Earliest Paper with Porous Silicon Based Micromachining Process

Tu Xiang Zheng

The present author published a paper with the title “Fabrication of Silicon Microstructures Based on Selective Formation and Etching of Porous Silicon” in J. Electrochem. Soc, Vol. 135, No. 8, in August 1988. It was the earliest paper that describes silicon microstructures formed based on porous silicon micromachining process.

The next paper with the title “Using porous silicon as a sacrificial layer” by P Steiner, A Richter and W Lang. was published in 1993, in Journal of Micromechanics and Microengineering  Volume 3, Number 1.

The silicon microstructures were tiny mechanical devices such as sensors, valves, gears, mirrors, and actuators embedded in silicon chips. Before porous silicon micromachining or 1988 year, all these devices were produced by bulk micromachining process or surface micromachining processing.

Silicon wafers can be anisotropically wet etched, forming highly regular structures. Wet etching typically uses alkaline liquid solvents, such as potassium hydroxide (KOH) or tetramethylammonium hydroxide (TMAH) to dissolve silicon. These alkali solvents dissolve the silicon in a highly anisotropic way so as to produce V-shaped grooves.

Surface micromachining process builds microstructures by deposition and etching of different structural layers on top of the silicon wafer. Generally polysilicon is commonly used as one of the layers and silicon dioxide is used as a sacrificial layer which is removed or etched out to create the necessary void in the thickness direction. Added layers are generally very thin with their size varying from a few microns.

The earliest paper provided a new micromachining process for silicon microstructures formation. The process consists of selective anodization of silicon in concentrated HF solution to form porous silicon and etching of the porous silicon in dilute KOH solution to form desired microstructures. In the process a starting material was n-type silicon wafer having resistivity in the range of 3.2 - 4.8 Ω-cm. Proton implantation with post-implantation annealing was employed to produce a high donor concentration layer in the wafer. Then nitrogen implantation was performed to create highly resistive region in the high donor concentration layer. The un-implanted regions provided the entrance windows through which the anodic current was able to reach the underneath layer. Since the donor concentration in the wafer was much lower than that in the proton implanted layer, the anodic reaction could be stopped automatically at the interface between the high donor concentration layer and the un-implanted regions.

The porous silicon micromachining incorporates the advantages of both bulk and surface micromachining:

• The porous silicon layer as a sacrificial layer can be formed in the silicon wafer and processed from the front side.
• Porous silicon is rapidly etched in dilute hydroxide solutions at room temperature.
• Sacrificial layer formation can be patterned both by selective substrate doping, as porous silicon formation is highly selective with respect to different dopant types and concentrations, and by masking of the substrate.
• Deep channels can be formed in the silicon wafer removing a formed porous layer.
• Porous silicon provides a planar sacrificial surface and is formed much more quickly than thermally grown or chemically deposited sacrificial layers.
• It can also be oxidized to form thick sacrificial oxide layers, thick oxide layers for thermal isolation or for SOI applications.
• Using porous silicon as a sacrificial layer also greatly reduces processing time and complexity, as well as device area, over bulk micromachining.
• It is possible to manufacture free-standing structures of high mechanical and electrical quality since the mechanical structures may be constructed from single crystal silicon. 

Using porous silicon micromachining process the present author had developed several MEMS sensors and actuators. Among them are piezoresistive pressure sensors, thermal flow sensors, thermal conducting (vacuum) sensors, capacitive pressure sensors, and ink jet printer heads. 

Porous Silicon Membranes for Removing Carbon Dioxide From Natural Gas

Tu Xiang Zheng

30 years ago I was assigned to prepare porous silicon membranes for study of removing carbon dioxide from natural gas based on Knudsen diffusion. Natural gas mainly consists of 79-84 mol % methane and 5-8% mol % Carbon dioxide. To meet specifications carbon dioxide must be removed before a natural gas can be delivered to the pipeline.

Knudsen’s diffusion occurs in a porous membrane, whose pore sizes are smaller than the mean free path of the gas molecules. The mass flux of a gas through the porous membrane can be expressed as:

J_k = D_k (∂ρ/∂L)              (1)

Where J_k is mass flux of the gas through the porous membrane, D_k is the Knudsen coefficient, ρ is the density of the gas, and L is the thickness of the membrane.

The Knudsen coefficient is defined as

D_k = d_p /3 (8R_gT/πM_g)^1/2         (2)

Where d_g is the diameter of the pores, R is the gas constant (8.3144J/mol k in SI units), M_g is the molecular weight of the gas (in units of kg/mol) and T has units of k.

Hence, for Knudsen diffusion, the square root of the inverse ration of the molecular weights of the gases will determine the mass flux of the gases through the porous membrane. As shown in the following table, the square root of the ratio of the molecular weights between the methane and the carbon dioxide is 1.66 that represents the mass flux ration between the carbon dioxide and the methane through the porous silicon membrane.

The porous silicon was prepared by anodization of silicon wafers in concentrated HF solutions. The used silicon wafers were p-type silicon wafer (0.01-0.001 Ω-cm), polished on one side and oriented along the crystalline direction. The used HF solutions were composed of HF wt 49% and ethanol and the anodization took place in a double tank cell. The anodization was carried out in the dark at a constant current density from 18 to 36 mA/cm². The obtained porous silicon membrane had a pore size from 6 to 10 nm and a porosity of about 50 %.

In order to form a porous silicon membrane first, applied current density was 36mA/cm² and anodization time was 30 minutes. As a result, a porous silicon layer with a thickness of about 30 microns was created. Then, the applied current density was abruptly increased in order to enable the porous silicon layer detached from the silicon wafer. The abrupt increase applied current density led to a high porosity layer and high released gas pressure. So the porous silicon layer was easy being detached from the silicon wafer.

