ADJUSTING FIRE DETECTION CRITERIA

Technical Field The present invention is in the field of early warning devices for fire detection.

Background of the Invention

Since 1975, the United States has experienced remarkable growth in the use of home smoke detectors, principally single-station, battery-operated, ionization- mode smoke detectors. This rapid growth, coupled with clear evidence from actual fires and fire statistics of the lifesaving effectiveness of detectors, has made the home smoke detector the fire safety success story of the past two decades.

In recent years, however, studies of the operational status of smoke detectors in homes revealed an alarming statistic that as many as one-fourth to one-third of smoke detectors are nonoperational at any one time. Over half of the nonoperational smoke detectors are missing batteries. The rest have dead batteries or nonworking smoke detectors. Homeowners' frustration over nuisance alarms was the principal reason for the missing batteries. Nuisance alarms are detector activations caused not by uncontrolled, harmful fires but by controlled fires, such as cooking flames. These nuisance or false alarms are also caused by nonfire sources, such as the moisture vapor that leaves a bathroom after someone has taken a

shower, dust or debris stirred up during cleaning in living quarters, or oil vapors escaping from the kitchen.

The reason the majority of smoke detectors, which are of the ionization type, is prone to these types of nuisance alarms is that these detectors are very sensitive to both visible and invisible diffused paniculate matter, especially when the fire alarm threshold is set very low to meet the mandated response time for ANSI/UL 217 certification for various types of fires. The size of visible paniculate matter ranges from 4 to 5 microns (although small panicles can be seen as a haze when present in high mass density) and paniculate matter is generated copiously in most open fires or flames. However, ionization detectors are most sensitive to invisible panicles ranging from 0.01 to 1 micron. Most household nonfire sources, as discussed briefly above, generate mostly invisible paniculate matter. This explains why most home smoke detectors encounter so many nuisance alarms.

The problem of frequent false alarms among ionization smoke detectors which results in a significant ponion of them at any one time being functionally nonoperational, led to the increased use in recent years of another type of smoke detector, the photoelectric smoke detector. Photoelectric smoke detectors work best for visible paniculate matter and are relatively insensitive to invisible paniculate matter. They are therefore less prone to nuisance alarms. However, the drawback is that they are very slow in responding to smoldering fires, in which the early paniculate matter generated is mostly invisible. To overcome this drawback, the fire alarm threshold of photoelectric smoke detectors must be set very low to meet the ANSI/UL 217 certification requirements. Such a low fire alarm threshold for photoelectric smoke detectors also lead to frequent false alarms. Thus the problem of nuisance false alarms for smoke detectors is seemingly unavoidable. Over the years the problem has long been recognized but has not been solved. A new type of fire detector is urgently needed to correct the dangerous ineffectiveness of present-day smoke detectors.

Another aspect of present-day smoke detectors that is often discussed but seldom addressed through innovation is the fire response slowness of these detectors. The current ANSI/UL 217 fire detector certification code was developed

years ago according to the fire detection technology available then, namely the smoke detector. Over the past two decades, the opinion of workers in the fire fighting and prevention industries has been critical of the speed of response of the smoke detector. Obviously, increasing smoke detectors' sensitivity by lowering their light obscuration detection thresholds will certainly speed up their response. However, that also drives up the nuisance alarm rates. It is clear that a better fire detector is needed.

Fire detectors that are currently available commercially can generally be classified into three basic categories - flame-sensing, thermal, and smoke detectors. The classifications are designed to respond to three principal types of energy and matter characteristic of a fire: flame, heat, and smoke.

The flame-sensing detector is designed to respond to the optical radiant energy generated by the diffusion flame combustion process, e , the illumination intensity and the frequency of flame modulation. Two types of flame detectors are commonly in use: the ultraviolet (UV) detectors, which operate beyond the visible at wavelengths below 4,000 A, and the infrared detectors, which operate in the wavelengths above 7,000 A. To prevent false signals from the many sources of ultraviolet and infrared optical radiation present in most hazard areas, the detectors are programmed to respond only to radiation with frequency modulation within the flicker frequency range for flame (5 to 30 Hz).

Flame detectors generally work well and seldom generate false alarms. However, they are relatively complex and expensive fire detectors that are not targeted for low-cost and mass-oriented usage. Instead they are mostly used in specialized high-value and unique protection areas, such as aircraft flight simulators, aircraft hangars, nuclear reactor control rooms, etc.

Thermal detectors are designed to operate from the thermal energy ouφut, the heat, of a fire. This heat is dissipated throughout the area by laminar and turbulent convection flow. The latter is induced and regulated by the fire plume thermal column effect of rising heated air and gases above the fire surface. There are two basic types of thermal detectors: the fixed temperature type and the rate-of- rise detector type. The fixed temperature type includes the spot type and the line

type. The spot detector involves a relatively small fixed unit with a heat- responsive element contained within the unit or spot location of the detector. With the line detector, the thermal reactive element is located along a line consisting of thermally sensitive wiring or tubing. Line detectors can cover a greater portion of the hazard area than can spot detectors.

Fixed temperature thermal fire detectors are reliable but not very sensitive. In modern buildings with high air flow ventilation and air conditioning systems, determining the placement of the fixed temperature detector is a difficult engineering problem. Consequently, this type of thermal fire detector is not widely used outside of very specialized applications.

A rate-of-rise thermal detector is usually installed in locations in which relatively fast-burning fires may occur. The detector operates when the fire plume raises the air temperature within a chamber at a rate above a certain threshold of operation, usually 15 ^* F per minute. However, if a fire develops very slowly and the rate of temperature rise never exceeds the detector's threshold for operation, the detector may not sense the fire.

A newer type of fire detector is called the rate-compensated detector, which is sensitive to the rate of temperature rise as well as to a fixed temperature level that is designed into the detector's temperature rating. Even with this dual approach, the most critical problem for the effective operation of thermal fire detectors is the proper placement of detectors relative to the hazard area and the occupancy environment. Consequently, this type of fire detector is seldom found in consumer households.

By far the most popular fire detector in use today is the smoke detector. Smoke detectors respond to the visible and invisible products of combustion.

Visible products of combustion consist primarily of unconsumed carbon and carbon- rich particles; invisible products of combustion consist of solid particles smaller than approximately 5 microns, various gases, and ions. All smoke detectors can be classified into two basic types: a photoelectric type, which responds to visible products of combustion, and an ionization type, which responds to both visible and invisible products of combustion.

The photoelectric type is further divided into a projected beam type and a reflected beam type. The projected beam smoke detector generally contains a series of sampling piping connected to the photoelectric detector. The air sample is drawn into the piping system by an electric exhaust pump. The photoelectric detector is usually enclosed in a metal tube with the light source mounted at one end and the photoelectric cell at the other end. This type of detector is effective due to the length of the light beam. When visible smoke is drawn into the tube, the light intensity of the beam received in the photoelectric cell is reduced because it is obscured by the smoke particles. The reduced level of light intensity causes an unbalanced condition in the electrical circuit to the photocell that activates the alarm. The projected beam or smoke obscuration detector is one of the most established types of smoke detectors. In addition to its use on ships, this detector is commonly used to protect high-value compartments of other storage areas and to provide smoke detection for plenum areas and air ducts. The reflected light beam smoke detector has the advantage of a very short light beam length, making it suitable for incorporation in the spot type smoke detector. The projected beam smoke detector discussed earlier becomes more sensitive as the length of the light beam increases, and often a light beam 5 or 10 feet long is required. However, the reflected light beam type of photoelectric smoke detector is designed to operate with a light beam only 2 or 3 inches in length. A reflected beam visible light smoke detector contains a light source, a photoelectric cell mounted at right angles to the light source, and a light catcher mounted opposite to the light source.

