Improvement of Olfactory Display Using Solenoid Valves

Improvement of olfactory display using solenoid valves

Takamichi Nakamoto and Hai Pham Dinh Minh

Graduate school of science and engineering, Tokyo Institute of Technology

ABSTRACT Recently, the research on olfactory sense in virtual reality has gradually expanded even though the technology is still premature. We have developed an olfactory display composed of multiple solenoid valves. In the present study, an extended olfactory display, where 32 component odors can be blended in any recipe, is described; the previous version has only 8 odor components. The size was unchanged even though the number of odor components was four times larger than that in the previous display. The complexity of blending was greatly reduced because of algorithm improvement. The blending method and the fundamental experiment using a QCM (quartz crystal microbalance) sensor are described here. CR Categories: H.5.1 [Information Interfaces and Presentation]: Multimedia Information Systems –Artificial, augmented, and virtual realities; Keywords: Olfactory display, solenoid valve, Pulse Width Modulation, QCM gas sensor, Movie with scents

1 INTRODUCTION Nowadays it is quite easy to deal with visual and auditory

information in a computer. We can acquire much visual and auditory information through Internet. However, we cannot obtain the complete sensory information when, for example, watching TV. When we see delicious food on TV, its smell is indispensable for reproducing much reality.

We have studied an odor sensing system using multiple sensors with different characteristics and pattern recognition technique [1-2]. Although there have been many reports about the artificial sensor called the electronic nose [3], a device for presenting olfactory information should also be studied in spite of the small population of researchers at the current stage.

We have also studied an odor recorder that reproduces as well as records smell [4]. Although odor sensing systems and olfactory displays have so far been studied separately, these two technologies should be closely inter-related. In this sense, therefore, the odor recorder, has an advantage over other devices. Since we describe this device elsewhere [5-7], we focused only on the olfactory display here.

There have been several works on the olfactory display. An olfactometer has been used for many years to give a human an olfactory stimulus so that human sense of smell or brain waves induced by olfactory stimulus could be studied [8]. However, the olfactometer is large and complex. Although a commercially available diffuser is simple, the smell cannot be changed quickly because the cartridge must be exchanged [9]. Although another PC-controlled scent diffuser that could present several smells was proposed, it has no blending function [10].

Kaye constructed a device to present plural smells according to

the state of the stock market [11]. Moreover, other researchers have reported an olfactory display for localizing a smell source in a virtual environment [12] and spotscents using an air cannon [13]. Several concepts of the olfactory display have been described by Davide et al. [14].

Although they are useful in certain situations, one of the most important functions of an olfactory display is to present a variety of smells. A variety of smells can be generated when the function of blending is introduced.

Buck and Axel reported the multigene family of G-protein-coupled ORs (olfactory receptors) in 1991 and, then, the molecular biology of olfaction rapidly progressed [15]. However, primary smells [16] are not known, unlike the primary colors in vision. In this situation, a device for blending as many odor components as possible is indispensable to cover a wider range of smells.

Even if primary smells have so far not been found, it is still important to blend smells because the blending process is currently essential in creating new smells, particularly in the flavor industry.

In addition to novel scent creation in the flavor industry, a variety of olfactory-display applications are feasible, such as a smell-presenting device in, for example, an odor recorder, a movie with scents, games, exhibitions, on-line shopping, restaurants, educational tools, medical-diagnostic tools, museums and art. In particular, a product in which scent is indispensable for evaluating its quality should be presented using an olfactory display in various situations.

Recently, we exhibited a cooking game with scents at several places in collaboration with artists, and 300-400 people tried this game [17]. In the cooking game, smells are essential to reproduce reality. People enjoyed the game using three senses, such as vision, hearing and olfaction in the virtual environment. Some of them said that they felt hungry after they had tried the game.

Olfactory information also contributes greatly to a sense of presence when it is presented synchronously with a movie. We have made a scented animation movie and a questionnaire survey revealed the contribution of smells to the sense of presence [18]. Moreover, olfactory information recorded by an odor-sensing system and visual information recorded by a digital video camera were reproduced synchronously [19]. In these applications of the olfactory display, the range of smells that can be presented should be further extended.

