Circuit noise analysis - The Sensing Circuit for Organic Gas Sensors

CHAPTER 3 The Novel Sensing Circuit for Organic Gas Sensors

3.3 The Sensing Circuit for Organic Gas Sensors

3.3.9 Circuit noise analysis

A general noise model for the output of readout circuit is shown in Fig. 24. The noise contributed by each of the transistors can be described by a noise current source di_n² for transistor M between the drain and source. Considering only the thermal noise we have _n

2 2 2 2 2

Fig. 24. Noise of readout circuit.

Obviously, reducing g will reduce the thermal noise. Since _m g is proportional to the _m ratio of width to length (W)

L , low thermal noise has dictated lower (W)

L ratio. We also model the flicker noise of readout circuit by a current source,

2 2 2 2

The output flicker noise voltage per unit bandwidth is therefore equal to

2 2 2 2 themselves do not help because the g scales up the same amount that the noise goes down. _m Thus we try to lower the g so that the flicker voltage at the gate produces less current at the _m drain.

CHAPTER 4 Simulations and Experimental Results

4.1 Simulation of the Designed Sensing Circuits

Fig. 13 shows the structure of readout circuit in the simulation. The design of the readout circuit block diagram is shown in Fig. 25. In order to test the readout circuit, the sensed current is created based on experimental data from VNJ-P3HT diode sensors. Then the sensed current is used in three places. First, the Auto-Reset circuit reset the initial current information in voltage signals of peak-detector-and-hold circuit according to the variation of the sensed current (see 3.3.5). Second, the peak-detect-and-hold circuit gets the initial current information in voltage signals from the sensed current (see 3.3.2). Third, the divider uses the sensed current and the initial current information in voltage signals to output the ratio of saturation current to initial current (see 3.3.3). In the next, the saturation detector detect when the response is saturated by the output of divider (see 3.3.4). Finally, the logic gate decides

Fig. 25. Functional block diagram of readout circuit.

Reset circuit.

Fig. 26 shows the simulation results of readout circuit for single ammonia concentrations.

Before sensing (0s~10s), the sensed current doesn’t change. Thus the saturation detector considers the response is saturated and the output of saturation detector is high voltage level.

Without a negative slope of the sensed current, the output of the auto-reset circuit is also high voltage level. When the VNJ-P3HT diode starts to interact with ammonia (10s~200s), the output of divider starts to decrease and the output of saturation detector switches to low voltage level. Due to the negative slope of the sensed current, the auto-reset circuit considers the response between sensor and ammonia start and the output of the auto-reset circuit also switches to low voltage level to hold the initial current information in voltage signals. So far, the final output has been low voltage level. After about 200 seconds, the response is saturated and the rate of change of sensed current is close to zero. Thus the saturation detector considers the response is saturated and the output of saturation detector switches to high voltage level.

In chapter 3.3.6, the final output show the result only when the output of auto-reset circuit is Fig. 26. The simulation results of readout circuit for single ammonia concentrations.

low voltage level and the output of saturation detector is high voltage level. The final output catches up with the output of divider to show the valuable data at this time.

Fig. 27 shows the simulation results of readout circuit for several ammonia concentrations (3ppm 、 1ppm 、 500ppb). When the response is saturated, ammonia is removed and the sensed current is on the increase. Thus the result is showed from final output for only a short time. The main principle of operation is the same as described previously. It is worth to mention that the auto-reset circuit automatically reset the initial current information in voltage signals of peak-detector-and-hold circuit (switches to high voltage level) when ammonia is removed and the sensed current is on the increase. The output of auto-reset circuit keeps the high voltage level until the next response start. This can prove the circuit can achieve automatic sensing.

3ppm 1ppm 500ppb

Reference voltage

Hold Initial Current

Concentration index

Saturation index

Saturation

Zero or Result

Fig. 27. The simulation results of readout circuit for several ammonia concentrations.

4.2 Experimental Validation for Sensing Circuits

Operation Amplifier

Peak-Detect-and-Hold Circuit Analog divider

(a)

Operation Amplifier

Peak Detect-and-Hold Circuit Analog divider

(b)

Fig. 28. (a) The layout (b) The photo of first IC.

Fig. 28 and Fig. 29 show the layout plot and the photo of the chip. Fig. 30 displays the layout plan of the third chip. The areas of ICs for the first, second and third chip are 1.083×1.04 mm², 1.082×1.082 mm² and 0.74×0.75 mm². The first IC has 24 pins, the second IC has 28 pins and

Analog divider Operation Amplifier

Peak-Detect-and-Hold Circuit

(a)

Operation Amplifier Analog divider

Peak-Detect-and-Hold Circuit

(b)

Fig. 29. (a) The layout (b) The photo of second IC.

the third IC has 18 pins. This study implements the proposed circuits by TSMC 0.35um Mixed Signal 2P4M process. 3.3V is the supplied voltage of the circuit. Keithley Series 2400 Digital Source Meter generates the control signals and the entire circuit is measured.

