,----------------------~-_ 254 .. _. -- --- PHILIPS TECHNICAL REVIEW VOLUME 29 Flame-failure control with a UV -sensitive cold-cathode tube I. Design of the tube and circuit Il. Statistical aspects of the detection process and choice of the alarm level Many heating systems - from the domestic gas water heater and the central heating boiler to large industrialfurnaces and steam boilers - require a simple but reliable safety device which monitors the flame to see that it is still alight. For this purpose there are particular advantages in making use of the ultra-violet radiation from the flame: the following two articles deal with this method of detection. The first describes a smal! gasdischarge tube which responds to UV radiation, There are statistical fluctuations in the output voltage of the tube, and therefore the maximum acceptable risk of false alarm is one of the factors determining the choice of the alarm level for the safety system. Since an exact calculation of the required alarm level is impracticable, a statistical method of approximation has been developed, and this is discussed in the second article. J. Design of the tube and circuit T. Poorter If the flame of a gas-fired furnace or boiler is accidentally extinguished with the gas supply left on, there is soon sufficient unburned gas accumulated to constitute an explosion hazard. Every furnace or boiler therefore requires a safety system which monitors the flames continuously and shuts off the fuel supply and gives an alarm signal if they go out. In oil-fired equipment, continued supply of the fuel . when the flame has gone out has consequences which are perhaps less Serious but are still highly undesirable, and here too it is standard practice to provide flame. failure control. Various methods may be applied to detect the presence of a flame; they differ considerably in their speed of response and the choice between them therefore depends to a large extent on the response time required in the safety system for a particular boiler. An exceptionally short response time can be achieved with a detector that responds to the ultra-violet radiation of the flame. A further advantage of this method is that it provides a simple means of distinguishing between the flame and the hot wall of the combustion chamber. For use in. this method Philips have developed a small Ir. T. Poorter is with the Philips Electronic Components and Materials Division (Elcoma), Eindhoven, cold-cathode gas-discharge tube which ignites when it is irradiated with ultra-violet light (wavelength between 200 and 290 nm). Combined with a simple electronic circuit, this tube gives a very reliable flame detector. Before describing the tube we shall first consider in more detail some aspects of flame-failure control in boilers and furnaces [11. Other safety systems, such as those which respond to the gas pressure and are sometimes used in addition to flame detection, will not be considered here. Some methods of flame-failure control For large boilers which can deliver several hundred kW, where the explosion hazard is relatively great and the possible consequences particularly serious, the safety system is required to have a very short response time, no longer than a few seconds. For small boilers a response time of a few tens of seconds is often permitted. The flame detector can then be a simple bimetallic strip or thermocouple, and is usually placed in the pilot flame. The advantage of these detectors is that no electronic amplification for energizing a relay . is required; the disadvantages are that they very quickly [1] A more general treatment can be found in: B. Maizier, C.R. Congrès Ass. tech. Ind. Gaz 81, 769-825, 1964. 1968, No. 8/9 FLAME-FAILURE CONTROL WITH COLD-CATHODE become sooted up, which -upsets the thermal contact with the flame, and the bimetallic strips have only a short life. In furnaces the situation is similar. For domestic cookers a bimetallic strip or thermocouple is considered sufficient, but for industrial furnaces a safety system with a much shorter response time is required. A short reaction time is obtained by using either the ionization in the flame or the radiation from the flame. In the first case several electrodes are placed in the flame; when a voltage is applied between these electrodes, a very small current (a few tens of microamperes) flows while the flame is alight. Electronic amplification has to be provided if a relay is to be energized while this ionization current flows. Disadvantages of the method are that the detected current is so small that any leakage currents due to dirt and moisture on the electrodes can cause considerable difficulties, and that the electrodes burn away slowly and therefore need regular replacement. The method is completely unsuitable for oil-fired equipment. With a device that detects the radiation from the flame the problems of wear and leakage currents due to contamination do not arise; a difficulty here, however, is that the detector, when used in a furnace, "sees" the hot wall of the furnace. Fig. 1 shows the spectra of the radiation, emitted by an oil flame, a gas flame and the wall of a furnace at different temperatures. If the detector is to be able to tell when the flame goes out, it must be