A Microflow-Based Differential Pressure Sensor

Microbridge Technologies Canada, Inc has developed a nanoliter-airflow sensor that uses the thermoanemometer flow sensing principle and is combined with a microscopic channel with very high flow impedance for accurate sensing of low differential pressures. The flow impedance is predefined at the die level, dramatically relaxing demands on subsequent packaging operations. Individual units require only offset and gain compensation, leaving barometric pressure and temperature correction to be implemented in standard post-sensor lookup tables. The resulting field-replaceable differential pressure sensors do not require field calibration; can tolerate variability of connection hoses, changing gas filter properties, and humidified air; and enable substantial price reductions and improved performance and ease of use.

Basic Thermoanemometer-Type Micro-Airflow Sensing
Thermoanemometer-type micro-airflow sensors have been well known for several decades. Air travels through a flow channel that guides the air over a central heating element, which then locally heats a small volume of gas (Figure 1). The heated volume is displaced by the flow and the resulting imbalance in temperature is measured by a pair of temperature sensors that are positioned symmetrically on either side of the heating element. The speed with which the air flows through the flow channel is determined by the difference in pressure between the two ends of the channel and by its flow impedance, measured in pressure difference divided by flow rate in ml/s.


Figure 1. In a thermoaneometer-type airflow sensor, air is guided over a central heating element (overhead view shown in A). The airflow shifts the warmed air, causing the temperature sensors flanking the heater to measure a temperature differential (side view shown in B)


Therefore, micro-flow sensors can be used as sensors for differential air pressure as long as the pneumatic impedance of the micro-flow sensor is (1) sufficiently consistent unit-to-unit, (2) high enough to avoid a large effect on the ambient pressures P1 and P2 at the two ends of the flow channel, and (3) high enough to minimize air leakage through the flow channel.

A Novel Integrated Nanoliter-Airflow Sensor
Figure 2 conceptually summarizes the integrated nanoliter-airflow sensor, so named because its minimum detectable airflow is typically <10 nl. A full scale pressure of 250 Pa means a maximum flow of 2.5 µl/s though the sensor with 100 kPa/(ml/s) impedance. The typical resolution of the sensor is better than 1/500 of full scale, therefore the minimal detectable airflow is <2.5/500 = 5 nl/s.


Figure 2. The chip diagram for the nanoliter-airflow sensor


As shown in the diagram, the chip consists of the thermoanemometer-type (TA-type) sensing element, an integrated flow-channel, and an analog sensor conditioning circuit with Rejustors (programmable analog resistors). The chip is approximately 2.0 mm by 3.0 mm and is powered by 5 VDC. It provides two analog outputs: Vout-a, an analog output that includes offset and gain adjustment; and Tsens, the ambient temperature sensed onchip.

The sensor conditioning circuit can be designed to provide low or high gain. Fine calibration of manufacturing variations is done using Rejustors embedded within the analog CMOS circuit. While the prototypes described in this article are powered by 5 V, the technology also allows nominal supply voltages from 2.5–5 V for both sensor and CMOS circuitry.

The Benefits of High Flow Impedances
The flow impedance of Microbridge's sensor die can be designed to be anywhere in a range from ~20 kPa/(ml/s) to ~600 kPa/(ml/s). In this article we consider two manufactured prototypes: Type A has a flow impedance of 60 kPa/(ml/s); Type B's flow impedance is 110 kPa/(ml/s). These flow impedances exceed the presently available ranges on the market, that are 1.75 kPa/(ml/s) maximum and 200–500 Pa/(ml/s) typical.

High flow impedance is an important feature of nanoliter-airflow sensors and it is generally useful in many applications:

  • Long tubing connections can be used without changing the sensor's calibration, even though the sensor is used with differing tubing lengths/diameters. This is because the overall flow impedance is dominated by the sensor flow channel geometry rather than the tubing characteristics.


  • It allows the use of gas filters whose flow impedance is prone to change over time and use, since overall flow impedance is dominated by the sensor flow channel geometry rather than that of the gas filter.


  • In shunt configurations, there is low flow through the sensor, making the sensor easier to protect from contaminants.

Most importantly—in the context of Microbridge's sensors—these high flow impedances substantially reduce demands on subsequent packaging operations, giving more flexibility with package types. In most sensors presently on the market, the flow impedance is largely determined or affected by the packaging, which is prone to dimensional and alignment variations. Figure 3 shows two generic examples of configurations suitable for packaging these differential pressure sensors.


Figure 3. Two possible packaging configurations


Raw Unamplified Response of the Prototypes
As part of basic characterization of manufactured devices, Figure 4 compares the raw, unamplified response (in mV vs. differential pressure in Pa), of the differential pressure sensor prototypes Type A and Type B, described earlier. Based on their lower flow impedance, Type A sensors (sensors #1 and #2) have higher pressure sensitivity than Type B sensors (sensor #3). All sensors have a monotonic pressure response over a wide pressure range, up to 1500–2000 Pa for Type A and up to >3000 Pa for Type B.


Figure 4. Graph of pressure response for both Type A and Type B sensors


Note that for Type A devices, the curves become flatter for differential pressures >2000 Pa. All of the sensors have nonlinear sensitivity, which is typical for thermoanemometers.

