PC board interconnection diagram: click for bigger.
|
|
The following circuits I drew 'out of my head'. They are theoretical designs mostly - individual sections have been tested on a breadboard.
Below is a proposed interconnection diagram, showing the different components and how they will interlink.
|
|
The power supply is a standard implementation of the LM317t and LM337t regulators. These are used because of their capability for external over-current sensing. A simpler 7805 is used for standby and the digital section's power.
I have selected a relay for controlling the on/standby switching. I rather prefer this 'hard break' in the current path in place of a solid state switch, although it's long term reliability may be lower. The microprocessor supply (+5) is always on, allowing IR remote power up, or other functions - such as talking the ReVox "Easy-Line" multiroom protocol.
Several readers have asked about the issue of 110V / 60Hz opertion (My own version is 220V / 50Hz). The circuitry I'm creating here requires a symmetrical DC supply of -15/0/+15, or thereabouts, plus an auxilliary supply for standby logic, of 5v. My power supply uses a centre tapped transformer (15-0-15) or a dual-secondary (15+15), with an additional low voltage winding. If you cannot find such a transformer with 3 windings, two independant transformers are quite OK.
The primary winding can be whatever you want it to be - whatever is appropriate to your area. The AC line frequency is of no consequence at all. Torroidal transformers are nice (especially since you can easily add your own auxilliary 5v winding), but there is no reason an EI or C-core cannot be used with equally good results.
The circuit will also run quite nicely on a symmetrial battery supply - four 12V lead-acid accumulators.
I have now (November 7th) combined all circuit components previously posted here, into a single circuit diagram (schematic) down below. This diagram corresponds to the PWM+power stage described on Page 6 - the contruction details page.
The drive circuit consists of 3 power op amps, which can be configured as inverting, or non-inverting, simply by omitting certain resistors. (This is why there seem to be so many resistors in the feedback paths).
Each power stage is preceded by a pair of clamping diodes, and then folows an LPF filter - the filter removing any residual square-wave component of the PWM drive signal and other HF noise. <NZFB> is a signal derived from the motor's star-point, and adjusts the 3-phase drive symmetry, compensating for any mismatch in the motor's 3 windings.
The phase pickup coils exciter oscillator is a 50kHz Wien bridge configuration. Germanium diodes are used for gain stability, rather than the more usual incandescent lamp or thermistor, as diodes are more readily available. The original SP10 uses a Hartley oscillator, which relies on a custom wound inductor - very difficult to source, and even to DIY wind since a suitable ferrite core is not widely available.
The AM envelope demodulator recovers the 3 120° AC sinusoids from the ultrasonic signals picked up by the motor's three phase pickup coils. Functionally, this is a precision full-wave rectifier followed by a peak-hold amplifier. The PWM modulator (3 phase variable gain cell) multiplies the 3-phase AC waveforms by the square-wave's duty cycle, to effectively achieve armature control of the DC motor's speed, in a precise and drift-free manner.
The PWM multiplier / gain cell is succeeded by an integrator to smooth out all square wave components. This can be configured as a single pole or two-pole filter - I have not yet decided which will work best.
Subsequently, there is the phase sequence reversal switching, which will allow counter-clockwise rotation of the platter and 'electronic' braking. The signal <PWM> is generated by the microprocessor, as a square-wave speed control signal, and <CCW> is the reverse rotation logic line.
The tacho coil voltage varies from around 8mVp-p up to 45mVp-p, as the speed changes from 16RPM to 78RPM. I had wanted to use the tacho voltage amplitude to derive a velocity feedback component, but the voltage is amplitude modulated with the rotational frequency, so that plan couldn't work. The tacho signal is amplified first (since 8mV is a small signal to work with) and then put through a comparator with hysteresis (Schmitt trigger) to convert it to a constant amplitude square wave. The output of the test circuit looks like this. (105Hz, 17vp-p)
|
|
The Wien bridge oscillator above gives very satisfactory results, but low distortion requires accurate gain control. This is easy enough to set with an oscilloscope, although a little more difficult without such test equipment. I looked at oscillators of the phase shift type, where no stabilising diodes are required, but even with gain trimming, I could not get as good a sinewave as the Wien Bridge circuit gives.
Interesting Trivia: Max Wien 'invented' the phase shift network - the Wien Bridge - in 1891. But the triode amplifier wouldn't exist until 1906, so Wien's bridge couldn't be used in a practical oscillator. Its widespread application came only in 1939 when William Hewlett designed the HP200 test bench oscillator, HP's first widely successful product.