sensitive in a spectral region in which the flame is a great deal "brighter" than the furnace wall. The ability to distinguish between flame and furnace wall can also be based on an entirely different criterion, the "flickering" of the flame. This calls for the use of a fast detector (some infrared detectors are suitable - the spectral region does not matter in this case) whose signal is applied to a selective a.c. amplifier, tuned to about 15 Hz. This method, however, is rather complicated and costly. At wall temperatures below about 1200 oe the oil flame is brighter than the wall in the visible range (above 400 nm), and in this case cadmium-sulphide cells, which are sensitive to visible light, can therefore be used. The brightness of the gas flame in the visible range is however too low to permit safe operation with a cadmium-sulphide cell. In the ultra-violet range, on the other hand, between say 200 and 300 nm, both the oil and the gas flame are much "brighter" than even the hottest furnace wall. Ultra-violet detection thus offers a universal means offlame-failure control. Moreover, as explained below, a UV detector can be given such a high intrinsic "gain" that its output signal may if required be used directly for actuating a relay, which simplifies the system considerably. What, is more, a TUBE, 255 I UV detector can monitor the ignition spark as well as the flame. Since large furnaces or boilers are always automatically ignited by means of an electric spark, this indicates the possibility of monitoring and safeguarding the entire ignition prograf!lme (a spark ignites a pilot flame which ignites the main burner). The radiation emitted by a flame becomes "redder", i.e, the spectrum shifts to longer wavelengths, as the distance from the base of the flame increases - the ultra-violet radiation is found mostly in the base of the flame. The UV detector must therefore preferably be situated in such a way that the radiation from the base of the flame is incident upon it. The best place for the detector is therefore in the air-supply opening; this has the additional advantage that the detector is cooled by the jet of air. 1250oC..······ .... .... .... .. .. ' / 10 .... ' 11500C...· .../,./ . .' .,/ . 400 . 500nm _À Fig. I. Spectral intensity distribution (in arbitrary units) of the radiation emitted by a gas flame (solid line), an oil flame (dashed line) and the furnace wall (black-body radiator) at different temperatures (chain-dotted lines). A cold-cathode tube as UV detector If a sufficiently high voltage is applied between the cathode and anode of a gas-discharge tube (higher than the "ignition voltage"), a self-sustaining discharge occurs in the tube as soon as there are sufficient numbers of free electrons between the electrodes. These electrons may be released from the cathode by thermionic emission or by incident radiation (photoelectric effect). In the latter case, which applies for the cold-cathode tube, the type of cathode material used determines the wavelength range of the radiation to which the tube responds. The current of the self-sustaining discharge is so very much greater than that of the originally generated electrons that the tube has an extremely high "internal gain". These are the principles underlying the UV detector described below. The tube, 'type 155 DG (fig. 2), contains two identical molybdenum electrodes. The work function of this metal (i.e. the minimum energy an electron must acquire in order 'to leave the solid) is 4.26 eV, corresponding PHILlPS 256 TECHNICAL REVIEW VOLUME 29 directly from the mains. The gas filling is a mixture of hydrogen and helium, giving an ignition voltage of 200 V, i.e. about 60 % of the peak voltage of the a.c. supply. The tube is operated in the glow-discharge region, in which the maintaining voltage for this tube is also about 200 V. An incident UV photon can thus initiate a gas discharge when the instantaneous value of the supply voltage is greater than 200 V; the discharge goes out immediately the voltage drops below the maintaining voltage. In order to prevent spontaneous reignition due to residual ionization from a preceding cycle, the de-ionization time of the gas mixture should be very short; the gas mixture which we have used meets this requirement. Fig. 2. The cold-cathode tube type 155 UG. If the voltage between the electrodes is sufficiently high, a self-sustaining gas discharge is produced when UV radiation is incident on the electrodes. Since molybdenum has been used as electrode material, the tube responds only to UV. to a photoelectric threshold of about 290 nrn ; only photons with a wavelength below this threshold are effective. Obviously, the bulb ofthe tube must be made of a material that transmits UV at these wavelengths; the glass used for the bulb is a type that transmits radiation at wavelengths greater than 200 nm. The tube thus responds to radiation between 200 and 290 n m. Fig. 3 shows the spectral distribution of the photoelectric quantum efficiency (i.e. the ratio of the number of electrons released to the total number of incident photons), the transmission characteristic of the glass, and the resultant spectral sensitivity for detection. The tube was designed for operation at 220 V a.c. and a frequency