Beogram 4000: New Main Capacitor Array with Integrated Efficient 24V Power Supply

If you ever experienced a Beogram 4000 you probably know that it gets pretty warm after a while! The main reason for this is the 1970s style voltage regulator based power supply, which is a bit of an energy hog.

So when I decided to replace my main capacitor replacement kit with a more modern PCB based design, I thought, why not integrate a modern buck converter based 24V supply that would smoothly replace the original 24V regulator?

This is the new design that resulted (it is available for purchase at the Beolover Store, there is also a version of this board for the AC motor Beogram 4002 Type 550x):

The many round capacitors are high-quality 105C rated Panasonic electrolytic capacitors that are connected as an array to match the original reservoir capacitor values. The row of small 'boxes' on the left is an array of Samsung X7R type ceramic capacitors that add up to the 150uF of the non-polar original electrolytic motor phase capacitor. Ceramic capacitors are much better for this application since they are inherently non-polar and they can take AC current much more easily than electrolytic capacitors. The circuit on the far end of the board is the buck converter based 24V power supply.

And here an impression of an installed board:

It bolts directly into the mounting holes of the capacitor clamps of the original setup. The solder pads for the wires are in approximately the same locations as the original connections to the big capacitor cans. This makes it straight-forward to replace the old capacitors. Simply unsolder them and then tack the wires to the pads according to position and color labeling.

This is an impression of the original setup that is being replaced by the board:

This schematic tries to make sense of the wiring around the capacitors and the voltage regulator:

For the installation of the new board this diagram does not need to be fully internalized. Simply match the colors of the wires with the labels next to the solder pads. This shows the board with all the connections in place:

Note that the 'thin blue wire' from the collector of the original voltage regulator (see below) needs to be moved over to the PCB. Solder it to the pad labeled "lgt. blue" up front where also the orange wire is connected. The blue wire connects the solenoid to the 45V coming from the rectifier.

Let's discuss the Beogram 4000 24V power supply a bit. The 24V rail is the main supply and energy provider of the Beogram. The motors, light bulbs as well as the circuitry that translates the commands from the 6V powered control logic beneath the keypad into actual behaviors of the Beogram are powered by it. The 6V supply that provides power to the control logic is only a marginal energy consumer since it only drives the logic chips producing control signals.

This is a clipping from the circuit diagram that shows the 24V setup (the red dotted rectangle indicates the parts replaced by the new Beolover board):

The rectifier 0D1 is fed about 46-49V RMS from the transformer secondary. The rectified voltage is smoothened by capacitor 0C3 into an unstabilized DC voltage of about 45V. This 'rough' DC voltage is fed into the collector of 0TR1, which is set up as an emitter follower controlled by a Zener stabilized ~24V voltage at its base. At the emitter of this voltage regulator a second big capacitor 0C4 removes most of the remaining DC ripple and a stabilized ~24V rail results.

This picture shows how this circuit is implemented in the Beogram 4000:

Since this regulator circuit produces a large amount of waste heat in the transistor, the transistor itself is bolted directly to the metal enclosure in its own metal compartment. This makes for an effective heat sink. The collector of the transistor is connected to the metal bar that clamps the transistor down (the screws go in from the bottom of the enclosure). Left and right of the collector bar the base and emitter leads poke out through insulating sleeves. The big maroon colored resistor on the left is 0R1, which pulls the Zener up to the unregulated 45V from the rectifier, which is connected to the collector bar via the red wire seen on top of the picture. The Zener itself is the metal can whose cathode is soldered to the base. The anode end of the Zener is anchored to GND, which is achieved by soldering the lead to the connection point of the two negative terminals of the biggest (3000uF) capacitor cans of the setup, which are also connected to GND via the black and green wires (see schematic above).

The blue wire that goes to the same solder spot like the red wire connects directly to the top of the arm lowering solenoid. In other words the solenoid directly gets the unregulated 45V and bypasses the 0TR1 regulator (this wire needs to be moved to the Beolover PCB during installation).

I think this explains why the B&O designers chose to regulate 45V down to 24V. The solenoid needs such a high voltage to actuate reliably, and apparently they did not see fit using a transformer with a 3rd dedicated secondary winding for the 24V rail. I guess in the 1970s it did not really matter that much if a device wasted almost 50% of its energy intake...;-), because this is about what this approach 'achieved'. 

Let's have a look at different ways that can be used to convert a voltage down:

This figure shows the three main ways to go from a higher voltage to a lower voltage:

The most straight-forward way is a simple voltage divider. In this figure the 'load' represents the entire 24V connected circuitry of the Beogram. If we want to go from 45V to 24V, we simply calculate R1 according to R1=((45V-24V)*R(load))/45V and as long as the load resistance remains stable we have 24V on the load. The problem with this setup is that the current is the same in R1 and through the load. In the case 45V->24V this means that R1 dissipates almost the same heat as the load, so close to 50% of the energy is directly wasted into heat.

The next step up is the voltage regulator like it is employed in the Beogram 4000. Here we replace R1 with an emitter follower and a zener/resistor divider that puts 24V on the base of the transistor. The great improvement with this approach is that it can keep the voltage pegged to 24V regardless of variations of the load resistance. In the Beogram example, the load resistance would for instance go down as soon as the platter motor turns on, which would draw more current. Whenever the carriage is moved the load resistance goes down a little, too. And so on, bulbs on/off etc....