Ionization smoke detectors detect both the visible and invisible panicle matter generated by the diffusion flame combustion. As indicated previously, visible paniculate matter ranges from 4 to 5 microns in size, although smaller panicles can be seen as a haze when present in a high mass density. The ionization detector operates most effectively on particles from 0.01 to 1 micron in size. There are two basic types of ionization detectors. The first type has a bipolar ionized sampling chamber, which is the area formed between two electrodes. A radioactive alpha particle source is also located in this area. The

oxygen and nitrogen molecules of air in the chamber are ionized by alpha particles from the radioactive source. The ionized pairs move towards the electrodes of the opposite signs when electrical voltage is applied, and a minute electrical current flow is established across the sampling chamber. When combustion particles enter the chamber, they attach themselves to the ions. Because the combustion particles have a greater mass, the mobility of the ions decreases, leading to a reduction of electrical current flow across the sampling chamber. This reduction in electrical current flow initiates the detector alarm.

The second type of ionization smoke detector has a unipolar ionized sampling chamber instead of a bipolar one. The only difference between the two types is the location of the area inside the sampling chamber that is exposed to the alpha source. In the case of the bipolar type, the entire chamber is exposed, leading to both positive and negative ions. In the case of the unipolar type, only the immediate area adjacent to the positive electrode (anode) is exposed to the alpha source. This results in only one predominant type of ions, negative ions, in the electrical current flow between the electrodes.

Although unipolar and bipolar sampling chambers use different .principles of detector design, they both operate by the combustion products creating a reduced current flow and thus activating the detector. In general, the unipolar design is superior in giving the ionization smoke detectors a greater level of sensitivity and stability with fewer fluctuations of cuπent flow to cause false signals from variations in temperature, pressure, and humidity. Most ionization smoke detectors available commercially today are of the unipolar type.

For the past two decades, the ionization smoke detectors have dominated the fire detector market. One of the reasons for this is that the other two classes of fire detectors, the flame-sensing and thermal detectors, are appreciably more complex and costlier than the ionization smoke detectors. Therefore, they are mainly used only in specialized high- value and unique protection areas. In recent years, because of their relatively high cost, even the photoelectric smoke detectors have significantly fallen behind in sales to the ionization types. The ionization types are generally less expensive and easier to use and can usually operate for a full year

with just one 9-volt battery. Today, over 90 percent of households that are equipped with fire detectors use the ionization type of smoke detectors.

Despite their low cost, relatively maintenance-free operation and wide acceptance by consumers, these smoke detectors are not without problems and are certainly far from ideal. There are a number of significant drawbacks for ionization smoke detectors to operate as successfully as early warning fire detectors.

One drawback to smoke detectors is the importance of placing the detector at die spot where the fire breaks out. Unlike ordinary gases, smoke is a complex, sooty molecular cluster that consists mostly of carbon. It is much heavier than air and thus diffuses much slower than the gases we encounter every day. Therefore, if the detector happens to be some distance from the location of the fire, significant time will elapse before enough smoke gets into the sampling chamber of the smoke detector to trigger the alarm. Another drawback is the nature of the fire itself. Although smoke usually accompanies fire, the amount of smoke produced can vary significantly depending on the composition of the material that catches fire. For example, oxygenated fuels such as ethyl alcohol and acetone generate less smoke than the hydrocarbons from which they are derived. Thus, under free-burning conditions, oxygenated fuels such as wood and polymethylmeϋ acrylate generate substantially less smoke than hydrocarbon polymers such as polyethylene and polystyrene. Indeed, a small number of pure fuels, such as carbon monoxide, formaldehyde, metaldehyde, formic acid, and methyl alcohol, burn with noniuminous flames and do not produce smoke at all.

However, as noted earlier, one of the biggest problems with ionization smoke detectors is their frequent false alarms. Because of its operational principle, any micron-sized particulate matter, in addition to smoke from an actual fire, can set off the alarm. Kitchen grease particles generated by a hot stove are one classic example. Overzealous dusting of objects and/or furniture near the detector is another. Frequent false alarms are not just a harmless nuisance; people may disarm their smoke detectors by temporarily removing the battery to prevent such annoying episodes. The latter situation can be dangerous, especially when such people forget to re-arm their smoke detectors by replacing the battery.

To lessen the problems associated with false alarms in ionization smoke detectors, such detectors are normally set to sound an alarm at a smoke detection threshold level that is higher than that required to detect a fire. By increasing the detection threshold, fewer false alarms will be triggered. Unfortunately, uiis reduction in false alarms does not come without cost. Because the detection threshold is increased, it takes longer for the smoke detector to sound an alarm during an actual fire. In other words, the response time of the device is increased in order to decrease false alarms. The competing considerations of preventing false alarms and minimizing the response time of ionization smoke detectors are balanced in industry standards that have been adopted to promote safety and establish reliability and performance characteristics for smoke detectors.

The present standard for common household fire detectors in the United States is UL 217 Standard for Single and Multiple Station Smoke Detectors (Third Edition), which has been approved as an American National Standard and is hereinafter refened to as ANSI/UL 217-1985, March 22, 1985. ANSI/UL 217- 1985, March 22, 1985 covers (1) electrically operated single and multiple station smoke detectors intended for open area protection in ordinary indoor locations of residential units in accordance with the Standard for Household Fire Warning Equipment, NFPA 74, (2) smoke detectors intended for use in recreational vehicles in accordance with Standard for Recreational Vehicles, NFPA 501C, and

(3) portable smoke detectors used as "travel" alarms.

Recognizing that different types of fires have different characteristics, ANSI/UL 217-1985, March 22, 1985 contains four different fire tests, tests for paper, wood, gasoline, and polystyrene fires. The procedure for performing tests characteristic of each of these fires is set forth in paragraph 42 of ANSI/UL 217- 1985, March 22, 1985. According to paragraph 42.1 of ANSI/UL 217-1985, March 22, 1985, the maximum response time for an approved fire detector is four minutes for paper and wood fire tests, three minutes for a gasoline fire test, and two minutes for a polystyrene fire test. Because the highest maximum response time is four minutes, it is common to refer to a maximum response time for a household fire detector of four minutes without reference to the paper or wood fire tests.

Although ionization flame detectors sold for household use could be set to have a lower response time than four minutes, most household detectors have a maximum response time of four minutes or just under four minutes to minimize the risk of false alarms. Thus, an inherent limitation of commercially available ionization smoke detectors is a response time that is not optimized. Because the response time of a fire detector can be critical to saving lives and fighting fires, any improvement in response time, assuming that it does not increase the risk of false alarms or come at a prohibitive cost, would represent a significant advance in the art of fire detection and help satisfy a great need for improved fire detectors that save additional lives and property.