of 50 or 60 Hz, so that it can be supplied e Fig. 3. General spectral distribution of the sensitivity s of the detector, derived from the quantum efficiency e(À) of the photoelectric effect at the electrode and the transmission t(},) of the glass bulb. Fig. 4. Cut-away view of the 155 UG, showing the electrodes. The effective part of the electrodes consists of closely-spaced straight pieces of wire at a short distance apart. The electrodes are identical and, since the supply is a.c. they serve alternately as cathode and anode. Fig. 5. Transverse cross-section through the effective part of the electrodes. The electrical field strength is greatest between the wires, and smallest at A. The probability that a UV photon will initiate a discharge therefore depends on the place where it strikes the electrode. The probability is greater at a higher voltage. Fig.4 illustrates the electrode configuration with which this value of the maintaining voltage is obtained. As an a.c. supply is used, the two electrodes are identical, each acting alternately as cathode and anode. The parallel parts of the wires constitute the active region of the electrodes. These wires are roughly circular in cross-section, so that the electric field strength is not the sarne at every point ofthe surface. The field strength is of course greatest between the wires (fig. 5). This 1968, No. 8/9 FLAME-FAILURE CONTROL WITH COLD-CATHODE TUBE, I 257 our explanation of this rnechanism we shall briefly conmeans that electrons released at different places on the sider some of the effects encountered in a cold-cathode electrode surface have a different probability of initiagas-discharge tu be. ting a self-sustaining discharge. Since, when the voltage During a self-sustaining discharge in such a tube the between the electrodes is increased, the area of the eleccathode surface is continuously bombarded by ions trode surface for which a released electron can initiate which are accelerated by the electric field in the cathode a self-sustaining discharge becomes greater, the sensitivfall. The electrons which maintain the discharge are ity of the tube (the number of ignitions per unit time released from the cathode by the bombardment, which for a given incident radiation) will increase with rising also releases particles ofthe electrode material from the voltage. The 155 UG does not therefore have the plateau so characteristic of the Geiger-Müller tube, ; surface (sputtering). These sputtered particles spread through the tu be by diffusion, perhaps forming cornwhich it closely resembles in operation. For this reason pounds with molecules in the gas, e.g. oxygen molethe 155 UG is not so suitable for measuring the intencules, and finally settle in one form or another on the sursity of the radiation; this tube is primarily intended to face of the electrodes or on the glass wall. The ion bombe a very sensitive device for detecting whether or not bardment thus has a decontaminating effect on the part radiation is present. of the cathode which is covered by the discharge (and The voltage between the electrodes is subject to an also on the gas atmosphere), and a contaminating effect upper limit. The maximum permissible value is deteron every part that is not covered. It is evident that the mined by the "free-running" voltage, that is to say the contarnination will be greatest on the part of the cathvoltage at which the tube no longer goes out when the ode surface which lies just outside the limits of the disirradiation is interrupted but spontaneously re-ignites charge. in each successive cycle of the supply voltage. The cause This effect occurs in the tube discussed here, as it of this effect is a declining weak ionization which perdoes in all other tubes designed for this kind of applicasists, in spite of the short de-ionization time of the gas tion. One of the forms which the contamination takes mixture, after the discharge has gone out. In the here is a spontaneous crystal growth of molybdenum 155 UG the free-running voltage is considerably higher oxide at the boundary of the discharge (fig. 6). The than the nominal supply voltage, so that in practice there is little danger of this effect being encountered. During the life of the tube, however, the value of the free-running voltage may decrease if the electrode surface becomes contaminated, e.g. by the growth of molybdenum-oxide crystals. Locally higher field strengths occur at the tips of such crystals, increasing the chance of re-ignition. The occurrence of such contaminations can be prevented, however, by applying a suitably shaped current waveform during the discharge. We shall now deal with this point at somewhat greater length. Effect of current waveform on the reliability of the tube Contamination or "poisoning" of the cathode surface is harmful not only because it lowers the freerunning voltage: if the contaminants are materials of relatively low work function, like molybdenum hydi ide or alkali metals from the glass, the tube might begin to respond to visible light, which is the very thing that has to be avoided. Fortunately, it so happens that the discharge itself removes the contaminants. Extensive life tests have shown that the electrode surface remains completely clean if at