We proposed olfactory displays in which several smells can be blended in an arbitrary recipe using mass flow controllers [20], inkjet devices [21], solenoid valves [22] and an autosampler [23]. A mass flow controller is expensive, an inkjet device requires skill to handle and an autosampler cannot achieve real-time blending even though it can accommodate a large number of odor components. We used solenoid valves here because they are cheap and easy to handle. Since they are stable and relatively small, they are suitable for integration in order to realize an olfactory display with many odor components in spite of compact size.

In the conventional olfactory display made up of solenoid valves, many empty bottles are required to maintain the symmetry of the flow system and, here, there is redundant space. Moreover, the dynamic range of odor concentration decreases as the number

2-12-1, Ookayama, Meguro-ku, Tokyo, 152-8552, JAPANE-mail: [email protected]

179

IEEE Virtual Reality Conference 2007March 10 - 14, Charlotte, North Carolina, USA1-4244-0906-3/07/$20.00 ©2007 IEEE

Authorized licensed use limited to: UNIV OF MASS-LOWELL. Downloaded on December 5, 2008 at 08:59 from IEEE Xplore. Restrictions apply.

of odor components increases. Thus, the purpose of this study is to extend the range of smells using more odor components while maintaining small size.

2 GENERAL DESCRIPTION OF OLFACTORY DISPLAY

There are several types of olfactory display as mentioned above.

The most typical one is an odor blender composed of several MFCs (mass flow controllers) [20]. The ratio of each odor component is determined by the flow rate of the corresponding MFC. A MFC is a device used to electrically control the flow rate without the influence of the pressure load. Although it is useful for controlling flow rate accurately, it is too expensive for blending many odor components.

Another blender makes use of an inkjet device which was originally developed for printers. It ejects tiny liquid droplets with the volume from one nanoliter down to one picoliter [21]. A heater is necessary to evaporate the liquid. When we use several inkjet devices, we can blend the liquid in an arbitrary composition. Since plumbing is not necessary to blend odor components, the influence of smell persistence is small. However, automatic priming is often difficult.

The third blender is an autosampler [23]. An odor component in the liquid phase is picked up from one vial and is transferred to another vial. Many odor components can be sequentially and automatically blended by repeating this procedure. However, it takes much time to blend odor components.

The final blender employs the method of using solenoid valves [22]. Since solenoid valves are much cheaper than mass flow controllers and easy to use, we adopted this blender at the current stage. It is also possible to blend multiple odor components in real time.

Table 1 Specifications required for olfactory display No Item 1 Number of smells 2 Adjustment of concentration

(Dynamic range of odor intensity) 3 Speed of odor variation in quality and intensity 4 Direction to object with smell

(Localization of smell source) 5 Area of smell diffusion

The specification of an olfactory display is tabulated in Table 1.

First, the number of smells to be generated is important. Although there has so far been no equipment to emit arbitrary smells, the region of smells can be extended if many odor components can be blended in an arbitrary composition.

Second, the adjustment of the concentration is necessary. A wide dynamic range of the odor intensity is preferable since human perception covers a wide range of concentration.

Third, the speed of odor variation in both strength and quality should be considered, since odor in the ambient air changes dynamically and irregularly. Smell persistence is also an important point. Odor generation synchronous with breathing is related to this point.

Moreover, the direction to an object with smell is one of the factors of olfactory display. The sense of direction is required when people localize a source of smell in a virtual environment.

Finally, the area of smell diffusion should be taken into account. It depends upon how many people simultaneously perceive the smell in the virtual environment. We focused on the first point in the present study although many other factors should be considered.

3 CONVENTIONAL METHOD The principle of olfactory display using solenoid valves is

shown in Fig. 1. Although a solenoid valve is a fluidic switching device with only two states, such as ON and OFF, high-speed switching enables any concentration. The frequency of the ON state corresponds to odor concentration. Switching noise does not appear at the outlet because of the fluidic LPF (low-pass filter). Since a solenoid valve is relatively cheap and small, it is easy to integrate many valves. Thus, it is possible to realize an olfactory display with many components. Some people are worried about wear of the valve due to the huge number of switchings. However, it is very robust and seldom breaks.