As an alternative, the porous silicon membranes can be obtained by combination of forming and etching of the porous silicon. First, a thick porous silicon layer was formed in a silicon wafer and then etched in a diluted KOH solution. Secondly, a thin porous silicon layer was formed by etching the leaved silicon layer in the silicon wafer.

After fabrication of the porous silicon membrane, the sample was rinsed gently with ethanol. Then a final rinse was carried out with hexane in order to minimize the possibility of shattering of the membranes due to strong capillary forces and thermal stresses exerted when ethanol evaporates from the pores. Finally, the membranes were dried in a nitrogen flow.

Low Power and Fast Response Carbon Dioxide Gas Sensors

Tu Xiang Zheng

Carbon dioxide can cause negative health affects to humans including drowsiness and at high enough concentrations suffocation. It has been recommended that the maximum time averaged exposure to the atmosphere containing 5,000 ppm carbon dioxide is not over eight hours. As such it is highly desirable to be able to measure carbon dioxide in order to control indoor air quality and in environmental monitoring.

The present author provides a MEMS carbon dioxide gas sensor as shown in the above figure. The sensor is based on a silicon wafer and fabricated utilizing CMOS technologies. Since the sensor is required to be operated at an elevated temperature a thermal insulating pad is formed in the silicon wafer which is used to support the sensor body. Both a resistor for heating and a thermopile for temperature sensing are formed on the thermal insulating pad. Then depositing an electrical insulating layer and laying a tin dioxide layer is formed thereon. By employing such device structure with good thermal insulation to the silicon wafer, the sensor presents a series of advantages such as miniaturized size, low power consumption, and fast response.

Reducing the size of the sensor is the most effective way to reduce overall power consumption. The size of the heater of the sensor can be reduced as small as 60 µm x 800 µm. This affords field operation on a single 9V battery for an acceptable time. In addition, shortening the thermal response time enables the sensor to be operated for a very brief period during measurement. This pulse heating provides a further reduction in the power consumption. All these capabilities make the sensors ideal for the applications ranging from low power wireless to cell phones and wearable and conformal sensing systems.

The thermopile of the sensor is not able to measure the absolute temperature, but generate an output voltage proportional to the temperature difference between the heater and the silicon wafer. This is suitable for the sensor to adapt a pulse width modulation (PWM) based temperature controlled circuit. It is a key to stabilize the temperature of the sensor since the response of the sensor increases and reaches the maximum at a certain temperature, and then decreases rapidly with increasing the temperature. In the modulation circuit a microcontroller is programmed to generate different modulating 8 bit digital signal which is converted to analog signal using digital to analog converter (DAC). The analog signal is then used to control a PWM based driver circuit which drives the heater of the metal oxide gas sensor.

The pulse width modulation (PWM) based temperature controlled circuit has been reported to have three advantages:

• A cyclic temperature variation can give a unique response for each gas as rate of reaction of the different analytic gases are different at different temperature;
• Low temperature may lead to the accumulation of incompletely oxidized contaminant, which may get removed during cyclic oscillating voltage; and
• Thermal cycling can lead to improvements in sensitivity because for each gas there is a heater voltage for which it shows maximum conductance-temperature characteristics.

Micromachined Vertical Vibrating Gyroscopes

Tu Xiang Zheng

As well known, micromachined vibrating gyroscopes have gained popularity in recent years. A main application of these sensors is to determine the yaw (vehicle rotation about its vertical axis) or roll (vehicle rotation about its lengthwise horizontal axis) angle of the vehicle. These sensors also play a key role in consumer electronics for platform stabilization in camcorders and video-game headsets.

The above mentioned US patent describes a micromachined vertical vibrating gyroscope that can be used to counteract the rolling effect on a vehicle, and thus, are a preferred stabilization tool for vehicles such as airplanes, ships, and cars. The emergence of micromachining technology has generated the possibility to produce gyroscopes that present many advantages over their counterparts, such as lower cost, small size, higher performance, and lower power consumption. These advantages explain why the gyroscopes have been widely used in smart phones.

The present gyroscope is implemented by vibrating the proof mass in a direction parallel to the substrate, either lateral or rotational, and sensing vertical displacement or torsional motion due to Coriolis effect.

The gyroscope consists of three single crystal silicon assemblies: an outer single crystal silicon assembly, an intermediate single crystal silicon assembly, and an inner single crystal silicon assembly. The outer assembly includes a plurality of arc-shaped anchors arranged in a circle and extending from a single crystal silicon substrate coated with an insulating annulus thereon. The intermediate assembly is a suspended wheel concentric with the arc-shaped anchors. The inner assembly is a suspended hub concentric with the circle formed by the anchors and having no axle at its center. The three assemblies are connected to each other through several flexures. The intermediate suspended wheel is driven into rotational vibration by lateral comb capacitors. Input angular rates are measured by two vertical capacitors. The gyroscope is fabricated utilizing a bipolar-compatible process comprising steps of buried layer diffusion, selective epitaxial growth and lateral overgrowth, deep reactive ion etching, and porous silicon processing.

In operation of the gyroscope a voltage is applied to the lateral driving capacitors. The intermediate wheel is then stimulated into rotational vibration about the coordinate z-axis that is set to be vertical to the substrate plane. For the rotational vibration of the wheel, the flexures provide flexible mechanical support. As the rotational angular becomes too large the stops begin to abate the vibration so as to prevent the flexures from damaging. The lateral monitoring capacitors is used to measure the frequency and amplitude of the rotational vibration of the wheel. When the substrate experiences an angular rate about the coordinate x-axis that is set to be perpendicular to the flexures a Coriolis force is induced. The Coriolis force exerts on the inner vibrating hub and causes the hub to be rotationally vibrated about the coordinate y-axis.

