Beogram 4002 (Type 550x): New Main Capacitor Array with Integrated Efficient 22.8V Power Supply

Immediately after offering the new Beolover Efficient 24V Power Supply and Main Capacitors for Beogram 4000 component, I received inquiries about whether this part could also be used in AC platter motor Beogram 4002s (i.e. Types 550x).

The earlier AC-motor 4002s have a fairly similar setup when it comes to their main power supply. But there are minor differences: The rail voltage is only ~22.8V instead of 24V, and they do not splurge on a continuously powered standby mode like the Beogram 4000. But they also waste a similar amount of energy during operation due to the fact that the 45V transformer voltage is regulated down to the specified 22.8V with a Zener stabilized power transistor. This results in a nearly 50% energy loss in the 22.8V rail in the transistor.

In my design for the Beogram 4000 I replaced this setup with a modern buck converter-based design that has a DC-DC conversion efficiency in the 90-95% range. This causes the Beogram 4000 to run much cooler due to the reduced heat load. I provided a basic analysis and explanation of this setup in my original post about this design.

While the voltage difference is not really an issue, unfortunately, the different turn-on method in the 4002s without stand-by prevents the direct use of the Beogram 4000 board in the 4002s. Therefore, I designed a dedicated board for 4002s that also directly replaces the main capacitors and the voltage regulator setup. This is how the new board looks (it is available for purchase at the Beolover Store):

The many round capacitors are high-quality 105C rated Panasonic electrolytic capacitor arrays that provide the new power supply with appropriate reservoirs and couple the motor to the Wien oscillator amplifier. 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 22.8V power supply.

It replaces is this original setup:

The two larger capacitors on the right (0C1/2) are the in parallel connected reservoirs for feeding the voltage regulator whose transistor (0TR1) is bolted directly to the chassis right of the motor. The reason that this transistor is not on the main PCB is its significant heat dissipation that needs to be sinked efficiently. A significant part of the energy going into the Beogram is leaving it as heat at this front corner. That is the main reason that this area gets pretty hot after playing a couple records. Motor and transistor pretty much divide maybe 40% of the total heat load of the deck between them. The rest is mostly dissipated from the transformer, the solenoid (when the arm is down), the Zener that controls the regulating transistor, the incandescent light bulbs and the electronics.

The other two capacitor cans are to couple power into the motor (0C3) and to shift the motor phase by about 90% for the second winding (0C4).

This is a snippet from the circuit diagram showing the setup of the original power supply:

The buck converter that is integrated on the board basically replaces the 0TR1 transistor eliminating most of its power dissipation.

Replacing this setup with the new Beolover board is straight forward. Simply remove the capacitors and then unsolder all the wires from them and the transistor:

The transistor can/should be left in place.

Then solder the wires that were connected to the transistor to the respectively labeled pads at the bottom end of the board:

Then connect the four wires from the motor according to their color:

Here a shot from a bit further away:

Then solder the red and black wires from the rectifier:

A detail photo:

Next are the wires that go towards the PCBs: The green wire goes to the pad next to the motor wires, and the two orange and black wires to the pads on the right of the rectifier wires:

And that is it: This shows the board fully connected and bolted in:

And with the main board replaced:

Beolovely!

Like for the Beogram 4000 setup, I also measured the temperatures and currents before and after. This is what I got (33RPM, 13.2V motor voltage, arm up and carriage at rest):

Similar to the results for the Beogram 4000, a significant drop in temperatures occurred: The motor temperature dropped from 47C to 38.4C, while the temperature at the transformer went down from 38C to 34.8C. As a consequence the deck does not feel unusually warm anymore to the touch.

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!

Beomaster 8000: Step One - Rebuilding the Output Amplifiers

While we wait for Nick's awesome specially made pulley for the Beogram 4002 (5501) that I just finished up, it is time to get started on the Beomaster 8000 that will also go on to the UK once restored. It recently arrived and I gave it an external inspection, which suggested this 8000 is an excellent starting point for a full restoration.

My first step in any Beomaster 8000 project is to rebuild the output amplifiers. They are the most crucial part of the restoration of the 'power' part of the unit. A failure of their often corroded quiescent current trimmers usually neatly kills all output transistors with a bit of smoke emission on that channel before the main fuse protects the transformer (and the breaker of the house grid on which the 8000 resides while this burnout happens). These old trimmers often go open circuit during transport, so my approach is to not even turn the unit on when I receive it, but I go straight to the amplifiers and rebuild and test them to make sure that the above does not happen.