Solenoid valve with high speedswitching

Component vapor

Air

Component vapor with arbitrary concentration

LPF

Solenoid valve with high speedswitching

Component vapor

Air

Component vapor with arbitrary concentration

LPF

Fig. 1 Principle of olfactory display using solenoid valves The entire structure of a conventional olfactory display is

illustrated in Fig. 2 [5]. It can blend up to 8 odor components using 16 three-way solenoid valves and 16 bottles comprising 8 for odor components and 8 empty bottles. An odorant liquid is put into the sample bottle and the vapor of the headspace over liquid surface is transported by carrier gas. Two paths are required: one for outputting smells and the other for exhaust. The latter is called the bypath. QCM sensors are used to monitor odors if necessary.

OSC

ComputerControl signals

Air

Air

Solenoid valvesQCM sensors(20MHz AT CUT)

Flowmeters with valves

Pump

1

1

1

1component1 P

Air

Air

component8

1

1

1

1OSCOSC

Freq. counterFreq. counter

Bypath8

8

8

8

Fig. 2 Schematic diagram of conventional olfactory display

180


In addition, the control method of the solenoid valves is much different. The comparison of the proposed method with the previous one is illustrated in Fig.5.

SensorOdor component

Air OutletSensor

Odor component

Air Outlet (a) (b) Fig. 3 Flow path control of conventional olfactory display, (a)

at the outlet and (b) at a sample bottle. Flow path control at the outlet and at a sample bottle are shown

in Figs. 3 (a) and (b), respectively. Flow rate at an outlet can be kept constant even when the odor component is not selected. We often put sensors at the outlet and flow rate at sensors should be held constant. The empty bottle is used for the path of air because the flow-path symmetry should be maintained to guarantee the same flow rate at each path.

The flow rate at a sample bottle should also be kept constant because the concentration at the bottle depends upon its flow rate. The concentration at the sample bottle should be maintained because the concentration of each odor component is determined on the basis of the concentration at the sample bottle.

In the olfactory display in Fig. 2, each odor component flows independently of others. Thus, the maximum concentration at the outlet is one-eighth of that at the sample bottle. This means that the dynamic range of the concentration decreases as the number of odor components increases.

The algorithm to control solenoid valves in the previous system is the delta-sigma modulation known as the 1-bit analog to digital conversion technique [24].

4 IMPROVED METHOD The proposed olfactory display is shown in Fig. 4. Thirty-two

odor components can be equipped with this system. The number of odor components is four times greater than that of the previous one whereas its size is still compact. In the previous olfactory display, a pair of an odor component bottle and an empty bottle was used and the corresponding two solenoid valves worked complementarily. On the other hand, only one empty bottle is used to supply air to the outlet. Other structure is similar to the previous one.

The significance of this advancement will be evaluated in the near future. However, the number of combinations of odor components becomes huge when 31 odor components are used (air is considered one odor component.). In the simplification in which each odor component is in either the ON state or the OFF state, 256 combinations are possible with 8 odor components. In contrast, approximately 2x109 combinations are possible in the case of 31 odor components. Even if we consider that smells of many combinations are not experienced under typical situations, it is still a large improvement.

OSC


Air

Air



Pump

component32 P

Air

Aircomponent2

OSC

Freq. counter

Bypath

component31

Empty bottle

1

2

31

32

OSC


Air

Air



Pump

component32 P

Air

Aircomponent2

OSCOSC

Freq. counterFreq. counter

Bypath

component31

Empty bottle

1

2

31

32

Sensor/outlet

BypathAir

Sensor/outlet

BypathAir

Fig. 4 Structure of 32-component olfactory display.

Conventional method

Odor component1

Odor component2

Time

Mix

AirConventional method

Odor component1

Odor component2

Time

Mix

Air Proposed

method

Time

Air AirOdor1 Odor2 Odor1 Odor2

Repetition cycle

Proposed method

Time

Air AirOdor1 Odor2 Odor1 Odor2

Repetition cycle

Fig. 5 Comparison of proposed method with previous one. In the previous method, each odor component is independently

generated and then all the components are blended. Although the concentration of each odor component is reliably controlled, the number of empty bottles required to maintain the symmetry increases. Furthermore, the maximum concentration of each odor component decreases.

In the proposed system, only one odor component is allowed to go to the output at a certain time and the time division multiplexing technique is adopted to blend multiple odor components. The dead space of the tube and connector acts as a fluidic low-pass filter in the same manner as that in Fig.1. The concentration of each odor component corresponds to its duty cycle ratio. The repetition cycle here is 1s. As a result, the

181


maximum odor concentration to be output reaches the concentration in each sample bottle and the problem of the dynamic range is solved. Moreover, only one empty bottle is used when the air is flowed at the outlet. Thus, the number of empty bottles is considerably decreased.