Low-Gain Amplified Sensor Response After Adjustment
Figure 5 shows sets of normalized and offset-adjusted response curves for low-gain amplified sensor prototypes. For gain adjustment of Type A sensors the curves were all normalized to their values at 500 Pa differential pressure and for Type B devices the curves were all normalized to their values at 2000 Pa. Although the sensitivity curves are nonlinear, they are very reproducible, up to at least ±1000 Pa for Type A and at least ±2000 Pa for Type B. The normalized full-scale voltages can be scaled by adjusting the design of the onchip signal conditioning circuit and Rejustors.


Figure 5. Graphs of pressure response, normalized and adjusted for offset


Such unamplified (or low-amplified) sensors are for applications—such as medical devices—that need a high dynamic range (~10,000 times and higher), and a nonlinear response approximately proportional to the square root of pressure at high differential pressures. In such cases, the characteristic nonlinear response can be linearized using standard digital correction.

To evaluate the noise performance, the raw, unamplified output signal of the Type A sensor was measured during 10 s and 100 s time intervals. The measurements are shown in Figure 6. With a discernible monotonic response from 0.1 Pa up to 1500–2000 Pa, the unamplified sensor has dynamic range greater than 10,000X.


Figure 6. Noise behavior of raw, unamplified Type A sensor output measured at 10 s and 100 s intervals


Amplified Sensor Response
As indicated in Figure 2, the manufactured differential pressure sensors have onchip CMOS signal conditioning circuitry that includes amplification. In this case, the signal can be amplified such that it saturates before any significant pressure nonlinearity (e.g. at ±50 Pa for Type A, or at ±250 Pa for Type B). The results measured at room temperature are shown in Figures 7 and 8. The zero offsets have been compensated by trimming the associated onchip offset Rejustors, and the sensitivities have all been adjusted to a common value, by trimming the onchip gain Rejustors. Within the five samples in each type, the corrected linear curves are the same to within 0.5%.


Figure 7. Response of the amplified Type A nanoliter-airflow sensor after offset and gain compensation


Figure 8. Response of the amplified Type B nanoliter-airflow sensor after offset and gain compensation


To evaluate the sensors' noise performance, the raw, amplified output signal of the Type A sensor was measured during 10 s and 100 s time intervals. The measurements are shown in Figure 9. With a discernible monotonic response from 0.1 Pa up to >50 Pa, the amplified sensor has dynamic range greater than 100X.


Figure 9. Noise behavior of raw, amplified Type A sensor output measured at 10 s and 100 s intervals



Field-Replaceable Nanoliter-Airflow Sensors
We believe that these sensors fall into two main categories of use:

  • Unamplified or low-gain amplified sensors for applications such as medical devices that need a high dynamic range (~10,000X), an intentionally nonlinear response, and where the output varies roughly as the square root of differential pressure at high differential pressures.


  • Amplified sensors configured such that the amplified signal saturates before any significant pressure nonlinearity, for applications that need a linear response and a high dynamic range (~100X), for example, in the automotive, industrial, and HVAC fields.

    In both cases, simple offset and gain compensation enable field-replaceable sensors with very repeatable performance. Post-correction (e.g., digital correction) can be standard for all sensors of a given category:


  • For low-gain amplified sensors, after simple offset and gain correction, the resulting curve is typical for thermoanemometers and does not vary significantly with minor manufacturing variations, up to pressures of ±1000 Pa or ±2000 Pa. Post-correction can consist of a lookup table for standard correction of pressure nonlinearity, barometric pressure, and ambient temperature, none of which vary significantly from unit to unit.


  • For amplified sensors, the linear curves are the same to within 0.5%. For these devices, post-correction can consist of a lookup table for standard correction of barometric pressure and ambient temperature, neither of which vary significantly from unit to unit.

For both categories, the sensors are field-replaceable without needing individual unit-by-unit field calibration.

Figure 10. Performance comparison

ParameterMicrobridgeCompetitor 1Competitor 2
Unamplified differential pressure sensor Type A (B)Amplified differential pressure sensor Type A (B)
Pressure range±2000 Pa
(±4000 Pa)
±50 Pa
(±250 Pa)
–5 to 125 Pa–100 to 3500 Pa±1000 Pa
Output voltage (FSO)±200 mV4.3 V4 V4 V20 mV at 600 Pa
Resolution0.1 Pa
(0.5 Pa)
0.1 Pa
(0.5 Pa)
0.1 Pa0.5 PaNA
Response time~ 1 ms~ 1 ms40 ms40 ms1–3 ms
Pneumatic impedance60 kPa/(ml/s)
(110 kPa/
60 kPa/(ml/s)
(110 kPa/
0.06 kPa/
1.75 kPa/
0.5 kPa/
Excitation (VDC)5 V5 V5 V5 V10 V


The sensor's pressure range is determined by the die-level flow impedance, and onchip amplification. Since the offset and amplification are onchip adjustable, this enables the creation of a precalibrated sensor die, which can be more easily packaged than typical micro-airflow sensors on the market today (Figure 10). This technology enables substantial price, performance, and ease-of-use improvements.