Signal conditioning for resistive strain gauge sensors in mobile applications Robert Wendlandt1, Christoph Warncke2, Jörg Müller2 1) University Hospital Schleswig-Holstein Biomechanics Laboratory, Ratzeburger Allee 160, D- 23538 Lübeck 2) Hamburg University of Technology Micro systems technology, Eissendorfer Strasse 42, D-21073 Hamburg Robert.Wendlandt@uk-sh.de Abstract—Resistive strain gauge sensors are used for precise measurement of mechanical forces. Although the sensor elements (strain gauge foils) are inexpensive, a sensor requires costly manual trimming. Signal conditioning and data acquisition ICs can render manual trimming dispensable. Different systems based upon a programmable system on chip and also dedicated sensor signal conditioning ICs are analyzed for absolute sensitivity and sampling rate. The medical application of strain gauge sensors in conjunction with a sensor signal conditioning IC in a robotic external fixator is shown to deliver accurate force measurements. acquisition are compared with the single-chip design using a reference force sensor. The designs are discussed for the use in a robotic hexapod external fixator [1] in terms of system complexity, flexibility and sufficient data rate. Index terms—strain gauge, sensor, mobile I. INTRODUCTION Resistive strain gauge sensors are commonly used for precise and reliable measurements of mechanical forces in different areas of application ranging from space and industrial applications to your bathroom scale. The usually small change of resistance of the strain gauge is related to the mechanical strain of a supporting object. Although the strain gauge foils are inexpensive to manufacture, complete sensors are costly. Time consuming manual processing steps, gluing the sensors to the object and especially trimming a Wheatstone-bridge with abrasive resistors for offset and thermal shift, are setting the price rather than material costs. Offset compensation with operational amplifiers or lock-in techniques and subsequent digital processing can render manual bridge trimming dispensable yet electronic complexity has to be considered as well. This paper describes the application of a singlechip signal conditioning and data acquisition system that connects resistive strain gauge sensors with a general purpose microcontroller for smart sensors. Different system architectures for data Fig. 1: Robotic hexapod external fixator The sensor element of the robotic fixator (fig. 1) is made from stainless steel tube with a diameter of 11mm. Two resistive strain gauge half-bridges (Vishay Measurements Group GmbH) are glued to the tube which is then integrated into the mechanical structure of an external fixator system for fracture treatment. The sensor shows a sensitivity of 440µV/V at 1000N while the zeropoint offset is typically in the same order of magnitude. II. MATERIALS AND METHODS Data acquisition systems (configuration A – E) for untrimmed resistive strain gauge sensors for mobile medical applications were designed using a programmable system-on-chip (PSoC) microcontroller (CY8C29x66, Cypress Semiconductor) and two different external sensor signal conditioning ICs with integrated ADC (PGA309, Texas Instruments and SX8723, Semtech Corporation). To compare the performance of the different designs a commercially available force sensor (8435, Burster Präzisionsmesstechnik GmbH & Co KG) was connected to the systems. The sensor has a sensitivity of 1 mV/V at the full load of 1 kN. It is assembled out of resistive strain gauge foils in a full-bridge configuration. The bridge resistance of the sensor is 350 Ω. The overall system design is similar for all of the five designs. The PSoC microcontroller is powered by 3.3 V low dropout regulator also powering the external sensor signal conditioning IC, if present, as well as supplying bias to the sensor. The PSoC uses its internal 12 MHz clock. A USB-to-serial converter is used for data transmission between the PSoC and a control PC. Sections A – E describe the design differences of the analyzed systems. Measurements are performed at room temperature. Weights from 0 kg up to 7 kg in increments of 1 kg are attached to the sensor and 1000 samples of data are taken. The data samples for the seven measurements are evaluated in Excel (Microsoft Corporation). Mean value, standard deviation and slope of the linear regression of the mean values are determined. The absolute sensitivity (slope of linear regression divided by standard deviation) is calculated to compare the different systems. A. Internal amplifier and ADC Only internal function blocks of the PSoC are used to implement an amplification of the small scale sensor signal. The instrumentation amplifier block (INSAMP) has a maximum gain of 93. A 14 bit analog to digital converter (ADC) digitizes the amplified signal. The data rate of the system is 40 Hz. Fig. 2: Configuration A (internal amplifier and ADC) B. Two internal amplifiers and ADC Configuration A and B are very similar. Yet the amplified sensor signal is fed out and in of the PSoC to a programmable gain amplifier (PGA) block. Gain of the PGA is 4 to stay within the system voltage supply resulting in a system gain of 372. A 14 bit analog to digital converter (ADC) digitizes the amplified signal. The data rate of the system is 40 Hz. Fig. 3: Configuration B (two internal amplifiers and ADC) C. Carrier frequency excitation Configuration C also uses internal blocks of the PSoC only yet carrier frequency excitation is used in this design. The sensor is supplied with