Only one sample bottle is connected to the outlet at a certain time and other sample bottles are connected to the bypath. Thus, the two flow paths are not symmetric and the flow rate at the bypath is much higher than that at the outlet.

The proposed olfactory display is shown in Fig. 6. The size of the olfactory display is still compact although the number of odor components has been increased fourfold. The manifold is also shown in Fig. 7. Since the two flow paths are linear, it is easy to design the manifold. Sample bottle No.32 is the closest to the outlet whereas Sample bottle No.1 (empty bottle) is the farthest from the outlet. Since the symmetry is lost in this structure, the flow rate at each sample bottle is not always the same. However, this does not cause a problem because we can calibrate the system even in such a case.

The manifold is made of brass and air is supplied from the other manifold below the sample bottles. Thirty-two solenoid valves are directly attached to the manifold without plumbing. Since the response speed of the solenoid valve is fast (1ms), the accuracy of blending increases when the repetition cycle of odor presentation is fixed. Moreover, the sound generated due to switching is small. All solenoid valves are controlled through a laptop computer using a digital I/O card (Contec, PIO-48D(CB)H) and driver circuits composed of discrete transistors.

Manifold

Sample bottles

No.32

Sample bottles No.1

Manifold

Sample bottles

No.32

Sample bottles No.1

Although the range of presented odors is greatly extended in the

proposed system, more labour is required to exchange sample bottles when odor components are changed. An easier way, such as the use of cartridges, to handle odor components will be required in the future.

Fig. 7 The photo of manifold.

5 EXPERIMENT 5.1 Experiment with single odor component First experiment was carried out to check the relationship

between a specified concentration and the sensor response to single odor component. Since the sensor response is almost proportional to odor concentration, we can confirm from that relationship whether the dilution of the odor is appropriately performed.

Olfactory display SensorsOlfactory display Sensors

Fig. 6 The photo of 32-component olfactory display.

-1000

100200300400500600700800

0 300 600 900 1200 1500 1800 2100 2400 2700 3000

Sens

or re

spon

se [H

z]

Time [s]-100

0100200300400500600700800

0 300 600 900 1200 1500 1800 2100 2400 2700 3000

Sens

or re

spon

se [H

z]

Time [s] Fig. 8 Sensor response to 2-hexanone at bottle No.2.

182


Fig. 9 Relationship between relative concentration and sensor response when 2-hexanone was set at bottles No.2, No.17 and No.32.

The 20MHz AT-CUT QCM coated with sensing film (Apiezon-

L) was used to measure relative concentration [25-27]. The response of QCM sensor is in the form of a frequency shift from the air-level value. The frequency shift occurs due to the mass loading effect and it is proportional to the mass adsorbed on the sensing film. The frequency shift due to the film coating was 17.4kHz. The sample used here was 2-hexanone. The output of the QCM sensor is the frequency shift due to vapor sorption at the sensing film. The response of a typical QCM sensor is relatively linear with concentration compared with other gas sensors such as a metal oxide gas sensor.

The concentration in the unit of [%RC] means that relative to the concentration at the sample bottle. The concentrations specified here were 0, 10, 20, 30, 40 and 50 [%RC]. The flow rate at the outlet was 600ml/min, whereas the flow rate at the bypath was 3.5l/min.

An example of a typical waveform of a sensor response to 2-hexanone is shown in Fig.8. The sample was set at bottle No.2. A zero sensor response is the air level. The clear response was observed in this figure.

The refresh rate, that is, the speed required to change from one smell to another, of the olfactory display is another important factor. As is shown in Fig. 8, smell can be exchanged within a short time when 2-hexanone is used. However, it takes more time for a human to perceive the odor because of the breathing cycle. The refresh rate should be longer than the period of breathing cycle.

The relationships between specified concentration and sensor response when the sample was set at bottles No.2, 17 and 32 are respectively shown in Fig. 9. It was found that the sensor response was proportional to the specified concentration even when the position of the sample bottle was changed. The same tendency was obtained at any position. The variation from sample bottle to sample bottle was within 10%. Thus, it can be said that the odor concentration is generated as specified in the case of a single odor component.

