- Email: [email protected]
New thermal pulse vacuum gauge
Vacuum/volume 38/numbers Printed in Great Britain
8-1 O/pages 897 to 899/l 988
New thermal Michael
Cole,
Genevac
0042-207X/88$3.00+.00 Pergamon Press plc
pulse vacuum Ltd, lpswich,
gauge
UK
A new gauge is described for measurement of pressures in the range 10e4mbar to 5 bar. The gauge uses a conventional Pirani sensor but pulses the filament temperature and measures filament heating and cooling rates. These parameters are functions of both conductivity and of the volume specific heat of the gas surrounding the filament rather than thermal conductivity only on which Pirani gauges depend. Thermal conductivity of gases is constant over much of this pressure range but volume specific heat varies sufficiently to allow measurements to be made over the whole of the above pressure range.
1. Introduction heat dissipation pressure gauges are widely used for vacuum measurement in the range l-lO-3 mbar. The most common types of heat dissipation gauge are the Pirani gauge, thermocouple gauge and thermistor gauge. In the Pirani gauge a tungsten or platinum filament is heated by applying a known voltage across or current through it. The resistance of the wire is measured after a steady-state value has been reached, usually by some form of Wheatstone bridge. Filament resistance depends on filament temperature which will, for a given heat input, be a function of gas composition and pressure. Gauges are normally calibrated for air to give direct pressure readings. In the thermocouple gauge a current is passed through a heater which heats a thermocouple and the output of the thermocouple is measured. This output is a function of temperature which again is a function of pressure and gas composition. The thermistor gauge is similar but a thermistor is substituted for the thermocouple. Heat dissipation from the heated element of these gauges is almost entirely by conduction over the pressure range 10A31 mbar. Below 10e3 radiation plays an increasing part. The conductivity of gases (Figure 1) is proportional to pressure in the molecular range (from zero to about 0.1 mbar depending on the dimensions of the system). In the intermediate range (typically from 0.1 to 10 mbar depending again on the dimensions of the system) the conductivity varies decreasingly with pressure until in the viscous range (typically above 10 mbar), conductivity Steady-state
0
10-q
I 10-s
I
10-Z Pressure
Figure 1. Gas thermal
conductivity
I
10-l
I I
(m bar)
K vs pressure.
I IO
I
I02
I
103
Pressure
Figure 2. Steady-state
(mbar)
element resistance R vs applied
5%x
voltage
and
pressure.
remains constant up to one atmosphere and above. Figure 2 shows filament resistance vs pressure for three different values of fixed heating current. It will be seen that the resistance does not vary sufficiently with pressures above 10 mbar to allow pressure to be measured by determining element resistance. This fact has meant that heat dissipation gauges have generally been limited in their operation range to pressures below I mbar. At low pressures low signal levels and drift have limited their use for measuring pressures below 10e3 mbar. Work has been done in recent years to extend the operation range of Pirani gauges by operating the element at constant temperature’ rather than with a constant voltage across the element or constant current through it and this has extended the useful range of this type of gauge to lo-“ at the low end and 10 mbar at the high end. Convection losses are pressure dependent in the viscous range and a number of gauges have been devised in which forced convection has been arranged. This has been done by a variety of means-using a loudspeaker to make the gas oscillate over the element’, using a magnetic field to make the element itself oscillate3 and using a separate heating element above or below the sensing element4. More recently people have used methods of enhancing natural convection. This normally involves making the long axis of the element horizontal’. Reviews of these topics can be found in the literature”.’ and these methods have generally 897
M Cole:
New thermal
pulse vacuum
gauge
been successful in increasing the useful range of the gauge up to I atmosphere. Although it has been noticed’ that the time constant of Pirani gauges is a function of pressure there appear to be no reports of attempts to base a pressure measurement device on this phenomenon. The rate at which an element increases in temperature for a given heat input and the rate at which it cools in the absence of any heat input depends on gas pressure and composition. The rate is affected by the conductivity of the gas as for the steadystate type of gauge but it is also affected by the volume heat capacity of the gas in the gauge. Thermal conductivity does not vary with pressure over the viscous range but volume heat capacity is proportional to pressure up to very high pressures. Heating and cooling times are, therefore, a function of pressure from below I OmmJ mbar up to many atmospheres. We have been experimenting with a simple Pirani type sensor (Figure 3). This has a single 7.5 pm tungsten filament 5 cm long. We have modelled the behaviour of this single strand element on a computer and are currently using the model to explore the infinite number of possible measurements on which a pressure gauge could be based. What has emerged is that cooling or heating rates can be selected to give a good signal separation over the entire range from 10 -’ mbar to 5 bar. Readings could be extended well above 5 bar but wc are not currently investigating that pressure region. Some interesting factors that have arisen from our work arc described in the following sections. 2. Heating rates with constant current A constant current is applied to the element and held until a predetermined temperature is reached. The temperature is calculated by measurement of the filament resistance. The time required to reach the given temperature is recorded and this can be used as a measure of gas pressure. Figure 4 illustrates how the time vs temperature curves vary with current for a given constant pressure. If T, is the chosen switch-off temperature. current 1 will never
Time
Figure 4. Filament heating current.