Here are a few impressions from this effort on this unit:

This shows the right channel as it came:

We see from the two clunky (white) emitter resistors that the above mentioned disaster already must have happened at some point. When the transistors burn out, the emitter resistors usually brown like a chicken in the oven due to the immediate heat emission. But usually this does not affect their performance down the road since they are wire wound types. An ugly replacement like seen here is definitely not an improvement in the Beolover's eyes! Also, whoever did this did not learn 'The Lesson'...he did not replace the trimmer! I always put in 25 turn precision encapsulated units to prevent this from ever happening (again). Multi-turn trimmers drift only very little and so the quiescent current adjustment is very stable over time. This is not the case for standard single turn trimmers, and that is one of the reasons that the 8000 often presents with one or two hot heat sinks, if it runs at all.

This shows the rebuilt board with new electrolytic 105C type capacitors and prettier resistors and the 25-turn trimmers:

Beautiful! I did the same to the left channel, which did not have new emitter resistors, but (slightly browned) original resistors. And then it was time to run these babies from external power supplies to test for any silicon failures, cracked traces and the like. This shows my hookup:

And here with the multimeter connected to the test points at the emitter resistors after powering the board up:

When installing a new trimmer make sure that the resistance is close to zero. This may be counterintuitive, but in this setup it turns off the output transistors preventing significant current flow between the +/- power rails. This shows my bench supplies at this point:

The left two are the - and + rails, which draw 60-70 mA while the quiescent current trimmer is close to zero Ohms. The right supply provides the 15V control voltage that is used by the Beomaster to enable the output by controlling the constant current source used for biasing the output transistors.

Adjusting the quiescent current means to ramp up the resistance in the potentiometer to set the working point of the output transistors that there is a 18mV voltage across the two emitter resistors:

At that point the power rails draw 0.1 (+) and 0.11(-) A:

I did the same for the right channel:

All good now in the output amplifier department! On to rebuilding the power supply board before giving this unit a first spin!

Beomaster 8000: Fault Switch Gets Triggered by Disconnected PTC Thermistor

Oh well...I guess I was a bit too enthusiastic when I declared victory in my last post about the output amplifiers of the Beomaster 8000 that I am restoring right now. It turned out that there was an intermittent issue that triggered the 'fault switch' and that in turn turned off the relays that enable power to the output stages. Here is the relevant circuit as shown in the Beomaster 8000 'Technical Product Information' booklet:

Here is how it works: TR16/17 form a latch that is triggered by a >~1.5V voltage coming from the "fault output" of the output stages between R254 and R253. When the latch is set it turns on TR15, and 15V apply to R68. That pulls up the base of TR11 and it turns on, which in turn turns on TR12. This pulls up the base of TR18 and the relay looses power. This cuts the power to the output boards, and the 8000 goes silent. All other circuits remain on, i.e. the display are functioning normally etc...

The two failure modes that can trigger the latch are:

1. A DC voltage at the 'AF OUTPUT' (this usually means that one or more TIP transistors burned out due to overload or failing quiescent current trimmers
2. The heatsinks get too hot. This increases the resistance of the positive temperature coefficient (PTC) thermistor that is connected to the junction between R353 and R252 and GND. When it is cool it has about 50 Ohms which pulls the fault output to GND. If the resistance increases, a DC voltage develops due to the pull-up to 55V and the latch is triggered.

In my particular case, the receiver came on, played for about 30 sec on the left channel and then it went silent. One could also hear three clicks in rapid succession right after starting it up, which is one too many. Normal are two clicks, which correspond to the start-up sequence during which the two relays that control the current into the toroidal transformer get switched in succession to limit the inrush current into the transformer. The third click meant that the fault mechanism turned the relays off again. In the beginning I was a bit confounded by the fact that the left output was continuing playing for half a minute. It turned out that this is possible due to the charge that is in the reservoir capacitors, i.e. it shut down when the voltage dropped close to zero in these capacitors. Why it did not do that on the right side, comes from the fact that the right supply also powers the preamplifier via an attached ±30V supply. This causes the charge on the reservoir caps on the right side to dissipate more quickly.

