Cosmetics and Physical Construction         views:

Some SP10 owners like to, or want to, dispense with the black cast alloy 'bathtub' that shields & protects the underside of the SP10. I'm keeping mine: I believe it adds helpful electromagnetic screening, and gives the top chasis added rigidity.

After acquring the real SP10 chassis from Shaun O. in Cape Town, I needed to restart my PCB layouts to make the boards fit inside the bathtub, and match the SP10 chassis' mounting holes. I changed my mind concerning the power supply: there is no space to fit a full power supply inside the bathtub, so I'm going with the original design concept of an outboard power box. As described previously, the power supply consists of a transformer and circuitry for five regulated DC outputs: +15V, -15V, +7.5V, -7.5V, and +5V TTL logic. All linear regulators - general switching regulators are too noisy for use here, and making them sufficiently quiet is just too much work & expense.

The power supply generates more heat than the other circuitry, so it is better off, if less convenient, in a separate enclosure where it will not heat up the platter.

The inboard circuitry consists of a 3-phase drive system (=power amplifier) and a PLL controller (2 printed circuit boards). The user interface (7-segment LED display and keypad) will be external, since I do not want to cut into the top chassis. The original 3 pushbuttons - start, 33, 45 & 78 - and speed LEDs will be retained.

Here is a photo of some mockup pc boards, cut from cardboard, to show how they will fit into the chassis.

Originaly, the SP10 came with a wired-in umbilical cord to connect it to its power supply box. I'm using a 25-pin "D"-connector to bring power into the chassis, and provide serial data in/out for control purposes. (N.B. European DIYers - the D connector hex-standoffs are NON-METRIC. Don't try to use a 3mm thread here!)

The power supply box can be any convenient or aesthetically pleasing instrument case or housing. I've had a broken 'Dantech' generic power supply in my junk box for years: its transformer is still good though and its housing will make a neat PSU for the SP10. A completely new circuit board is required (9"x4") The PCB layout is shown below (as yet un-built).

Sharp-eyed readers will notice the 'psu-on' line on the I/O connector. This allows remote on/off switching of the PSU from the SP10 pushbutton pad - or any other exotic control method that comes to mind (IR?!).

Note that this power supply is NOT a drop-in replacement for the original SP10 with its factory electronics. It doesn't have the appropriate grounding system - although it could be modified to be compatible with the original if someone required a replacement PSU for original SP10 electronics.

The three-phase power amplifier, exciter oscillator and phase envelope detectors, plus PWM and phase sequence reversal switching PCB layout is shown below, (as yet un-built).

These boards have been specifically laid out (manually routed, no autorouting!) with a view to DIY contruction & manufacture: single sided, wide traces, wide component spacing, no surface mount components, no traces between IC pins except in 'emergencies'.

The board size is limited (the left edge of the circuit board is blank) so it will fit on less than one-half of an A4 sheet of 'Press-&-Peel' toner transfer film in portrait aspect.

The heatsink for the power amp stage can be any convenient aluminium panel or strip that is physically compatible with the chosen case or plinth. I selected a 50mmx50mm "L" extrusion since I had it left over from another project. 28cm (11 inch) lengthwise just fits in the SP10 bathtub. The L165s will not dissipate a great deal of heat, probably less than 5 watts under normal running, but my philosopy is that with any heatsink, bigger means cooler and hence better, so "L" angle is better than a flat strip. This allows good thermal headroom for coping with 'problem' situations, such as a stalled motor.
A crescent portion in the centre will have to be cut away to accommodate the curve of the motor casing.

A second length of 20mm "L" extrusion - bolted to the heatsink - forms a mounting flange for the circuit board, and mounting surface for the three L165V semiconductors' TO-220 cases.

The PC board I made using Press-'n'-Peel Blue laser printer film. I prefer phenolic board because it's easier to work than fibreglass.
The pins that mate with the MJX-12A motor connector are harvested from a 25-pin male "D" connector: these are soldered directly into the circuit board.

February 13^th 2010:

Below, the the drive system is complete!
The white flylead is a temporary link to make up for a ground commoning that exists in the custom power supply, but not in the test-bench power supply. The yellow/blue twisted wiring carries supply voltage to the 50kHz oscillator so it can be powered and tested separately from the power op amps.
The tiny Veroboard on pink & blue flyleads is for temporary DC balancing: until I get to measure the motor running, I have no idea what sort of DC offset magnitudes will exist, and so component values are unknown. Correct automatic DC balancing will allow the elimination of the three small blue electrolytic capacitors, which are there to protect the motor from DC latchup during testing/prototyping.
The power op-amps are on temporary individual heatsinks, and the bigger supply decoupling capacitors are not yet installed as the temporary heatsinks get in the way. Three 10R 1W resistors temporarly substitute for wire links, to protect the motor from overcurrent during testing.