Initially, we had concerns that the asymmetry of the manifold structure might cause a dependence on the sample bottle position. However, the experimental result shown in Fig. 9 reveals that the asymmetry did not influence the result since the sensor response

was almost independent of the sample bottle position. The concentration inside the sample bottle depends upon the flow rate of the supplied air. However, the flow rate does not actually depend upon the location of the sample bottle since fluidic resistance at the manifold might be much smaller than that of the valve at the flowmeter.

5.2 Experiment with plural odor components Since the sensor characteristic for a single odor component was

confirmed, the sensor was next simultaneously exposed to two odor components at various concentrations. We first checked whether the linear superposition was valid. The

additive property is approximately valid for a QCM sensor. If the additive property were not valid here, it would mean that each component concentration is influenced by those of other odor components. It would be too complicated to design the output of the olfactory display under such a circumstance. In contrast, it is quite simple to design the output of the olfactory display if the linear superposition is valid.

0

200

400

600

800

1000

0 10 20 30 40 50 6Concentration [%RC]

Sens

or re

spon

se [H

z]

0

No.2 No.17 No.32

0

200

400

600

800

1000

0 10 20 30 40 50 6Concentration [%RC]

Sens

or re

spon

se [H

z]

0

No.2 No.17 No.32

In the case of a non-volatile compound, pure air is supplied because no odorant is added at the sample bottle. It is not meaningful to use a non-volatile compound because it does not emit a smell. In the case of a low-volatility compound, linear superposition is still valid. The same sample of 2-hexanone as in the first experiment was

set at two bottles (No.3 and No.11). First, a single component at each position with a concentration between 0 and 50[%RC] was measured. Then, two same components with the same concentration between 0 and 25[%RC] were blended and the blended odor was measured. Furthermore, the sensor response was also obtained with the concentration at No.3 being kept at 5[%RC] while the concentration at No.11 was changed from 5 to 45[%RC]. The result is shown in Fig. 10. It was found that the concentration of the odor from bottle No.3

was the same as that from bottle No.11. The sensor responses in the two cases were almost linear with the specified concentration. It was also found that the superposition of the sensor response was valid when the two same samples were used.

0

100

200

300

400

500

600

700

800

900

0 10 20 30 40Concentration at outlet [%RC]

Sens

or re

spon

se [H

z]

50

Odor component 3Odor component 11Odor 3 5-25[%RC] Odor 11 5-25[%RC]Odor 3 5[%RC] Odor 11 5-45[%RC]

0

100

200

300

400

500

600

700

800

900

0 10 20 30 40Concentration at outlet [%RC]

Sens

or re

spon

se [H

z]

50

Odor component 3Odor component 11Odor 3 5-25[%RC] Odor 11 5-25[%RC]Odor 3 5[%RC] Odor 11 5-45[%RC]

Fig. 10 Superposition of sensor response to odors from bottles No.3 and 11.

Next, a binary mixture of 1-hexanol and 2-hexanone was used. 1-Hexanol was set at bottle No.7 and 2-hexanone was set at bottle No.31. The sensor used here was the QCM coated with Apiezon L. The frequency shift due to coating was 20kHz.

First, the concentration of 1-hexanol was 5[%RC] and the concentration of 2-hexanone was changed from 5 to 25[%RC].

183


Then, the concentration of 1-hexanol was changed to 15 and 25[%RC] and the concentration of 2-hexanone was again changed from 5 to 25 [%RC]. The results are shown in Fig. 11.

The constraint of 10ms comes from the operating system. The dynamic range can be extended since the actual response time of the solenoid valve is faster than 10ms.

It is well known that Weber-Fechner’s law explains the relationship between the concentration and the intensity perceived by a human. The sensor response here is almost linear with concentration. Thus, the perceived intensity might be proportional to the logarithm of the sensor response. Moreover, the perceived intensity is different from person to person. Thus, it is useful to have the function of intensity adjustment according to a user’s preference in a similar manner as the volume adjustment in the case of auditory sensing.