In the balance state the two vertical capacitors and are designed to be completely equal. When the hub rotates about the coordinate y-axis, the two vertical capacitors are no longer equal. If the hub rotates counterclockwise, the capacitance of the vertical capacitor will increase and the capacitance of the vertical capacitor will decrease. If the rotation direction reverses, the difference of the capacitance also reverses. Since the difference of the capacitance of the two vertical capacitors depends upon the input angular rate, the input angular rate can be determined by measuring the difference of the capacitance of the two vertical capacitors.

The measurement circuit of the gyroscope can be adopted in open loop or in close loop. In open loop the amplitude of a carrier signal can be modulated by the difference of the capacitance of the two vertical capacitors. After demodulation with the carrier frequency and the driving signal frequency a DC voltage proportional to the input angular rate can be yielded as the output of the measurement circuit. In close loop the yielded signal is first fed to a rebalance circuit. The rebalance circuit then provides a rebalance voltage applying to the vertical capacitors to null the rotation of the inner vibrating hub about the coordinate y-axis. The rebalance voltage is proportional to the input angular rate. So the measurement circuit can be implemented as a Σ∆ interface circuit. 

1986 Year’s Transparent Silicon Membranes Used for Ion Implantation Self Annealing

Tu Xiang Zheng

 

The above picture shows a thin silicon membrane supported by a thick silicon frame. In order to indicate the very small thickness of the thin silicon membrane a business card is placed under the silicon membrane. A letter “POSIFA” of the business card can be seen clearly through the thin silicon membrane, which means that the membrane is very thin so as to be transparent. Actually the thin silicon membrane is n-type single crystal silicon with a 3 Ω-cm resistivity and has 1 micron in thickness and 1cm in diameter. In general when the thickness of a silicon membrane is less than 5 micron the membrane can transparent part of the visible light.

The thin silicon membrane was fabricated by the present author in early 1986 year. At that time the present author was asked to fabricate a thin single crystal silicon membrane for the studies of single crystal silicon ion implantation self annealing. Thermal annealing in a furnace is the technique normally used to remove lattice damage and restore the electrical properties of ion implanted single crystal silicon. As an alternative, the annealing can be done utilizing the heating effect produced by the same ion beam during the implantation process. The heating temperature should be very high so that the ion implanted silicon layer can enter epitaxial regrowth phase. On the other hand the cooling of the silicon layer is slow enough allowing epitaxial regrowth to be carried out. This is why the thin single crystal silicon membrane is required, which can provide an excellent thermal insulation.

Etching stop technology enables the formation of thin single crystal silicon membranes. An early etch stop technology is based on P⁺ etch stop layers. This is because the rate of silicon etching depends on the boron concentration and decreases so dramatically that anisotropic etchants, especially KOH, barely attack boron doped (P⁺) silicon layers with boron concentrations around 10^-19cm^-3. Unfortunately, heavy boron doped silicon membrane can not be used for the studies of single crystal silicon ion implantation self annealing. The base resistivity of the silicon to be implanted should be higher than 1 Ω-cm so as to be able to fabricate semiconductor devices.

A lightly doped p-n junction can be used as an etch stop by applying a bias between the wafer and the etchants. But the etchants are limited to be alkali such as KOH. It has long been known that metal impurities including alkali metals (Na, K, and Li) can affect a variety of silicon device characteristics, including junction leakage, surface and bulk recombination, emitter to collector shorting, and gate oxide integrity. In addition, in order to apply a positive voltage to the silicon wafer a metal contact needs to be made thereon, which is not accepted by the silicon device fabrication due to the same reason.

The present author did a lot of literature survey and did not find any useful reference materials. The present author knew that he faced to a tough target and a difficult challenge, but he did not give up. He tried many new ways and finally found the best one is selective formation and selective etching of porous silicon.

Porous silicon is a material which is formed by anodization that is electrochemical oxidation of single crystal silicon in concentrated hydrofluoric acid (HF) solutions. The formation reaction is highly dependent on the type and level of silicon doping, and the material can be selectively formed on particular regions of a wafer that present appropriate doping characteristics obtained by diffusion or ion implantation.

The process started with an n/n ⁺ epitaxial silicon wafers with heavily boron doped substrates. The silicon wafer is patterned on the frond side and inserted into an etching cell for electrical contacting. The silicon wafer serves as anode. Two types of etching cells are used: double-cells and single-cells. By the use of a double cell the wafer is contacted with the aid of electrolyte on both sides. That means the electrodes of HF-resistant material (platinum or silicon) are placed into electrolyte and the induced current /applied voltage goes through the electrolyte to contact the wafer. The positive potential is applied on the back side and the negative potential on the front side of the silicon wafer, where the porous layer is generated. Then the porous silicon is etched selectively so as to leave a thin silicon membrane suspending by a silicon frame of the wafer. The etchants such as a mixture of HF and H₂O₂  may be used for selective etching of porous silicon.

Image Sensing Using Micromachined Ultrasonic Sensor Arrays

Tu Xiang Zheng 

Ultrasonic sensors convert ultrasound waves to electrical signals or vice versa. They detect a wide range of materials, are not influenced by problematic surfaces, and are largely immune to environmental influences. They have many uses in medicine as well as in other various advanced technologies including electronics, chemicals and construction. As well known, an ultrasonic sensor applied to the abdomen of a pregnant woman sends ultrasonic waves into the body and receives the echoes back from the inside, which are used for making visual images. These real time images showing the appearance and movement of the fetus allow observation of the development of the fetus.

Silicon based capacitive ultrasonic sensor arrays bring revolutionary improvement in performance and represent a major advance in ultrasonic sensor technology.

