MXPA96003693A - Method to turn on and operate a high pressure discharge lamp and a circuit to delete the met - Google Patents

Abstract

The present invention relates to a method for operating a gas discharge lamp, characterized in that it comprises the steps of: a. operating the gas discharge lamp at a plurality of operating frequencies b. detecting and taking a plurality of samples of a parameter of the electric lamp, at each of a plurality of operating frequencies; calculate a deviation to each of the plurality of frequencies, of the samples taken at each of the frequencies, and d. evaluate the calculated deviations and select a lamp operating frequency, based on the evaluation

Description

Method to turn on and operate a high pressure discharge lamp and the circuit to execute the method. The invention relates to a method for igniting and operating a high pressure gas discharge lamp, in particular at high frequencies, and also a syrup for executing the method. In general, a circuit for turning on and operating a high pressure discharge lamp is called a ballast. In particular, the invention relates to a method that detects arc instabilities in the lamp and adjusts the operating frequency of the ballast to avoid visible situation during the operation of the lamp. High-pressure discharge lamps (HID), for example, mercury vapor, metal-alurous lamps and high-pressure sodium lamps, are typically operated with a magnetic ballast at normal supply line frequencies or slightly above of normal, for example, 50-100 Hz. It would be desirable to provide an electronic ballast that operates HID lamps at frequencies above about 20 kHz. High frequency ballasts are becoming increasingly popular for low pressure mercury vapor fluorescent lamps. High frequency operation allows the magnetic elements of the ballast to be of a very small size and weight compared to a conventional low frequency magnetic ballast. High-frequency operation also provides substantial increases in lamp performance of the order of 10-15% for fluorescent lamps due to reductions in cathodic fall. A similar decrease in size and weight for HID lamps would be desirable, especially for the lower-wattage metal halide lamps used for shop and runway lighting, because it would provide more flexibility in the design of aesthetically pleasing appliances for such uses. The performance of the lamp would also increase a few percentage points, but not as much as in the case of fluorescent lamps. A major obstacle to the use of high frequency electronic ballasts for HID lamps, however, is formed by arc instabilities due to acoustic resonances that can occur in high frequency operations. The acoustic resonances, at least, cause fluctuation of the arc that is very annoying for human beings. In a more severe case, acoustic resonance can cause the arc discharge to extinguish, or even worse, permanently deviate against and damage the wall of the discharge vessel, which causes the discharge vessel to explode. The frequencies at which acoustic resonance occurs depend on many factors, including * - dimensions of the arc tube (ie, length, diameter, shape of the final chamber, presence or absence of tabulation), density of the filling gas, the operating temperature and the orientation of the lamp. For high frequency ballasts, the operating frequency of the lamp current is fj will generally be selected above the audio frequency range (fj> 20 kHz), but may be lower. For the typical balasta that works with sinusoidal waves (distorted), the frequency of power fp is twice the frequency of the current, so that fp is greater than 40 kHz. The arc tubes, or discharge vessels, of the high pressure sodium lamps and some of the newer metallic aluro lamps are ceramic and cylindrical in shape. The arc tubes of mercury vapor lamps and metallic aluro lamps are made of quartz glass, typically with a cylindrical body and rounded end chambers. The feeding frequencies from which longitudinal acoustic resonance occurs for these generally cylindrical arc tubes can be roughly determined from the formula: I c1-1, / t = l, 2.3 ,. where L represents the typical length of the arc tube, n denotes an enter number, denotes an average speed of sound in the longitudinal direction of the burner and equal to approximately 450 m / s. The radial-azimuthal modalities are given by: where * denotes an average speed of sound in the radial direction, R denotes the typical radius of an arc tube, and a-m denotes the zeros of the first derivative of the Bessel functions. The full resonance spectrum? \ M? It is calculated from: F? Lnn? * Ü (3) If the length of the arc tube is substantially greater than the radius, the frequency at which the fluctuation occurs can be estimated by formula (1) for the longitudinal resonant frequencies. For the specific case of a 100W metal halide lamp with a 15 mm arc tube length, for example, it is expected that the minor longitudinal resonant frequencies will occur at 15 kHz power frequencies. Therefore, resonances of higher magnitude will occur at power frequencies fp greater than 30 kHz, which correspond to the frequencies of the current fx over the audible range. Therefore, an approximation of the resonant frequencies can be made by calculations and / or observed through experiments by operating the lamps at different frequencies and visually observing the fluctuation > visible. For a specific lamp type under specific operating conditions, an operating frequency may be selected at which no visible fluctuation occurs and there is a ballast designed to operate the lamp at this previously selected frequency. However, the ballast would be limited to a specific lamp from a specific manufacturer. In addition, modifying the operating conditions, for example, changing the environmental conditions or the blackening of the lamp through its life which would alter the temperature and / or operating pressure, could change the resonant frequencies in such a way that resonance would occur. in the frequency of operation of the previously selected ballast. Alternatively, especially in the case of quartz glass arc tubes where the control of the dimensions is difficult, even lamps of the same manufacturer would have different resonance points in such a way that it is possible that a considerable percentage of the lamps would fluctuate in the operating frequency of the selected ballast In addition