0

100

200

300

400

500

600

0 5 10 15 20 25Concentration [%RC]

Sens

or re

spon

se [H

z]

1-haxanol (No.7) 5[%RC]1-hexanol (No.7) 15[%RC]1-hexanol 25[%RC]

0

100

200

300

400

500

600

0 5 10 15 20 25Concentration [%RC]

Sens

or re

spon

se [H

z]

1-haxanol (No.7) 5[%RC]1-hexanol (No.7) 15[%RC]1-hexanol 25[%RC]

Fig. 11 Relationship between concentration of 2-hexanone and

sensor response with 1-hexanol concentration as a parameter. It was found that the concentration of 2-hexanone was

linear with the sensor response when the concentration of 1-hexanol was constant. Moreover, the lines of 1-hexanol concentrations were parallel and equally spaced. Thus, the linear superposition of the binary mixture was confirmed to be valid from Fig. 11.

0

2

4

6

8

10

12

14

16

0 20 40 60 80 100 120

Time [s]

Sens

or re

spon

se [H

z]0

2

4

6

8

10

12

14

16

0 20 40 60 80 100 120

Time [s]

Sens

or re

spon

se [H

z]

Fig. 12 Sensor response to 2-hexanone with 1% relative

concentration.

5.3 Range of odor concentration

5.4 Sensor response to low-volatility odor component We should pay attention to low-volatility compounds

from the viewpoint of smell persistence. 2-Hexanone used in the experiment described in the previous section is volatile since its saturated vapor pressure is 4.1mmHg at 300[K]. In this case, the speed of exchanging a smell is rapid. However, it is expected that this speed might be slow when a low-volatility compound is used. Low-volatility compounds cannot be ignored since they typically emit strong smells. Thus, a low-volatility compound generated by the current system was measured.

The low-volatility sample used here was citral. Its saturated vapor pressure at 300[K] is 0.053mmHg, which is two-orders lower than that of 2-hexanone. It is commonly known as the smell of lemon. The sensor used here is the same as that described in section 4.1. The specified concentrations were 0, 10, 20, 30, 40 and 50 [%RC]. Citral was set in sample bottle No.32.

In this subsection, we confirmed whether the dynamic range of odor concentration was improved compared with the previous olfactory display. In the previous one, the provided odor concentration was, at most, one-nth of the odor concentration at the sample bottle when the number of odor components was n. On the other hand, the maximum provided concentration of each odor component is 100% of the concentration at each bottle in the current olfactory display.

The repetition cycle here was 1s, as is described in section 4. The concentration of each odor component is determined by the duration of valve opening which should be a multiple of 10ms. Thus, the minimum concentration of each odor component to be realized is 1% relative to the headspace concentration at each sample bottle since the minimum time of the valve opening is 10ms. Then, the single odor concentration of 1[%RC] was evaluated using the QCM sensor.

The sensor response is shown in Fig. 13. The relationship between concentration and sensor response is also shown in Fig. 14. Although the sensor response to citral was slower than that to 2-hexanone, it is still within an acceptable range. Since a low-volatility compound such as citral can be exchanged within a few tens of seconds, it can still be used in many situations. However, users should take such delay into account. It was also found from Fig. 14 that the sensor response to citral was linear with its concentration.

The sensor used here was the same as that described in section 5.2 and the sample was 2-hexanone. The sample was placed at sample bottle No.2. The waveform of the sensor response is shown in Fig. 12. The resolution of the sensor response is 1[Hz] owing to the performance of the frequency counter. The sensor response to 1% 2-hexanone was 14Hz, while the response to 50% 2-hexanone was approximately 700Hz in Fig. 9. It can be said, from the sensor response, that the dilution is appropriately performed. We also confirmed that the sensor did not respond when the concentration was 0%. Thus, it can be said that the current olfactory display has a dynamic range of two orders.

Scent persistence occurs to some extent. It takes 5-8s to exchange the smell in the case of 2-hexanone. It takes slightly longer for of a low-volatility compound such as citral. In Figs. 12-13, the odor exposure times were between 2 and 10 minutes. These are too long for a human to smell. However, most of the

184


duration corresponds to the steady state of a sensor response. The exposure time can be greatly reduced for a human sniffing smells. Fig. 13 Waveform of sensor response to citral. Fig. 14 Relationship between concentration and sensor response to citral.

