The design uses some useful chips from Analog Devices which convert a sine-cosine resolver signal into digital position. These chips essentially do all the work for you. However, there is some weirdness which is described in the Design Problems section.
The coarse position system is as given in the data sheets for the converter chip. An AD2S99 oscillator provides the reference signal which drives the synchro rotor directly. The stator signals are then converted to resolver style sine/cosine if needed and fed into the AD2S90 converter. The signals are also fed back into the AD2S99 to synthesize the reference signal. The converter outputs 12-bit serial position data. The process is diagramed in Figure 4.
Figure 4 - Course Position System
The fine position system is very similar to the coarse system but includes circuitry to deal with the Inductosyn. The reference signal is given power amplification and the stator signals are preamplified to regain the losses in the fine synchro. The AD2S82A converter outputs 16-bit parallel position data. Figure 5 illustrates this system.
Figure 5 - Fine Position System
The coarse and fine digital position data are read by a controller. The controller runs a program which reads the data and sends them to a fiber optic transmitter serially.
The fiber cable carries the data from the encoder to the Data Receiver/Buffer (DRB) near the ACU. The DRB decodes the data stream, combines the coarse and fine, runs a small display, and provides the data to the ACU.
The coarse synchro outputs to drive the
mechanical position displays on the ACU will no
longer be available. These displays are unreliable and
inaccurate. Their function has been replaced by the
position display on the DRB.
3.1 Design Problems
The main difficulty in using an Inductosyn is the small rotor impedance combined with the low Voltage Transfer Ratio (VTR). This combination requires a compromise between driving power and usable output signal.
The VTR is the ratio of the driving voltage on the rotor to the signal out of the stator. On a typical brushless resolver the VTR is between 0.5 and 2. Due to the special nature and construction of the Inductosyn the VTR is typically between 1/100 to 1/500. The output signal must be amplified to 2Vrms before it can be used by our converter chip.
The VLA Inductosyns have a measured VTR at 10KHz of 1/177. This is the peak VTR obtainable. It drops rapidly as the driving frequency is raised and drops slowly as frequency gets below 5KHz.
The impedance of the rotor is small, about 0.5 at 10KHz. A relatively large current is required to produce a usable voltage across the rotor. This large current requires that the driving amplifier dissipate a lot of power, more power than the rotor receives. Thus, to drive the synchro directly, the power amplifier must be chosen with many times the power handling capacity than the power that is delivered to the rotor.
However, there is another way to push a lot of current through the rotor. We build a tank circuit with a natural frequency of 10KHz around it, and use the amplifier to excite the oscillation. We can get away with this because our converter synthesizes its own reference. Controlling the phase of the rotor oscillation is not important. We can let it resonate and reconstruct the reference from the output.
By doing this we can use inexpensive power amplifiers and run them cool and at low current levels. The amplifiers and power supplies can be smaller and cheaper. They will be more reliable due to the lower operating temperatures.
There are some considerations for the amplification of the stator signal. The conversion scheme is somewhat resistant to noise, but to get 16 bits of resolution we need about 85dB SNR. When the signal gets through the Inductosyn it is reduced by the VTR to a small level, typically 8 to 50mV. Thus we want to keep the noise down to 0.4 to 2.8µV.
As the gain of the preamplifiers is increased, the noise level rises. This limits the number of usable bits of resolution, and forces the compromise between rotor drive power and preamplifier gain.
We used some techniques to reduce the noise level. The Inductosyn stator is a low impedance source, and the amplifier has very high input impedance. The preamp circuit is therefore very sensitive to magnetically coupled noise, and almost immune to electric field noise. The manufacturer of the Inductosyn knows this and provides twisted pair wire for connections to minimize magnetic coupling. This is a balanced two-wire signal and to maximize noise immunity we must preserve the balance.
We also need to limit the bandwidth of the signal. The low stator impedance makes it difficult to simply strap a capacitor across it for low-pass filtering. I had to add two 47 ohm resistors in series (both sides to preserve balance) to artificially increase the impedance and to dampen the Q of the stator-capacitor circuit. Using a good low-inductance 0.1µF capacitor I achieved very good noise rejection starting at 12KHz. The inductance in the input circuit destroys the rejection above about 5MHz, but the frequency compensation and limited bandwidth of the preamplifiers rejects any signal above about 1MHz. The output of the amplifier is AC coupled to reduce low frequency noise (for example, 60Hz line noise) and DC error.
To limit the magnetic coupling sensitivity we should use a relatively low input impedance amplifier. Hence the choice of a bipolar input amplifier rather than a FET input. Differential amplifiers preserve the balance of the input signal. An instrumentation amplifier on a chip is a good choice for a differential amplifier due to the high common-mode rejection, low bandwidth, simplicity of use, low external parts count, and low price.
To improve stability and noise immunity I placed both instrumentation amplifiers close together on the PCB and minimized the signal path length. I used linear power supplies to try to limit the nearby noise sources. However, the worst noise problem will always be the reference drive to the rotor. It is physically close to the preamps, it is not affected by the filtering, and has large currents creating magnetic fields.
I managed to find compromise between the drive power and preamplifier gain. The rotor drive amplifiers would run cool to the touch but provided enough signal-to-noise on the output to give 23 usable bits of resolution. I did this because our primary goal is reliability, and we won't be able to use the extra bits until the ACU is upgraded. If the full 25 bit resolution is desired then the rotor drive amplifiers will require heat sinks, but will still run well within the Safe Operational Area (SOA).