In an attempt to provide such an advance, efforts have been made to develop a new type of fire detector. In this regard, it has been known for a long time that as a process, fire can take many forms, all of which involve a chemical reaction between combustible species and oxygen from the air. In other words, fire initiation is necessarily an oxidation process because it invariably involves the consumption of oxygen at the beginning. The most effective way to detect fire initiation, therefore, is to look for and detect end products of the oxidation process. Wim the exception of a few very specialized chemical fires (i.e.. fires involving chemicals other than the commonly encountered hydrocarbons), there are three elemental entities (carbon, oxygen, and hydrogen) and three compounds (carbon dioxide ("C0 ₂ "), carbon monoxide, and water vapor) that are invariably involved in the ensuing chemical reactions or combustion of a fire.

Of the three effluent gases that are generated at the onset of a fire, CO ₂ is the best candidate for detection by a fire detector. This is because water vapor is a very difficult gas to measure as it tends to condense easily on every available surface, causing its concentration to fluctuate wildly depending upon the environment. Carbon monoxide, on the other hand, is invariably generated in a lesser quantity than CO _: , especially at the beginning of a fire. It is only when the fire temperature reaches 600 ^* C or above that a significant amount of carbon monoxide is produced. Even then, more CO ₂ is produced than carbon monoxide,

according to numerous studies of fire atmospheres. In addition to being generated abundantly from the start of the fire, C0 ₂ is a very stable gas.

Although it has been known in theory for many years that detection of CO ₂ should provide an alternative way to detect fires, CO ₂ detectors have not yet found widespread use as fire detectors due to their high cost and general unsuitability for use as fire detectors. In the past, CO, detectors have traditionally been infrared detectors that have suffered drawbacks related to cost, moving parts or false alarms. However, recent advances in the field of Nondispersive Infrared (NDIR) techniques have opened up the possibility of a viable CO ₂ detector that can be used to detect fires.

In U.S. Patent No. 5,053,754 by Jacob Y. Wong entitled "Simple Fire Detector," a fire detector using NDIR techniques is proposed. A beam of 4.26 μ light is directed through a sample of room air to measure the concentration of CO ₂ in this air, because CO ₂ has a strong absorption peak at this wavelength. Both the concentration and the rate of change of concentration of the CO ₂ are measured, enabling an alarm to be generated whenever either of these measured values exceeds a respective threshold value. Preferably, an alarm is sounded only if both of these values exceed their respective threshold values. The device is considerably simplified by the use of a window to the sample chamber that is highly permeable to CO ₂ but keeps out particles of dust, smoke, oil and water.

In U.S. Patent No. 5,079,422 by Jacob Y. Wong entitled "Fire Detection System Using Spatially Cooperative Multi-Sensor Input Technique," a set of N sensors are spaced throughout a large room or unpartitioned building. Comparison of data from different sensors provides information that is unavailable from only a single sensor. The data from each of these sensors and/or the rate of change of such data is used to determine whether a fire has occurred. The use of data from more than one sensor reduces the likelihood of a false alarm.

In U.S. Patent No. 5,103,096 by Jacob Y. Wong entitled "Rapid Fire Detector," a blackbody source produces a light that is directed through a filter that transmits light in two narrow bands at the 4.26-micron absorption band of CO, and at 2.20 microns, at which none of the atmospheric gases have an absorption band.

A blackbody source is alternated between two fixed temperatures to produce light directed through ambient gas and through a filter that allows only these two wavelengths of light to pass. To avoid false alarms, an alarm is generated only when both the magnitude of the ratio of the measured intensities of these two wavelengths of light and the rate of change of this ratio are both exceeded.

In U.S. Patent No. 5,369,397 by Jacob Y. Wong entitled "Adaptive Fire Detector," a fire detector that includes a CO ₂ sensor and a microcomputer is described that can alter the threshold detection level for CO ₂ before an alarm is sounded to compensate for variations in the background concentration of CO ₂ . Because virtually all fires generate CO ₂ , CO, detectors should be able to be used as fire detectors. However, there are three practical limitations that have to be dealt with in designing a fire detector that uses a CO, detector.

First, although fires generate copious amounts of CO,, there is one other commonly encountered source, albeit relatively weaker (namely, people) that also has to be taken into account. Because of this, the concentration level and rate of increase thresholds for alarm for CO ₂ sensors used as fire detectors cannot be set arbitrarily low. Otherwise CO ₂ generated by people's respiration in an enclosed space might be misinteφreted as a real fire. In practice, the rate of C0 ₂ generation by a typical fire can exceed that of human presence by several orders of magnitude. Thus, this limitation does not impair in any significant way the speed of response to the onset of real fires by CO ₂ fire detectors.

Second, because of the fact that CO ₂ concentration level and rate of increase thresholds cannot be set arbitrarily low because of human respiration, fires that generate very small amounts of CO ₂ , such as some types of smoldering fires, cannot be optimally detected by CO ₂ fire detectors.

Third, until such time as the manufacturing cost of an NDIR CO, detector is reduced to an economically attractive level, the consumer is unwilling to purchase this new and improved fire detector because of hard-nosed economics. The concomitant effort to simplify and cost-reduce an NDIR C0 ₂ detector is therefore equally important and relevant in introducing the currently disclosed practical and improved fire detector.

In U.S. Pat. No. 5,026,992, the present inventor began a series of disclosures on the novel simplification of an NDIR gas detector with the ultimate goal of cost-reducing this device to the point where it can be used to detect CO ₂ gas in its application as a new fire detector. In U.S. Pat. No. 5,026,992, a spectral ratioing technique for NDIR gas analysis using a differential temperature source was disclosed that leads to an extremely simple NDIR gas detector comprising only one infrared source and one infrared detector.

In U.S. Pat. No. 5,163,332, the present inventor disclosed the use of a diffusion type gas sample chamber in the construction of an NDIR gas detector that eliminated virtually all the delicate and expensive optical and mechanical components of a conventional NDIR gas detector. In U.S. Pat. No. 5,341,214, the present inventor expanded the novel idea of a diffusion type sample chamber of U.S. Pat. No. 5,163,332 to include the conventional spectral ratioing technique in NDIR gas analysis. In U.S. Pat. No. 5,340,986, the present inventor extended the disclosure of a diffusion type gas chamber in U.S. Pat. No. 5,163,332 to a

"re-entrant" configuration, thus simplifying even further the construct of an NDIR gas detector. Further simplification is required if CO ₂ sensors are to gain acceptance in low-cost household fire detectors and thus fulfill the long-felt need for an improved fire detector with a lower response time that still minimizes the occurrence of false alarms.

Summary of the Invention

The deficiencies of present-day smoke detectors can be substantially and effectively overcome in accordance with the present invention by the union of a smoke detector and a CO ₂ sensor. By combining a conventional smoke detector (photoelectric or ionization) with a CO ₂ detector into a new "dual" fire detector, it is possible to eliminate most commonly encountered false alarms. Furthermore, this dual fire detector is also significantly faster in detecting all types of fires, from the slow-moving, smoldering kinds to the almost smoke-free, fast-moving varieties.

Contrary to the common practice of increasing the sensitivity, or lowering the obscuration detection threshold, of a smoke detector to speed up its fire

detection response (thereby invariably decreasing its false alarm immunity), the new dual fire detector uses CO, as an additional input to minimize false alarms.

