Crescendo_complete_1_small.jpg  The Crescendo  Crescendo_complete_1_small.jpg

The Crescendo is a fully symmetrically built MOSFET poweramp with a substantial amount of output power, that was published in 1982 by the monthly magazine Elektor. This amplifier was part of the Elektor audio-XL series, that consisted of a number of high-quality audiocomponents. The design of this amp is still up-to-date and very 'buildable', despite the fact that it's from 1982. With a few small adaptions, the notorious oscillation problems from back then are history.

In this project, the circuit boards are assembled with high-end audio components, something that most certainly contributes to the excellent sound quality. During the gathering of the components, the availability of the 2SJ50 and 2SK135 MOSFET's from Hitachi that were used in those days, is a problem. In this amplifier, the original types from Hitachi are used for the 2SJ50's, of wich two are 2SJ49. The 2SK135 are from Magnatec. Luckily, good replacements are available in the form of the BUZ900 and BUZ905, also from MagnaTec. These are pin-compatible and can be used on the board without any adaptions.

For most of the components that were current in 1982, better replacements can be found nowadays, and so they've been used as widely as possible in this project. Special audio-components were chosen for a number of parts. Also, a few components were changed to improve stability and reliability. Encapsulated MKT capacitors are used instead of the open-frame MKM/MKH capacitors that were common back then. Most of the electrolytics were replaced by Black Gate's and in the input and feedback circuit N/Nx types are used. For the standard resistors, Holco H4 or the newer PRP resistors are used.


Transistors T1 to T6 were replaced by other types. Because of the big sized power supply, the supply-voltage is a little over ±75V when the amp is idle. Considering there's a drop of only a few volts across the collector resistors, almost the entire supply-voltage is present across these transistors. The maximum collector-emitter voltage (Uce) of the BC546B and BC556B is only 65V, so actually they are being overloaded. Looking at the advertisements in the old Elektor, there were most likely no types available with a higher Uce in 1982! In the many Crescendo's that were built sincethen, this doesn't seem to be a problem, but nonetheless other types were chosen here, just to be sure. The 2SA970 and 2SC2240 are good replacements. These are very known types with a higher max. Uce of 120V and they are also low-noise. Unfortunately, their pinning does not match, but by rotating them about a quarter of a turn, they fit on the PCB without any problems. The BC550/560 have a max. Uce of 45V, but they can be used without problems. They have to deal with only a few volts, because T8 and T10 take on the largest part of the voltage.

BC546_pins.jpg 2SA970_pins.jpg
pinning BC546/BC556                           2SA970/2SC2240

For trimmer P1 a cermet-type has been used, instead of the low-cost part from 1982. The value is lowered to 100Ω to enable a more accurate setting of the idle-current, which was quite tricky with 250Ω in this case. This strongly depends on the characteristic of the used MOSFET's, so it might not be necessary in all cases.

The cheap wirewound 5W cement-resistors are best to be avoided in the output stage. Due to their high self-inductance, they will almost certainly cause oscillations, a problem that was haunting this amplifier from the beginning. On the next page this is discussed more widely. A better choice are the low-induction MPC71 metalfilm resistors for example. These are radial (standing) types, but with some creative bending they fit on the board perfectly. Another option are the axial 5W MOX resistors, which are often used in speaker-building. But due to some experience with these resistors in the Vifa Vivace '94 loudspeakers they are not recommended.

The input capacity of the N- and P-MOSFET's is different and this leads to the driver stage not being loaded symmetrically. To make the loads equal, a few extra capacitors can be placed at the 2SK135's. These can be soldered directly at the pins of the MOSFET's, at the back of the PCB. An extra advantage of this is that it also reduces the oscillations a bit. Between the drain and the gate an extra 33pF is fitted, and between the source and the gate a 330pF. For the BUZ900/901 MOSFETs the values are 15pF and 220pF. It's best to use ceramic C0G or NP0 or polystyrene capacitors for these.

Overview changed components
820n MKM
BG-Nx Hi-Q
820n MKM

220p cer.
C4, C5
1x 100u/16V
C6, C7
220n MKM
470n MKT

BG Std.
F1, F2

P1 250Ω 100Ω cermet

R5, R6

0,22Ω/5W wirewound 0,22Ω/5W metal film MPC71
T1, T2, T6 BC546A 2SC2240
T3, T4, T5 BC556A 2SA970
T8, T10 BF169/BF170 MJE340/MJE350

The arrangement of the case largely has been taken over from the Elektor article. If the connectors are placed on the rear of the case, this is probably the most practical setup when using a 19" case. The power supply consists of a big 800VA toroidal transformer, two bridge rectifiers and four 15,000uF/100V smoothing caps. With a few pieces of aluminium sheet these are made into a solid construction that can handle a substantial amount of current. Four MKT capacitors of 2,2uF are placed in parallel with these caps. The transformer and the electrolytics were switched places, because the transformer doesn't fit between the front and one of the amp PCB's; the depth of the case that's used here is 30cm. With a 35cm. deep case, the original arrangement can probably be applied.

The heatsinks are model SK85 from Fischer and they are 160mm wide and 150mm long. The thermal resistance at this length is 0.65K/W. That's a little higher than specified, because the original SK53 heatsinks have a thermal resistance of 0.55K/W. But, with 180mm these are a lot wider, and fitting two pieces of this size on the back of a standard 19" case will be difficult. The SK85-150mm fit exactly on the back of a 4HE case. This way, enough space remains between the heatsinks for the connectors. At about 100mA quiescent current for each MOSFET-pair, the dissipation is a bit over 30W per channel; the heatsinks are already getting quite warm with this. During the endurance test at full output power, the SK85 shows to suffice perfectly. More cooling is always better of course: one can choose to mount large heatsinks on the front instead and make a real power-stage of the amp. The volume-pot can be omitted if a pre-amp is present in the audioset and the mains switch can be mounted on the rear side.

SK53.jpg SK85.jpg

In the middle, between the output PCB's, a PCB is mounted with an auxilary 12V supply and two relays for the DC detector and power-on delay. A 100Ω/7W resistor is fitted in series with the primary winding of the transformer to limit the inrush current. It is bridged by the first relay, a few seconds after power-up, and makes sure the smoothing capacitors have time to charge. The speaker connectors are connected to the amp's PCB's though a double-throw relay. In case of a defective output-stage, which causes a dangerous DC voltage on the output, this relay will de-energize quickly to protect the speakers.

On the left side of the transformer, the driver PCB that drives the relays and two LED's on the front is mounted. A red LED above the mains switch comes on when the power-up relay is energized. Next to it, is a green LED that comes on as soon as the output relay is energized, and this indicates the amplifier is ready to use. Located a bit more towards the front of the case are two PCB's with a clip-detection circuit mounted on top of each other. The two most righthand LED's are connected with these and they indicate for each channel if a pre-set output level is reached. With the help of a scope, the circuit can be adjusted so the LED's come on just before the amp's clipping level.

A sturdy power-supply leads to more output power. Hence, the maximum power is substantially higher than indicated in the specifications. With this 800VA supply, the amp delivers 200Wrms at 8Ω and 280Wrms at a 4Ω load, measured with a 1kHz~ input signal and driven just before the clipping-point on the scope. The supply voltage, which is ±75.3V in idle, drops to ±73.4V when loaded with 8Ω and to ±67.5V at 4Ω.
During the test with a 4Ω load, the on-board fuses blew several times. Due to the higher output-power, obviously, more current runs through the output stage. That's why it's better to raise the fuses to 4A.

To page 2: "The Crescendo - tamed..." >>