Technology Nugget—Controlling all those Segments
  Part 2: Sensors
  Mark Sirota
  TMT Telescope Controls Group Leader

In the previous edition of Technology Nugget we introduced the complex system responsible for maintaining the overall shape of the TMT primary mirror which consists of 492i segments. We call this system the M1 Control System (M1CS) and it consists of 1,476 actuators, 2,772 sensors, thousands of electronic assemblies, and many sophisticated control algorithms (software). The system that is used to determine the proper "set-points" for the M1CS using starlight is known as the Alignment and Phasing System (APS). We can think of the M1CS as a system that "stabilizes" the 492 mirror segments and the APS as the system that "aligns" the 492 segments by determining the set-points for use by the M1CS. In particular, last month we described the actuators that are responsible for moving the segments. In this edition of Technology Nugget we will discuss the M1CS edge sensors. The edge sensors provide very precise measurements of the height differences between adjacent segments.

Recall from our discussion on actuators that motions within the plane of each segment are minimized passively by a complex structure called a Segment Support Assembly (SSA). The remaining out-of-plane motions of each segment are controlled by three actuators symmetrically positioned so as to enable motion of piston, tip, and tilt. The sensor height measurements and the "set-points" determined by the APS are utilized by a software control algorithm to generate actuator commands. The actuator commands are calculated to position the 492 individual segments so as to imitate a nearly perfect monolithic mirror.

The accuracy with which the mirror array can be held in alignment depends to a large part on the precision and stability of the sensors. Sensor readings are influenced by both random and systematic errors including those caused by temperature variation, gravitational deformation, and electronic noise. Systematic and repeatable errors caused by thermal, gravitational and inter-segment motion effects will be corrected by the "set points" generated by the APS. Random errors result in imperfections of our nearly perfect segmented mirror surface. The random noise in the sensors needs to be less than 5 nanometers in order to support our performance requirements.

An additional requirement is that the sensors be insensitive to in-plane motions. As mentioned above in-plane motions are minimized by the SSAs and the telescope structure but despite our best efforts a small amount of in-plane motion will always remain. To the extent that the sensors misinterpret these motions as changes in segment height differences the primary mirror segments will not be properly aligned. As you will see when we discuss the TMT sensor in detail this is a particularly thorny problem.

In addition to the performance characteristics identified above the sensors need to be small, light weight, low power, reliable, easy to maintain, not interfere with the precisely controlled motions of the segment, and since we need nearly three thousand sensors they must be inexpensive!

The sensor design used on the successful Keck Telescope was the starting point for the design of the TMT sensorsii. To keep things simple we'll start off with a description of the Keck sensors and then describe the necessary and important differences that are being incorporated into the TMT sensor design. A picture of a Keck sensor is shown in Figure 1.

The sensors are mounted on the rear of the mirror and do not interfere with the imaging of stars. Two sensors are located along each inter-segment edge. Orphan segment edges (edges with no neighbors) do not have sensors. The assembly to the left in Figure 1 is known as the sense body and the assembly to the right is known as a drive paddle. The sense body is attached to one segment and the drive paddle is attached to the adjacent segment. The drive paddle extends across the gap separating the two segments. When the drive paddle is inserted into the drive body there is approximately 4 millimeters of clearance between the drive paddle and sense body on both top and bottom. The drive paddle and sense body are gold coated so as to form two capacitors; one between the top of the drive paddle and the sense body and one between the bottom of the drive paddle and the sense body. The drive paddle moves relative to the sensor body when the two segments move relative to each other, changing the gap between the coated areas that form the two capacitors. Since capacitance is proportional to the gap between the plates, the difference in capacitance between the two capacitors is proportional to the relative displacement of the two segments. The difference in capacitance is measured by the associated sensor electronics and converted into a height difference by the control software.

A key point to remember is that the Keck sensors use a change in the gap between two plates that form a capacitor to determine segment height differences. Keep this in mind for it is one of the key differences between the Keck and TMT sensors. Figure 2 is an illustration of a Keck sensor mounted on two adjacent segments.

The TMT sensors are also based on the capacitive principle but two key design modifications were necessary. First, due to the large number of TMT sensors it was desirable to reduce the cost of the sensors by nearly a factor of 10 relative to the cost of the Keck sensors. This was achieved by reducing the number of parts from eight plus attachment hardware for the Keck sensor to only two for TMT. The prototype sensors for TMT are shown in Figure 3.

The other important design change is a result of recognizing that the large number of TMT segments requires a more efficient segment exchange process than is achieved at Keck. When a Keck segment is removed the interlocking drive paddle needs to be swung away so as to allow the segment to be lifted out of the telescope. This is a delicate and time consuming operation. The TMT sensors have been designed to be non-interlocking and operate in the orientated illustrated in Figure 4.

As one might expect these changes didn't come for free. Although the cost of the sensor has been dramatically reduced and the efficiency of the segment exchange process increased; the complexity of the sensor has increased. The TMT design relies on the fact that the capacitance between two plates is a function of the overlapping area of the two plates. As in the Keck design the drive block moves relative to the sensor block when the two segments move relative to each other. The difference in the overlap area between each of the drive plates and the sense plate gives rise to a capacitance difference and a measurement of segment height difference. This is in contrast to the Keck design that uses the capacitance change with gap to provide a measure of the segment height difference.

The complexity with TMT arises because the gaps between segments change with gravity and temperature. Since capacitance is a function of gap, as well as area, the sensor electronics can erroneously interpret this gap change as a relative segment height change. The Keck design doesn't have this problem as long as the drive paddle and sensor body are parallel.

We have fortunately discovered that there are a number of workarounds, beyond the scope of this article, which can be used to reduce this undesired sensitivity but this is where we pay the price of additional complexity. As usual, there's a price for everything.

So far we've described the actuators that move the segments and the sensors that measure segment height differences. Only two pieces of the puzzle remain; the control algorithm which converts the sensor measurements and set-points into actuator commands and the APS, the system that determines the set-points. We'll address these issues in a future edition of the Newscast.

ⁱIf you are wondering why we are now talking of 492 segments while the November/December edition of Technology Nugget spoke of 738 mirror segments; the reason is that the individual segment size has been increased resulting in fewer segments and control components in total. See this month's edition of the Project Manager's Corner for an in depth discussion.

ⁱⁱDisplacement Sensors for the Primary Mirror of the W. M. Keck Telescope , Minor, Arthur, Gabor, et al., SPIE Vol 1236 Advanced Technology Optical Telescopes IV (1990).