temperature
R)
7’ vs time
f for four Mfcrent valuesot
allow the filament to reach it, and current 4 will reach it very rapidly. If too large a current is used the element reaches T, too quickly to allow an appreciable volume of gas to be heated up so that the time is not as strong a function of heat capacity as is desirable. At a current equivalent to 2 in the diagram the time would be so long that the steady-state is reached and again the influence of heat capacity is not maximised. At current values around 3 a considerable amount of gas has been heated and the filament is not approaching the steady-state. This is the region which gives the best separation of heating time with pressure. With the 7.5 pm filament good separation is obtained with currents which give a heating time in the region of IO-20 ms. A possible measurement is that of the current required to give a specified heating time for the chosen temperature. Figure 5 illustrates this approach. The chosen temperature in this case is 80°C and the current plotted is that required to raise the element temperature to 80°C in 12 ms. It can be seen that this parameter varies with pressure over the range in which the steady-state resistance is almost constant. 3. Cooling rates Cooling rates from a given temperature at a given pressure depend on the extent to which the gas in the gauge has been heated before the cooling cycle is started. If the filament is taken quickly up to temperature and immediately switched off, cooling
- Ceramic metal seat
I 1 Nickel
Tungsten
filament
_ Vacuum
Kgurr 898
3. Hot wire sensor
support
envelope
;
180
-
170
;
160
2
I50
E 140 L z 130 E” 120 2 w
110 ioc 90
Figure 5. Current I which heats filament Compared with steady-state resistance.
to X0 C !n I7 ms YS prcs\urc.
M Co/e: New thermal pulse vacuum gauge rates are faster than those which obtain when the element has been taken more slowly up to temperature or held there for some time. They are also less dependent on gas heat capacity and therefore do not give good separation with pressure. Cooling rates appear to give better separation with pressure than heating rates but are more difficult to measure because the small amount of power required to measure element resistance can interfere with the cooling rate. 4. Effect of gas composition In common with other heat dissipation gauges the performance of this type of gauge is very dependent on gas composition. With conventional Pirani gauges it is necessary to know gas composition to obtain meaningful pressure readings. With the heating/cooling rate gauge, however, it seems likely that by taking measurements at different heating or cooling rates information may be obtained about the thermal properties of the gas in the gauge. For instance, the conductivity can be derived from the steady-state heat loss rate and, with this knowledge, the use of different heating or cooling rates may allow estimation of the specific heat. This would allow a gas-independent pressure
measurement to be made and, in the case of simple gas mixture, possibly also gas composition to be calculated. 5. Construction of a practical gauge We have not yet designed a gauge in detail but will be doing so when we have completed our analysis of the system on the computer model. In principle the gauge would probably use a conventional Pirani type head similar to that illustrated above and a microprocessor. The microprocessor would generate the pulses and analyse the results to give pressure readouts. References
’ H Van Ubisch, Appl Sci Res Hague, A2, 363 (1951). ’ G E Iglesias, R R Santochi and M Puparelli, Rev Sci Instrum, 40, 129I (1969). ‘H Mosta and P Herrwirth, U S Patent, 3,071,968 (1959). 4W J H Moll and H C Burger, 2 Tech Phys, 21, 199 (1940). ’ W Steckelmacher and B Fletcher, .I Sci Instrum, 5,405 (1972). ’ W Steckelmacher, Vacuum, 23,307 (1973). ‘A Berman, Total Pressure Measurements in Vacuum Technology, Academic Press Inc. (1985). “C J Milner, Rev Sci Insrrum, 54, 890 (1983).