each discharge the glow is made to cover the entire active area of the electrodes by sending a short pulse of current which is sufficiently large through the tube. Since an a.c. supply is used, both electrodes are continuously decontaminated in this way. To assist Fig. 6. Molybdenum-oxide crystals, which can form on the electrodes of the 155 UG if the peak current through the tube is too low. Magnification 30x. oxygen needed for this reaction comes from the glass wall. This release of oxygen from the wall is also one of the reasons for the gradual decline in the sensitivity of the tube, since the oxygen absorbs UV photons produced in the discharge and thus hinders the transition from non-self-sustaining to self-sustaining discharge, in which these photons play an important role. Although the decline in sensitivity does not reduce the margin of safety - if the sensitivity is inadeq uate the tube gives the alarm while the flame is still burning (it is "fail-safe") -, this deterioration should obviously be avoided as far as possible for the sake of a reasonable tube life. 258 .:1 PHILlPS TECHNICAL VOLUME REVIEW 29 The way to stop cathode contamination by crystal quired waveform and produce an output signal that growth has already been indicated: in each discharge indicates whether or not the flame is alight. We have there should be a pulse of current which is large enough seen that two types of circuit are possible: one that to make the discharge cover the entire active surface of gives a very small dissipation in the tube and one that the cathode. During the remaining part of the half- gives a high dissipation. In the first circuit a short high-current pulse must cycle the current may not drop to such a value that the discharge contracts to a small part of the cathode sur- flow in the tube at each discharge, and the current must face: in this part of the cycle the current must be either then drop right down to zero. This is done by making zero or so high that the whole cathode remains covered. the discharge self-quenching. The circuit is shown in Both alternatives are possible, and the circuits which fig. 7. Vs is the mains voltage with an r.m.s. voltage can be used will be discussed below. In the first case the of 220 V, the load resistance R2 has a high value and mean current is very Iow and there is thus very little RI has a very Iow value. We start from the situation dissipation; in the second case the dissipation is rela- in which the discharge is out; the full voltage then appears across the tube. When Vs increases to a value tively high. It is interesting that this difference is reflected in all kinds of ways in the sensitivity behaviour of the tube. With low dissipation the glass wall remains relatively 155UG~ cold and gives off little oxygen. Nevertheless, oxygen pressure is gradually built up. The glass wall is also 3L.. ,...., ----+ .gradually darkened by the UV radiation from the discharge. The overall result is a slight decline in sensitiv- V. =FCt ity, about to half the original value in 10000 hours C2==' VrtJ RI, of operation. With high dissipation the wall gets much R2[ [JRt hotter and the emission of oxygen is greater. However, the oxygen pressure remains lower than it does for low 0--........ - ...... -----........ -0---'- ----- dissipation, since the higher current causes greater Fig. 7. Circuit for the cold-cathode tube in which the discharge is cathode sputtering and therefore better decontaminaself-quenching: at each discharge a short pulse of current passes tion of the gas atmosphere. The higher mean current through the tube, but the average value of this current - i.e. the also gives rise to greater UV irradiation of the wall, but dissipation in the tube - is low. The supply voltage to the circuit is 220 V from the a.c. mains; the output voltage V(t) is a measure here again there is ample compensation; since the wall of the intensity of the UV radiation incident on the tube. temperature is higher, the glass does not darken so quickly. For one reason and another the decline in sensitivity is even smaller for high dissipation than it greater than the ignition voltage VI, an incident UV photon can initiate the discharge. Since RI is Iow, a is for low dissipation. What are the consequences for the practical applica- high current first flows through the tube via Cl and RI. tion ofthe tube? Since the sensitivity of the tube grad- Cl becomes charged, the voltage across R2 rises and ually decreases, while the free-running voltage does the voltage across the tube falls. The component values not decrease when the tube is operated with the kind of are so chosen that the tube voltage then falls to below current waveform we have described (in practice it may the steady-state maintaining voltage Vrn which correeven rise for various reasons), the tube gives a reliable sponds to the load resistance R2. Now when the current flame-failure control. The behaviour of the tube out- through Cl and RI stops flowing because Cl is charged lined above has been established from life tests of up up, the tube can no longer continue to operate and to 25 000 hours and a total of one and a half million cuts out. The capacitor Cl now discharges through hours of tube life. The end of the useful life