Foreword
This allows designers to have 26-bit resolution, or less than 5.36 × 10⁻⁶ degrees of rotational position error accuracy when implementing the solution.
Is it rotating at the expected speed? Closed-loop motor control systems will continue to answer this question, as closed-loop systems are implemented wherever there is motor rotation—this is a trend. Whether the end system is a car (using computer-controlled steering for parallel parking), a satellite (adjusting satellite angles to lock onto specific signals), or factory machinery (handling machines), position feedback sensors are an inherent component of the overall motor control system. There are many types of motor control; this article will discuss two control schemes that implement analog signal chains around position sensors: resolvers and encoders.
Decomposer
Before discussing the resolver signal chain solution, let's first consider its basic operating principle, as shown in Figure 1. The resolver (here, a transmitter unit) consists of three distinct coil windings: a reference winding, a sine (SIN) winding, and a cosine (COS) winding. The reference winding is the primary winding, which is energized by an AC voltage applied to the primary side of a transformer called a resolver. The resolver then transmits the voltage to the secondary side of the transformer, thus eliminating the need for brushes or collars. This improves the overall reliability and stability of the resolver.
The reference winding is mounted on the motor shaft. As the motor rotates, the voltage output of the SIN and COS windings changes with the shaft position. The SIN and COS windings are mounted at angles 90° apart relative to the shaft. As the reference winding rotates, the angle difference between the reference winding and the SIN/COS windings changes, which can be represented as the rotation angle θ or θ in Figure 1. The voltage induced in the SIN and COS windings is equal to the reference voltage multiplied by the angle θ of the SIN and COS windings.
The induced output voltage waveform is shown in Figure 2. The figure shows the normalized voltage output signal obtained by dividing the SIN and COS windings by the reference voltage. The conventional reference voltage is typically between 1 and 26V, while the output frequency range is 800Hz to 5kHz.
The requirements for the appropriate signal chain components can now be determined. The signal chain must be bipolar because the signal will swing below ground (Figure 2). It must sample both channels simultaneously, convert signals up to 5kHz, and provide AC voltage to the reference winding for the resolver. The optimal solution is to implement a Δ-Σ modulator for each channel. Δ-Σ modulators can sample at extremely high frequencies (in the 10 to 20MHz range), therefore the output after Δ-Σ modulation needs to be balanced and filtered to achieve acceptable resolution.
When providing a reference voltage or AC excitation power supply, the preferred approach is to apply a pulse width modulation (PWM) signal directly to the resolver. Texas Instruments (TI) offers a recommended solution for this implementation. Data converters (such as the ADS1205 or ADS1209) are preferred for Δ-Σ modulators because both can be directly connected to the SIN and COS windings of the resolver. Additionally, the data converter can be connected to a four-channel sinc filter/integrator to generate a PWM signal output for the reference winding, such as the AMC1210. Finally, a digital signal processor (DSP) or real-time controller is needed to handle various signals other than those from the motor control system. TI's C2000/Piccolo/F2806x microcontroller, based on the C28x series, is a suitable option. Figure 3 shows a typical signal chain solution.
In summary, the resolver is a very stable position sensor for control systems, supporting not only high accuracy but also a long service life. The drawback of the resolver is its maximum rotational speed. Since the resolver signal frequency is typically less than 5kHz, the motor speed needs to be less than 5,000 revolutions per minute.
encoder
Similar to the case of resolvers, before discussing signal chain implementation schemes, it's essential to understand the physical characteristics and signal output properties of encoders. Encoders are generally of two types: linear and rotary. Linear encoders are used for schemes involving movement in only one dimension or direction, converting linear position into an electronic signal; they are typically used in conjunction with actuators. Rotary encoders are used for schemes involving movement around a central axis, converting rotational position or angle into an electronic signal. Since rotary encoders are used in conjunction with motors (which rotate around a central axis), linear encoders are not discussed in this paper.