This additional input functions as a flag or a status switch for the new dual fire detector. When the CO, detector of this dual fire detector senses a preselected high level of CO, (e.g.. 3,000 ppm) and/or a preselected high rate of increase of CO ₂ , (e.g.. 200 ppm min.), the status switch is set to positive or "Ready to Go. " Once this flag is set ready to go, the dual fire detector can use its low light obscuration alarm threshold for smoke (which theoretically could be as low as the smoke detector would allow, typically a few tenths of a percent) to announce the onset of a fire with minimum delay while still minimizing the possibility of false alarms. (Light obscuration per foot is a standard unit for smoke concentration. It is frequently used even when a smoke detector that does not measure light obscuration is used. It is also frequently abbreviated to a simple "percent light obscuration. ") On the other hand, if the flag has not been set, the dual fire detector will not sound an alarm even if the normal light obscuration alarm threshold is reached or exceeded. During this normal alarm-sounding smoke condition, it waits for the flag to go positive before it announces the onset of the fire. This explains how most of the conditions for false alarms, whose obscuration times are usually much shorter than real fires such as the smoldering types, can be neutralized, thereby rendering the dual fire detector virtually false alarm resistant.

In order to safeguard against the occurrence of smoldering fires, the dual fire detector will sound an alarm if the smoke obscuration reaches a normal preset threshold such as that mandated by ANSI/UL 217-1985, March 22, 1985 for a predetermined period of time of up to an hour. Since most common household false alarm episodes last at best a few minutes, this alarm-sounding ability by the dual fire detector will at least equal that for die conventional smoke detector. However, it is faster than the conventional smoke detector to indicate a smoldering fire since it also detects the C0 ₂ level and/or rate of increase thresholds. Once the CO ₂ flag is set at ready to go, it will immediately sound the alarm and does not have to wait for the maximum period of up to an hour to do so.

Skilled persons will readily recognize this represents a dynamic adjustment to the smoke detector output signal fire detection criteria.

Another aspect of the dual fire detector takes full advantage of the fact that certain types of fast-moving fires generate a tremendous amount of CO ₂ but a relatively small amount of smoke. Thus, for these types of fires, the dual fire detector will quickly sound the alarm when the rate of CO ₂ increase exceeds an abnormally high threshold, such as 1,000 ppm/min., irrespective of whether any smoke obscuration has been reached. This particular fire detection capability of the dual detector for fast-moving fires is new and unique in the present invention and has not been realized or implemented by presently available fire detectors.

Although the CO, detector side of the dual fire detector could either use the concentration level and/or the rate of increase as a threshold condition to set the flag, the rate of increase alone suffices, and such a carbon dioxide detector can be implemented in the simplest and lowest cost fashion. Accordingly, detecting all types of fires, including the smoldering kind, with a shorter response time and virtually false alarm resistant and without prohibitively increasing cost would represent a significant advance in the art of fire detectors that could save lives and reduce property damage caused by fires.

The present invention discloses a number of the simplest possible embodiments of a combined NDIR C0 ₂ gas detector with a conventional smoke detector to achieve a practical and improved fire detector that is low in cost yet faster than presently available smoke detectors while still minimizing false alarms.

The present invention describes a practical and improved fire detector having a fast response time that detects common fires, including smoldering and fast- moving types, while still minimizing false alarms through the combination of a smoke detector and a CO, detector. In particular, the present invention uses novel design configurations, both mechanical and electrical, to implement the combination of a smoke detector and an NDIR CO ₂ gas detector as a low-cost, practical, and improved fire detector. In a first, separate aspect of the present invention, a smoke detector is used to detect smoldering fires when light obscuration exceeds a reduced threshold level

for longer than a second preselected time. If either of these conditions occurs, an alarm signal is generated in response to a smoldering fire. In addition, a CO ₂ detector is used to rapidly detect fires by monitoring the rate of increase in the concentration of CO,. When the rate of increase in the concentration of CO ₂ exceeds a second predetermined rate, an alarm signal is generated.

In another, separate aspect of the present invention, the maximum response time of the fire detector is lowered by relying upon the decreased maximum response time of the C0 ₂ detector. False alarms attributable to the smoke detector are minimized because there is no significant CO ₂ production in nonfire sources. Finally, false alarms attributable to the CO ₂ detector are rninimized by alarm logic, which responds to the detecting output of both the smoke detector and the CO ₂ detector.

Accordingly, it is a primary object of the present invention to provide a low- cost, practical, and improved fire detector with a reduced maximum response time that still minimizes false alarms.

This and further objects and advantages will be apparent to those skilled in the art in connection with the drawings and the detailed description of the preferred embodiment set forth below.

Additional objects and advantages of this invention will be apparent from the following detailed description of preferred embodiments thereof which proceeds with reference to the accompanying drawings. Brief Description of the Drawings

Fig. 1 is a logic diagram for a signal processor used in the preferred embodiment of the present invention; Fig. 2. is a block diagram for the preferred embodiment of the present invention;

Fig. 3. is a flow diagram implementing the logic of a signal processor in accordance with an alternative embodiment of the present invention;

Fig. 4 is a block diagram for an alternative embodiment of the present invention;

Fig. 5 is a schematic layout of a preferred embodiment of the current invention for a practical and improved fire detector showing a combination of a photoelectric smoke detector and an NDIR CO ₂ gas detector and their respective signal processing circuit elements and functional relationships; Fig. 6 is a schematic layout of a first alternate prefened embodiment of the current invention for a practical and improved fire detector;

Fig. 7 is a schematic layout of a second alternate preferred embodiment of the current invention for a practical and improved fire detector;

Fig. 8 is a schematic layout of a third alternate preferred embodiment of the current invention for a practical and improved fire detector; and

Fig. 9 is a schematic layout of a fourth alternate preferred embodiment of the cunent invention for a practical and improved fire detector;

Fig. 10 is an exploded isometric view of an infrared detector assembly exemplary for use in the present invention. Fig. 11 is an enlarged bottom view of substrate 450 of Fig. 10 showing thermopiles manufactured thereon.

Detailed Description of Preferred Embodiment

Fig. 1 is a logic diagram for a signal processor used in the preferred embodiment of a practical and improved fire detector. In the preferred embodiment of the present invention shown in Fig. 2, fire detector 100 combines a smoke detector 300 with a C0 ₂ detector 200. and the detection ouφuts of the smoke detector and the CO, detector are fed to a signal processor 40 to determine whether an alarm signal 51 should be generated and sent to alarm 500. The CO, detector 200 generates an output signal 210 representative of the CO, rate of increase in accordance with known principles of NDLR gas sensor technology. Skilled persons will readily recognize that a simple stream of CO, concentration samples is representative of the rate of change of C0 ₂ because the stream of C0 ₂ samples contains the rate of change of CO, information. Moreover, skilled persons will recognize that whether CO, detector 200 or signal processor 40 extracts the CO, concentration information makes no difference to the actual functioning of smoke detector 100.

The smoke detector 300 generates an output signal 310 representative of light obscuration in accordance with known principles of smoke detector technology. The signal processor 40 uses alarm logic to determine whether alarm signal 51 should be generated. Although it is preferred that a single signal processor 40 be used, multiple signal processors can be used; alternatively, portions of the alarm logic used to determine if an alarm signal 51 should be generated can be implemented as part of smoke detector 300 or C0 ₂ detector 200.

Fig. 1 is a flow diagram implementing alarm logic 400 of signal processor 40 shown in Fig. 2. The exact components used to accomplish the logical functions are not critical, nor are the pathways critical, as long as the same data will lead to the same results. Thus, for example, OR gate C ₄ could be replaced by multiple OR gates or other equivalent logic devices for accomplishing the same result. Similarly, although this diagram used AND and OR gates, the AND and OR gates can all be replaced be decision boxes. Accordingly, use of AND and OR gates is not meant to be restrictive and is done solely for ease of comprehension and illustration.