4.2.1 The Sensing System

Because the readout circuit has to detect the difference between various signals about ammonia concentration, it is tested for sensing function first. The sensed current is created based on experimental data from VNJ-P3HT diode sensor to test the readout circuit. After constructing our own sweep by specifying the number of measure points and the source level at each point, Keithley Series 2400 Source Measure Unit (SMU) Instrument generates simulated sensing current. The analog output current of source meter are similar to sensed current in forms. The source meter is designed specifically for test applications that demand tightly coupled sourcing and measurement. It provides precision voltage and current sourcing as well as measurement capabilities. It is both a highly stable DC power source and a true

Analog divider Peak-Detect-and-Hold

Circuit

Operation Amplifier

Logic Gate

Fig. 30. The layout of third IC.

instrument-grade 6½ -digit multi-meter. The power source characteristics include low noise, precision, and read-back. The multi-meter capabilities include high repeatability and low noise. The result is a compact, single-channel, DC parametric tester. In operation, it can be used as a voltage source, a current source, a voltage meter, a current meter, and an ohmmeter.

Manufacturers of components and modules for the communications, semiconductor, computer, automotive, and medical industries will find the Source Meter SMU instruments invaluable for a wide range of characterization and production test applications.

The MCU is used to process the output of the readout circuit and estimate concentrations of ammonia. In order to search the corresponding concentration for output of the circuit, the programs create the data base. The analog-to-digital converter (ADC) is needed to transform analog output of circuit into digital input of MCU.

After presenting part of this system, the overall view of the sensing would be introduced.

Keithley Series 2400 Source Measure Unit (SMU) Instrument

Power Supply

Laptop

MCU

Fig. 31. The test environment of the sensing system.

The system configuration of test environment is exhibited in Fig. 31. The sensing system is composed of source meter, the sensing circuit, MCU and ADC devices. The following describes the process of the sensing system.

1. Keithley Series 2400 Source Measure Unit (SMU) Instrument acquires the experiment data based on ammonia sensor.

2. The proposal readout circuit processes the input signals related to the ammonia concentrations.

3. The ADC obtains the output signals of the readout circuit and transforms output signals of the sensing circuit into digital input of STM32.

4. In order to compute the corresponding concentrations, the STM32 processes the

MCU

ADC

LCD Laptop

USB

Power

IC

Sensor

Fig. 32. The system configuration in a diagram.

digital input by programming.

Fig. 32 constructs the configuration of sensing system. The MCU board get its power (5V) from USB ports contained in laptops, cars, aircraft or even wall sockets. Then the MCU board provides power (3.3V) to the readout circuit. The LCD of MCU board real-time reveal the concentration of ammonia, which transformed from output of readout circuit by MCU.

4.2.2 Experimental Results of the Peak-Detect-and-Hold Circuit

It requires to verify that the peak-detect-and-hold circuit (shown in Fig. 16) can detect the initial current information in voltage signals and hold it for 200 seconds without the leakage current of capacitor first. This is an important difference to previous design which use MCU to hold the initial value. A current source dropped from 100μA~10μA created by Keithley Series 2400 Source Measure Unit (SMU) Instrument.

Fig. 33 shows the experimental results of the peak-detect-and-hold circuit. After resetting the output (the reset signal is “1”), the input is higher than the output. Thus the peak-detect-and-hold circuit start to detect the peak of input and hold it. Because our sensing time is 200

— Input Signal

— Reset Signal

— PDH Output

— Input Signal

— Reset Signal

— PDH Output

Fig. 33. The experimental results of the peak-detect-and-hold circuit.

seconds, the super capacitor is used to hold the peak of the input without the leakage current.

From Fig. 33, the peak-detect-and-hold circuit is workable and the super capacitor hold the peak for 200 seconds successfully. It can be used to get an initial value of the sensed current.

4.2.3 Applying Experimental Data of Ammonia Sensor and Experimental Results

With the peak-detect-and-hold circuit in function, the input voltage of divider is correct. The experimental setting is shown in Fig. 13. The VNJ-P3HT diode sensor is changed to the

(a)

(b)

Fig. 34. (a) Post-simulation and (b) experimental results of analog divider with 10ppb ~ 3ppm of ammonia.

Keithley Series 2400 Source Measure Unit (SMU) Instrument instead, which creates I . In _S previous chapters, Fig. 7 shows the original data of response in two hundred seconds. Eight ammonia concentrations for test are 10ppb, 30ppb, 50ppb, 0.1ppm, 0.3ppm, 0.5ppm, 1ppm and 3ppm. If the response multiplied by the initial current ( I_O), values of the output current I will be decided. Fig. 34 shows the post-simulation and experimental output voltages of S

analog divider (V ) in 0.01 ~ 3ppm ammonia. Table II shows the differences between _O experimental results with post-simulation results in 10ppb ~ 3ppm of ammonia. The differences are within 7.24%.