The above circuitry shows the motor speed control by the signal "PWM", which ultimately will be derived from a microprocessor. That implies that this section cannot be tested (or used) without the microprocessor and its software. Ugh! To make the power drive stage immediately usable, with no 'digital' circuitry attached, a quick and simple PWM modulater can be put together using an NE556 (= dual NE555) timer IC.
|
|
The NE555 circuit will be used to generate the motor speed control voltage. I may even decide to retain this right to the end in one variant of the entire drive system in order to have a controller system that does not use any microprocessor - for those purists who want no frills or exotic ICs such as a microproceesor: few chips are more common, cheap & widespread than the 555!
This circuit is newly updated as of Feb 19th, 2010. The original circuit I posted here came direct from the Philips 1988 Databook Application Note 170, but I'm pretty sure it will never work as published; modifications didn't give satisfactory results, so I came up with the circuit shown above.
The pulse generator circuit comprises two 555 blocks in a single NE556 DIP IC. The first section based around Ic1A / R1 / R2 / C1 is a square-wave oscillator that generates a fixed frequency (5kHz), fixed duty-cycle (around 0.65) pulse train.
The second section (Ic1B / Ic2) is a linear voltage controlled monostable (=one-shot) which varies the duty cycle in response to the voltage developed on the wiper of potentiometer R4. The usual PWM configuration of the 555 is not very linear (because of the exponential charging character of a capacitor), but replacing the charge resistor with a voltage dependant current source straightens out its response curve.
The signal 'pwm-out' is clamped by a Zener diode to constrain it to TTL level.
EXPERIMENTAL RESULTS
I'm liking this circuit a lot! Varying the reference voltage (IC2 pin 3) gives a variable PWM ratio of around 7,5x. This is adequate for speed control over the entire range of 14RPM up to 100RPM. BUT, it's pushing the stability limits of the current source at the extremes. My prefered solution is to select a different value of R16 for each speed (CD4016), use R4 as a calibration trim, and drive Ic1B's pin 11 (Control Voltage) with the servo/feedback error signal. Since the CV pin modifies the reference point of the 555's internal comparators, it provides a very linear means of PWM control.
Below is the Pulse Width Modulator circuit including the phase-frequency detector (PFD). The PFD generates a DC control voltage proportional to the speed error. This voltage is converted into a fixed-frequency, variable duty-cycle pulse train which is passed to the CMOS FET switches on the Power Amplifier Board, where it modulates the 3 phase drive voltages.
The signals are as follows:
<QUARTZ> is the divided down squarewave signal from the quartz crystal, <TACHO> is the processed tacho head pickup signal, present on the power driver board, <LOCKED> is a logic signal that shows when the PLL is locked, can be used to drive an LED or be monitored by the microproessor.
The signals <TTL 33>, <TTL 45>, are logic inputs from the speed selector block. These shift the control system DC level into appropriate operating regions: the nominal speed is hereby set, and the PLL controls the system about this nominal value. This method puts the least burden on the servo system.>
|
"Standard" speeds for records are an approximation. A record marked 45RPM is generally only cut at 45.00RPM in 60Hz line frequency countries (USA, Japan, etc.) In 50Hz line frequency current countries, (Europe, Australia, South Africa, etc) the cutting speed is 45.11RPM Similarly "78" in 60Hz countries is 78.26RPM, and in 50Hz countries, 77.92RPM. Early "78s" were not even cut at a standard speed, and different companies used different speeds - anything from 72 to 86. Victor standardised at 76RPM, Columbia at 78 and Edison at 80. This is somewhat academic, since the speed accuracy of many cutting lathes is not that accurate. If the disc is cut from an analogue master tape, the speed error of that also adds to the overall error. Lastly, a quartz referenced turntable for playback cannot either EASILY be made to run at EXACTLY 33.333RPM, 45.0RPM or 78.26RPM, since one is obliged to use standard frequency crystals, and divide their frequency by a whole number. Nonetheless, the errors are small - less than 0.01% Luckily, for 33.33RPM on which most good quality records are available, the errors are smallest, being typically 0.002%. Even if one wants to make the speed absolutely accurate, no 'ordinary' person has easy access to measurement equipment that can reliably show such small timing errors.
For my prototype, I had thought of the Texas Instruments CD4095A divider chip - designed for synthesizer communications receivers. It allows frequency division by any number - within certain restrictions. This chip is still widely available, but it will one day become obsolete since programmable microcontrollers can do the same job and do it with more versatility. The 4059 provides an 'instant' solution with no programming required. One annoyance is that at usual crystal frequencies, for 16.66RPM it does not count high enough, and an extra divider stage would be required. It's also a relatively costly chip.
|
|
|