The circuit in operation! Since the PWM and direction-control chips were not yet inserted when this movie was shot, the rotation was CCW, and speed was uncontrolled.

MOVIE - 1.2MB

May 17, 2010
The PLL subsystem design is now complete, in prototype form (see page 4 for basic circuit and its description). I changed my original plan, and the microprocessor is now included on this board (U5) along with the quartz crystal (X1). *IF* this prototype board works, then it together with the power output board shown above, will form a complete working SP10 system. The PCB layout is shown below (click for full sized):

- a monumentous occasion - I have achieved stable, quartz PLL lock!

The problem I've been having during the past few days (the PLL hunting either side of locked, appearing as if it was about to lock, and then suddenly surging) turned out to be noise on the tacho signal - spikes of extremely short duration, far too brief to show up on my oscilloscope with a timebase suitable for the 105Hz tacho signal. A single 1nF capacitor across the tacho comparator output fixed that.

The oscillograph shows the quartz oscillator signal (bottom) at 105.55Hz, from which the trace is triggered. Above it is the processed tacho signal. Not obvious from the static photo is that these two pulse trains remain absolutely motionless under undisturbed conditions. The relative phase-offset of the two traces gives rise to the error signal that drives the servo into corrective action. Applying drag to the platter by touching it results in the top wave (tacho) shifting slightly to the side, and then, when the finger is removed from the platter, the wave returns to its shown position. If too much drag applied over-burdens the motor, the PLL unlocks and the trace runs uncontrollably.

The design is not complete though.

These are the aspects I still need to work on:

• Currently I am using the NE555's control votage input pin to vary the motor drive power. This works by shifting the ¹/₃V_cc and ²/₃V_cc reference points at which the two comparators switch. Because they are internally referenced within the 555, there is a limited span over which the PWM duty cycle can be artificially pulled, that being around 0.65x up to 1.35x nominal. This is totally OK for speed regulation, but it restricts the servo loop from pulling the motor drive voltage right up against the supply rail for fast, high torque start. So, outside of lock range, the servo idealy needs to be able to push the PWM duty cycle right to the upper limit.
  
  This change is illustrated here, on page 4. Modifications are in red.
  (1) The PLL error signal from the loop filter now becomes the offset voltage for the constant current source (the CCS determining the charge rate for the PWM timing capacitor). At this node, the PLL error signal has much greater effect on the charge rate.
  (2) The PWM offset point is now injected as the DC operating point of the loop filter.
  (3) The control voltage input of the NE555 is not used.
  Scratch that above in small print! It was an interesting theoretial diversion, but in reality, the maths shows that controlling the NE555 by means of its charge-pump current is inheremtly too non-linear. I'll go back to the original very linear method of modulating the comparator threshold voltage, and achieve the necesary range by upping the 555 supply voltage to +15V. And I have a few other level clamping 'tricks' in mind to give fast start.

• The time constants of the loop filter and lead/lag conpensation are not optimised, degenerating the lock-up time to some degree, an allowing a small degree of oscillation about the nominal speed. The oscillation is small, a fraction of one tacho pole ( about 10% of ¹/₁₉₀ ^th of a revoltion, or 0.18 °), but I'd like it eliminated.

• It may be posssible to eliminate the PWM oscillator altogether and drive the motor directly from the square-wave derived from the phase detector. My current PWM drive operates at 5kHz, and the phase detector only generates a 105Hz signal. I don't know if 105Hz is going to be fast enough for direct excitation of the motor - the steep filter required may compromise the PLL loop bandwith, which was exactly why I chose such a high freqency of 5kHz for test purposes.

• My phase detector uses a 4-bit shift register, the CD4035. While not exactly rare, this is now an 'old' IC. The same function can be performed by a CD4013 D-flip-flop, which is a cheaper, more ubiquitous component. The advantage of the shift-register is that it can directly give a 'PLL locked' status signal, to illuminate an LED.

• Pitch control needs to be added. This is essentially just adding a programmable divider to the quartz oscillator, but it may also require that the DC levels and loop gain be changed when the setpoint speed is changed, to keep the PLL well within its lock range.

• I have no accurate test disc for measuring the wow/flutter performance of this system. I have noted a degree of jitter on the tacho signal, that is inherent in the tacho pickup itself, owing to non-uniformity of the teeth in the pole-piece stamping. I thought this would mean measuring W&F directly from the tacho signal would be of little value, but quantifying mesurements show the tacho error is way below anticipated wow/flutter figures of the best turntables, and even lower still that the wow/flutter achievable on calibration LPs.