a sinusoidal signal generated by a pulse width modulator (PWM) block generating a rectangular signal of 5 kHz which is then filtered by a band pass filter (BPF) block with a center frequency of 5 kHz an a bandwidth of 300 Hz. The sinusoidal output voltage of the sensor is amplified by an INSAMP block with a gain of 93 and also filtered by a BPF. A mixer is used for phase-sensitive rectification. The rectified signal is sampled by the integration ADC over 125 full periods effecting a low pass filtering of the digitized signal. The data rate of the system is 40 Hz. the ADC is set to 15 Hz for a filter notch close to 50 Hz. Fig. 6: Configuration E (SX8723 external amplifier and ADC) Fig. 4: Configuration C (carrier frequency excitation) D. PGA309 external amplifier A PGA309 [2] sensor signal conditioning IC is used in addition to the PSoC in this configuration. The sensor’s output is connected to the PGA309 and a common power supply is used for the sensor, the PGA309 and the PSoC. Gain of the PGA309 is set to the maximum of 1152 and offset registers are programmed for 0 V output at no sensor load. The sensor output is digitized by an internal 15 Bit ADC of the PGA309, which is read out via an OneWire Bus. Sampling time of the ADC is 100 ms. The low data rate on the OneWire Bus limits the effective data rate to approximately 8 Hz. III. RESULTS For each of the configurations A – E seven measurements with different weights are made and 1000 samples of data are taken for each measurement. Absolute sensitivity is calculated from the slope of the linear regression divided by the standard deviation. The results are given in table 1. Tab. 1: Measurement result of configurations A – E Configuration A B C D E Slope [N-1] 1.77 6.63 1.68 42.79 66.53 Standard deviation 2.26 8.92 1.46 2.99 5.88 Absolute sensitivity [N-1] 0.78 0.74 1.15 14.31 11.31 IV. DISCUSSION Fig. 5: Configuration D (PGA309 external amplifier and ADC) E. SX8723 external amplifier The SX8723 [3] data acquisition IC is used in addition to the PSoC in this configuration. Gain of the SX8723 is set to the maximum of 1000 and offset registers are programmed for a near 0 output at no sensor load. The sensor’s output is digitized by the internal 16 Bit ADC of the SX8723, which is read out via I²C-Bus at 400 kHz. Sampling rate of The different system designs differ largely in their absolute sensitivity. The designs with an external sensor signal conditioning ICs show an absolute sensitivity 10 times better than the designs realized with the internal PSoC blocks. Zero point offset of the sensor is a large concern for the configurations A – C. Although a commercially available sensor with a very low zero point offset was used, it was still the limiting factor for the gain thus limiting the absolute sensitivity to the given values. The PGA309 sensor signal conditioning IC has the highest system gain of 1152 although only a 15 Bit ADC is used. The slowest data rate of the compared systems also gives the highest absolute sensitivity. System complexity and board space required is highest in this design as many passive components are used and a leaded package is available only. The internal ADC is intended for calibration purposes only therefore the data rate on the OneWire Bus is low as a polling method has to be used to monitor an end of conversion flag. The absolute sensitivity of the SX8723 is slightly worse than of the PGA309. But system complexity and extra board space is minimal in this design. Only one capacitor and two external resistors are used for the I²C Bus. Data rate was chosen with 15 Hz in this analysis but can be configured with same standard deviation and accuracy to 60 Hz. With digital filters the standard deviation can be reduced while the data rate is still sufficient for the intended application. Offset compensation range is highest in this design with up to 15 times the full scale input range of the sensor. V. CONCLUSION A data acquisition system for untrimmed resistive strain gauge sensors for mobile medical applications was designed using a microcontroller and the SX8723 (Semtech Corporation, CA) signal conditioning system. The SX8723 provides a complete low power acquisition path and uses least extra board space. The sensor element of the robotic external fixator is made from stainless steel tube with a diameter of 11mm. Two resistive strain gauge half-bridges (Vishay Measurements Group GmbH) are glued to the tube which is then integrated into the mechanical structure of an external fixator system for fracture treatment. The complete sensor system has a board space of 10 mm * 20 mm on a one side, two layer PCB. The complete system is housed in polyurethane/composite enclosure made with a rapid prototyping system (ZPrinter 450 Z Corporation). The sensor shows a sensitivity of 440µV/V at 1000 N while the zero-point offset is typically in the same order of magnitude. With an ADC data rate of 60 Hz and digital filtering the absolute sensitivity of the system is 16.26 N-1 allowing a resolution better than 0.1 N with a data rate of 7.5 Hz. VI. REFERENCES [1] K. Seide, U.-J. Gerlach, R. Wendlandt, N. Weinrich, J. Müller, C. Jürgens: “Intelligent external fixator for fracture treatment and deformity correction” in Trauma und Berufskrankheit, Volume 9, Number 2, 2007, pp. 109–116 [2] [3] Texas Instruments: “PGA309 Datasheet (SBOS292B)”, 2005, http://www.ti.com/lit/gpn/pga309 Semtech Corporation: “SX8723 Datasheet (V1.8)”, 2009, http://www.semtech.com/pc/downloadDocument.do?navId=H0,C 1,C193,C195,C199,P3361&id=2652