Since the output stages showed perfect current values during my test with external power supplies, I first thought that the fault switch itself was at fault. So I tested it by grounding the fault output from the output stages, and this caused the issue to disappear.

Note: This test needs to be performed from a cold start, since the latch in the fault switch only resets when the 15V rail turns off, i.e. the Beomaster needs to be turned off for that. However, the Beomaster runs a turn-off sequence after pressing the off bar on the keypad, where the 15V remain alive for another 60-90 sec. This means that one needs to wait a couple minutes before turning it on again to see if the grounded fault output disables the fault condition.

In my case this test yielded a functioning fault switch, and therefore the issue had to be in one of the outputs! It turned out that the left stage had a PTC that was intermittently disconnected from R253/R252, which caused it to loose its ability to pull the fault output to zero at normal temperatures. The picture shows the broken white-blue striped lead that connects the PTC (which is mounted on the heatsink) to the fault switch voltage divider. The tricky part here was that the lead looked perfectly o.k. since it was attached by the insulation to the solder terminal and the lead was only broken off in the inside.

After restoring this connection, the Beomaster started working normally.

Beomaster 4400: Full Recap and Output Trimmer Replacement

Today I completed the replacement of all electrolytic capacitors in the Beomaster 4400 (2419) that I am currently restoring. I also replaced the quiescent current trimmers of the output stages. This is probably the most important item to do in these models due to the design of the output stages. If these trimmers go open circuit due to old age, the output transistors immediately die, i.e. whenever one opens up this model, these trimmers should be replaced and the quiet current adjusted properly.

But first I removed the input/output board that is bolted underneath the main PCB. Here you see it still bolted on:

It is easy to remove it. It is only held with the four screws next to the Tape 1 and Phono jacks. Take the screws out and the board can be removed:

It is connected with one wire-to-board connector to the system, so it is easily taken out. I really like the models they designed after they discovered the existence of such connectors. It makes it so much easier to work on them. The 4400 design is somewhere in between..still a lot of directly soldered connections, which can make id difficult to work on certain areas in these units. Anyway. Here it is taken out:

While I had it out, I decided to do the caps first. So I replaced everything with quality Japanese made 105C types for a long lasting restoration:

The next step was to remove the shield of the power switch. It makes it hard to get to a couple capacitors that are located underneath the switch. I did not want to touch the switch, and it seemed that after removing the shield, there would be enough space to get the task done. This shows the switch as I found it:

And here after unsoldering the resistor and taking the shield out:

Then I replaced all the electrolytic caps on the main board and put in new trimmers. Here some impressions of the original condition:

Left channel and a view on the tone control and preamp section:

And the right channel:

A few capacitors had been already replaced by a previous tech, but he used low grade 85C types, so I decided to bring everything up to 105C spec. Here are a couple pics:

This shows the left channel. I replaced the original single-turn 100 Ohm trimmers with 25 turn cermet trimmers. These are encapsulated and allow a precise adjustment of the quiet current that will remain without drifting over time. My experience shows that regular single turn trimmers can drift through repeated temperature changes, while the worm gear mechanism in a multi-turn trimmer will effectively prevent this, since the gear cannot turn as long as the 'worm' is not turning. This is the left channel:

And here is the right:

The next step was to adjust the quiescent current in both outputs. The manual prescribes to adjust it to 10-15mV across both emitter resistors in the output. The way the resistors were soldered in in this unit, it was more convenient to do the measurement just across one of them, and adjust to 50% of the above range. Here is how I did that:

I prefer to adjust to the lower end of the range, since this will make it run much more cooly. I checked with the oscilloscope, and there was no measurable crossover distortion in this unit with this adjustment.

After all this I fired the unit up and I listened a bit to the radio. It sounded very nice, and the heatsinks stayed very cool. When I received the unit it it ran pretty hot, indicative of misadjusted or problematic trimmers. Even after extended periods of low-volume listening the heat sinks should never get warmer than 30-35C. 

After the radio got boring, I decided to hook up the Beogram 4000 that I recently restored to try out the Phono input. And this was disappointing. It worked briefly, and then started cutting out intermittently on both channels. Still some issues I guess...time to explore the circuit diagram...I always enjoy doing that. This is Beolove!