As illustrated in Fig. 1, fire detector 100 generates an alarm signal 51 when any of four conditions are met. First, an alarm signal 51 will be generated if the ouφut signal 310 form smoke detector 300 exceeds a threshold level A, for greater than a first preselected time A ₂ . Second, an alarm signal 51 will be generated if the output signal 310 from smoke detector 300 exceeds a reduced threshold level B , for greater than a second preselected time B ₂ . Third, an alarm signal 51 will be generated if the rate of increase in the concentration of CO, exceeds a first predetermined rate of C, and light obscuration exceeds a reduced threshold B,. Skilled persons will readily recognize that the third condition, when compared with the second condition, represents a dynamic adjustment to the smoke detector output signal fire detection criteria. Fourth, an alarm signal 51 will be generated if the rate of increase in the concentration of CO, exceeds a second predetermined rate C ₃ .

To increase the maximum response time, the preferred embodiment relies upon a CO, detector to allow the fire detector to measure rate of increase in the concentration of CO,. If the rate of increase exceeds a first predetermined rate of C , , and the smoke detector output signal 310 indicates that light obscuration also exceeds a reduced

7/27571 PC17US97/01264

18 threshold level B, as indicated by the AND gate C,, an alarm signal 51 is generated. Alternatively, if the CO, rate of increase exceeds a second predetermined rate C ₃ , an alarm signal is generated.

In accordance with the preferred embodiment, the first predetermined C0 ₂ rate of change C, is between approximately 150 ppm min. and 250 ppm/min., and the second predetermined CO, rate of change C ₃ is approximately 1,000 ppm/min. The first predetermined rate of change was obtained based upon fire tests for paper, wood, gasoline, and polystyrene fires performed in accordance with ANSI UL 217-1985, March 22, 1985 using an NDIR sensor in which the following averaged rates of change indicated a fire during each of the four tests: 300 ppm/min. for the paper fire test, 150 ppm min. for the wood fire test.250 ppm/min. for the gasoline fire test, and 170 ppm/min. for the polystyrene fire test. Using the foregoing rates of change to detect a fire, the average response time for detecting fires in each of these tests was 1.5 minutes. Under normal circumstances, a first predetermined CO ₂ rate of change between approximately 150 ppm/min. and 250 ppm/min. should not trigger false alarms, absent a sudden, localized fluctuation measured by the C0 ₂ detector, because it is well above the rate of change that should be encountered assuming proper ventilation. In this regard, HVAC Standard 62-1989 for a confined space states that the maximum rate of increase of CO, should be between 30 and 50 ppm/min. Thus, even if ventilation is not in compliance with this standard, a rate of change of 150 to 200 ppm min. still leaves a margin of error to prevent false alarms.

However, there may be situations in which there is a faulty ventilation or a sudden, localized fluctuation measured by the CO, detector. It is conceivable that the CO ₂ sensor could detect a sudden, localized rate of change int he range of 150 to 250 ppm/min. if it is located too near a potential source of CO ₂ , such as one or more persons exhaling directly into the CO, sensor. To prevent false alarms attributable to such unlikely situations, the fire detector logic of the preferred embodiment is configured such that an alarm signal will not be generated unless the rate of increase in the concentration of CO, exceeds the range of 150 to 250 ppm min. C, and light obscuration detected by the smoke detector exceeds a reduced threshold level B , . With

both of these conditions required to sound an alarm, the chance of false alarms is minimized.

Because the reduced light obscuration threshold can be set well below thresholds currently being used in smoke detectors designed for home use and still function as an inhibitor of a false alarm, the maximum response time is significantly less than that of current smoke detectors. This is because the reduced threshold is not being used in this application as an indication of a fire per se. Instead, it is being used as a test of the accuracy of the fire indication attributable to the CO, detector. Thus, the reduced threshold is set at a rate lower than that which would be acceptable in a smoke detector by itself, because it would be too susceptible to false alarms. Because light obscuration above the reduced threshold will not trigger an alarm signal absent a rate of change of CO, concentration that exceeds the first predetermined rate, false alarms attributable solely to the reduced threshold will not be caused by the fire detector. As a result, if a rate of change of between approximately 150 and 250 ppm min. is used as the first predetermined rate, the maximum average response time to detect a fire under each of the paper, wood, gasoline, and polystyrene tests of ANSI/UL 217-1985, March 22, 1985 can still be less than 1.5 minutes and in some instances actually less than 1 minute.

If the rate of change of C0 ₂ exceeds the second predetermined rate, it is unlikely that such a change would not be caused by a fire, assuming that the second predetermined rate is set high enough, that the fire detector is correctly positioned, and that there is no intentional attempt to set off the fire detector, such as a person deliberately and rapidly exhaling directly on the fire detector. Moreover, even if there is no fire, such an alarm will not be wasted because it can still identify a potentially dangerous condition that needs immediate attention. By including this option in the fire detector logic, the preferred embodiment detects fires with a very high rate of change in the concentration of CO ₂ , indicative of a fast-moving type of fire, earlier. In addition, this option helps to avoid problems associated with the incorrect placement of smoke detectors, because C0 ₂ gas molecules diffuse much faster than smoke particles. Although a CO, detector is very good at rapidly detecting fires, it is not very good at detecting smoldering fires in accordance with the test set forth in paragraph 43

of ANSI/UL 217-1985, March 22, 1985 using an NDLR sensor, it was found that the rate of CO, concentration necessary to detect a smoldering fire was approximately 10 ppm/min. Unfortunately, this rate of change is too low to be very useful in the types of applications covered by ANSI/UL 217-1985, March 22. 1985, such as household smoke detectors, because such a rate of change is below the acceptable rate of increase that can be encountered under normal conditions and thus would lead to false alarms.

To detect smoldering fires, the preferred embodiment includes a smoke detector to detect smoldering fires when light obscuration exceeds a smoldering fire detection level for greater than a preselected time. This can be accomplished in one of two ways: when light obscuration exceeds a threshold level A, for greater than a first preselected time A, or when light obscuration exceeds a reduced threshold level B, for greater than a second preselected time B,.

The first option for detecting smoldering fires relies upon a threshold level of obscuration that would detect wood, paper, gasoline, or polystyrene fires in accordance with ANSI/UL 217-1985, March 22, 1985 and still minimize false alarms but avoid the problem of false alarms by suppressing the alarm until a sufficient time has passed to rule out the possibility of a false alarm. In a preferred embodiment, the threshold level is the ANSI/UL 217-1985, March 22, 1985 threshold level, which originally was approximately 7 percent, and the first preselected time is 5 minutes. The second option for detecting smoldering fires relies upon a reduced threshold level of obscuration that is less than the threshold level and a second preselected time that is greater than the first preselected time. In this option, lower levels of obscuration are detected, but false alarms are avoided by requiring this condition to be met for a longer period of time. In a preferred embodiment, the reduced threshold level is substantially less than 7 percent, and the second preselected time is greater than 5 minutes but less than 60 minutes. In selecting the reduced threshold level, the reduced threshold level should not be set so low that it will produce false alarms due to the inherent sensitivity of the smoke detector; accordingly, the sensitivity of the smoke detector establishes a minimum level beneath which the reduced threshold should not be set. In selecting a reduced threshold level above this minimum, empirical test data can be used to optimize the desired results.