These sensors benefit from the economies of scale found in semiconductor manufacturing and are well suited for high-volume applications that demand high-performance sensors at low costs. A similar sensor array can be found in the US Patent 6,359,276 B1, which is used for sensing infrared image.

As shown in the above figure, each sensor of the array comprises two electrodes facing each other, one of which is fixed and the other is movable. The two electrodes are separated by an insulating layer and an air gap. It can operate on transmit and receive mode, by converting electrical energy into acoustic energy or vice versa through the displacement of the movable electrode.

When a voltage is applied between both electrodes and the membrane is pulled down to the bottom electrode by electrostatic forces. The membrane moves until the electrostatic force has equilibrium with internal force of the membrane. AC signals cause vibrations of the thin diaphragm and generate ultrasonic waves. Furthermore, the receiver can detect an ultrasonic wave using the change of capacitance when displacement of the membrane is caused by the pressure of an arriving ultrasonic wave.

In according to this patent, the sensor array is disposed in a silicon substrate in which there already exists a CMOS circuit with readout electronics. Each sensor includes a silicon nitride membrane bridging a cavity recessed into the substrate. The membrane has four beams. The distal ends of the beams are anchored to the substrate, so that the membrane is supported by the substrate and the surface of the beams is aligned with the plane surface of the substrate. The surface of the membrane and beams is coated with silicon dioxide film. A metal layer is disposed on the surface of the silicon dioxide film. The end portions of the metal layer are disposed on the beams and keep in contact with the proximal end portions of electrical conductors which already exist on the surface of the substrate. The cavity is a narrow gap, so that the membrane can touch the bottom of the cavity without damage as it is forced to bend downward. The trenches between the membrane and the beams and between the beams and the edges of the substrate are also narrow, so that the membrane and beams can touch the edge of the substrate, as they are forced to bend in the lateral directions. 

It was reported by Andrew et al. that MEMS technology makes it possible to produce air and gas ultrasonic sensors that can operate at higher frequencies (200 kHz to 5 MHz). The ability to make microscopic structures with MEMS technology permits the fabrication of very small sensors that emit high-frequency ultrasound. The smaller the sensor implies the higher the frequency of the ultrasonic signal. This was realized by Lemmerhirt et al. They developed a CMOS based ultrasonic 32 X 32 sensor array. The array was built with CMOS process for 3D image acquisition. Each sensor has 100 µm diameter membrane with 60 µm diameter top electrode and 0.6 µm gap. The center frequency of each element is 1.8 MHz.

Wireless Meter of Methane Number and Mass Flow of Natural Gas

Tu Xiang Zheng

  

Natural gas is used in an amazing number of ways. Although it is widely seen as a cooking and heating fuel in most households, natural gas has many other energy and raw material uses that are a surprise to most people who learn about them. In 2012 about 30% of the energy consumed across the United States was obtained from natural gas. It was used to generate electricity, heat buildings, fuel vehicles, heat water, bake foods, power industrial furnaces, and even run air conditioners.

Natural gas is a gaseous mixture chemically composed by methane, smaller fractions of higher molecular weight hydrocarbons and inert gases (mainly N₂ and CO₂). The different components ratio in the gas mixture determines its physical and chemical properties and consequently, its quality. Concretely, composition fluctuations affect to properties such as methane number.

The methane number is the parameter used to quantify the quality of the natural gas.  A 100 methane composition is given 100 methane number and as the higher hydrocarbons and inert gases percentage increases the methane number decreases. It is assigned that a 100 methane composition is used as the knock resistant reference fuel. Every natural gas engine has a higher than a 65 methane number to prevent engine knocking.

For methane number measurement many different sensor techniques are available in a variety of classes. Sensor principles include electrical techniques, like electrochemical detection, or electrical detection of adsorption by induced capacitance changes, optical techniques; for instance infrared (IR) adsorption or Raman spectroscopy, chromatography, calorimetry and acoustic analyses. What most sensor techniques have in common is that their applicability into real time monitoring systems is limited, either because sensors are hard to integrate based on practical considerations like, size, cost or response time, or because the sensors rely on principles that generally do not apply to all gasses.

The present paper proposes a new method to determine on line the methane number of natural gas. The method is based on the measurement of the gas density and the correlation between the density and the methane number of natural gas. The gas density can be measured with a thermopile flow sensor combining a differential pressure sensor. These two sensors are installed in an orifice plate. When natural gas flows through the orifice plate the mass flow rate and the pressure drop of the gas flow can be measured simultaneously. Then the density of the natural gas can be calculated based on the Bernoulli equation which states that there is a relationship between the pressure drop and velocity of the natural gas flow.

The correlation between the specific gravity and the methane number of natural gases is shown as the following table. The table gives methane number: 48.1, 66.2, 76.4, 80.8, 91.4 and 100 and the volume percent of their corresponding compositions. Using the specific gravity of each composition the specific gravity of each methane number can be calculated which is also given in the table. The specific gravy of each composition of the natural gas is shown in another table. It can be seen that the methane number increases and the specific gravity of the different methane number gases decreases. It is not surprising because the specific gravity of methane is lower than all other composition of the natural gas. 

A proposed wireless natural gas meter is shown in the above figure. The meter can measure both the methane number and wirelessly send the data to a smart phone for the user to monitor the consumption and the quality of the natural gas precisely. In order to do so a key component of the meter is a thermopile flow sensor developed based on a mix of integrated circuit manufacturing and micro-machining process. Some of the advantages of the thermopile flow sensors can be listed as

• Direct mass flow sensing;
• Large dynamic range;
• Fast response;
• Excellent low flow sensitivity;
• Low power consumption;
• Small size, mass, volume;
• low cost; and
• Easy to integrate in gas or fluid transport networks.