5.5 Comparison with conventional method The current system was compared with a conventional one

using the same blending algorithm. It is possible to implement the algorithm of the delta-sigma modulation used in the previous system into the current system. Thus, the comparison of the current system with the previous one was possible. The sample was 2-hexanone and the QCM coated with Apiezon L was used as the sensor. The sample was set at bottles 1, 3, 5 and 7 and each concentration was set at 10[%RC]. Bottles No.2, 4, 6 and 8 were used as empty bottles when the delta-sigma modulation was applied. Thus, the total concentration was 40[%RC]. In both methods, the concentrations at the outlet were set to be the same.

The waveforms of the sensor responses are shown in Fig. 15. It was found that the sensor response to the sample in the previous method was almost the same as that in the current method. The symmetry of the flow path is not considered in the current system because the solenoid valves are linearly aligned. The experiment revealed that the same concentration was obtained even when the flow path symmetry was not strictly satisfied.

0

100

200

300

400

500

600

700

800

0 20 40 60 80 100 120 140

Time [s]

Sens

or re

spon

se [H

z]

Previous system

Current system

0

100

200

300

400

500

600

700

800

0 20 40 60 80 100 120 140

Time [s]

Sens

or re

spon

se [H

z]

Previous system

Current system

0

100

200

300

400

500

600

700

800

900

0 100 200 300 400 500 600

Sens

or re

spon

se [H

z]

50[%RC]

40[%RC]

30[%RC]

20[%RC]

10[%RC]

Time [s]

0

100

200

300

400

500

600

700

800

900

0 100 200 300 400 500 600

Sens

or re

spon

se [H

z]

50[%RC]

40[%RC]

30[%RC]

20[%RC]

10[%RC]

Time [s]

Fig. 15 Responses of current and previous systems.

0

100

200

300

400

500

600

700

800

900

0 10 20 30 40

Concentration [%RC]

Sens

or re

spon

se [H

z]

6 CONCLUSION In the present study, the number of odor components used in

a olfactory display was extended from 8 to 32 even though the device size was still within the size of a laptop computer. The linear relationship between the specified concentration and the actual one at each bottle was confirmed in the case of a single odor component. The variation from bottle to bottle was within 10%. For a binary mixture, it was confirmed that the superposition theorem was valid. It was found that the fundamental capability of the current olfactory display was sufficient for presenting smells to people.

In the case of a low-volatility component, the linear relationship between the specified concentration and the sensor response was also confirmed in spite of slightly slow response.

Moreover, it was confirmed that the characteristic of the current system was equivalent to that of the previous one by comparing the two methods. Thus, it can be said that an olfactory display with a large number of odor components was successfully established in the present study.

The actual evaluation of the number of odor components will be performed in the near future. However, we selected about 90 compounds contained in fruit flavors when we searched for a flavor database (ESO, BACSIS). The number of odor components can be reduced when compounds with overlapping smells are eliminated. Although the required number of odor components depends upon the accuracy of odor approximation, 8 odor components might be insufficient for general purpose. However, there is the possibility that we can express fruit-flavor smells using 32 or fewer odor components.

Even if the number of odor components is 8, it can be used for the certain purposes. We have already demonstrated a movie with scent generated using the previous olfactory display [18]. We have also demonstrated other contents at several exhibitions [28]. The improved olfactory display is highly useful for enhancing movie experience with scents.

The application of the current olfactory display to an odor recorder is also promising. Although many odor components can be blended using an autosampler [23], such a blending process cannot be performed in real time. Since it is possible to blend many components in real time by using the current olfactory

500

100

200

300

400

500

600

700

800

900

0 10 20 30 40

Concentration [%RC]

Sens

or re

spon

se [H

z]

50

185


display, it will be useful in realizing an odor recorder with large recordable range. Acknowledgement This work was partially supported by SCOPE of Ministry of Internal Affairs and Communications, Japan.

REFERENCES [1] T.Nakamoto and T.Moriizumi, Artificial Olfactory System Using

Neural Network, H.Yamazaki (Ed.), Handbook of Sensors and Actuators, vol.3, Elsevier, 1996, 263-272.

[2] T. C. Pearce, S. S. Schiffman, H. T. Nagle and J. W. Gardner, Eds., Handbook of machine olfaction, 2003, Wiley-VCH.

[3] Proceedings of 11th International Symposium on Olfaction and Electronic Nose, Barcelona, Spain, 2005.

