Eliminating the reel motor position sensor in a tape drive.

This article is the second in a two-part series. The first part appeared in the December issue of CTR.

There are several methods that can be used to eliminate the sensor. Unfortunately, none of them work very well for tape drives. Some of these methods include back-EMF sensing, the addition of a sense winding in the motor, or measuring the impedance change of the motor windings by injecting current pulses or a high frequency sense carrier. All of these techniques have been suggested and some have actually been used in a few applications. In the following paragraphs, however, we describe these procedures and discuss the reasons why none of them are suitable in tape drives.

Back-EMF sensing measures the voltage across the unused winding in a typical brush-less DC motor with a three-phase, Y-stator configuration (Figure 1). Current is switched to two of the three windings at any time. As the motor rotates the drive windings and the sense windings are switched. The signal amplitude measured at the unused winding is proportional to the motor speed. This simple method works well when the motor turns at a constant speed. It does not work at all, however, when the motor turns slowly or is stopped. For tape drives, where the speed and motor position must be controlled at low speed as well as when the motor is stopped, back-EMF sensing is not at all useful. Thus far it has been used in disk drives and fans, but not in tape drives (Figure 2).

[FIGURES 1-2 OMITTED]

There are several patents and proposals that involve adding a sense winding to the motor. This technique does not meet with our stated goals because it adds manufacturing cost. Nor does it meet with our preference that the sensor-less motor driver be able to function with any motor without requiring any modification of said motor. We know of no practical application where this method could be employed.

The next unsuitable manner to eliminate the sensor is to measure the impedance change of the motor windings by injecting a high-frequency sense carrier, or probing with current pulses. The motor winding impedance varies with the rotation of the motor. It should be possible to measure this impedance change and thereby get an indication of the motor rotation.

A high frequency sense carrier operates by imposing a high frequency signal into the drive current. There has been some academic discussion of this method, but it has never been implemented. There are several problems with this approach so it should probably remain in the realm of theory.

The rise and fall time of current pulses varies with the winding inductance. In addition to measuring the rotational position of the motor, this current pulse probing method can also detect the polarity of the magnetic motor pole when high currents are used. It can therefore be used to distinguish between the two 180-degree half-cycles (Figure 3).

[FIGURE 3 OMITTED]

Current pulse probing also works well when the motor is stopped and when the motor is turned off. In a tape drive application we can do this when tape is not loaded and when it is not under tension. Use of current pulses when the motor is running is unacceptable in a tape drive, however, because it disturbs the motor torque and produces undesirable speed and tension variations.

In a well-designed servo system the injection of current pulses can easily be implemented in firmware.

For initialization on power-up when a tape is present, the method of pulse probing still can be used, but we must be careful not to spin the motors. Motor movements to any significant degree could damage tape. We have found that the combination of two modifications works quite well. First, we keep the width of the pulses short, to about 300[micro][s.sup.21] Second, we reverse the polarity for every other pulse through the same winding, which cancels any small motion that may result from the first pulse.

Eliminating The Sensor--The Right Way

We have shown above that there are several methods with which to sense the rotational position of a motor without using a sensor. But none of these methods work well for tape drives. Does this mean we cannot do what the disk drive industry is doing?

In the following discussion we will show that we can indeed use a sensor-less method. But, as usual, things are a bit more complex with tape drives, so we need to be a bit smarter to achieve success.

The solution to the above problems requires a new technology. Our new approach measures the motor windings impedance ratio. Our design includes use of a reference network to simulate the voltage across an ideal winding with no impedance variation. The two driven motor windings form a voltage divider. Knowing that the impedance variations will cause the center node to vary, our technique compares this center node to the reference voltage. This comparison now yields our sense signal. The sense signal for each phase is roughly sinusoidal (Figure 4).

[FIGURE 4 OMITTED]

The sense signal depends on the ratio of the impedance of the two windings. Using the measurement of the impedance ratio is very advantageous. It compensates for variables that affect both windings, such as the drive current, the drive voltage, and temperature variations.

The impedance pattern is repeated every 180 electrical degrees. (One turn of the motor = 360 electrical degrees x the number of pole pairs). It is not possible to distinguish between the two half-cycles unless the motor is initialized. Once initialized, the position tracking is incremental, and therefore initialization only needs to be done once, on power-up. We have found that current pulse probing is quite effective for this purpose.

At higher speed, the back-EMF influences the center node voltage. The effect of the back-EMF is predictable and can be compensated for. It is, however, easier to simply revert to using the back-EMF for the positioning measurement once the motor reaches medium-to-high speeds.

The impedance ratio measurement method requires that a current run through the motor windings at all times. There is sufficient current both when tape is moving and when tape is stopped and is under tension. When tape is unloaded, a 'trickle' current must pass through the windings. This current is miniscule and will not spin the motor.

Our new sensor-less motor driver uses current pulse probing to initialize the motor after power-up. The current pulses do not rotate the motor, so this method can be safely used even if a tape was loaded on power-up. Once the tape has been accelerated to moderate-to-high speeds, we use back-EMF sensing to determine the motor speed.

We use impedance ratio to measure rotational speed of a motor at all other times. This includes the acceleration time, the deceleration time, when tape is stopped and is under tension, and when no tape is loaded.

By combining two tried and proven technologies--back-EMF sensing and current pulse probing--with our inventive impedance ratio measurement technique, we have safely eliminated the motor position sensor without sacrificing performance. Our three goals--increased reliability, reduced size, and reduced cost--have all been achieved.

This technique also has general utility to other industries and need not be a narrow application restricted only to tape drives. It can be used in any situation where a motor position needs to be measured at variable motor speeds. Printers and robotic applications are only two examples of a potentially versatile technology.

Our approach required the addition of intelligence to the motor driver electronics. Adding more intelligence and removing mechanics is always more efficient in the world of high technology.

By eliminating a central mechanical part our "Tin man" has become more reliable and in the process also a bit more intelligent. Less tin but more intelligence is the way of the future for tape drives.

Peter Groel is president of Mountain Engineering II, Inc. (Longmont, CO).

www.mountainengineering.com