Thermopile Flow Sensors with Differential pressure sensors for Measurements of Mass Flow Rate, Density and Void Fraction of Gas-Liquid Two-Phase Flow Fluids

Tu Xiang Zheng

Gas-liquid two-phase flow exists broadly in chemical, petroleum and metallurgical industries. The measurements of gas-liquid two-phase flow parameters in real time without separating the phases is desirable in order to reduce costs, increase production and reach excellence in oil and gas transport. Although many measurement techniques have been developed, it is yet difficult to measure some flow parameters because of the complexity of the two-phase flow. It is necessary to explore new measurement techniques.

This paper proposes a measurement technique of gas-liquid two-phase flow parameters which has the advantages of low cost, simple structure and non-intrusiveness. This technique is based on the combined use of a thermopile flow sensor and a differential pressure sensor. The setup is shown in the above figure which consists of a venture, a bypass tube, a thermopile flow sensor, and a differential pressure sensor. The thermopile flow sensor is installed at the central point of the bypass flow tube and the differential pressure sensor is used to measure the pressure difference between the inlet and the outlet pressure of the bypass tube. The mass flow rate measured by the thermopile flow sensor can be used to derive the main mass flow rate passing through the venture according to a ratio dependent on the setup structure. Knowing the mass flow rate and pressure difference the density of the gas-liquid two-phase fluid can be obtained. The void fraction of the gas-liquid two-phase fluid can be further calculated using know pure liquid density and pure gas density.  

As can be seen from the above figure, there are two flow loops: the venture loop and the bypass loop. In accordance with the nature of these two parallel loops, it is reasonable to have their pressure drop to be equal.

According to the homogeneous model of two-phase flow two phases travel at equal velocities and mix well; therefore, they can be treated as if there is only one phase.

Using Bernoulli's equation in the special case of incompressible flows (such as the flow of water or other liquid, or low speed flow of gas), the theoretical mass flow rate through the venture can be given by:

m_v = C_vA₂{2(p_up-p_down)/ρ[1-(A₂/A₁)²]}^1/2   (1)

where m_v is the mass flow rate of the fluid, C_v is the discharge coefficient = (actual flow rate) / (theoretical flow rate), A₁ and A₂ is the cross-sectional area of the venture in the indicated section of the venture, p_up, p_down are the fluid's static pressure in the indicated section of the venture, and ρ₁ is the fluid's density passing through the venture.

Similarly, the theoretical mass flow rate through the bypass tube can be given by:

m_b = C_bA₃[2(p_up-p_down)/ρ₂]^1/2                       (2)

where m_b is the mass flow rate of the fluid, C_b is the discharge, A₃ is the cross-sectional area of the bypass tube, p_up, p_down are the fluid's static pressure in the indicated section of the bypass tube, and ρ₂ is the fluid's density passing through the bypass tube.

In arriving at the homogeneous model for two-phase flow, area averaging is performed for both phases and the density ρ₁ and ρ₂ must be equal a similar ρ. Forever the void fraction α of the two-phase fluid must satisfied the relation:

  ρ = (1-α)ρ_l + αρ_v                                           (3)

where ρ_l and ρ_v are the densities of the pure liquid and the pure gas, respectively.

It is clear that a thermopile flow sensor with a differential pressure sensor can be used to measure the mass flow rate, density and void fraction of a gas-liquid two-phase flow. A proposed setup comprises a thermopile, a differential pressure sensor, a venture tube and a bypass tube. Using the venture equation and the homogeneous model form the mass flow rate measured by the thermopile flow sensor and the pressure difference measured by the differential pressure sensor the density and the void fraction of the two-phase fluid can be derived respectively. This technique requires the gas and the liquid mixes so well that the mixture can be seen as one single phase approximately. The mixing degree of the gas and the liquid affects the measuring accuracy. For un-complete mixing two-phase fluid correction is needed to improve the accuracy.

Thermopile Flow Sensors and Differential Pressure Flow Meters

Xiang Zheng Tu

 

Air conditioning can refer to any form of technology that modifies the condition of air including heating, cooling, (de-)humidification, cleaning, and ventilation. In order to do so air movement needs to be created and air flow needs to be measured. This is hwy differential pressure flow meters are popular for a long time. But in recent years things have changed. Differential pressure flow meters have been gradually and irreversibility replaced by thermopile flow sensors.

It is not surprising in view on the working mechanism of the differential pressure meters and related limitations to the flow measurements. The working mechanism is based on Bermoulli’s Equation. Bernoulli’s equation states that the pressure drop across the constriction is proportional to the square of the flow rate, as shown in the following figure.

 

It can be seen from the above figure that using this relationship, 10 percent of full scale flow produces only 1 percent of the full scale differential pressure. At 10 percent of full scale flow, the differential pressure flow meter accuracy is dependent upon the meters being accurate over a 100:1 range of differential pressure. The meters accuracy is typically degraded at low differential pressures in its range, so flow meter accuracy can be similarly degraded. Therefore, this non-linear relationship can have a detrimental effect on the accuracy and turndown of differential pressure flow meters. Remember that our interest is the accuracy of the flow measurements instead of the differential pressure measurements.

In addition, the flow rate measured by the differential pressure flow meters is not mass flow rate that is required by many applications. According to ideal gas law, gas pressure changes with its temperature and volume. To obtain a mass flow rate, it is necessary to measure additional parameters: differential pressure; absolute pressure; and absolute temperature. These measurements with the differential pressure measurement then sent to a computer for calculating the mass flow rate.