Further, the first and second options for detecting smoldering fires can both be used in the same fire detector to optimize results as is shown in Fig. 1. The signal processor can use alarm logic to trigger an alarm signal when either the first or the second option is met. Thus, for example, the threshold level could be set at approximately 7 percent. The reduced threshold level could be set at substantially less than 7 percent, the first preselected time could be set at 5 minutes and the second preselected time could be set at greater than 5 minutes but less than 60 minutes.

In accordance with a preferred embodiment, it is now possible to construct a fire detector that will meet ANSI/UL 217-1985, March 22, 1985, including the smoldering fire test, and also trigger an alarm within a maximum average response time of approximately 1.5 minutes when subjected to Tests A through D described in paragraphs 42.3 through 42.6 of ANSI/UL 217-1985, March 22, 1985.

In another aspect of the present invention, it is possible to build a fire detector with a very fast maximum response time in which a CO, detector is used to detect fires and a smoke detector is used to prevent false alarms. In this embodiment, shown in

Fig. 3, alarm logic 4A does not use the ouφut 310 from the smoke detector 300 to detect smoldering fires; instead, it is used solely as a test of the accuracy of the fire indication attributable to the CO, detector. Although this embodiment is not as preferred as the embodiment already described, it still represents a significant advance over the state of the art.

As illustrated in Fig. 3, fire detector 100 generates an alarm signal 51 when either of two conditions are met. First, an alarm signal 51 will be generated if the rate of increase in the concentration of C0 ₂ exceeds a first predetermined rate C, and light obscuration exceeds a reduced threshold B,. Second, an alarm signal 51 will be generated if the rate of increase in the concentration of C0 ₂ exceeds a second predetermined rate C ₃ .

As for the actual construction of a fire detector in accordance with the principles of the present invention, the components of the fire detector can be contained in a single package; alternatively, and less preferably, the individual components need not be contained in a single package. The fire detector can contain an alarm that is audible or visual or both: alternatively, the fire detector can generate an alarm signal

that is transferred to a separate alarm or an alarm signal can be used in any suitable device to trigger an alarm response or indication.

The CO, detector is preferably an NDIR gas detector. Suitable NDIR detectors could incoφorate the teachings of NDIR detectors disclosed in U.S. Patent No. 5,026,992 to Jacob Y. Wong entitled "Spectral Rationing Technique for NDLR Gas

Analysis" or U.S. Patent No. 5,341,214 to Jacob Y. Wong entitled "NDLR Gas Analysis Using Spectral Rationing Technique." For those CO, detectors used to measure CO ₂ concentration levels in parts per million, from which the C0 ₂ rate of change is derived, they should be stable and capable of accurate detection over long periods of time. To ensure accuracy and reliability, the drift of this type of CO, detectors should preferably be limited to less than approximately 50 ppm/5 years.

A simpler type of NDLR CO, detector is disclosed in U.S. Patent No. 5.163,332 to Jacob Y. Wong entitled "Improved Gas Sample Chamber." The reader's attention is directed to this reference for a description of this type of NDLR CO, detector. This patent discloses an NDIR C0 ₂ detector, the ouφut of which is directly indicative of and proportional to the C0 ₂ rate of change. This type of so-called "single beam" ND R gas detector is simpler, and hence easier, to implement and is consequently among the lowest cost NDIR gas sensors.

Smoke detector 300 can be an ionization type detector, but a photoelectric type of smoke detector is preferred.

The above discussion of this invention is directed primarily to the preferred embodiment and practices thereof. Further modifications are also possible in alternative embodiments without departing from the inventive concept. Thus, for example, the fire detector can be constructed so as to be programmable for different functions or to meet different requirements. In such a fire detector, any or all of the following can be programmable: the threshold level and the first preselected time, the reduced threshold level and the second preselected time, and the first and second predetermined rates of change. In another modification of the preferred embodiment, the fire detector logic can be altered to provide a first reduced threshold used to generate an alarm signal for detecting a smoldering fire and a second reduced threshold used as a test of the accuracy of the fire indication attributable to the CO, detector. In

another modification of the preferred embodiment, a different alarm or alarm signal can be generated for different types of fires. Such a detector is depicted in Fig. 4, in which fire detector 100 contains a CO, detector 200. a smoke detector 300, a signal processor 40, a fire alarm 500, and a smoldering fire alarm 600. Of course, the same result could be obtained by using fire alarm 500 to produce different alarms depending upon the type of fire.

In the preferred embodiment shown in Fig. 5, the pulsed ouφut of the silicon photodiode 1 of the photoelectric smoke detector 2 is pulsed by driver 5 at a frequency of typically 300 Hz and a duty factor typically of 5 percent. Under normal operating conditions, i^, in the absence of a fire, the AC output of photodiode 1 is near zero because no light is scattered into it from the LED source 4. During a fire condition, in which smoke is present in the space between the LED 4 and the photodiode 1 , an AC ouφut signal, the magnitude of which depends upon the smoke density, appears at the input of sample and hold integrator 3. The ouφut of the sample and hold integrator 3, which is a DC signal, is fed into high and low obscuration threshold comparators 6 and 7, respectively. The reference voltage at the high obscuration threshold comparator 6 represents a signal strength of scattered light at the silicon photodiode 1 where the obscuration due to the smoke condition is approximately 7 percent. Thus, when the smoke obscuration is equal to or exceeds 7 percent a the photoelectric smoke detector 2, the ouφut of comparator 6 will be at a HIGH logic state. Similarly, reference voltage at the low obscuration comparator 7 represents a signal strength of scattered light at the silicon photodiode 1 where the obscuration due to the smoke condition is less than 7 percent, e.g.. 2 percent. Thus, when the smoke obscuration is equal to or exceeds 2 percent at photoelectric smoke detector 2, the output of comparator 7 will be at HIGH logic state.

The outputs of comparators 6 and 7 are connected, respectively, to timers 8 and 9. Timer 8 is set at approximately 5 minutes and timer 9 is set at approximately 15 minutes. Timers 8 and 9 will be activated only when the ouφut logic states of comparators 6 and 7 are HIGH respectively. The outputs of timers 8 and 9 form two of the four inputs to OR gate 10. The ouφut of OR gate 10 is buffered by amplifier 1 1

before connected to the input of the siren alarm 12. The siren alarm 12 will sound whenever the ouφut of the OR gate is TRUE or HIGH.

The ouφut of low obscuration threshold comparator 7 also forms one of the two inputs to AND logic gate 26. The output of AND gate 26 forms the third input to OR gate 10.

The infrared source 13 of the NDLR C0 ₂ gas detector 14 is pulsed by current driver 15 at the typical rate of 1 Hz. The pulsed irifrared light incidents on infrared detector 16 through a thin film narrow bandpass interference filter 17 that allows only 4.26 microns radiation through to the detector. The filter 17 has a center wavelength of 4.26 microns with a full width at half maximum (FWHM) pass band of approximately 0.2 microns. CO, gas has a very strong infrared absoφtion band located spectrally at 4.26 microns. The amount of 4.26 microns radiation reaching the detector 16 depends on the concentration of C0 ₂ gas present between the source 13 and the detector 16.

The detector 16 is a single-channel micromachined silicon thermopile with an optional built-in temperature sensor in intimate thermal contact with the reference junction. The sample chamber area 18 of the NDIR C0 ₂ detector has small openings on opposite sides that allow ambient air to diffuse through the sample chamber area between the source 13 and the detector 16. These small openings are covered with a special fiberglass supported silicon membrane 20 to allow C0 ₂ to diffuse and to prevent dust and moisture-laden particulate matter from entering the sample chamber area 18.

