This project started due to my curiosity about the gyroscopic effect - where a rotating object generates a torque if a torque is applied along an axis perpendicular to the rotation axis. The general property behind this is that a rotating object tends to maintain its rotation axis in an absolute sense, and thus can be used to detect motion such as the turning of a vehicle, or even revolution of the earth (Foucault pendulum), and furthermore can be used to find true north by aligning with the earth's rotation axis (gyrocompass). Energy, linear momentum, and angular momentum are conserved in this process, so it is not possible to use it to make objects levitate. However, it is curious why this effect arises. Angular momentum is the only representation of an absolute physical space I know of, since all observers can agree on whether an object is spinning based on measurements of their centrifugal acceleration. In this view, once a gyroscope is spun up, it becomes 'locked' in a particular absolute orientation and can be used to track future angular rotation (including rotation of the earth, such as true north finding gyrocompass). This seems to be valid even in gravitational (curved space) interactions, since gyroscopes are routinely used on satellites for navigation. To explain why this 'locking' happens, it is insufficient to be content with "conservation of angular momentum" - the real explanation must be in the deeper nature of matter and space, and beyond the scope of this article. Instead what I attempted to achieve with this experiment was an electrically operated gyroscope that could be used for future experimentation to test the unintuitive physics behind the gyroscopic effect. The starting point was an old navigational gyroscope I bought on ebay.
The gyroscope was quite well made, with no magnetic parts in it (using aluminum bolts for instance) since it was used as a base for a magnetic compass (flux gate compass). Mounted on a gimbal, the gyroscope itself consisted of a soft magnetic core (I guess the motor noise would have to be filtered out from the magnetic compass) and a fairly heavy nonferrous metal rotor. The motor is a cage-less induction motor (that is, the rotor is a solid piece of metal) which helps with manufacturing a highly balanced assembly as required for operation at high speed. Even with this, there were indents drilled in specific locations on the bottom and top of the rotor, assumed to be for dynamic balancing of the rotor after initial machining.
The problem, however, was powering up the gyroscope. Four of the wires coming out of the assembly were for the magnetic compass, and three were for the gyroscope motor. The motor had two capacitors (in parallel, so acting as one capacitor) always connected, which I assumed meant a capacitor run motor. Checking resistance of the phases confirmed this, and at this point I knew which two wires to connect to a one-phase AC supply to power the motor. However, as reasoned from its intended marine application, the induction of the phases, and the capacitance, the motor required 400 Hz and 120 V to operate. The closest thing I had was an industrial motor controller which went up to 240 Hz, and the motor worked successfully at 120 V. However it was clearly operating below the rated speed.
It turned out that finding an adequate power supply was quite a challenge. My initial attempt at making one involved using two permanent magnet motors for frequency conversion - one was a brushed DC motor, operating from a regular DC power supply, while the other one was a brushless DC motor which would output a sinusoidal waveform directly from one of its three phases. The two motors were coupled without gearing (1:1), and the 400 Hz frequency was achieved by regulating the DC voltage applied to the brushed motor and a factor of 14 multiplication from mechanical to electrical frequency due to 14 magnetic poles on the brushless motor. Thus the shaft needed to spin at about 29 Hz.
A motor-motor coupling to convert from DC to AC voltage mechanically (this is still done in some industrial applications where transistors to handle high currents would be more expensive!) Brushed motor is on the left, connected to the brushless motor on the right. Two of the three wires are used to select a single phase out of 3 on the brushless motor. Incidentally, this is not a good mechanical design for a coupling because it is too stiff (requiring perfect shaft alignment), and thus the motors were mounted with foam padding.
The output of the brushless motor would be hooked up to a transformer, to step up the voltage to 120 V as required. This approach worked, to some extend, but was nonetheless troublesome. The first issue was that the brushless motor output, being based on the strength of permanent magnets which was not adjustable, had a voltage proportional to frequency and also affected by the load current. Since the frequency requirement was already set, this meant that a step-up transformer (and the gyroscope motor itself) would have to be properly matched to operate at a particular input voltage. Of the large number of transformers that I tried, none would output a high enough voltage under load (although I was able to get 120 V at no load). This points to the second issue, which is that most transformers I was able to find were designed either for 60 Hz operation or some kHz in high frequency converters. Thus the magnetic properties of the 60 Hz transformers would tend to attenuate the 400 Hz signal, while the high frequency transformers would present an effective short circuit at the lower frequencies. This is likely why the voltage dropped significantly under load. It was necessary to find a transformer that would be lower inductance than typical 60 Hz but higher than the ferrite ones. A variable autotransformer (variac) can be used to manually set the inductance of the output voltage, with a fixed inductance primary. This can be used in reverse, where the lower-inductance and lower-voltage secondary is connected to the generator output, while the primary is used to output a higher voltage. This setup allowed adjusting the primary/secondary ratio to get the most power transfer, which achieved 50V under load on the primary output. This was sufficient to spin up the motor slightly, but since it relies on the strength of the magnetic field to spin up (this increases the amount of magnetic 'drag' between the stator and rotor) the low voltage output significantly limited the effectiveness of the system. Furthermore, this approach was inefficient, due to losses in the brushed motor windings as well as hysteresis losses in the transformer. As a demonstration of the losses typical in a poorly chosen transformer, I connected a 4:1 transformer to the primary output of the variac, and the output voltage of that under load was even lower than connecting the variac directly! So, while it would have been very satisfying to have a mechanical frequency converter, this option did not seem to be working properly.
