JVC DD-5 capstan servo fix

[Studio263]

I recently acquired a JVC DD-5 cassette deck which was going for scrap otherwise. It was complete and lit up when switched on but would not play or record as the capstan wasn’t turning – the winding modes functioned normally. This model is fitted with a direct drive capstan motor, the basic layout of which is shared wit the DD-7 and DD-9. The others have 3 heads and Quartz Lock as well, but the basic motor is much the same. The circuit diagram for the DD-9 motor is attached; the Quartz Lock circuit is in the dashed box at the bottom of the diagram and is not present in the DD-5.

Looking at the motor, the supply was absent to the hall sensors because the servo was completely cut off. Applying some current to the base of transistor X10 made the motor spin up wildly, indicating that the servo IC was faulty. This is a special chip made just for this motor and not available separately, hence the machine was now not repairable by conventional means. This seemed a shame, direct drive cassette decks are an unusual breed and this one has JVC’s excellent Sendust Alloy head too, far too nice a machine to let stand idle due to unavailable replacement parts.

Making a velocity servo isn’t that difficult, the trouble is that to give the level of performance that this machine is capable of the design needs to be approached with care, particularly in respect of both instantaneous and long term stability. I set about the task of designing something suitable, using junk-box components which are easily obtainable. Although the characteristics of the circuit are closely matched to the JVC motor, by changing a few component values it could be adapted for use with other motors in other things (turntables for instance).

The first thing to do was to determine some basic parameters. A single +13.5V DC supply was available at the motor and by manually varying the drive to the transistor X10 while playing a 1KHz test tape and monitoring the off-tape frequency and the output from the motor frequency generator (FG) I determined that at the correct tape speed the FG signal was around 620Hz. The FG pre-amplifier in the servo IC (IC1 in the JVC circuit) was still working with a healthy waveform available at pin 3, it made sense to use this to keep the wiring carrying the low level signals from the pick-up coil as short as possible. To isolate the rest of the circuit C4 was then removed. The waveform was basically sinusoidal and swung between +0.8V and +4.8V, although how stable this would be over time was not clear.

With only a single positive supply available and op-amp IC with rail to rail outputs was mandated. The LM324 is used in the tracking servo of the B&O Beogram 8000 (etc) as well as some Philips CD motors and in the cassette interface of the BBC micro, so I always have some around – this fitted the bill perfectly. The first stage of the circuit turns the sine wave into a square one, thus yielding a signal that carries information about frequency only. R1, R2 and C1 form an automatic comparator which compares the instantaneous voltage of the incoming signal with its average level, meaning that however the input signal may drift the comparator’s reference voltage is always at a central point. The square wave output drives transistor Q1, which clamps the voltage which builds up on C2 via R4 once every FG pulse. As the motor slows, the voltage on C2 has the time to rise to a higher peak before it is clamped, conversely as the motor speeds up the voltage is clamped earlier and therefore achieves a lower peak value. This is the heart of the servo system, a voltage which is inversely proportional to the motor speed. The whole system aims to keep the peak voltage across C2 constant. Since C2 is charged at a fixed rate it follows that if the peak voltage is constant and controlled, the motor speed must be constant and controlled too.

The ideal way to charge C2 is from a current source but to make an accurate and stable one is surprisingly difficult in a simple circuit, so a constant voltage source was chosen instead. This comes from the last section of the LM324 IC, a 5.6V zener diode forms the basis for a buffered output. Using a constant voltage means that the ramp across C2 isn’t linear, but that is of no importance in this circuit. The main thing is that the ramp is consistent, so C2 is a film capacitor in order to achieve a good level of thermal stability. A ceramic capacitor would not be suitable here. Elsewhere in the circuit the capacitor values are not critical, so I selected the resistors so that the same type of capacitor could be used throughout (47nF X7R ceramic). From experience, there is always a bigger choice of resistors on hand than there are capacitors in the typical junk box!

