Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.apo.nmsu.edu/Telescopes/eng.papers/SPIEpapers/temperature/temperature.html
Дата изменения: Wed Sep 21 18:50:46 1994
Дата индексирования: Sun Apr 10 07:26:31 2016
Кодировка:

Поисковые слова: п п п п п п п п п п п п п п п п п п п
Temperature Measurement System For A 3.5-Meter Borosilicate mirror

Temperature Measurement System For A 3.5-Meter Borosilicate Mirror

Charles Hull, Walter Siegmund and Dan Long

Abstract

Optimizing the performance of a telescope requires the ability to accurately measure and monitor the spatial variation of temperature in critical components. Surfaces near a telescope may warm or chill ambient air and cause image degradation. It is desirable to monitor the temperature of such systems.

The design, fabrication and testing of a reliable, low-cost, multiplexed temperature measurement system with a resolution and stability approaching 0.01° C is described. This system, with 176 temperature sensors, will be used to monitor the performance of a 3.5 m borosilicate mirror ventilation system. It is applicable to a broad range of telescope and telescope enclosure temperature measurement problems

Introduction

Temperature gradients can degrade the performance of telescopes and instruments in a number of ways. Gradients in optics may change their power and/or cause optical aberrations unless they are fabricated from very low expansion coefficient materials. Gradients in a telescope may affect its pointing and gradients in spectrographs can cause changes in wavelength calibration. The temperatures of surfaces near the telescope should track the ambient air temperature to about 0.1° C to prevent locally produced image degradation. To understand and monitor these effects a simple and robust temperature measurement system capable of measuring temperature differences of about 0.01° C at up to several hundred locations is needed.

Description of the Temperature Measurement System

Design

The AD590 (Analog Devices Inc.) is a two terminal integrated circuit temperature transducer. When properly biased, it acts as a low impedance current source. The magnitude of the current is proportional to absolute temperature with a nominal scale factor of 1mA/°K. The basic operating principle of the device is that the base-emitter voltage difference of two identical transistors, operating at a constant collector current ratio, is proportional to absolute temperature.

The AD590JF is packaged in a 1.3x2.4x5.8mm hermetically sealed ceramic case with a Kovar cover. Before assembly, the sensors are burned in at 140°C, both forward and reverse biased, for one week each direction. This is intended to increase reliability by causing early failure of marginal sensors, and to improve temperature stability of the sensors. To assemble these devices into a temperature measurement system, cyanoacrylate adhesive is used to attach the package to a 10x20x1.5mm Kovar rectangle Figure1. Kovar matches the coefficient of thermal expansion (CTE) of the package and minimizes stress on the IC due to thermal expansion of the surface to be measured. Kovar has good thermal conductivity compared to glass and helps couple it thermally to the sensor. Ribbon cable conductors are soldered directly to the leads and the joint is insulated with heat shrink tubing. Strain relief is provided by attaching the ribbon cable conductors to the Kovar with epoxy. Assembly is quick and the assembled transducers are robust.

Figure 1. This 2:1 scale drawing shows an AD590 temperature sensor attached to a Kovar block with cyanoacrylate. The ribbon cable conductors are soldered to the sensor leads and insulated with heat shrink tubing. A bead of epoxy provides strain relief. This mounting scheme provides good thermal coupling to glass. The Kovar block matches the CTE of the sensor package and minimizes stress on the IC that might cause changes in device response.

The ribbon cables are mass terminated and plug into a small printed circuit (PC) board called a temperature distribution board. Up to 16 of these boards can be connected along another ribbon cable that connects them all to a multiplexer card. Crimp-on ribbon connectors are used to simplify wire harness assembly. A jumper on each temperature distribution board defines the bank number of the ribbon cable and the associated sensors plugged into the board.

The AD590 temperature sensors have negligible conductance when reversed biased and very low impedance so they can be easily multiplexed. The multiplexer board uses two DG460DJ (Siliconix Inc.) 16 to 1 analog multiplexers Figure2. These devices have an "on" resistance of 50ohm. The uniformity of the coefficient of resistance change with temperature from channel to channel, coupled with the insensitivity of the sensors to bias voltage, causes negligible offsets for reasonable temperature changes. One multiplexer applies the bias voltage to one of sixteen banks of sensors. The other multiplexer selects one of sixteen sensors in the selected bank to be read out.

Figure 2. The temperature multiplexer board multiplexes up to 256 AD590 integrated temperature sensors. Two 8-bit digital I/O ports control the multiplexer. One D/A channel is used to subtract a bias from all sensors. The sensor readings are digitized by a single A/D channel. Since most of the components are common to all of the sensors, drifts in component properties may cause an offset but not differences between sensors.

Since the sensors produce a current that is proportional to absolute temperature, it is useful to subtract a current corresponding to the nominal operating temperature of the system thereby reducing the dynamic range of the signal and the resolution of the analog to digital converter (A/D) required. The output of the multiplexer is connected to the input of a low noise, low input offset voltage operational amplifier (Precision Monolithics Inc., OP-490), that acts as a current to voltage converter. A digital to analog converter (D/A) output is connected to the node by a Vishay precision resistor with a very low temperature coefficient. This allows any desired offset current to be subtracted by programming the D/A appropriately. The output of the current to voltage converter is connected to a two pole, low-pass, antialissing Bessel filter with a corner frequency of 83Hz. The output is digitized by a 12-bitA/D.