With the available transformer choices, the only way of getting a high enough output voltage was to have separate control of the frequency and the inverter voltage going into the transformer. This can be achieved by a transistor-controlled oscillator. I had an extra 60V MOSFET from a brushless motor controller I was working on, and decided to build an oscillator based around that. I used a simple 555 timer circuit to generate a nearly 50% duty cycle 400 Hz square wave, which was fed directly to the gate of the MOSFET. The particular 555 chip handles up to 15V, and was run at 10.5V to allow for overvoltage during the switching transients (confirmed by oscilloscope). With 10V at the gate the MOSFET is well into the conduction region, although the 555 is fairly slow at switching, taking around a microsecond. The rest of the oscillator is an LC tank made from a transformer that seemed to work best at 400Hz and two 10uF capacitors wired in series for 5uF, since this combination seemed to draw the least amount of current when loaded (resonant frequency close to the operating frequency of around 450 Hz, I have no measurement of the transformer inductance so no way to actually check).
Prototyping the inverter on a solderless breadboard, and then soldering it onto a small perforated board. The huge capacitors on the board help stabilize the 30V input voltage and the 10V control voltage (which are subject to transients not only from driving the gate of the MOSFET, but also from driving the LC tank, and from the full-wave rectifier).
A full wave rectifier is connected to a 4:1 control transformer, resulting in 120/4=~30 VDC. (There is even a fuse in there! This is totally up to code!) This powers the oscillator and drives the motor. About 0.5 A is drawn while spinning up, which is approximately 15W dumped into the motor. This compares favorably with the 0.1 A at 120V done with the industrial motor controller.
One side of the tank is connected to a 30V DC power supply, while the other side is connected to the MOSFET source. When the 555 timer outputs a high signal, the MOSFET connects the tank to 0V, and on a low signal that is disconnected and the tank allowed to oscillate. The oscillation creates an approximately 34V AC waveform (some voltage rise due to the resonant frequency of the tank, which is desirable), which the transformer converts into approximately 92V AC at 450 Hz. While this is still not 120 V, it is close enough to operate the gyroscope quite effectively, and the faster frequency (450 Hz instead of 400 Hz) will likely lead to a higher attainable speed. Strangely, the voltage as measured on an oscilloscope is over 200V peak to peak, and the discrepancy on the meter may be due to the square wave shape. In any case, the motor accelerates similarly to when it was connected to the proper motor controller at 120V (but attains a twice higher maximum speed - note theoretically it should accelerate slower from a standstill at the higher frequency), showing that the circuit is effective in its application. The circuit diagram is shown below:
The circuit diagram for the single MOSFET based inverter.
The final assembly was mounted on a board and all connections soldered together:
Testing all the components by connecting everything with alligator clips. The parts will subsequently be mounted on a wood board.
Inverter final assembly, stage 1 image on left includes (left to right) the 120VAC to 30VDC converter, oscillator board, and two huge 5uF (2 kV? They were lying around (probably since the 80s) so I used them.) capacitors used in the LC tank. Stage 2 image on the right shows the huge capacitors again, then the 400 Hz 1:4 transformer used in the LC tank and increasing voltage to ~100VAC at 400Hz. This is connected to two capacitors in series which makes an out-of-phase waveform to drive an auxiliary coil in the motor (a capacitor-run induction motor). On the right is the gyroscope motor itself, with the rotating parts enclosed.
Another view of the completed assembly.
The spinning gyroscope behaves in an interesting way, as can be seen in this video. The inverter has worked successfully for many spin-ups, but the MOSFET does require a heatsink to not overheat. This is strange given that it is running at only a fraction of its rated capacity, and perhaps indicated a problem with switching speed or switching voltage; it is also possible that the transient spikes (as well as resonant voltage rise in the LC tank) on all pins of the MOSFET exceed its voltage ratings and cause increased heating. However everything seems to be working fine for now, and taking into account that most of the components used here were found in various scrap piles, I would consider this project a success.