All these limitations with the differential pressure flow meters can be eliminated by the thermopile mass flow sensors. The thermopile flow sensors use the thermal properties of the fluid to measure the flow rate. A measured amount of heat is applied to the heater of the sensor. Some of this heat is lost to the flowing fluid. As flow rate increases, more heat is lost. The amount of heat lost is measured using the thermopile(s) in the sensor. The output of the thermopile(s) represents the fluid flow velocity or flow rate.

The thermopile flow sensors are fabricated using micromachining techniques in a CMOS production line. They offer many advantages over the differential pressure meters, including but not limited to:

• Large dynamic range
• High accuracy
• Excellent low flow sensitivity
• Direct mass flow sensing
• Low pressure drop
• Very low power consumption
• Miniaturization and small device footprint
• Manufactured in CMOS production line and low cost

The thermopile flow sensors are not only used for air flow measurement in air conditioning, but also for monitoring flow in clean room, in fan/filter units, controlling flow in production facilities in the pharmaceutical, food processing and semiconductor industries; and monitoring flow in glove boxes, insulators, medical equipment such as anaesthetic machines and respirators in order to maximize energy efficiency, and also increase the accuracy of gas flow control.

Low Detectable Range Thermopile Flow Sensors

Xiang Zheng Tu

Thermal flow sensors are inevitably influenced by natural convection. An operating thermal flow sensor is a hot source surrounding by air. The air receives heat from the sensor, becomes less dense and rises. The cooler air then moves to replace it. This cooler air is then heated and the process continues, forming the convection current around the sensor. The temperature of the sensor will changes due to a part heat of the sensor is carried away by the convection current. Therefore the temperature change will add to the signal of sensor as a part of offset of the sensor.

But for thermopile flow sensors the natural convection can be reduced as low as being ignore. The installation of the thermopile flow sensors in a tube is commonly required to have their hot surface facing down to the Earth’s surface. In this case the heat transferred from the sensor chip by natural convection is the lowest. This is not surprising, since the hot air is “trapped” under the sensor chip and can not move away from the sensor chip easily. As a result, the cooler air in the vicinity of the sensor chip will have difficulty reaching the sensor chip, which results in a reduced rate of heat transfer.

Even the hot surface of the sensor intersects the Earth’s surface at a specific angle, the formed natural convection current is small and the influence still can be ignore. The reason can be explained as follows. Natural convection is characterized by Grashof number G_r which expresses the ratio between buoyancy forces due to spatial variation in air density to viscous forces acting on air. It is given as:

Gr = gβ(T_sensor - T_air )L³ /  ע² ,                   (1)

where g is the acceleration due to gravity; β is the volumetric thermal expansion coefficient; T_s and T_air are temperature of the heater of the sensor and the surrounding air, respectively; L is the characteristic length and the ע is the kinematic viscosity of the surrounding air. With g = 9.8 m/s², β = 3.25x10^-3 /k,   ע = 1.65 x 10^-5, T_s - T_air = 21.9 ^oC, and assuming the characteristic dimension L = 1x10^-3 m, we have

Gr = gβ(T_s - T_air )L³ /  ע² = 2.56    (2)

As well know, the ratio Gr/Re₂ defines the importance of natural convection in respect to a forced convection. The R_e is Rayleigh Number which represents forced convection and can be expressed as R_e = vL/ ע. The thermopile flow sensors provided by POSIFA Microsystems have been widely used for air mass flow meters. The characteristic length of the house tube of the air mass flow meter is 6mm. If assuming the velocity of air flow is 0.1mm/s, the Re is calculated to be 3.64 and the ratio Gr/Re₂ to be 0.19. This is lower than the lower bound of Gr/Re2 =0.3, which means that compare with the forced convection the contribution of the natural convection to the heat transfer can be negligible. (E. M. Sparrow, R. Eichhorn, and J. L. Gregg, “Combined forced and free convection in a boundary layer flow,” Phys. Fluids, vol. 2, no. 3, pp. 319–328, 1959.)

In addition to natural convection, the noises of the thermopile flow sensors are also limiting the detectable flow velocity. The noises of the thermopile flow sensors are basically the temperature noise and the thermal noise. The temperature noise is caused by temperature fluctuations in the surrounding atmosphere. This noise can be cancelled because the thermopile flow sensors are responsive to the temperature difference between a hot place and a cold place instead of a temperature. The thermal noise or the Johnson noise is an electrical noise source caused by random motion of electrical charges in the material. It is determined by the following equation:

V_noise = (4k_bT_eR∆f)^1/2               (3)

where k_B is the Boltzmann’s constant; T_e is the absolute temperature in kelvin ; R is the electrical serial resistance and ∆f is the frequency bandwidth.

Thermal noise calculation can be carried out online. The following results are given by http://www.sengpielaudio.com/calculator-noise.htm.

In the calculation T = 45.9 ⁰c representing the operating temperature of the thermopile flow sensor, R = 150 kΩ representing the serial resistance of a thermopile of the sensor. The bandwidth is calculated using equation ∆f = f_cutoff – 20 Hz, where f_cutoff is determined by measuring the response time of the sensor, as shown in the following figure.

 

It can be obtained from Fig.1 that the response time of the sensor is 0.74ms corresponding cutoff frequency 1351 Hz. With these parameters the thermal noise of the sensor is 1.88µV.

In conclusion, the thermopile flow sensors provided by POSIFA Microsystems can detect as low as 0.1mm/s velocity for air mass flow meter applications. This is resulted by the inherence low Grashof number of the sensors, which is down to 2.56. The temperature noise of the sensor can be canceled since the sensor is responsive to the temperature difference instead of temperature. The thermal noise or the Johnson noise is calculated to be 1.88µV, which is very low and is suitable for medical applications.  Furthermore the cutoff frequency of the sensors is up to 1331Hz, which is required by several medical instruments such as spirometers.

