, this in turn make demands on heat sinks and cabinet. The size of the power supply, the cabinet and heat sinks explain why class A amplifiers simply not are cheap to build (or buy). You may assume a price of NOK 5000 for this amplifier, but in return you will have a particularly well playing amplifier. The amplifier is stable and have really fine data, which make the price reasonable.
The class A amplifier is not easy to build, and beginners should get help from persons with more experience within electronics before they set to work.
Why really bother with class A when the amplifier manufacturers continually are advertising that their amplifiers have all the benefits of the class A principle, but none of its drawbacks?
A class A amplifier may be made from nearly all ”usual” amplifiers simply by increasing the DC current (the quiescent current) in the transistors to suitable values. It is however another case that only a few amplifiers would endure this intervention. The class A principle in itself, in other words, does not require any special circuit diagram.
Class A operation means that all transistors in the amplifier draw current all the time, until the amplifier start to clip, caused by a too large input signal. When such an amplifier shall drive low impedances, e.g. 4 or 8 ohms, it is necessary with a considerable current consumption. This causes heat that has to be removed.
The large current consumption gives us, however, one of the benefits of class A operation, namely a distortion that is low and kind to our hearing. A transistor initially is a very nonlinear component, and one way to reduce the influence of these non-linearities, is to limit the drive of the transistor.
Let us assume that the quiescent current in the transistor is 10 mA, and that the largest signal causes a maximum peak current of 1 mA. Accordingly the collector current of the transistor will vary between 9 and 11 mA, corresponding to a maximum drive of 10 % from the quiescent value. Assuming we have chosen the right transistor, this form of conservative design will create small distortion of the signal, since we are moving very little around in the transistor characteristic. If we are designing an amplifier using a similar design procedure, firstly we are making a class A amplifier, secondly the amplifier will have very little distortion. It is not necessary to use this form of oversized design to make a class A amplifier, but it is a good idea if one by simple means wants to reduce static and dynamic distortion. In a normal class B (A/B) amplifier at least 2 and often more transistors will work on part time, which results in heavy and non musical distortion, unless special care are taken in the design work. It is the reason why one often would prefer a class A amplifier’s way of playing music, supposing it is well designed.
The class A principle also gives some other advantages, which at the end all have to do with distortion. One way to design class A/B amplifiers, is to use a form of sliding bias for the output stage. This method is meant to remove the distortion arising from the transistor switching on and off. In a class A amplifier, however, we are ensured that switching distortion not can arise. Here we are preventing the distortion from arising instead of trying to cure it.
Regarding working with loudspeakers with very nonlinear impedance graphs, the class A principle again is superior, due to the fact that the amplifier is so well dimensioned. Therefore it is not unusual that a 25 W class A amplifier sounds as heavy as a normal amplifier with a much higher output power, maybe up to 100 W. In addition the class A amplifier sounds better in situations where it has to deliver large currents, due to the fact it is designed to do this without any significant increase in the distortion level, in other words an indirect reduction in the distortion figure.
Of further advantages we can mention the low output impedance, especially without feedback, compared to “normal” amplifiers, and the good temperature stability without any ”thermal feedback”. A normal class A/B amplifiers temperature is arising when the output level is increased, which causes the current in the output transistors to arise, which in turn causes more heat to be generated, and so on. Unless one is counteracting this effect (e.g. with a design allowing some thermal feedback), the result is thermal runaway.
This kind of risk is not a problem with class A amplifiers, since the temperature is falling when the output level is increased. This is due to the fact that the power handled by the transistors alone without any signal now has to be shared between the transistors and the load. Therefore there is no need for thermal feedback, which eases the amplifier assembly. In addition the low output impedance is a result of the high quiescent current itself. When the amplifier is designed for an output power of 25 W RMS into 8 ohms, the minimum quiescent current for class A operation is 1,25 A.
The class A principle has, however, some disadvantages, which all coincide with the large current consumption at quiescent condition. Some we have mentioned: the large expenses to cabinet and heat sinks. If these are not sufficiently rated, one has to allow for reduced life time for the components, especially the electrolytic capacitors.
Principally the amplifier is based on voltage feedback, see figure 1. This is one of the reasons why this amplifier is called classical. The other reason is the schematic design itself, see later on.
Fig. 1.
However, it is only partly designed as a traditional operational amplifier. This is due to the fact that the author aims at an open loop bandwidth larger than the audible range. This is to ensure that the fundamental and harmonics of an instrument are treated in the same manner and thus have the same distortion figures. This implies that the instruments characteristic is not changed when amplified.
Compared to the operational amplifier, the current capacity of course also is much larger. This is achieved by using a three stage current amplifier. This implies that the voltage amplifier practically is isolated from the load, which is varying strongly (among other factors as a function of the frequency). The output transistors consequently also are voltage driven, which reduces the distortion.
The power supply is designed with a minimum of 2 ohm load impedance in mind. Even though this amplifier is not working in class A at this load, this design ensures that the characteristic of the class A amplifier is easily recognized.