of a tube RI and R2, which causes the voltage across the tube to in a particular application is determined entirely by the rise again. Spontaneous re-ignition is not possible beRI)CI has been made minimum sensitivity required for that application; this cause the time constant (R2 depends on factors such as the distance from the flame, so large that the voltage across the tube rises more the size of the flame and the minimum level required slowly than the re-ignition voltage (which rises again to actuate the electronic circuit in which the tube is rapidly because of the de-ionization). The discharge thus remains out and can only be re-ignited again when incorporated. the voltage across the tube has risen again sufficiently The circuit and a new UV photon strikes the (acting) cathode. The circuit in which the tube is incorporated must be The required value of the time constant has been found designed to give the current through the tube the re- to be small enough for several discharges to take place if '---' .e, !IW + 1968, No. 8/9 FLAME-FAILURE CONTROL in this way during a half-cycle of the supply voltage. Each discharge produces a voltage pulse across R2. The use of an a.c. supply voltage means that both positive and negative pulses appear across R2 (in the positive and negative half-cycles) with respect to the reference potential in fig. 7. Only the positive pulses are passed via Ra and the diode D to the capacitor C2; this charges up and discharges again between two pulses through the load resistance R4, giving rise to a fluctuating voltage Vet) across R4; see fig. 8. This output voltage Vet) is a measure of the number of discharges that have taken place during a number of cycles (or, to be more exact, during their "positive" halves) before time t, and is thus a measure of the intensity of the UV radiation. The average magnitude of Vet) at a particular radiation intensity is defined as the sensitivity of the tube in the circuit. The incident radiation is adjusted such that Vet) is normally 10 to 15 V; ifthere are no discharges, then Vet) is of course zero. The rate at which C2 discharges between two pulses is determined, for a given R4, by the capacitance of C2; if the capacitance has a high value (i.e. the circuit has a high time constant), Vet) decreases slowly, so that the average value is relatively high. A smaller capacitance gives a faster decrease in Vet), and therefore a greater speed of response and a lower output voltage, which in turn gives a lower sensitivity. The fluctuations of Vet) have a statistical character. In the first place, the incidence of the UV photons at the electrodes is a statistical effect and in the second place not every photon incident at the negative electrode causes a discharge, the probability of this depends on the location. As already explained, this probability increases with increasing supply voltage; mains voltage fluctuations therefore affect the output voltage as well. When the circuit is used in a flame-failure control system it is necessary to take these different fluctuations into account; this point will be dealt with below. In the circuit shown in fig. 7 the available current I in the load resistor R4 is of the order of some tens of microamperes, so that the current has to be amplified to energize a relay in a flame-failure control system. In the second circuit, where a high dissipation is tolerated in the tube, there is no need to amplify the output signal; a relay winding is incorporated directly in the circuit. The circuit is given in fig. 9; the discharge here is not self-quenching. Vs is the mains voltage, RI has a high value and R2 a low value. The relay Rel is shunted across the capacitor C. The current surge begins here in the same way as in the circuitin fig. 7, but the discharge goes out only when Vs falls below the maintaining voltage. The average current flowing through the tube in this circuit is considerably higher than in fig. 7; the peak value can be as high as WITH COLD-CATHODE I v t 1 TUBE, I I ill I."i--._ !;: I I 1 I: I I I I I 1 I " ---+: 1 . .1 I I I I I I 259 I I I : I I Irt! r I II I"-t---:.. I 1 I I I. 1 I 1 1 I 1 1 I I I 1 I I I I 1 1 I I I I I I I _Q I-t I I I I I 1 1 I I I I Fig. 8. Illustrating the variation of the output voltage V of the circuit in fig. 7 as a function of time (a). The waveform of the supply voltage Vs(t) is shown in (b). A UV photon incident on the tube can initiate a discharge in those parts of the cycle in which Vs is greater than the ignition voltage Vi of the tube. On account of the statistical character of this process there is a fluctuation in V(t). several hundred milliamps. The tube has been found capable of withstanding this high dissipation without suffering any damage. The operation of this circuit differs from that of the circuit described earlier in that there are fewer ignitions per unit time, and because of this the tube does not discriminate so sharply between flame and background radiation. These details cannot be dealt with here. The remainder of this article will be based on the circuit of fig. 7, and the same applies for the following article. The choice of alarm level When the UV detection circuit in fig. 7 is used in a flame-failure control system, the output voltage Vet) is 155UG ! 