To understand the principle of a rotary encoder, we must first consider a basic optical rotary encoder. An optical encoder has a disk that supports specific patterns, mounted on the motor shaft. The patterns on the disk can either block or allow light to pass through. Therefore, a light transmitter and a photodetector are also needed. The signal output of the receiver can be correlated with the rotational position of the motor.
There are three common types of rotary encoders: absolute position value, incremental TTL signal, and incremental sine wave signal. For absolute position value rotary encoders, the patterns on the disk can be divided into very specific patterns according to their position. For example, if an absolute position encoder has a 3-bit output, then it will have eight different patterns evenly distributed (Figure 4). This is on the disk and is evenly distributed, so the spacing between each pattern is 360°/8 = 45°. Now, for a 3-bit absolute position value rotary encoder, the position of the rotary motor within a 45° range can be determined.
The output of the absolute position rotary encoder has been optimized for a digital interface, so no analog signal chain is required.
For incremental TTL rotary encoders, the mode output on the disk is either a high or low digital signal, i.e., a TTL signal. As shown in Figure 5, the mode of a TTL output disk is simpler compared to that of an absolute position rotary encoder, as it only needs to represent a high or low digital signal. In addition to the TTL signal, there is a reference mark that is crucial for determining the current rotational position of the motor. This reference mark can be considered as a 0° angle. Therefore, the exact rotational position of the motor can be determined by simply counting the digital pulses.
Figure 5 shows multiple cycles in one rotation of the motor shaft. Encoder manufacturers offer incremental TTL rotary encoders (and incremental sine rotary encoders) with 50 to 5,000 cycles per revolution. Similar to absolute position rotary encoders, the output is already in digital format, thus eliminating the need for an analog signal chain.
For incremental sine rotary encoders, the output and disk mode are very similar to those of TTL signal encoders. As the name suggests, its output is not digital, but a sine wave output. In fact, it has sine and cosine outputs as well as a reference marker signal, as shown in Figure 6. These outputs are all analog signals, thus requiring an analog signal chain solution.
Similar to incremental TTL output, there are multiple signal cycles in one rotation. For example, if an encoder with 4,096 cycles per rotation is connected to a motor rotating at 6,000 rpm, the resulting sine and cosine signal frequencies are calculated as follows.
The signal chain solution in this example requires a bandwidth of at least 410kHz. Since this is a closed-loop control system, latency must be minimized or eliminated entirely. Typically, the encoder output is 1Vp-p, and the sine and cosine outputs are differential signals.
Typical requirements for analog signal chain solutions are:
Two analog-to-digital converters (ADCs) that sample simultaneously: one for sine wave output and one for cosine wave output.
No system latency: Requires bandwidth of 400kHz or higher, therefore
The ADC must be able to handle a minimum rate of 800kSPS per channel.
• Supports 1-Vp-p differential input with a full-scale range of approximately 1V, which can optimize the full-scale range of the ADC or amplify the input signal within the full-scale range of the ADC.
• A reference marker signal comparator.
TI's best solution is the ADS7854 series of successive approximation register (SAR) ADCs (Figure 7). This SAR-ADC features two synchronous sampling channels, an internal reference, and a 1-MSPS output data rate per channel to meet specific needs. It can be used in conjunction with comparators and fully differential amplifiers to drive the ADC.
The ADS7854 is a 14-bit ADC. If a sinusoidal incremental rotary encoder has 4,096 cycles in a single rotation, the total number of measured steps can be calculated as follows.
This provides designers with 26-bit resolution, or less than 5.36 × 10⁻⁶ degrees of rotational position error accuracy when implementing the scheme.
in conclusion
There are two common implementation schemes for rotary/position sensors in motor control feedback paths: resolvers and encoders. We evaluated the feedback paths and output signal characteristics of several control systems from an analog signal chain perspective, focusing on resolvers or encoders, to ensure signal integrity and optimal performance.
