Advanced Downstream Pressure Control Increases Throughput

Semiconductor cluster tool throughput and yield rates can be significantly affected by the performance level of process and transfer chamber vacuum control. Faster step transitions and more precise pressure control are often desired or needed to enhance tool productivity or increase production yields. In fact, the advanced pressure control technology covered here has, in some instances, been identified as enabling technology in leading-edge 300 mm wafer processing.

The technology revolves around a new method of operating a variety of stepper motor-driven downstream control valves (such as butterfly, poppet, gate and pendulum). In essence, it combines closed-loop stepper motor control with advances in pressure control algorithm development, allowing such valves to be controlled in ways never before possible. As a result of the exceptionally fast valve actuation and ultrahigh position resolution, throughput increases up to 15% coupled with significant yield improvements, and 100-fold scrap reduction have been realized.

One common design criterion for many semiconductor process tools is the need for fast, accurate and repeatable chamber pressure control. Though several methods for pressure control exist, downstream pressure control is used most. In this method, a throttle valve is installed in the exhaust piping between the vacuum chamber and the vacuum pump. A throttle valve controller continuously adjusts the angle or position of the throttle valve, in effect varying the conductance in the foreline to maintain the desired pressure in the chamber.

Downstream pressure control generally works well over a wide range of conditions, but its effectiveness can be challenged by "external" factors such as the sudden changing of inlet gas flow rates or events involving the turning on or off of plasma. Furthermore, certain flow-and-pressure combinations can force the throttle valve to operate in a position at or beyond the limit of its intended control range. In such instances, neither accurate nor repeatable pressure control may be feasible. Alternatively, pressure control may be feasible, but not in a fast and efficient manner. As a result, semiconductor wafer yield and throughput suffer.

At present, throttle valves are available from a host of manufacturers and tool OEMs. Many different types of throttle valves exist, with the most common being butterfly, pendulum, gate and poppet designs. As different as the various valves may be, almost all possess one common characteristic: They are driven by stepper motors. Another shared trait among existing throttle valves is that some means of mechanical advantage is always inserted between the motor and the valve. The type and degree of mechanical advantage varies by manufacturer and valve type, but it is generally comprised of either a gear box or a belt-and-pulley system. Regardless of the kind used, the combined stepper motor and means of mechanical advantage together serve two important purposes: 1) to increase the torque available to drive the valve, and 2) to increase the position resolution of the valve drive. This way, conventional 1.8° stepper motors having a relatively compact form factor (typically NEMA-17 or NEMA-23) can then be used.

As far as pressure control performance is concerned, however, the design of the valve or valve drive mechanism is of secondary importance to the way in which the stepper motor is driven. Conventional motor drive technology involves sequencing the stepper motor through a prescribed combination of motor winding currents designed to guide the motor to move in a given direction using the desired number of steps. In Table 1, we can see a typical sequence for a bipolar full-step moving sequence.

From any given position (step), the motor can be moved to an adjacent position by changing the current going to the four respective drive phases (A, A', B and B'). Sequencing, or pulsing, the current values in the proper manner therefore results in having command over the actual position of the motor at all times. Knowledge about the actual position in these cases is done by incrementing a step or pulse counter. The position of the valve is therefore derived through calculation, not by observation. This is referred to as open-loop motor control. Similar schemes are employed when half-stepping or micro-stepping a motor. The major difference is in the relative amount of current delivered to the motor drive phases. Whereas full-stepping only requires equal magnitudes of current of the same or different polarity be delivered to phases A and B simultaneously, finer steps can be achieved by delivering different-magnitude currents to the two phases. The position of the motor will be determined by the ratio of currents delivered.

Unfortunately, the speed and resultant position accuracy with which conventional open-loop stepping can be done is negatively influenced by non-linear effects from the valve and the motor drive assembly. Examples of such effects include inertia, friction and backlash. For example, if a valve assembly exhibits a high degree of inertia (as some large valves do), the rate at which the motor can be accelerated is limited. Too high an acceleration, or pulse rate, would put the step sequencer out of synchronization with the actual motor position — ultimately to the point where an entire cycle, or four full steps, might be missed. Similarly, when a valve assembly exhibits a high coefficient of friction, the speed with which the valve can be operated is limited. In short, the fact that actual valve or motor position is not positively or continuously known forces the user to operate the motor in a conservative manner. As a result and in comparison to what it could be, open-loop motor operation and positioning is by design sluggish.

Motor control performance can be greatly improved by employing some means of true position feedback. By accurately tracking position, the user is no longer forced to be as conservative with respect to the acceleration or speed used in operating the motor. In addition to using the position feedback signal to determine the actual position, a position error term (target position less observed position) can easily be calculated, monitored and used to alter the amount of current delivered to the motor to overcome variations in external inertia and friction. This is what is referred to as closed-loop motor control, and it enables the motor to be driven to its full torque-speed potential.