Even though the opposite maybe seem to be the fact, there has not been done any extra efforts to keep the static distortion figures especially low. It is however done what is assumed to give a well sounding amplifier. With the technology of today, it is not any problem to design an amplifier with very low harmonic distortion, but this is not any guarantee to achieve an amplifier with very good sonic properties.
It is used FETs in the differential input stage (J105 and
106), see the schematics in figure 2. This has two advantages, one is that we
have a negligible input offset current, the other is that the distortion consists
mainly of well sounding even harmonic components. The value of the input impedance
of the amplifier is chosen to 100 kohm (R101). The relatively high value is
chosen because the author is not using any preamplifier, and the volume control
is placed at the input of this amplifier.
Fig. 2.
The current generator of the differential stage is realized by use of LM134 with the current set to about 3.6 mA. The gain is set by R107, and the offset adjustment is made by P108. The local feedback in the differential stage reduces both static and dynamic open loop distortion.
Following the input stage, is two differentially coupled common emitter stages (Q116 and Q117), which also have some local feedback. These have a quiescent current of about 5 mA. The output of the voltage amplifier (the sum point) is at the collector and emitter of Q126. Here the currents from Q116 and 117 are summed by means of the current mirror (Q119 and 120) across the resistors R127 and 128. These are effectively in parallel at the signal frequencies, since Q126 is coupled as a virtual zener diode and makes the bias generator. The quiescent current is set by P125.
R127 and 128 set the open loop gain to 53 dB. As R135 together with R112 set the closed loop gain to 30 dB, the feedback factor is 23 dB, a very modest value. The capacitors C129 and 130 set the open loop bandwidth to 100 kHz and the slew rate limiting to about 70 V/us, which are quite respectable values.
With the given value for C136, the closed loop bandwidth is about 650 kHz, and the phase margin is about 90 degrees.
Before the actual output stage an extra buffer, made up of the two transistors Q140 and 141, are added. In addition to reducing the distortion with a factor of 5-10 times, this involves that we may use small signal transistors in the voltage amplifier. The voltage swing is also increased, since the transistors are placed in opposite direction, and are using current generators (D131 and 132), with a current of about 2 mA. These current generators also will limit maximum output current to about 15 A.
Regarding the output stage, it is not much to add, since this is quite conventional. There is no voltage amplification here, but the current amplification is very high.
The choice of transistors always cause some appraisals, so it is relevant to state the reason for the choices that are made.
We are using plastic transistors for the power transistors, which make the mounting assembly very easy. These transistors also are more stable than transistors housed in TO3 package because the connection is printed on the circuit board and the leads are short, thus high frequency instability is not any problem.
We are using the complementary pair 2SA1216 and 2SC2922. These are very fast, but also very tough power transistors that withstand a power loss up to 200 W and has a large safe operating area. The heat sinks should have a size corresponding to 0.35 K/W or better. Heat sinks at this size are unfortunately expensive, but they indeed are necessary. If one try to use smaller heat sinks, the life time of the output transistors is reduced, since the junction temperature will arise. In the long run there is no good idea to try to save money here.
As drivers for the output transistors are chosen the types 2SA1358 and 2SC3421. These can withstand the necessary currents. These transistors also need heatsinking, but to make this simple, they are mounted on the same heat sink as the output transistors, not after the text book, but…The rest of the bipolar transistors in the design is the pair 2SA872 and 2SC1775. They are relatively noise free types with good linearity even for large changes in the collector current, which gives low distortion.
With the values given in the parts list, the amplifier will perform 25 W RMS into 8 ohm. This demands a quiescent current of 1.25 A. With the heat sinks used (0.35 K/W), the quiescent current may be increased beyond this value. In the prototype this is set to 1.6 A, this corresponds to class A working for full output voltage down to a load impedance of 6 ohms.
|
Performance |
The layout (note that this is mirrored) and component placement are shown as images. The printed circuit board measures 160x105 mm. Two boards are needed for a stereo version. The parts list applies for one channel (figure 2 and 3). |
|
|
Output power: |
2x25 W RMS into 8 ohm |
|
The power supply, see figure 3, is placed on the same circuit board as the rest of the amplifier.
Fig. 3.
The transformer secondary windings is thus connected to the circuit board. There are at least a couple of advantages arising from this mounting practice worth mentioning. Firstly, the impedance in the power leads will be low, and secondly, the stability will be irreproachable. It is not used separate supplies for the voltage and the current amplifier. The supply for the voltage amplifier is low pass filtered with the aid of a RC-filter with a low corner frequency. In addition the overall feedback and the design of the amplifier itself prevent noise on the power supply to reach the amplifier output. However, a mains filter with DC current filtering is highly recommended.
Shown in figure 4 is a filter and power on delay circuit. This is inserted between the power switch (and the mains filter) and the mains transformer (primary side). C313 and C314 with the diodes in parallel makes the DC blocking filter. The components 303-309 reduce and rectify the mains voltage with a slight delay, in such a way that the relay is shortening the resistors R311 and 312 after a few ms. The voltage at the output thus increases gradually. Thus a large current inrush is avoided when the amplifier is switched on. The varistor R319 is attenuating mains transients, while the components 320 and 321 provide for additional filtering.