899
8-1 O/pages 897 to 899/l 988
New thermal Michael
Cole,
Genevac
0042-207X/88$3.00+.00 Pergamon Press plc
pulse vacuum Ltd, lpswich,
gauge
UK
A new gauge is described for measurement of pressures in the range 10e4mbar to 5 bar. The gauge uses a conventional Pirani sensor but pulses the filament temperature and measures filament heating and cooling rates. These parameters are functions of both conductivity and of the volume specific heat of the gas surrounding the filament rather than thermal conductivity only on which Pirani gauges depend. Thermal conductivity of gases is constant over much of this pressure range but volume specific heat varies sufficiently to allow measurements to be made over the whole of the above pressure range.
1. Introduction heat dissipation pressure gauges are widely used for vacuum measurement in the range l-lO-3 mbar. The most common types of heat dissipation gauge are the Pirani gauge, thermocouple gauge and thermistor gauge. In the Pirani gauge a tungsten or platinum filament is heated by applying a known voltage across or current through it. The resistance of the wire is measured after a steady-state value has been reached, usually by some form of Wheatstone bridge. Filament resistance depends on filament temperature which will, for a given heat input, be a function of gas composition and pressure. Gauges are normally calibrated for air to give direct pressure readings. In the thermocouple gauge a current is passed through a heater which heats a thermocouple and the output of the thermocouple is measured. This output is a function of temperature which again is a function of pressure and gas composition. The thermistor gauge is similar but a thermistor is substituted for the thermocouple. Heat dissipation from the heated element of these gauges is almost entirely by conduction over the pressure range 10A31 mbar. Below 10e3 radiation plays an increasing part. The conductivity of gases (Figure 1) is proportional to pressure in the molecular range (from zero to about 0.1 mbar depending on the dimensions of the system). In the intermediate range (typically from 0.1 to 10 mbar depending again on the dimensions of the system) the conductivity varies decreasingly with pressure until in the viscous range (typically above 10 mbar), conductivity Steady-state
0
10-q
I 10-s
I
10-Z Pressure
Figure 1. Gas thermal
conductivity
I
10-l
I I
(m bar)
K vs pressure.
I IO
I
I02
I
103
Pressure
Figure 2. Steady-state
(mbar)
element resistance R vs applied
5%x
voltage
and
pressure.
remains constant up to one atmosphere and above. Figure 2 shows filament resistance vs pressure for three different values of fixed heating current. It will be seen that the resistance does not vary sufficiently with pressures above 10 mbar to allow pressure to be measured by determining element resistance. This fact has meant that heat dissipation gauges have generally been limited in their operation range to pressures below I mbar. At low pressures low signal levels and drift have limited their use for measuring pressures below 10e3 mbar. Work has been done in recent years to extend the operation range of Pirani gauges by operating the element at constant temperature’ rather than with a constant voltage across the element or constant current through it and this has extended the useful range of this type of gauge to lo-“ at the low end and 10 mbar at the high end. Convection losses are pressure dependent in the viscous range and a number of gauges have been devised in which forced convection has been arranged. This has been done by a variety of means-using a loudspeaker to make the gas oscillate over the element’, using a magnetic field to make the element itself oscillate3 and using a separate heating element above or below the sensing element4. More recently people have used methods of enhancing natural convection. This normally involves making the long axis of the element horizontal’. Reviews of these topics can be found in the literature”.’ and these methods have generally 897
M Cole:
New thermal
pulse vacuum
gauge
been successful in increasing the useful range of the gauge up to I atmosphere. Although it has been noticed’ that the time constant of Pirani gauges is a function of pressure there appear to be no reports of attempts to base a pressure measurement device on this phenomenon. The rate at which an element increases in temperature for a given heat input and the rate at which it cools in the absence of any heat input depends on gas pressure and composition. The rate is affected by the conductivity of the gas as for the steadystate type of gauge but it is also affected by the volume heat capacity of the gas in the gauge. Thermal conductivity does not vary with pressure over the viscous range but volume heat capacity is proportional to pressure up to very high pressures. Heating and cooling times are, therefore, a function of pressure from below I OmmJ mbar up to many atmospheres. We have been experimenting with a simple Pirani type sensor (Figure 3). This has a single 7.5 pm tungsten filament 5 cm long. We have modelled the behaviour of this single strand element on a computer and are currently using the model to explore the infinite number of possible measurements on which a pressure gauge could be based. What has emerged is that cooling or heating rates can be selected to give a good signal separation over the entire range from 10 -’ mbar to 5 bar. Readings could be extended well above 5 bar but wc are not currently investigating that pressure region. Some interesting factors that have arisen from our work arc described in the following sections. 2. Heating rates with constant current A constant current is applied to the element and held until a predetermined temperature is reached. The temperature is calculated by measurement of the filament resistance. The time required to reach the given temperature is recorded and this can be used as a measure of gas pressure. Figure 4 illustrates how the time vs temperature curves vary with current for a given constant pressure. If T, is the chosen switch-off temperature. current 1 will never
Time
Figure 4. Filament heating current.