[4] T.Nakamoto, Y.Nakahira, H.Hiramatsu and T.Moriizumi, Odor recorder using active odor sensing system, Sensors and Actuators B, 76(2001) 465-469. .

[5] T.Yamanaka, R.Matsumoto and T.Nakamoto, Fundamental study of odor recorder for multi-component odor using recipe exploration based on singular value decomposition, IEEE Sensor Journal, 3 (2003) 468-474.

[6] T.Nakamoto, Odor recorder, Sensor Letters, 3 (2005) 136-150. [7] T.Nakamoto, Record of dynamical change of odor using odor

recorder, Sensors and Materials, 17, 7 (2005) 365-384. [8] M. Kendal-Reed, J. C. Walker, W. T. Morgan, M. LaMacchio and R.

W. Lutz, Human responses to propionic acid. I. Quantification of within- and between-participant variation in perception by normosmics and anosmics, Chem., Senses, 1998, 23, pp. 71-82.

[9] http://www.scentair.com [10] J. Messager, The diffusion of fragrances in a multimedia

environment, 3rd Aroma Science Forum, 2002, Sep 13 2002, Tokyo, pp. 2-7 (in Japanese).

[11] J. J. Kaye, Making scents, Interactions, Jan.+Feb. 2004, pp. 49-61. [12] T. Yamada, T. Tanikawa, K .Hirota and M. Hirose, Wearable

olfactory display: Using odor in outdoor environment, Proc. VR2006, pp. 199-206.

[13] F. Nakaizumi, Y. Yanagida, H. Noma and K. Hosaka, SpotScents: A novel method of natural scent delivery using multiple scent projectors, Proc. VR2006, pp. 207-212.

[14] F. Davide, M. Holmberg and I. Lundstorm, Virtual olfactory interfaces, electronic noses and olfactory displays, Communication through virtual technology: Community and technology in the Internet age, Eds., G.Riva and F.Davide, IOS Press, Amsterdam, 2001, pp. 193-220.

[15] L. Buck and R. Axel, A novel multigene family may encode odorant receptors: Molecular basis for odor recognition, Cell, 65 (1991) pp. 175-187

[16] J. E. Amoore, Molecular basis of odor, Charles C Thomas Publisher, 1970.

[17] http://www.ceatec.com/2006/en/visitor/ [18] T.Nakamoto, K.Yoshikawa, Movie with scents generated by

olfactory display using solenoid valves, Proc. VR2006, 2006, 291-292.

[19] T.Yamanaka, N.Nimsuk and T.Nakamoto, Concurrent recording and regeneration of visual and olfactory information using odor sensor, Presence, accepted.

[20] T.Nakamoto, S.Utsumi, N.Yamashita, T.Moriizumi and Y.Sonoda, , Active gas sensing system using automatically controlled gas

blender and numerical optimization technique, Sensors and Actuators B, 20 (1994) 131-137.

[21] T.Nakamoto, H.Takigawa, T.Yamanaka, Fundamental study of odor recorder using inkjet devices for low-volatile scents, Trans. on IEICE, 2004, E87-C (2004) 2081-2086.

[22] T.Yamanaka, R.Matsumoto, T.Nakamoto, Study of odor blender using solenoid valves controlled by delta-sigma modulation method, Sensors and Actuators B, 87(2002) 457-463.

[23] P.Somboon, B.Wyszynski and T.Nakamoto, Novel odor recorder for extending range of recordable odor, in press.

[24] S. R. Norsworthy, R. Schreier and G. C. Temes, Delta-sigma data converter, IEEE Press, 1997, pp. 1-41

[25] G. Sauerbrey, Verwendung von Schwingquarzen zur Mikrowagung, Z.Phys., 155 (1959) pp. 206-222.

[26] W. H. King, Piezoelectric sorption detector, Anal. Chem., 36 (1964) pp. 1735-1739.

[27] T.Nakamoto and T.Moriizumi, A theory of Quartz Crystal Microbalance based on Mason Equivalent Circuit, pn. J. Appl. Phys. Vol.29, No.5, 963-969.

[28] Pham Dinh Minh and T.Nakamoto, Evaluation of animation with smell using virtual olfactory display, Proc. IEICE general conference, 2006, A-16-15 (in Japanese).

186


Documents

Improvement of Olfactory Display Using Solenoid Valves