Stan: What about the precision of control needed?
Mark: The temperature of the wafer for some applications needs to be held within 0.1 °C in order for the gases to react and the substrate layers to grow. For example, the temperature must reach 999.9 °C to exceed the activation energy for the reaction, but must be kept from exceeding 1000 °C. An improper temperature causes not only an inconsistent uniformity, but also different layer thicknesses that create problems for feature size and other key electrical properties.
Greg: What are the temperature measurement requirements?
Mark: Besides the need for contactless sensors, the resolution must be better than 0.1 °C. Optical pyrometers without any emissivity errors have a resolution that ranges from 0.1 °C for standard devices to 0.02 °C for special devices. The response time must be faster than sample time. For rapid thermal processing, the response time must be less than 10 ms. Optical pyrometers can be as fast as 2 ms.
Stan: What about measurement noise?
Mark: The prongs holding the rotating wafer can cause a 40 °C to 60 °C drop. The EMI noise from a zero-crossing SCR can also cause noise of 1% to 3%. Intelligent filtering is needed, as well as an incredibly fast model-based control system to smooth out noise and push out the bandwidth.
Greg: Besides reducing the effect of process and EMI noise, why is model-based control important?
Mark: Model-based control reduces the adverse effect of resolution on the signal-to-noise ratio by relying less on feedback control. The multivariable nature of model-based control provides optimal coordination that eliminates interaction in wafer zone control. The radial zones must be controlled at different temperatures. The outermost radial zone must be kept at the highest temperature due to heat loss at the edge.
Stan: What type of model-based control technology do you use?
Mark: We use the best controller for the system under control given the process characteristics and performance requirements. This includes the sensors, actuators and the process dynamics. The controller might be a multivariable PID or one of the many possible model-based controller methodologies with a Kalman filter to overcome system noise issues.
Greg: We conclude with the first half of my "Believe It or Don't" list.
Believe It or Don’t
- A young engineer told the CEO of a large corporation that there were too many presidents and not enough process control. The CEO thanked him profusely and immediately transferred the presidents to plants to install automation systems.
- An accountant found a major design flaw in the control system after carefully reviewing the drawings and immediately authorized the purchase of instruments to solve the problem.
- An accountant said we need the best automation system regardless of cost.
- After major problems meeting product quality specs, a process engineer said, “The transmitters are accurate; it must be the process that is screwed up.”
- A vendor refused a purchase order and recommended the purchase of transmitters from a competitor because the vendor’s instruments had an excessively high failure rate.
- A P&ID was drawn with control strategies that didn’t resemble anything ever done before.
- The resolution of the automation system was actually known and its impact evaluated.
- A controller successfully used the original Ziegler Nichols tuning.
- A sample rate was chosen that was not too fast or too slow.
- A person read a whole book on control theory.