Table II. The experimental results compared to post-simulation results.

Concentration(ppm) Simulation result (V) Experimental result (V) Error (%)

0.01 0.359 0.375 4.266667

0.03 0.337 0.363 7.162534

0.05 0.319 0.339 5.899705

0.1 0.301 0.321 6.23053

0.3 0.265 0.267 0.749064

0.5 0.247 0.237 4.219409

1 0.222 0.207 7.246377

3 0.181 0.183 1.092896

Fig. 35(a) summarizes the relationship between the output voltage and concentration.

With sets of experimental data collected on readout voltage and corresponding concentration levels in ppm, a linear relation is validated between logarithms of gas concentration level and output voltages of the readout circuit, as shown in Fig. 35(b). The resolution of the sensor achieves 86.8mV/log (ppm). The linear region is between -2~0.5ppm in log scale. The curves

(a)

(b)

Fig. 35. Post-simulation and experimental output voltage in (a) 10ppb ~ 3ppm of ammonia, (b) 10ppb ~ 3ppm of ammonia in log scale.

can be approximated to linear equations, which based on simulation result is ( ) 0.0729log 0.2227

y x   x (29)

and experimental result is

( ) 0.0868log 0.2195

y x   x , (30) where y is output voltage of the analog divider, x is concentration of ammonia.

4.2.4 Reliability of Integrated Circuits

In previous charpters, this study verifies that the readout circuit can calculate the ratio of saturation current to initial current, which presents the specific gas concentration. Under manufacturing process variation, not all chips perform and maintain its functions. Before mass-production, the reliability of integrated circuits is very important. Fig. 36 shows the experimental output voltage of eight chips in 10ppb ~ 3ppm of ammonia in log scale, lines of eight results and the square of the residuals of the data after the fit. There are eight trend lines because of eight chips. The variation exist between chips.

Fig. 36. The experimental output voltage of eight chips in 10ppb ~ 3ppm of ammonia in log scale.

To analyze the variation, Fig. 37 shows the distribution of the parameters of the lines (y_i c x b_i  _i). The range of b and _i c are 0.0047 and 0.0062, which can be ignored. _i Because not all points of data are on the lines, this study also tries to analyze the reliability of the lines. Fig. 38 shows the distribution of r-squared (R ) which is the square of the residuals ² of the data after the fit. The R² value is basically a measure of how good the correlation is.

The closer the R² value is to 1, the better the correlation. A good correlation between concentration in log and the output of divider indicates that the methodology is sound, and

Fig. 37. The distribution of the parameters of the lines (y_i c x b_i  _i).

Fig. 38. The distribution of r-squared (R ). ²

that the readout circuit is working well. In other words, the higher the R², the better.

y are defined as the observed values and i f are called the predicted values. In the _i

where n is the number of observations. The "variability" of the data set is measured through different sums of squares,

tot ( _i ) .

SS 



y y (32)

The total sum of squares (proportional to the sample variance) is

reg ( _i ) .

SS 



f y (33)

The regression sum of squares, also called the explained sum of squares is

res ( _i _i) .

SS 



y  f (34)

The sum of squares of residuals is also called the residual sum of squares. The most general definition of the coefficient of determination is

2 res reg

tot tot

1 SS SS .

R  SS  SS (35)

Generally speaking, an R² of 0.75 means that a reasonable correlation between the output of divider and concentration in log. 0.9 or above is very good. Fig. 36 shows a pretty good correlation. Because the line fits very nicely (R-squared over 0.9), this study can use it to estimate concentrations of ammonia.

4.2.5 Experimental Results of the Saturation Detector

Having verified that the readout circuit can detect ammonia concentration accurately, this study tries to prove that the automatic readout circuit is workable. Fig. 26 shows the simulation results of readout circuit which can validate the automation of readout circuit. The experimental results of the saturation detector is shown in Fig. 39. The experimental setting is shown in Fig. 13 without auto-reset circuit, logic gate and buffer because the final version of the chip still in fabrication by National Chip Implementation Center. The VNJ-P3HT diode sensor is also changed to the Keithley Series 2400 Source Measure Unit (SMU) Instrument instead, which creates I . _S

In the beginning, the reset signal is “1” and the super capacitor of peak-detect-and-hold circuit discharged to ground voltage. Thus the denominator of divider get smaller and the output of divider get larger. When the reset signal switch to “0”, the capacitor starts to charge and hold the new peak. Until after this, the output of divider makes sense. Then the output of divider starts to decrease and the saturation detector judge that it is not saturated. Therefore the output of saturation detector switch to “0”. After about 200 seconds, the rate of change of

— Reset Signal

— Sat. Detector

— Divider Output

— Reset Signal

— Sat. Detector

— Divider Output

Fig. 39. The experimental results of the saturation detector.

divider output is close to zero and the output of saturation detector switch to “1”. The experiment of second chip verify that the readout circuit can automatically detect when the response is saturated.