Low Detectable Range Thermopile Flow Sensors

 Xiang Zheng Tu

Thermal flow sensors are inevitably influenced by natural convection. An operating thermal flow sensor is a hot source surrounding by air. The air receives heat from the sensor, becomes less dense and rises. The cooler air then moves to replace it. This cooler air is then heated and the process continues, forming the convection current around the sensor. The temperature of the sensor will changes due to a part heat of the sensor is carried away by the convection current. Therefore the temperature change will add to the signal of sensor as a part of offset of the sensor.

But for thermopile flow sensors the natural convection can be reduced as low as being ignore. The installation of the thermopile flow sensors in a tube is commonly required to have their hot surface facing down to the Earth’s surface. In this case the heat transferred from the sensor chip by natural convection is the lowest. This is not surprising, since the hot air is “trapped” under the sensor chip and can not move away from the sensor chip easily. As a result, the cooler air in the vicinity of the sensor chip will have difficulty reaching the sensor chip, which results in a reduced rate of heat transfer.

Even the hot surface of the sensor intersects the Earth’s surface at a specific angle, the formed natural convection current is small and the influence still can be ignore. The reason can be explained as follows. Natural convection is characterized by Grashof number G_r which expresses the ratio between buoyancy forces due to spatial variation in air density to viscous forces acting on air. It is given as:

Gr = gβ(T_sensor - T_air )L³ /  ע² ,     (1)

where g is the acceleration due to gravity; β is the volumetric thermal expansion coefficient; T_s and T_air are temperature of the heater of the sensor and the surrounding air, respectively; L is the characteristic length and the ע is the kinematic viscosity of the surrounding air. With g = 9.8 m/s², β = 3.25x10^-3 /k,   ע = 1.65 x 10^-5, T_s - T_air = 21.9 ^oC, and assuming the characteristic dimension L = 1x10^-3 m, we have

Gr = gβ(T_s - T_air )L³ /  ע² = 2.56     (2)

As well know, the ratio Gr/Re₂ defines the importance of natural convection in respect to a forced convection. The R_e is Rayleigh Number which represents forced convection and can be expressed as R_e = vL/ ע. The thermopile flow sensors provided by POSIFA Microsystems have been widely used for air mass flow meters. The characteristic length of the house tube of the air mass flow meter is 6mm. If assuming the velocity of air flow is 0.1mm/s, the Re is calculated to be 3.64 and the ratio Gr/Re₂ to be 0.19. This is lower than the lower bound of Gr/Re2 =0.3, which means that compare with the forced convection the contribution of the natural convection to the heat transfer can be negligible. (E. M. Sparrow, R. Eichhorn, and J. L. Gregg, “Combined forced and free convection in a boundary layer flow,” Phys. Fluids, vol. 2, no. 3, pp. 319–328, 1959.)

In addition to natural convection, the noises of the thermopile flow sensors are also limiting the detectable flow velocity. The noises of the thermopile flow sensors are basically the temperature noise and the thermal noise. The temperature noise is caused by temperature fluctuations in the surrounding atmosphere. This noise can be cancelled because the thermopile flow sensors are responsive to the temperature difference between a hot place and a cold place instead of a temperature. The thermal noise or the Johnson noise is an electrical noise source caused by random motion of electrical charges in the material. It is determined by the following equation:

V_noise = (4k_bT_eR∆f)^1/2       (3)

where k_B is the Boltzmann’s constant; T_e is the absolute temperature in kelvin ; R is the electrical serial resistance and ∆f is the frequency bandwidth.

Thermal noise calculation can be carried out online. The following results are given by http://www.sengpielaudio.com/calculator-noise.htm.

  

In the calculation T = 45.9 ⁰c representing the operating temperature of the thermopile flow sensor, R = 150 kΩ representing the serial resistance of a thermopile of the sensor. The bandwidth is calculated using equation ∆f = f_cutoff – 20 Hz, where f_cutoff is determined by measuring the response time of the sensor, as shown in the following figure.

It can be obtained from Fig.1 that the response time of the sensor is 0.74ms corresponding cutoff frequency 1351 Hz. With these parameters the thermal noise of the sensor is 1.88µV.

In conclusion, the thermopile flow sensors provided by POSIFA Microsystems can detect as low as 0.1mm/s velocity for air mass flow meter applications. This is resulted by the inherence low Grashof number of the sensors, which is down to 2.56. The temperature noise of the sensor can be canceled since the sensor is responsive to the temperature difference instead of temperature. The thermal noise or the Johnson noise is calculated to be 1.88µV, which is very low and is suitable for medical applications.  Furthermore the cutoff frequency of the sensors is up to 1331Hz, which is required by several medical instruments such as spirometers.

Thermopile Flow Sensors Operating at as low as 45.9⁰C

Tu Xiang Zheng

There are two types of popular micromachined thermal flow sensors: resistive flow sensors and thermopile flow sensors. Both the thermal flow sensors work by heat convection transfer away from a heated resistor. As the resistor cool, the corresponding change in voltage or current can be calculated to fluid flow. The major difference between the resistive flow sensors and the thermopile flow sensors is the heat sensing element. Their heat sensing elements are unheated resistors and thermopiles, respectively.

An infrared camera is used to take the temperature image of a thermopile flow sensor. To do this, a DC voltage is applied to the resistor of the sensor. A typical temperature image is shown in Fig.1.The brightest (warmest) parts of the image are customarily colored white, intermediate temperatures reds and yellows, and the dimmest (coolest) parts black.