Since the motor is heavily damped by the drag of the tape the servo need only be one that adds power – there is no need for a braking function to control deceleration after an overshoot. Therefore a simple sample and hold circuit can be employed where the voltage on C2 is sampled through a diode arranged as a peak detector. The forward voltage drop of a diode is temperature dependant and so it is cancelled by the action of the second LM324 stage. These ICs shouldn’t be used to drive capacitive loads directly so a 100R resistor (R5) is inserted in series with the output to ensure stability. The sampled voltage is then stored across C3, which discharges through R6. It would be possible to select values for C3 and R6 so that the voltage across them in respect to time was an almost pure DC value that was virtually free of ripple, however for a number of reasons this is not a desirable mode of operation for this circuit and this motor. Introducing too long a time constant at this point would introduce a large amount of “lag” into the system, leading to instability and extended settling times. The waveform across C3 is in fact a triangle wave superimposed onto a DC offset, the triangle form originating from the tops of the rising ramp coming through the peak detector and the decay of the charge held in C3 through the action of R6. The magnitude of the DC offset is roughly equal to the speed control set point defined by the following stage.

The voltage across C3 is compared to a fixed DC reference voltage by the third LM324 stage in order to set the operating point of the circuit. This gives the opportunity to set the tape speed accurately with a simple variable resistor; I used a 10 turn trimmer type. The reference voltage potential divider (R7, R8, VR1) is fed from the zener stabilised source and is further decoupled by C4 so that variations in the supply do not affect the set-point. When gross errors are present, for example at start-up, the comparator saturates and gives a constant “high” DC output. Similarly, if the motor speed where to be greatly in excess of its correct value, the comparator output would be a constant “low”, although this is an improbable situation in this application. However, once the motor nears the correct speed the speed control voltage from VR1 begins to move to within the envelope of the triangle wave produced by the previous stage. The exact position depends on the motor loading and on how much power is required to maintain the correct speed, therefore in this mode the comparator produces a Pulse Width Modulated (PWM) output whose duty cycle tends to become more positive as the loading on the motor increases. This arrangement gives excellent fine control of the motor speed and effectively cancels the noise that would occur around the comparator balance point were the circuit to be used to compare two more gently undulating DC voltages. The PWM signal is integrated back into a DC current to bias the transistor X10 by R9 and a capacitor (C2) in the original motor circuit. R10 restricts the gain of the servo as a whole and is chosen to keep the stage in PWM mode at all times – too much gain causes the circuit to try and act as a variable linear voltage source which degrades the stability of the system as a whole. R15 (JVC circuit) was removed from the motor to prevent the previous circuit upsetting the operation of the new one.

Since the output of the circuit is saturated for some of the time any noise present on the supply rail will also be present in the output. A volt or so of ripple from the motor coils was visible on the ‘scope and so to avoid an unintentional feedback loop forming R12 and C6 are used to filter the incoming supply.

In use the servo gives near instant lock with no instability at start up, as well as good immunity to load changes. Playing the 1KHz test tape at its beginning and at its end gives less than a 0.1% change in speed, an excellent result which many quite elaborate machines cannot match. Another fine cassette deck saved!

[Mikey405]

Hi Tim.

An incredibly interesting article. Or at least I'm sure it would have been if I'd read it. I managed right up to the word "Sinusoidal" before keeling over. Well done that man.

Kind regards.

From Mike.

PS.

[Lloyd 1985]

An excellent repair! This sort of thing will become more and more common as such machines start giving up and obscure parts are unavailable. I nearly bought a Technics cassette deck with direct drive motor that didn't run, but chickened out and bought a Technics M245X instead, which turned out to be a pain in the backside with a slipping belt!

I've got a similar 'direct drive' problem with a Sony PS-Q7 turntable, one of the hall effect sensors has given up, and I can't find a suitable replacement anywhere, unless I buy another expensive machine to strip for parts... I did try some that I'd pulled from some DC brushless fans, but they weren't quite right. Borrowing some from a Technics SL-5 made it run well though! One day I'll fix it...

Regards,
Lloyd.

[Nickthedentist]

You're a very clever man, Tim, that's amazing.

Nick.

[ITAM805]

An excellent and insightful repair, well done !