temperature
R)
7’ vs time
f for four Mfcrent valuesot
allow the filament to reach it, and current 4 will reach it very rapidly. If too large a current is used the element reaches T, too quickly to allow an appreciable volume of gas to be heated up so that the time is not as strong a function of heat capacity as is desirable. At a current equivalent to 2 in the diagram the time would be so long that the steady-state is reached and again the influence of heat capacity is not maximised. At current values around 3 a considerable amount of gas has been heated and the filament is not approaching the steady-state. This is the region which gives the best separation of heating time with pressure. With the 7.5 pm filament good separation is obtained with currents which give a heating time in the region of IO-20 ms. A possible measurement is that of the current required to give a specified heating time for the chosen temperature. Figure 5 illustrates this approach. The chosen temperature in this case is 80°C and the current plotted is that required to raise the element temperature to 80°C in 12 ms. It can be seen that this parameter varies with pressure over the range in which the steady-state resistance is almost constant. 3. Cooling rates Cooling rates from a given temperature at a given pressure depend on the extent to which the gas in the gauge has been heated before the cooling cycle is started. If the filament is taken quickly up to temperature and immediately switched off, cooling
- Ceramic metal seat
I 1 Nickel
Tungsten
filament
_ Vacuum
Kgurr 898
3. Hot wire sensor
support
envelope
;
180
-
170
;
160
2
I50
E 140 L z 130 E” 120 2 w
110 ioc 90
Figure 5. Current I which heats filament Compared with steady-state resistance.
to X0 C !n I7 ms YS prcs\urc.
M Co/e: New thermal pulse vacuum gauge rates are faster than those which obtain when the element has been taken more slowly up to temperature or held there for some time. They are also less dependent on gas heat capacity and therefore do not give good separation with pressure. Cooling rates appear to give better separation with pressure than heating rates but are more difficult to measure because the small amount of power required to measure element resistance can interfere with the cooling rate. 4. Effect of gas composition In common with other heat dissipation gauges the performance of this type of gauge is very dependent on gas composition. With conventional Pirani gauges it is necessary to know gas composition to obtain meaningful pressure readings. With the heating/cooling rate gauge, however, it seems likely that by taking measurements at different heating or cooling rates information may be obtained about the thermal properties of the gas in the gauge. For instance, the conductivity can be derived from the steady-state heat loss rate and, with this knowledge, the use of different heating or cooling rates may allow estimation of the specific heat. This would allow a gas-independent pressure
measurement to be made and, in the case of simple gas mixture, possibly also gas composition to be calculated. 5. Construction of a practical gauge We have not yet designed a gauge in detail but will be doing so when we have completed our analysis of the system on the computer model. In principle the gauge would probably use a conventional Pirani type head similar to that illustrated above and a microprocessor. The microprocessor would generate the pulses and analyse the results to give pressure readouts. References
’ H Van Ubisch, Appl Sci Res Hague, A2, 363 (1951). ’ G E Iglesias, R R Santochi and M Puparelli, Rev Sci Instrum, 40, 129I (1969). ‘H Mosta and P Herrwirth, U S Patent, 3,071,968 (1959). 4W J H Moll and H C Burger, 2 Tech Phys, 21, 199 (1940). ’ W Steckelmacher and B Fletcher, .I Sci Instrum, 5,405 (1972). ’ W Steckelmacher, Vacuum, 23,307 (1973). ‘A Berman, Total Pressure Measurements in Vacuum Technology, Academic Press Inc. (1985). “C J Milner, Rev Sci Insrrum, 54, 890 (1983).
899