As can be seen, the highest temperature is 45.9⁰c at the central region of the sensor chip and the lowest temperature is 24.4⁰c at the surrounding region of the sensor chip. It can be seen from this image that the thermopile flow sensor can be operated at 45.9⁰c. This operating temperature is much lower than the operating temperature of any resistive flow sensors, which is usually over 100⁰c.

The circuit module of the thermopile flow sensor is also shown in Fig.1. The module includes a thermopile flow sensor, a microcontroller, a regulator, and an amplifier. The regulator may input a buttery voltage and output a regulated voltage to the microcontroller. The microcontroller may create a pulse width modulation (PWM) voltage to the heated resistor of the thermopile flow sensor. The thermopile of the thermopile flow sensor may provide a static output voltage to the amplifier. The microcontroller may process the static output voltage for adjusting the PWM voltage so as to set an original offset of the amplifier to be as close to zero as possible.

In order to operate the thermopile flow sensor at 45.9⁰c, the following settings should be made. The buttery voltage is 5V and the regulated voltage is 3V. The resistance of the resistive heater is 240Ω. The PWM applied to the resistive heater is 2.54V. These settings result in a heating power of 26.9mW, an operating temperature of 45.9⁰c, and a 20mV static output of a thermopile. The 20mV static output is the original offset of the thermopile, which may drift over time. It is necessary to be able to maintain the original offset by adjusting the PWM output accordingly. For this reason, the PWM is set a duty cycle of 60% yielding 1.8V and 90% yielding 2.7V so that a duty cycle is 84.7% can yield 2.54V PWM output.

The low temperature operating thermopile flow sensors can provide several advantages over high temperature operating resistive flow sensors. An outstanding advantage is that oil droplets can be avoided to form around the sensor chips when they use for automobile air mass flow meters. The reason for this can be explained using Fig. 2 that is a graph showing the relationship between partial pressure and temperature of gasoline vapor in a gaseous mixture of air. 

Reference to Fig.2, P_ms represents the saturated vapor pressure curve and P_m represents the un-saturated vapor pressure curve. The yellow star with 24.4⁰c indicates the temperature and un-saturated vapor pressure of the gaseous mixture and the green star with 45.9⁰c represents the temperature and saturated vapor pressure of the gaseous mixture in the temperature boundary layer over the heated sensor chip. As can be seen, when the gaseous mixture in the temperature boundary layer over enters its surrounding space it still keeps un-saturated. But if the sensor chip is heated up to higher than 60⁰c as indicated by the red star, the gaseous mixture entered the surrounding space will become saturated and condense to be gasoline droplets.

It should be noted that a laminar flow is supposed to form over the heated sensor chip with a temperature boundary layer built up thereon. Since the gaseous mixture of air in the temperature boundary layer is heated, its volume increases and its partial vapor pressure decreases correspondingly. This will result in a partial vapor pressure difference between the temperature boundary layer and the surrounding space which drives vapors diffuse from the surrounding space to the temperature layer until reach the balance between these two spaces. This is why the vapor pressure in the temperature boundary layer on the heated sensor chip is indicated by the green star vapor pressure instead of the corresponding yellow star vapor pressure.

No Oil Droplets Formation in Thermopile Flow Sensors

Used in Internal Combustion Engines

Tu XianZheng

Hot-film air mass sensors are commonly used for measuring an air-mass flow, which include a resistor for heating and one or two other resistors for temperature sensing. It has been reported that when hot film air mass flow sensors are used directly in the intake tract of the internal combustion engine or in a bypass channel to the intake tract of the internal combustion engine, oil may deposit on the sensor chips and, in particular, on the sensor diaphragms during operation or shortly after the internal combustion engine has been shut off. This oil deposit may result in undesirable effects on the measuring signal of the sensor chip, in particular because an oil film affects the thermal conductivity of the sensor chip surface, which results in corruption of the measuring signals or a signal drift.

As well known, condensation can produce water droplets on the outside of soda cans or glasses of cold water. When warm air hits the cold surface, it reaches its dew point and condensed. As result droplets of water leave on the glass or can.

It is the same that when a thermal air mass flow sensor is operated at the border region of the heated measuring areas oil accumulates and over time results in oil droplets. The air flow drives the oil droplets on the surface up to the boundary of the heated measuring area, at which a stronger temperature gradient appears. The strong temperature gradient exerts a force opposite to the force exerted by the air flow. Oil droplets thus accumulate on the boundary line, which, when they reach a certain size, may be entrained again by the air flow to then contaminate the surface of the measuring area. In addition to the oil droplets, other contaminants also reach the surface of the measuring area due to this effect.

How to solve this problem? Back to the above mentioned condensation. Condensation of water occurs when water vapor within the air cools enough in order to change into the liquid state. A good example of condensation often occurs in the home during winter time, when water droplets form on the surfaces of cold windows. If open the window to let the cold air enter the room there will no any water droplets form on the surface of the window. So the only way to solve the problem is to reduce the operation temperature of the sensor. This can be done using a thermopile air mass flow sensor instead of a hot film air mass flow sensor.

A thermopile is an electronic device that converts thermal energy into electrical energy.

It does not respond to absolute temperature, but generates an output voltage proportional to a local temperature difference or temperature gradient. A thermopile air mass flow sensor is constructed with a heater for heating and several thermocouples for temperature sensing. The thermopiles are in series and so the output voltage due to temperature change is summed and increased over that of a single thermocouple.

With this advantage, POSIFA has developed thermopile air mass flow sensors with two group thermopiles positioned at the two side of the resistor heater and each group consisting of 40 thermopiles. Since the summed output of each group is so great that the operation temperature of the heater can be reduced as low as no condensation to take place. For example, the operation temperature can be set 10 to 20 degree Celsius higher than the air temperature. In this case condensation of oil vapor in the air is almost impossible.