0 L Rel Fig. 9. Circuit in which the mean tube current is sufficiently high to energize directly a relay Rel. The 155 UG tube can withstand this high dissipation without damage. 260 PHILTPS TECHNICAL applied to a circuit which closes the gas valve when the voltage Vet) drops below a critical value, the alarm level. The gas valve is opened again when Vet) rises above this level. There are some advantages in using different levels for closing and opening the gas valve, but here we shall confine ourselves to the simple case of a single level. Even if the flames are not alight, discharges can still be caused by cosmic radiation, or by ultra-violet radiation from sunlight or a fluorescent lamp shining in through a furnace window. As a result, even when the flames are out, the voltage Vet) can rise to a small positive value. The alarm level must therefore lie well above this value - the higher the better for safety. It mayalso happen that, while the flame is alight, there are no discharges at all during one or two positive halfcycles of the supply voltage, so that the voltage Vet) drops below the alarm level and the burner is unnecessarily shut off. The chance of such a false alarm is of course greater the higher the alarm level has been chosen. On the other hand, a higher alarm level implies a faster response. The correct alarm level therefore represents a compromise between the speed of response and the chance of a false alarm. The main parameters which determine the circumstances under which the compromise is chosen are the time constant of the detector REVIEW VOLUME 29 circuit and the radiation level at the detector. By varying these parameters, for example in positioning the detector in relation to the source, one has to ensure that a compromise can be found which meets the minimum requirements specified by the user in both respects (speed of response and avoidance of false alarm). . Since by its nature, the output signalof the UV detector is subject to fluctuations, the relation between alarm level and the probability of a false alarm can only be investigated by means of a statistical analysis. This is the subject of the second article. Summary. The flame-failure control system for a furnace or boiler must shut offthe fuel feed the moment the flames are extinguished for one reason or another. There are various methods of detecting the presence of a flame. After an introduetion to this subject, a small cold-cathode tube is discussed which is ignited by the UV radiation from a flame provided the voltage across the tube is greater than the ignition voltage. The tube operates on a 220 V a.c. supply; there are two identical molybdenum electrodes, alternately acting as cathode and anode. If steps are taken to ensure that the glow covers the entire cathode during each discharge, there is no contamination of the electrodes, which could reduce the reliability of the tube. The effects which determine the life of the tube are dealt with at some length. Two circuits are discussed which give the current through the tube the required form and deliver an output current which is a measure of the intensity of the incident UV radiation. There are statistical fluctuations in the output signal, and a result of this is that a certain chance of false alarm has to be accepted when establishing the alarm level, i.e. the level at which the fuel supply is shut off. Il. Statistical aspects of the detection process and choice of the alarm level R. P. Adriaanse and P. van der Laan The first of these two articles described a UV detector for use in a flame-failure control system; there are statistical fluctuations in the output voltage of the detector. In this second article the output signal will be mathematically analysed and described with the aid of a statistical model. A calculation will then be given which shows how the alarm level can be chosen in such a way that a' false alarm will occur on an average no more than twice in a year of continuous operation. The problem is of a complicated nature since the statistical model describes a "Markov" process with dependent increments. One of the implications of this is that the Ir. R. P. Adriaanse, is with the Technical University of Delft, and Drs. P. van der Laan is with Philips Information Systems and Automation (Research Dept.), Eindhoven; both were formerly with Philips Research Laboratories, Eindhoven. fluctuations in the output voltage can only be characterized by probability distributions. The analysis has been carried out for the circuit in fig. 7; a typical curve of the output voltage Vet) from this circuit when the flame is alight is shown in fig. 8. A false alarm occurs if Vet) drops below the alarm level while the flame is alight: we are therefore interested in the lowest values that Vet) can reach. The various components of the circuit of course each have a' spread about a nominal value. These deviations have an effect on V(t); in our calculations we must always use the particular combination of values that yields the lowest value of Vet) ("worst-case" combination). Since the calculations will be illustrated with numerical examples, Table I gives the nominal value of various components and the worst-case value for. a particular version of the