The ouφut of the detector 16, which is a modulated signal, is first amplified by preamplifier 21 and then rectified to a DC voltage by rectifier 22 before being differentiated by differentiator 23. The ouφut of differentiator 23, which is proportional to the rate of change of CO, concentration in the sample chamber area 18, is fed into a pair of comparators 24 and 25. Comparator 24 is a low rate-of-rise comparator and its reference voltage corresponds to a rate of change of CO, concentration of approximately 200 ppm/min. When this rate of change for CO, is detected or exceeded, the output of the low rate-of-rise comparator 24, which is connected to the second input to the AND gate 26. will go HIGH or TRUE.

Comparator 25 is a high rate-of-rise comparator and its reference voltage corresponds to a rate of change of CO, concentration of approximately 1,000 ppm/min. When this rate of change for CO, is detected or exceeded, the ouφut of the high rate- of-rise comparator 25, which forms the fourth input to the OR gate 10, will go HIGH or TRUE.

The power supply module 27 takes an external supply voltage V^, and generates a voltage V ₊ for powering all the circuitry mentioned earlier. A backup power supply using standard batteries can also be derived from module 27 in a straightforward manner. The logic for the signal processor for the present invention of a practical and improved fire detector, as shown in Fig. 1, is implemented by the schematic layout of the preferred embodiment, as shown in Fig. 5, and the accompany description above.

In the first alternate preferred embodiment shown in Fig. 6, all the circuit elements described and shown in Fig. 5, with the exception of the module 27 and the siren alarm 12, are integrated using standard application specific integrated circuit

(ASIC) technique into a single ASIC chip 28. All the functions for this first alternate preferred embodiment are exactly the same as the preferred embodiment shown and described in Fig.2.

In the second alternate preferred embodiment shown in Fig. 7, the single- channel silicon micromachined thermopile infrared detector 16 (see Fig. 5) is replaced by a dual-channel silicon micromachined thermopile detector 30. As implemented, the C0 ₂ gas detector in this second alternate preferred embodiment is a full-fledged double-beam or dual-channel NDIR gas detector. Filter 31 is a thin film narrow bandpass interference filter having a center wavelength at 4.26 microns and a FWHM of 0.2 microns. Filter 32 has a center wavelength at 3.91 microns and a FWHM of 0.2 microns. It establishes a neutral reference channel for the gas detector as there is no appreciable absoφtion by common gases in the atmosphere in this particular neutral pass band.

In addition to the ASIC chip 28 in this second preferred embodiment, a microprocessor section 29 is added to the overall signal processor (SP) chip 33. With the use of a dual-channel CO, sensor, the gas concentration is first determined by

measuring the ratio between the ouφuts of the two detector channels within the dual- channel thermopile detector 30. The calculation of the ratio and the subsequent determination of the rate of change for CO, are performed in the microprocessor section 29 of the SP chip 33. As in the first alternate preferred embodiment shown in Fig. 6, all the logical functions are performed by the ASIC chip 28 as before.

In the third alternate preferred embodiment shown schematically in Fig. 8, the C0 ₂ gas detector is implemented with a special gas analysis technique known as "differential source" as disclosed in U.S. Pat. No. 5,026,992 by the present inventor. In this embodiment, the SP chip 33, comprising the microprocessor section 29 and the ASIC chip 28 used in the second alternate preferred embodiment (see Fig. 7), is retained. The microprocessor section 29 generates the necessary pulsing wave forms, namely two power levels alternately, to drive the infrared source 13. Meanwhile, the infrared detector 16 needs only to be a single-channel silicon micromachined thermopile with a dual pass band filter that has two nonoverlapping pass bands. One band is at 4.26 microns (C0 ₂ ), and the other is at 3.91 microns (neutral). The rest of the embodiment is the same as those already described.

In the fourth alternate preferred embodiment shown schematically in Fig. 9, the photoelectric smoke detector 2 and the NDLR C0 ₂ gas detector 14 of the previous four embodiments (see Fig. 5) are combined in a single device or detector assembly contained within a housing 36. Detector 34 housed within housing 36 can be a special dual -channel detector; one channel is a thermopile detector 35 with a CO, filter 37 and the other is a silicon photodiode 1 fabricated in its vicinity on the same substrate. Both are optically isolated from one another. Alternatively, housing 36 can contain a single- channel thermopile detector 35 with a C0 ₂ filter 37 and a separately packaged silicon photodiode 1.

In housing 36, there is a physical light-tight barrier 55 separating the two detector channels. On the CO, detector side, two or more small openings 38, made on one side of the container wall opposite the barrier 55. allow ambient air to diffuse freely into and out of the sample chamber area 39 of the CO, detector. Furthermore, these small openings 38 are covered with a special fiberglass silicon membrane 20 to

screen out any dust or moisture-laden particulate matters from area 39. CO, and other gases can diffuse freely across this membrane 20.

On the photoelectric smoke detector side 101, the light-tight barrier 55 sets up a scattering mode of operation for the infrared source 13 and the silicon photodiode 1 to detect smoke-caused obscuration due to fire. The microprocessor section 29 of the SP chip 33 processes the signals in nearly the same manner as in the preferred embodiments shown and described in Fig. 5. The rest of the signal processing for this fifth alternate preferred embodiment is exactly the same as that for the previously disclosed embodiments. As those skilled in the art will readily recognize, there are a number of ways to manufacture or configure a single-channel infrared detector 16, a dual-channel thermopile detector 30, and the dual channel detector 34, which is comprised of a thermopile detector 35 and a photodiode detector 1. With respect to detectors 16 and 30. however, preferably the detectors) and corresponding band pass filter(s)~ depending on whether the detector is a single- or dual-channel infrared detector— are combined in a single platform such as a TO-5 can to form an infrared detector assembly.

An exemplary detector assembly 403 is now described in connection with FIGS. 10 and 11. Although, as illustrated in FIGS. 10 and 1 1, the detector assembly 403 includes three thermopile detectors 404, 405, and 406, the physical configuration of each thermopile detector and its supporting elements is generalizable to the infrared detector assemblies of the embodiments shown in FIGS. 5-9. Thermopile detectors 404, 405, and 406 have been formed on substrate 450 mounted within detector housing 431. Detector housing 431 is preferably a TO-5 can. comprised of a housing base 430 and a lid 442. Lid 442 includes a collar 407 into which a gas permeable top cover 420 is set and bonded.

Thermopile detectors 404, 405, and 406 are supported on a substrate 450 that is made out of a semiconductor material such as Si. Ge, GaAs, or the like. Interference band pass filters F„ F ₂ , and F ₃ are bonded with a thermally conductive material, such as thermally conductive epoxy, to the top of raised rims 482 surrounding apertures 452.

An advantage of securing the filters to the raised rims 482 with a thermally conductive

material is that it improves the thermal shunting between the filters and the substrate 450, which is the same temperature as the reference, or cold, junctions of the thermopile detectors 404.405, and 406. As a result, the background noise from the interference filters is minimized. In the present embodiment, thermopile detectors 404, 405, and 406 are preferably thin film or silicon micromachined thermopiles. Thermopiles 404, 405, and 406 each span an aperture 452 formed in the substrate 450. Apertures 452 function as windows through which the radiation that is passed by band pass filters F„ F ₂ , and F ₃ is detected. As is well known in the art, thin film or micromachined thermopile detectors 404, 405, and 406 are manufactured on the bottom side of the substrate 450 and may employ any of a number of suitable patterns. Fig. 11 is an enlarged view of the bottom side of substrate 450 and illustrates one suitable pattern that could be employed for thin film or micromachined thermopile detectors 404, 405, and 406.