Fig. 4.
The circuit in figure 4 is laid out on a small printed circuit board, measuring about 95x65 mm. The layout (notice that this is mirrored) and the component placement are shown as images. The parts list is also available.
With exception of the 6 W power resistors, 1/2 W metal film resistors with 1 % tolerance can be used for the amplifier (figure 2). Instead of the JFET's 2SK170GR for J105/J106, 2SK147GR may be used. This has higher internal capacitance and transconductance and is slightly more linear. In return the price is quite much higher. For both types one may use the BL-types that have higher saturation current (IDSS), but they are quite equivalent beyond that. A replacement for 2SA872/2SC1775 for Q116, Q117, Q140/Q119, Q120, Q141 may be 2SB 716/2SD756 or 2SA970/2SC2240. 2SC3421/2SA1358 used for Q126, Q142/Q143 may be replaced by 2SC4793/2SA1837. Regarding all replacements, be sure that the 'pinning' is correct when the transistors are mounted on the board.
Q126, Q142 and Q143 are mounted on the large heat sink. The bias generator (Q126) can favorably be mounted on one of the large power transistors (Q147 or Q148). This reduces the thermal time constant and gives a better measure of the junction temperature for the power transistor. The transistors used for Q147/Q148, are the well known pair 2SC2922/2SA1216 from Sanken. They are relatively linear, fast and at the same time very rugged; a rare combination. These transistors have been on the market for many years and are used relatively much in commersial amplifiers, and they are relatively cheap. They are available in a quite unconventional plastic housing, and they are attached to the heat sink together with the drivers Q142 and Q143 (and eventually Q126). A better solution would be to let the drivers have separate heat sinks, but the quiescent point is moving relatively little when the temperature is stabilized. In addition extra heat sinks are saved in this manner.
To get a 2x25 W RMS amplifier, a 2x18 V transformer has to be used. A transformer for each channel should be used. The transformer should have a minimum power ranking of 200 VA, a toroidal transformer of 250-300 VA is appropriate. The secondary windings of the transformer are attached to the circuit board at the points X159/X165 and to the points X166/X160. From the circuit board (X158) it should be good connection to chassis. The phono socket ground terminal should be connected to chassis only at one end (e.g. near the input). The (ground) shield of the phono cable should not be connected to the circuit board (at X100) if it is attached at the input. The inner conductor of the phono cable should be connected to the circuit board at X100 (pin 2). The loudspeaker output minus socket should be connected to the chassis at the output. The loudspeaker output plus socket should be connected to X157 on the circuit board. Make all connections as short as possible. If any form of instability or noise should arise, the probability is high that bad wiring scheme (e.g. ground loops) is to blame.
It is recommended to use a variable transformer or variable DC voltage generator first time the amplifier is started up. When the power supply voltage is increased, adjust the output-offset voltage by means of the potentiometer P108 to be close to 0 V DC. Also adjust the quiescent current to initially be at a minimum, and increase this slowly by means of the potentiometer P125. If possible, look at the output with an oscilloscope, there should not be anything but noise here if everything is OK. When the temperature is increasing, it is necessary to re-adjust both offset voltage and quiescent current (min. 1.25 A). The offset voltage at the output varies, but should not exceed 50 mV.
The amplifier is not provided with any servo coupling. If one has loudspeakers which not at all can withstand any offset voltage (rarely a problem), one may add a high quality plastic capacitor in series with R112. For a cut off frequency of 5 Hz, a value of 33 µF is needed. If it is not possible to get a high quality plastic capacitor, it is possible to use a bipolar (or bipolar coupled) electrolytic. It is recommended to place a high quality plastic capacitor (e.g. with a value of 100 nF) in parallel with the electrolytic capacitor. It is not made any room to these capacitors on the circuit board.
For 25 W RMS output power about 0.65 V peak input voltage is required. This should be sufficient for the most modern signal sources without being forced to use a preamplifier. If higher gain is wanted, R112 may be reduced. This implies that the feedback factor is reduced. This may be counteracted by also reducing R107. Then none of the good properties of the amplifier, like bandwidth, distortion and Slew Rate, are deteriorated worth mentioning. This assumes a reasonable change of the gain. The amplifier has also been tried with R112 equal to 750 ohm and R107 equal to 130 ohm. This gives a sensitivity of about 0.50 V peak.
The amplifier may be supplied with a voltage up to +/- 34 V. This should make it possible to increase the output power to 50 W RMS. The quiescent current for class A is then min. 1.8 A. The cooling requirement is large and should not be underestimated. But som loudspeakers may demand higher output power than 25 W. It is the author's belief that 25 W RMS real class A is sufficient for domestic use in most cases. A class A amplifier generally is perceived more powerful than a class B (or A/B).
|
Copyright©2002
|