To perform closed-loop motor control well, however, is not as easy as it may appear on the surface. The first attempt might involve combining a stepper motor with a position feedback sensor. Many different types of position sensors are commercially available. But, because the knowledge of position can only be as accurate and timely as the means by which the true position is obtained, it is important to use a feedback sensor with a high enough resolution and accuracy. Second, it is imperative to synchronize the reading of position with the commanded position, for fear that the position error term cannot be accurately calculated. It is because of the challenges associated with the achievable resolution and synchronization that a preferred method of position feedback was developed for use with throttle valves.

	Semiconductor process equipment in the service chase of a 200 mm fab.

By closely monitoring the back EMF (electromotive force) generated from these windings as the motor is moved, one can uniquely determine the exact motor position within a phase. This is done by performing an inverse trigonometric function on the ratio of the back EMF values for the two phases. In essence, this relationship can be expressed as the following general formula:

Position = trig^-1 (EMF_A ÷ EMF_B)

Using the back EMF signals from the motor accomplishes three primary tasks. First, it establishes the motor itself as the means of feedback, thereby eliminating the need for (and cost of) additional position sensors. Second, it eliminates the need for the aforementioned motor-to-position sensor synchronization. Third, it places no limits with which one can theoretically resolve the true position, though some practical limits do exist.

Over and beyond simplifying the task of obtaining true position information, the back EMF signal can serve toward another important benefit — namely increasing the motor-step position resolution to the point where a means of mechanical advantage is no longer necessary.

As described previously, a means of mechanical advantage is usually necessary, in part, to give the motor drive assembly enough positional resolution to perform satisfactory pressure control. Take a butterfly valve, for instance, in which a plate or flapper turns 90° from open to close. Then consider that, when a conventional 1.8° stepper motor is driven in the typical manner described earlier, it offers 50 full steps, or 100 half-steps in 90° of movement.

Therefore, if directly coupled to the motor, the valve would only provide that many discrete plate positions. One familiar with downstream pressure control will immediately recognize that this is insufficient for almost all pressure control tasks.

Now consider the case in which back EMF is used for position sensing and accurate position placement of the motor. It has been demonstrated that, when a standard 12-bit A/D converter is used in the position sensing loop, the true motor position in 90° of movement can be resolved into more than 100,000 discernable steps. This is a huge increase in comparison to the 50 or 100 steps attainable from conventional motor control, and a significant increase over the valve drive assemblies that typically make use of gear or pulley reducers with ratios ranging from 15:1 to 100:1.

On a related note, consider the relative difference in valve actuation speeds. Again, using the butterfly valve as an example, a standard butterfly valve assembly using conventional motor control usually can be actuated from open to closed in no less than 1.5 sec. A direct-drive valve using the back EMF as the position sensing means can travel the same distance in 0.125 sec.

The inherent improvements to a direct-drive valve as made possible by using back EMF position sensing can most easily be seen from the results in Table 2.

	2. In this test, three different valve closing technologies were compared. The valve with back EMF position sensing (V1) closes faster, enabling the set-point pressure to be reached more quickly.

The enhancements in the motor and valve drive technology discussed up to this point would have little importance if they could not be tied to quantifiable improvements in pressure control. A live test was designed and conducted to illustrate and quantify any possible benefits of using higher-speed and higher-position resolution valve drive technology. In this case, a direct-drive butterfly valve using back EMF position sensing (V1) was compared with two alternative techniques (V2 and V3). A multi-step wafer recipe was executed in a 35 L chamber outfitted with throttling valve in the downstream position.

As seen in Figure 2, the first notable event occurs in the pressure transition step in which the pressure set point is suddenly increased. As each of the three controllers drives its respective valve completely closed, the chamber pressure rises accordingly. The valve with back EMF position sensing closes completely in 0.125 sec, compared with 1.7 and 2.0 sec for the other valve setups tested. The result of the faster closing is a faster onset of pressure rise, allowing the set point to be reached more quickly.

	3. The use of back EMF as the sensing mechanism in V1 results in smooth pressure control at 1000 mTorr. In contrast, a significant amount of “hunting” is evidenced by the open-loop motor control technique, which in this case translates into 50-60 mTorr pressure swings.

In summary, it has been shown how open-loop motor control differs from closed-loop motor control, and specifically how back EMF can be used to provide an unparalleled method for motor position feedback. Because of the high precision and resolution of such feedback mechanisms, ordinary stepper motors can be employed in ways not possible by conventional means. The advancements in motor control capabilities were then substantiated by demonstrating how cluster tool throughput and wafer yield can be positively impacted by the resultant improvements in pressure control.