As is typical in the art. the hot junctions 460 of each of the thermopile detectors 404, 405, and 406 are preferably supported on a thin electrically insulating diaphragm

454 that spans each of the apertures 452 formed in substrate 450 and the cold junctions 462 are positioned over the thick substrate 450. Alternatively, diaphragms 454 may be absent and the thermopile detectors 404, 405, and 406 can be self-supporting. To improve the sensitivity of thermopiles 404, 405, and 406 to incident radiation, the top side of the electrically insulating diaphragm 454 can be coated with a thin film of bismuth oxide or carbon black during packaging so the aperture areas absorb incident radiation more efficiently. If the thermopile detectors 404, 405, and 406 are self-supporting, the side of hot junctions 460 upon which radiation is incident can be directly coated with bismuth oxide or carbon black. By positioning the cold, or reference, junctions 462 over the thick substrate 450, the reference junctions of each of the detectors are inherently tied to the same thermal mass. Substrate 450, therefore, acts as a heat sink to sustain the temperature of the cold junctions 462 of each of the detectors at a common temperature. In addition, substrate 450 provides mechanical support to the device. The present embodiment has been described as a single substrate 450 with three infrared thermopile detectors 404, 405, and 406 formed thereon. As one skilled in the

art would recognize, two or three separate substrates each having one infrared thermopile detector manufactured thereon could be used in place of the substrate 450 described in the present embodiment.

Electrically insulating diaphragm 454 may be made from a number of suitable materials well known in the art, including a thin plastic film such as Mylar® or an inorganic dielectric layer such as silicon oxide, silicon nitride, or a multilayer structure composed of both. Preferably, the diaphragm 454 is a thin inorganic dielectric layer because such layers can be easily fabricated using well-known semiconductor manufacturing processes, and, as a result, more sensitive thermopile detectors can be fabricated on substrate 450. Moreover, the manufacturability of the entire device is improved significantly. Also, by employing oniy semiconductor processes to manufacture thermopile detectors 404, 405, and 406, substrate 450 will have on-chip circuit capabilities characteristic of devices that are based on the full range of silicon integrated circuit technology; thus, the signal processing electronics for thermopile detectors 404, 405, and 406 can, if desired, be included on substrate 450.

A number of techniques for manufacturing thermopile detectors 404, 405, and 406 on the bottom side of substrate 450 are well known in the thermopile and infrared detector arts. One method suitable for producing thermopile detectors 404, 405, and 406 using semiconductor processing techniques is disclosed in U.S. Patent No. 5,100,479, issued March 31, 1992.

Output leads 456 are electrically connected using solder or other well-known materials to the ouφut pads 464 of each of the thermopile detectors 404, 405, and 406. Because the reference junctions of thermopile detectors 404, 405, and 406 are thermally shunted to one another, it is possible for the reference junctions for each of the thermopile detectors 404, 405, and 406 to share a common output pad. As a result, only four, rather man six, output leads would be required to communicate the ouφut of the detectors. The ouφut leads 456 typically connect the thermopile detectors 404, 405. and 406 to signal processing electronics. As mentioned above, however, the signal processing electronics can be included directiy on substrate 450, in which case ouφut leads 456 would be connected to the input and output pads of the signal

processing electronics, rather than the ouφut pads from the infrared thermopile detectors 404, 405, and 406.

A temperature sensing element 453 is preferably constructed on substrate 450 near cold junctions 462 of thermopile detectors 404, 405, and 406. The temperature sensing element monitors the temperature of substrate 450 in the area of the cold junctions and thus the temperature it measures is representative of the temperature of the cold junctions 462. The ouφut from the temperature sensing element 453 is communicated to the signal processing electronics so the signal processing electronics can compensate for the influence of the ambient temperature of the cold junctions of the thermopile detectors. Temperature sensing element 453 is preferably a thermistor, but other temperature sensing elements, such as diodes, transistors, and the like, can also be used.

In FIGS. 10-11, interference band pass filters F„ F ₂ , and F ₃ are mounted on the top of substrate 450 so they each cover one of the apertures 452 in substrate 450. Because the interference filters cover apertures 452, light entering detector assembly

403 through window 444 must first pass through filter F„ F ₂ or F ₃ before reaching thermopile detector 404, 405, or 406, respectively. Thus, by employing three separate apertures in substrate 450, light passing through one of the filters is isolated from the light passing through one of the other filters. This prevents cross talk between each of the detector channels. Therefore, the light reaches thermopile detectors 404. 405, and

406 from the passive infrared source 408 is the light falling within the spectral band intended to be measured by the particular detector. This construction is generalizable to the two-channel case shown in Fig. 7. Infrared light source 413 works as infrared source 13 works, as described in the text that refers to FIGS. 5-9. Interference band pass filters F„ F ₂ and F ₃ are mounted on the top of raised rims

482 so they each cover one of the apertures 452 in substrate 450. The center wavelength and FWHM of band pass filters F„ F ₂ , or F ₃ may be set as described, in connection with Fig. 5-9 above, with two or more of exemplary filters F,, F,, or F ₃ absent. Because the interference filters cover apertures 452, light entering detector housing 431 through window 444 must first pass through filter F,, F ₂ , or F, before reaching thermopile detector 404, 405, or 406, respectively. Thus, by employing three

separate apertures in substrate 450, light passing through one of the filters is isolated from the light passing through one of the other filters. This prevents cross talk between each of the detector channels. Therefore, the light that reaches thermopile detectors 404, 405, and 406 from infrared source 413 is the light falling within the spectral band intended to be measured by the particular detector.

Substrate mounting fixtures 486 are connected using solder or other well-known materials to the output pads (not shown) of each of the thermopile detectors 404, 405, and 406 at bonding regions 488. As the reference junctions of the thermopile detectors 404, 405, and 406 share a common ouφut pad in the present embodiment, only four substrate mounting fixtures 486 are required to communicate the ouφuts of the detectors. Substrate mounting fixtures are insulated from the housing base 430 of detector housing 431 because they are mounted on an electrically insulative substrate 490. which is preferably made out of a material selected from a group consisting of aluminum oxide and beryllium oxide. The output signal from thermopile detectors 404. 405, and 406 is communicated through substrate mounting fixtures 486, via wire bonds 494, to signal processing electronics 492. The signal processing electronics 492 can comprise a plurality of microchips or a single microchip diebonded to insulative substrate 490. Ouφut leads 456 are connected to the input and output of the signal processing electronics 492 via wire bonds 496. Similarly, in connection with dual-channel detector 34 described in connection with Fig. 6, the same principles of construction are equally applicable to the micromachined thermopile detector 35/C0 ₂ filter 37 combination. Further, as one skilled in the art would readily recognize, it is possible to fabricate silicon photodiode 1 on the same silicon substrate as thermopile detector 35. It will be readily apparent to those skill in the art that further changes and modifications in the actual concepts described herein can readily be made without departing from the spirit and scope of the invention as defined by the following claims.