The multiplexer board is controlled by a STD bus computer. Two 8-bit parallel digital input/output ports are used to control the multiplexers and switches. A 12-bitD/A port is used to supply the offset current to the be subtracted from the sensor currents. A 12-bitA/D port is used to digitize the sensor signals. The data are transferred to a Macintosh computer over an RS232 serial interface for further analysis.

The sensors are read out and digitized continuously. The duty cycle of each sensor is 1/256 and average dissipation is 16µW. The thermal resistance from the chip to the Kovar mounting block is 10°C/Watt so self heating is 0.0002°C. Self heating during readout is minimized because each sensor is only forward biased for 10ms. Readout is timed so that is occurs at a consistent time interval after it is turned on. Since each sensor is read out at the same point on the self heating curve, differences due to this effect are small. Generally, six successive read outs are averaged and the result recorded.

Calibration

Even the most precise AD590's available do not have the precision that we require. Consequently, we buy the lowest precision, and lowest cost grade and calibrate their offsets and slope. For calibration, the sensors are assembled with their Kovar blocks and the Kovar blocks are attached to a 0.1x0.3x0.025m aluminum calibration block using hot-melt adhesive. Copper tubing 25mm in diameter is attached to both of the long edges of the calibration block using thermally conductive silicon rubber. The tubing is plumbed so that an ethylene glycol-water mixture flows up one side and back the other.

The calibration block must be in a thermally controlled environment to get stability at 0.005°C. This is provided by an aluminum tube 0.3m in diameter and 1.5m long that is wrapped with a helix of copper tubing. The aluminum tube is surrounded by insulation. The calibration block is placed at the center of the aluminum tube in a lightly insulated inner chamber. The multiplexer board is located just outside the inner chamber but within the aluminum tube so that it is at approximately the same temperature as the sensors.

Temperature controlled fluid is produced and circulated through the copper tubing by a Neslab HX-150 recirculating chiller. This chiller has excellent stability characterized by a slow drift of less than 0.06°C/day. Since both the calibration block and aluminum tube are directly coupled to the temperature controlled fluid, the temperature can be changed and stabilized it in less the 2hours, although is takes about 5 hours for the multiplexer board to stabilize since it is not well coupled thermally to the temperature controlled fluid.

We have found that it is necessary to determine two parameters, zero point and slope, to get differential accuracy of 0.05°C. The absolute accuracy of the system is not important for our application, but is probably a few tenths of a degree for each sensor.

Operation

The goal for this system is to be able to measure small temperature differences between sensors. The temperature itself is less important. This circuit is very simple and straightforward to analyze. Only the multiplexers and the sensors affect the temperature differences directly. Effects due to variations in the bias voltage, D/A voltage, and variations in other components, can only affect temperature differences after averaging several readings if they occur in phase with readout. The intervals between successive reads of a particular sensor is 2.56s, which is not a simple multiple of the AC power period of 16.7ms.

Figure 3. Drift in temperature readings for a set of 176 sensor over the two week period from February 22 1994 to March 3 1994. During this run the temperature of the calibration system started at 15°C and was raised to 25°C for 2 hours and returned to 15°C for the remainder of the run. Short term stability is excellent. All 176 sensors track to 0.07°C peak to peak with a standard deviation of 0.008°C.

The short term stability of this system is excellent. Figure3 shows that the drift after two weeks is about then 0.07°C peak to peak for the 176 sensors. Occasionally, sensor to sensor offsets that are associated with temperature cycling of the system occur. However, we have not yet determined their source.

Large transient voltages are produced when sensor banks are switched by the upper multiplexer (U02). It is likely that this is due to charge stored in the capacitance's of the multiplexers and the sensors. Currently each bank of sensors is read out twice. By the second read out, the transients have decayed. Only the second readout is saved. The analog switches around the current to voltage converter op amp are intended to solve this problem, although this has not yet been tested. The switch across the 750Kohm feedback resistor reduces the gain of the current to voltage converter and prevents saturation of the op amp. The switch on the output of the current to voltage converter disconnects the active filter from the converter. During bank switching, the output switch is opened and the feedback switch closed. Once the transient voltage has decayed, the feedback switch is opened and the output switch closed.

Results

We have assembled and tested a simple and reliable temperature measurement system with 176 sensors and a differential drift over two weeks of less than 0.07°C peak to peak. Only two or three days are necessary to calibrate the system. It appears that this system is suitable for monitoring the temperature variations in large optics and for other demanding temperature measurement tasks.

Acknowledgments

We thank Jim Fowler and Mark Klaene of Apache Point Observatory who helped assemble and test the temperature measurement and calibration systems. Our design owes much to a somewhat similar design of Dave Dryden and Earl Pearson of the National Optical Astronomy Observatories.

References

1. D.M. Dryden and E. Pearson, Multiplexed precision thermal measurement system for large structured mirrors. Advanced Technology Optical Telescopes IV 1236, p.825-833, 1990.