4.2.6 Experimental Results of MCU

With the readout circuit in function, the output of analog circuit connect to the ADC of MCU.

Then the data is processed using a MCU to estimate concentrations of ammonia. Finally, the concentration of ammonia is displayed on the LCD. Fig. 40 contrasts the experimental results of LCD with waveform. At the bottom of Fig. 27, the output show the result only when sensor

3ppm 1ppm 500ppb 300ppb 100ppb

Fig. 40. The experimental results of LCD in 0.1 ~ 3ppm ammonia.

react with gas and the current saturate, otherwise the output display zero because of the logic gate. The valid output is varied in the range between 0.15V and 0.4V (see Fig. 34). As show in the Fig. 27, if the output of readout circuit is lower than 0.15 voltage, “----ppb” and

“Sensing……” are displayed on the LCD. This situation represents an unsaturated result.

On the other hand, the LCD accurately shows the results in 0.1ppm, 0.3ppm, 0.5ppm, 1ppm and 3ppm ammonia, as shown in Fig. 40. Even though the initial current changes, the readout circuit obtains the new initial value and outputs the correct results. This indicates the readout system is workable. Table III shows the overall specifications of readout circuit.

Table III. The specifications of readout circuit.

Specifications Values

Power Supply (VDD) 3.3V

Power consumption 4.6mW

Input Range of Sensing Current 0μA~100μA

Input Range of Sensing Voltage 1.5V~3.3V

Range of Bias Voltage 1.5V~3V

Resolution 86.8mV/ log(ppm)

Process of Manufacturing 0.35μm 2P4M 3.3V mixed‐signal CMOS process

Die Area 0.74×0.75 mm²

CHAPTER 5 Conclusions and Future works

An automatic readout circuit for gas sensor is proposed in this study. These analog IC are realized by the TSMC 0.35um Mixed-Signal 2P4M process. The circuit is designed, implemented and validated by the experimental output currents of the readout circuit which varies with sensed currents.

The circuit is connected externally to the sensors that output the current varying with the concentration of target gas, although the ammonia sensors based on VNJ-P3HT diode are not integrated into a single chip with the readout circuit. In order to perform experiments in a more convenient and efficient way, Keithley Series 2400 Source Measure Unit (SMU) Instrument is chosen to simulate the output current of the front-end VNJ-P3HT diode device in experiments. Finally, the sensing system is able to detect minimum ammonia concentration of 10ppb, while the maximum one reaches around 3ppm.

Without the AD/DA of MCU in the previous design, the peak-detect-and-hold circuit is used to holds the input value of the analog divider. The ratio of saturation current to initial current can be calculated by the analog divider and present the specific gas concentration.

Thus the problem that the initial current varies from time to time is solved. The auto-reset circuit for detecting transitions between different gas concentration levels is designed and implemented in this study for sensing automation. The design of the proposed saturation detector and logic gate can provide more convenience for real-time detection of gas concentration.

analog divider’s output voltage versus concentration of ammonia in log scale are determined.

The resolution of the sensor achieves 86.8mV/log (ppm). This relation is utilized further in a micro-processor for calculating gas concentration level in ppb. Finally, the gas concentration level in ppm is displayed on a LCD panel. The experimental results show that errors are within 7.24%.

The size of MCU can be further reduced in future works. Then the entire system (analog IC, MCU and LCD) can be fabricated to a compact module. The imaginary figure is shown in Fig. 41. Finally, the resolution and the linear range of the readout circuit’s output about the ammonia concentration can be further improved.

Charging

1Power on

2Blowing 3Reading

MCU(ADC) +LCD

IC+Sensor

Fig. 41. The imaginary figure.

APPENDIX

Table IV. The design parameters of readout circuit.

Transistors Aspect ratio (μm/μm)

m Transistors Aspect ratio (μm/μm)

M1-M6 10/4 1 M3p,M4p 10/1 2

M7-M10 25/5 2 M5p 10/1 9

M10 20/2 1 M6p 1/1 10

M1p,M2p 2/1 1 M7p, M8p 3/1 1

Resistance Ω Capacitance F

R0 8k C0 1p

REFERENCES

[1] M. Kikuchi, K. Omori, and S. Shiratori, "Quartz crystal microbalance (QCM) sensor for ammonia gas using clay/polyelectrolyte layer-by-layer self-assembly film,"

Proceedings of IEEE Sensors, pp. 718-721 vol.2, 24-27 Oct. 2004.

在文檔中基於影像辨識之移動式機械臂抓物控制設計 (頁 50-0)