Since loads are rarely constant, the voltage regulator is a popular and simple way to achieve a regulated stabilized voltage rail.

But since the emitter follower basically only replaces R1 and essentially still acts as a resistor, albeit a variable one, we get a similar heat dissipation like in R1 in the simple divider. In fact, it gets even worse, since we have to feed the Zener divider, which constantly carries a current to peg the base to 24V. This current is actually fairly substantial in the Beogram setup since the regulator is a simple bipolar power transistor with a low gain, maybe in the 25-50 range. This means that the base needs to constantly carry a current that is a few percent of the collector-emitter current through the transistor into the 24V rail, proportionally increasing the heat load. Hence the big 5W power resistor!

While this approach was maybe the best approach in the 1970s, it is not used much anymore in modern designs since we have now MOSFET transistors, which can switch at high speeds with a very low heat loss due to their very low ON resistance. This enabled the development of the so called buck converter, which is able to step down voltages with high efficiency. Modern designs can achieve 90-95%, minimizing heat loss dramatically.

A very simple schematic model of such a converter is shown in the above figure. At its heart the R1 resistor is replaced with an appropriately dimensioned inductor. This inductor acts like a resistor plus energy storage due to the fact that a coil upon turn-on has a high resistance due to the EMF that is acting agains the inrush current. As the magnetic field builds up in the coil the resistance drops asymptotically towards zero. So in the end a coil is basically like a straight wire with some residual Ohmic resistance.

So how do we go from 45V to 24 with this process? This is achieved by chopping the 45V input into a high-frequency 'pulse width modulated' (PWM) signal. This results in a continuous partial ramping up of the current in the coil during the ON 'duty' cycle, and a release of the stored energy in the coil into the load during the OFF (-duty) period when the voltage is cut.

As a consequence, most of the energy in the 45-24=21V drop that is wholly dissipated in the divider and regulator circuits is instead fed into the load!

Pretty cool (in the literal sense of the word!...;-)! Of course, nothing is free in nature, and the disadvantage of this setup is that there is a bit of noise on the output voltage caused by the incessant chopping of the input voltage. This is dealt with by connecting a reservoir capacitor across the load, which stabilizes the voltage. 

The above basic circuit also omits the necessity of a feedback based control circuit that adjusts the PWM duty cycle depending on the load resistance to keep the voltage on the load constant. So this setup is in reality quite a bit more complicated than the simple regulator circuit of the Beogram. Luckily, one does not really have to deal with this 'complication'. Nowadays integrated circuits take care of all of this and one simply has to select the right external components (i.e. resistors, capacitors and the inductor) for the desired output voltage and current needs.

The two traces at the bottom of the above figure show the result of the simulation: The chopped input voltage (green) and the resulting ripple on the output voltage (blue). With a fairly modest 100u cap the ripple comes out to about 27mV, or ~0.1% of the output voltage. The Beogram has 3000uF at the output (0C4), and so it is much smaller, probably easily in the range of the original regulator output.

Let's have a look at the actual 24V supply circuit on the new Beolover board:

The big 'box' in the back is the inductor and the small 8-legged integrated circuit in front is the buck converter. The passive components around it take care of the feedback to keep the voltage constant.

Alright! On to some measurements! I measured the current draw into the 24V rail and the temperatures of the transformer and at one of the screws that bolt the motor down before and after the implantation of my new board.

It is important to state the motor voltage at which these measurements were made since this voltage directly influences the current draw of the 24V rail. This is a oscilloscope trace measured at the purple lead (the nice sine wave also demonstrates that my ceramic capacitor array works happily as motor phase capacitance):

The 13.3V amplitude I chose balances enough motor torque for sweeping dust off the record with a still reasonable current draw.

The temperature measurement location at the motor made sense since the voltage regulator is directly next to it and so I would measure the heat in the front left corner, coming from the motor and the regulator. This shows my measurement setup (Beolover PCB already installed):

The temperature meter above shows the two temperatures and the multimeter is hooked into the 45V line as current meter. The picture shows the current draw in ON condition with running platter. About 190mA go into the 24 power rail. This table shows the measurement results for the original setup vs. the new Beolover 24V supply:

The current difference in OFF condition is 70mA vs. 15mA (original vs. new) and 360mA vs. 190mA. We see that the power consumption of the 24V rail goes down by almost 50% in ON condition and about 80% in OFF (standby) condition. The better performance in standby comes from the absence of the Zener diode voltage divider, which is absent in the buck converter circuit on the Beolover board. In OFF condition the Zener divider represents the main power draw of the 24V rail.

The standby power draw of the original setup of 3.15W is pretty impressive. Considering the hours in a year (8760), this means that over a year about 27kWh are dissipated by a Beogram 4000 just by sitting on the shelf being plugged in. This is reduced to about 6kWh with the new Beolover circuit. Just to put it in perspective, 1kWh allows driving a typical electrical vehicle for about 2-3 miles. So this is a pretty substantial power waste for a standby audio device. 

This more frugal power consumption of the Beolover circuit became immediately evident in my temperature measurements. The original setup reached 47C at the motor and 41C at the transformer in ON condition (after about 1 hr runtime). In contrast, with the Beolover board installed much lower temperatures of 37C and 34C were measured, respectively. 

So in conclusion one can say that the Beolover board reduces power intake of the 24V rail by almost 50% while playing the deck, and by almost 80% while it sits in standby.

This Beogram 4000 has arrived in the modern age in power supply terms!