This is a beautifully simple circuit. The oscillator’s amplitude is demodulated, and the op-amp closes the loop by maintaining a fixed amplitude of oscillation. The op-amp is figuratively a comparator, but not literally. The oscillator never stops oscillating in normal operation. I have never looked at a device like this up close, but the oscillator-AM demodulator-error amplifier scheme is quite widely used.
Yes but the amplitude is much smaller than the amplitude present when the plate is far away from the coil. I think that this value was chosen for best performance. The position of the lever corresponding to that level should be stable against ageing and temperature.
Certainly more complicated than I would have imagined. An analog computer to measure acceleration using a galvanometer. Nowadays sensing electronics have become good enough that a pair of strain gages can do just as good a job if not better.
A pair of strain gages, yes, but also a compensating coil. The trick for making this sensor stay in calibration in spite of rather crude electronics is that it’s a nulling servo. Absolute amplitude doesn’t matter and can (and will!) drift with temperature. The calibration constant is entirely due to the amp-turns of the coil, and is affected a bit by its resistance changes slightly. Same with a flexure with strain gages: as long as there’s an actuator to zero out the strain, the electromagnetic nulling force-to-voltage relation will maintain its calibration well over time. Nonlinearities of the magnetic actuator leave only higher-order error terms in the output - AC distortion during dynamic load changes. When the load/acceleration is static, the nonlinearities are fully canceled. A very useful principle. Sensing electronics needed to do this using a strain gages sensor were no problem even in the 60s. AC excitation was used, and DC offset drift and gain drift in the strain gages amplifier were inconsequential. The operating point of the system was at zero net strain. You could make it with vacuum tubes and it wouldn’t be much worse than a modern implementation. The coil parameters determine the output scaling. With a current readout, even the coil resistance changes are of no effect. Configuring this thing for 4-20mA output, driving the coil through the current loop, and the thing is very much rock stable. The only DC calibration error sources are mechanical then. As long as the system keeps its geometry, mass and number of turns constant, the DC calibration is constant as well. Gain changes due to drift, current loop impedance, and so on, can change the AC response of course. But that’s usually easier to deal with.
There is no connection effectively and this is not a mistake! Why do you want such connection ? Everything is explained in details, at 01:45 for example.
- *Intro (**0:05**-**0:31**)*
- Linear servo accelerometer overview (0:05)
- Range: -5G to +5G (0:12)
- Manufactured by Sperry (0:16)
- Smaller than expected (0:19-0:25)
- Difficult to disassemble (0:31)
- *Teardown (**0:35**-**1:30**)*
- Cover removed (0:38)
- Internal electronic board (0:50)
- Galvanometer and zero detector (1:04)
- Copper plate for Eddy currents (1:15-1:26)
- Torque compensation mechanism (1:33-1:39)
- *Reverse Engineering (**1:54**-**3:27**)*
- Schematic explained (1:54-1:59)
- Balancing force via galvanometer (2:03)
- Zero detector with Colpitts oscillator (2:11)
- AC voltage measurement (2:26-2:29)
- Error amplifier (2:50-2:53)
- Equilibrium state (3:10-3:12)
- Output as an image of acceleration (3:19-3:27)
- *Repairing the Coil (**3:32**-**3:51**)*
- Coil connection issue (3:35)
- Attempt to solder repair (3:43-3:51)
- *Open Loop Test (**4:04**-**4:47**)*
- Galvanometer disconnected (4:05)
- Oscillation test (4:18-4:21)
- Oscillation response to plate distance (4:35-4:47)
- *Closed Loop Test (**5:22**-**5:39**)*
- 1G = approx 2 volts (5:22-5:30)
- Voltages for +1G and -1G (5:39)
- *Conclusion (**6:00**-**6:04**)*
- Thanks for watching (6:03-6:04)
This is a beautifully simple circuit. The oscillator’s amplitude is demodulated, and the op-amp closes the loop by maintaining a fixed amplitude of oscillation. The op-amp is figuratively a comparator, but not literally. The oscillator never stops oscillating in normal operation.
I have never looked at a device like this up close, but the oscillator-AM demodulator-error amplifier scheme is quite widely used.
Yes but the amplitude is much smaller than the amplitude present when the plate is far away from the coil. I think that this value was chosen for best performance. The position of the lever corresponding to that level should be stable against ageing and temperature.
Cute device! Well done as usual…
Avec un petit appareil ou un grand c'est toujours aussi intéressant, un grand merci pour le partage c'est à chaque fois un réel plaisir.
Hello, astucieux !
For an electromechanical system, it seems quite responsive.
Certainly more complicated than I would have imagined. An analog computer to measure acceleration using a galvanometer. Nowadays sensing electronics have become good enough that a pair of strain gages can do just as good a job if not better.
A pair of strain gages, yes, but also a compensating coil. The trick for making this sensor stay in calibration in spite of rather crude electronics is that it’s a nulling servo. Absolute amplitude doesn’t matter and can (and will!) drift with temperature. The calibration constant is entirely due to the amp-turns of the coil, and is affected a bit by its resistance changes slightly.
Same with a flexure with strain gages: as long as there’s an actuator to zero out the strain, the electromagnetic nulling force-to-voltage relation will maintain its calibration well over time.
Nonlinearities of the magnetic actuator leave only higher-order error terms in the output - AC distortion during dynamic load changes. When the load/acceleration is static, the nonlinearities are fully canceled. A very useful principle.
Sensing electronics needed to do this using a strain gages sensor were no problem even in the 60s. AC excitation was used, and DC offset drift and gain drift in the strain gages amplifier were inconsequential. The operating point of the system was at zero net strain. You could make it with vacuum tubes and it wouldn’t be much worse than a modern implementation. The coil parameters determine the output scaling. With a current readout, even the coil resistance changes are of no effect. Configuring this thing for 4-20mA output, driving the coil through the current loop, and the thing is very much rock stable. The only DC calibration error sources are mechanical then. As long as the system keeps its geometry, mass and number of turns constant, the DC calibration is constant as well. Gain changes due to drift, current loop impedance, and so on, can change the AC response of course. But that’s usually easier to deal with.
Looks like a mistake, the output of the opamp has no DC connection to the oscillator biasing
There is no connection effectively and this is not a mistake! Why do you want such connection ? Everything is explained in details, at 01:45 for example.
I mistaked which wire is the output. The galvo coil and electrolytic would smooth any output pulsing
If you make reference to the 39k and 6k8 resistors and associated capacitors they permit a damping of the servo control loop.
@@lelabodemichel5162 Ok, that makes sense thanks. Quite a minimalist circuit that works good.