This one did not have less than 42 transistors in the amplifier section, despite its performance, it was a system too complex. In short, it's actually a montage inspired by the 20 W Class A Before returning to this circuit, other tests, still under way, related assemblies including single-transistor outputs, type germanium. Power limited to 5 W, the difficulty of finding good germanium power transistors have been that this project has not yet been completed. Other tests, which did not lead to satisfactory results concern several montages briefly described in Figure 1.
Floor of entry: even quieter transistors, high gain, but linear
- Low input leakage current
- Input impedance higher
- Circuit to reduce the Miller effect, to reduce the amount of distortion at high frequencies
- Input stage can be overdriven without risk of saturation.
° Etage driver:
- Circuit autocompensation distortion of linearity
- Low output impedance
- Low distortion
- Output level higher
- Bandwidth.
° Floor Power:
- Similar to the 20 W Class A
- Choices facing other output transistors, less powerful, but higher in subjective quality.
For the desired improvements in subjective terms, they have been described previously. Some seem quite contradictory, but, apart from the result that proves it, how to proceed in the selection of parameters shows how this is possible. Besides the unpredictable, it would be to tailor the sound. Listening to final should not come as no surprise, except, perhaps, very small details.
Figure 2 shows the general circuit, which recognizes the output stage "Darlingnot" in inverted Darlington. Note that the old combination 2SC1096/2SA634 2SD188/2SA627 passes and a new combination, a little less powerful but more efficient. The choice of drivers is both subjective and objective. The value of the Cob. 75-pF on 2SA634 passes only 1.8 pF On the 2SB716. By cons, there is a PC much lower (only 750mW) on the new driver, however, sufficient to value the driver output stage. The output pairs 2SD844 and 2SB754 are type cast in new housing. This complementary pair has a Pc of 60 W, which is sufficient for work in class A with a modulated power from 8 to 15W. This pair can work at a voltage input two times lower than the pair 2SD188/2SA627, which explains the use of a smaller driver stage. Figure 3 shows the differences between these transistors. Note that for work in class A up to 20 W, these transistors could not agree. The output stage and mounted with 2SB7l6/2SD756 2SD844/2SB754 and provides, over and 2SC1096/2SA634 2SDl88/2SA 627:
- A little less distortion between 0.1 and 3 W, high frequency (Cob effect of lower
of drivers);
- Set treble;
- Further lower midrange;
- Even more serious given (Rbb transistor output instead of 3.2 ohm 7ohm);
- The more open (lower rate of CR);
- Medium more "hot" but also detailed.
Fig. 2: Circuit of the amplifier 
ator 8 watts Class A
ator 8 watts Class A The circuit stability is unconditional, even loaded 1uF 8ohm in parallel (see photos). The set provides a very large bandwidth (more than 4 MHz), an extremely fast rise time (less than 0.5 uS). Noted that such performance MOS-FETs could be as stable on capacitive load. Another advantage is the possibility of reducing the length of the link driver / power transistor. From about 18 cm on the 20 W Class A, this time it is direct, the power transistors can be mounted directly on the PCB. Which reduces the binding capacities and potential risk of instability.
As mentioned previously, we see that jective conform exactly to what was desired and the disadvantage of a limited output power of about 8 W.
As previously mentioned, there exist very close relationship between subjective performance and configuration schema used. A systematic and rigorous and helps achieve the goal, however, the sacrifice of a parameter which is, in this case, the limited power around 8 W.
Transistors | V CBO V | V EBO V | I C m A | P C W | M F | V CE V | I C A | V CB V | I E mA | T F mhz | ON R ohm |
2SD188 | 100 | 7 | 7 | 60 | 60 | 2 | 3 | 10 | -200 | 10 | 7.5 |
2SD844 | 50 | 5 | 7 | 60 | 70 ~ 240 | 1 | 1 | 5 | -1A | 15 | 3.5 |
Fig. 3: Comparison Chart transistors 2SD188 2SD844 and
The input stage
It is not at all like the one that was used on the 20 W Class A
In this circuit, the choice of the input stage was paramount. As strange as it may seem, it was found by a sound similar to a tube driver known in Japan for its subjective qualities: the WE310A, a pentode tube absolutely remarkable voice, guitar, piano, exceptionally short in the band 200-5 000 Hz. The use of bipolar transistors can easily produce the odd harmonic distortion while a pair complementary field effect will produce a bit of odd harmonics (harsh, unpleasant sound, which is explained briefly in Figure 4. In the If the circuit of 20 W, the compromise was to use bipolar transistors to very good subjective quality, the 2SA872 (E) and 2SC1775 (E) which provided a fitting distortion rate higher, but degraded to a particularly good harmonic distortion . The second floor also attacked the driver to the limit of saturation, which fortunately did not ask too much of a problem, according to various settings (see n 0 15) and adjusting the supply voltage to + / - 21V . 
Fig. 4 (a): Spectrum of distortion on my Tage cascade FET-Bipolar.
Fig. 4 (b) Spectrum distortion FET complementary pair.
- Output low impedance;
- Very high gain;
- Low distortion;
- Low leakage current input;
- Circuit with very low Miller effect;
- Input saturation level.
This is a complementary pair cascode "variegated" FET / bipolar transistor in which the election will be done carefully in order to obtain the desired result.
Without this additional cascode, these results could not be obtained any other way.
The cascode arrangement makes it possible to obtain a very high gain and the risk of instability in the case of this assembly are virtually nonexistent. In the case of high gain triode tubes, it probably would not happen. Then, combining FET / bipolar combination produces a characteristic similar to a pentode tube. This is equivalent to a spectrum of distortion with a predominance of odd harmonics. This is voluntary, as the push-pull assembly will take care to reduce these where a combination of all to bring a good result.
A cascode arrangement of this type, low impedance output to bring the desired improvements in subjective, that is to say, more scale in the mid-bass, but also a serious and well run farm (also due to supply circuits ). But its advantage is an important gain in transparency. But obtaining these results is highly dependent on the choice of transistors. A mandatory requirement: use an input field effect transistor with very high Gm, 20 to 30 times higher than that of a MOSFET of its kind 2SK30AGR. Used alone, this type of transistor, very low noise could not agree as to pre-preamp and preamp. Only the Fet used here, the complementary pair 2SKl7O/2SJ74 including the only two benefits have a very low noise
(En = 0.9 nV / √ Hz)
and a high Gm: 2.2 mmho. But the faults of these transistors are numerous:
- Gate leakage current of high (loss of sound transparency);
- Stray capacitance Ciss and CRSS (entry and return) key: 30 pF and 6 pF (instead of 8 and 2 pF
sure about the 2SK30AGR);
- Gate leakage current increases rapidly when the working voltage Vds increases;
- Saturation voltage input low, due to the high gain (0.2 V approximately).
A cascode mounting significantly improves these characteristics. We could have climbed in cascode FET, as shown in Figure 5 (a) but the combination of bipolar NPN / Fet N-channel is better (b). The advantages are:
- Considerable reduction in parasitic capacitance CRSS (ability to "return" drain-gate) that goes to 1/10th of its initial value, or 0.06 pF instead of 6 pF, a significant reduction of the Miller effect (Figure 6);
- Lowering the working voltage Vds (assembly being in series), IGX consequent reduction (gate leakage current), as shown in Figure 7 th.
- Input saturation level higher (about 1V year instead of 0.2 V).
8 shows schematically the input circuit and the electrical equivalent.
Fig. 5: cascode.
Fig. 6: Reduction of the Miller effect, through the use of cascode
Fig. 7: Reduction of leakage current IGX by the use of cascode,
compared to that of a single FET.
Fig. 8 (a) equivalent circuit diagram of a cascode complementary.
Fig. 8 (b) complementary cascode circuit.
In this arrangement, the input impedance, which is charged by high and 47 kohm resistor 1.2 kohm stop is mounted in series in the input circuit. The complementary cascode circuit is loaded by only 47 kohm, the current being in the range of 0.9 to I mA. The bases are polarized by the four 2 kohm resistors and various control trials (zener diodes) have been less than attentive. The choice of a combination 2SK170-2SJ74/2SC1775-2SA872
yet been made on subjective criteria, depending, of course, the overall result.
In the next issue, assembling and other possible settings are detailed, as well as the imposing power of + / - 14 V, lead acid battery mounted in buffer. The reader will find in the figure by 10 against the circuit board of this assembly.
Measurement and listening
This circuit has been carefully developed to the extent as to listening, in April 1982. He had been "set aside" as a matter of transistors whose choice brought results surpassing expectations, in terms of e taient listen but still very difficult to obtain in the Japanese market. The pair 2SD844/2B754 was particularly hard to find, the Hfe not corresponding to the desired values, this Hfe, 60 2SA627/2SD188 on here including the following lots (K, L, M, N, O) between 70 and and only 240 lots K and L (2SD844K and 2SB754L) may agree. As for 2SK170/2SJ74, the transistors are still quite difficult to find as recent and only in small anufacturing f s eries by the firm Toshiba.
For listening, the result also depends on food, but you get the odd surprise compromise triodes tubes / amplifier Hiraga 20 W class A, where only the output power is a small shadow on the performance table. Overall, we get a particularly his set, ventilated, sound reverberation, echoes more free, while direct s ons are even more present, more and more stamped "hot". The paradox lies in the grave, with the impressive power, can finally be compared to that of class A amplifiers Kaneda 30 W and 50 W: firmness exceptional combination of sounds extremely firm on its infinitely soft and light. Super position of sounds on sounds infinitely vague contours finem ent chiseled.
Even on systems of average yield, this amplifier has performed very well, print equivalent of space, power reserve that can normally be obtained only with a few amplifiers mentioned above.
Dubbed "The Monster" because of its unusually large size compared to its output power, he could still be called "Tube Memory," because of his own stamp on a few triode tube amplifiers that were previously used in fixtures so-called "high definition". This device will find its ideal place in low-medium, medium, or in the acute in systems bi, tri or quad amplified.
Fig. 9 (a): Response of square wave at 20Hz. Top,
amplifier output, down, output generator.
Fig. 9 (b): Response to 20 kHz square wave on
capacitive load, 0.47 uF in parallel on 8 ohm.
Fig. 9 (c): Allure of the rising edge to 10 kHz. Time
rise is less than 0.5 uS. It is difficult
measured with the tester to stay employed.
Fig. 10: PCB and implementation
Thank you for the article. On the picture Le Classe A amplifier above...do you use 2SD844 and 2SB754?
ReplyDelete