to not being fault-free, manufacturing a ballast for a specific lamp from a certain manufacturer is expensive in view of its limited market and is inflexible for the user. Therefore, it would be desirable to provide a ballast for a greater range of lamps that detects arc instabilities during operation and select the frequency of operation to avoid arc instabilities due to acoustic resonance. The article "An Autotracking System for Stable Hf Operation of HID Lamps", F. Bernitz, Sy p. Light Sources, Karlsruhe 1986, discloses a controller that continuously varies the operating frequency of the lamp around a center frequency over a sweep range. The sweep frequency is the frequency at which the operating frequency is repeated through the sweep range. The controller detects the lamp voltage to evaluate the arc instabilities. A control signal is derived from the voltage of the detected lamp to vary the sweep frequency between 100 Hz and a few kHz to achieve stable operation. However, this system has never been commercialized. JP 4-277495 (Kamaya) reveals a ballast that detects the impedance of the discharge lamp. If the impedance of the lamp is below a specified level, the ballast reduces the high frequency oscillating components in the lamp current. A disadvantage of this design, however, is that the specified level is fixed, and as noted previously, the resonant frequencies actually vary from lamp to lamp. Traditionally, even though the high frequency components are reduced in the current of the lamps, there is no guarantee that the operation will not change to another resonant frequency in which arc instabilities occur. Accordingly, it is an object of the invention to provide a method for detecting arc instabilities in gas discharge lamps, which is universally applicable despite the power, type, dimension, or physical or chemical composition of the lamp. It is another objective to provide a method that can be implemented in a wide range of ballast topologies. It is yet another object to provide a method for operating HID lamps at high frequencies to detect avoiding frequencies in which acoustic resonance occurs for a wide range of lamps, or at least lamps. It is still another object to provide a lamp ballast or controller, to put this method into practice. In the method according to the invention, the arc instabilities are detected by evaluating the deviations in an electrical parameter of the lamp. The operating parameters of the lamp can then be modified based on an evaluation of the deviations to avoid additional arc instabilities. The invention is based on the recognition that arc instabilities are accompanied by deformations in the arc that change its length, which in turn vary the electrical parameters of the lamp, for example the voltage, current, conductance, or so Inverse, the impedance of the lamp. The operating frequencies at which the arc instabilities caused by the acoustic resonance occur are determined by evaluating the deviations determined in each of a plurality of operating frequencies. A particularly suitable embodiment of the invention for detecting the frequencies under which arc instabilities due to acoustic resonance occur includes the steps of: operating the gas discharge lamp at a plurality of frequencies; detecting and taking a plurality of samples of a selected electrical parameter in each frequency; calculating a deviation, in each of the plurality of frequencies, of the electric parameter samples taken at each frequency; evaluate the calculated deviations; and selecting a lamp operating frequency based on the evaluation of the deviations. Favorably, the deviations are evaluated to determine the frequency that has the smallest deviation in the parameter of the sampled lamp and the frequency of operation is adjusted to this frequency. Iterating the previous steps with different frequency sections and scanning speeds allows to determine quickly and precisely the frequencies at which the global and then local minima occur in the deviations and therefore the arc deviations. Favorably, the parameter of the detected lamp is the conductance of the impedance, which in general allows the precise detection of arc deviations much smaller than is possible with either the voltage or the current alone. In assessing deviations in conductance, it has been discovered that arc deviations due to acoustic resonance can be detected at levels that are not perceptible by the human eye. The use of conductance or impedance, which are reasons for the voltage and current of the lamp, allow simpler sampling techniques than is possible with either the voltage to the current alone. Additionally, the use of conductance or impedance overcomes the problems of detecting voltage or current alone, which usually does not work properly. For example, voltage detection alone is sensitive to line variations such as jolts, drops in the main line and other external conditions. In accordance with another embodiment, the deviation detected is the standard deviation. This has the advantage that all arc deviations detected in a sampling interval contribute to the effective deviation instead of, for example, only the maximum deviation detected in a sampling interval. According to another embodiment, the detection of the frequencies at which acoustic resonance occurs is achieved by performing a symmetric sweep of frequencies around a central frequency, including sweeping a portion with a set of frequencies where the frequencies increase and a portion that includes the same set of frequencies where the frequencies decrease. It was found that there is hysteresis in the deflection response of the arc, such that arc deviations at any arc frequency depend not only on the sweep rate at which the frequencies are changed but also whether the frequencies decrease or increase. By measuring the deviations of each frequency in an asymmetric way, for example, sampling each frequency both when the frequencies increase and when they decrease at the same scanning rate, the effects of the hysteresis are eliminated and the detection of the frequencies at which the resonance occurs is eliminated. . Advantageously, the fluctuation-free frequencies section and the selection of the operating frequency are carried out during a step of starting the operation of the lamp, between the lighting of the lamp and the steady-state operation. In the starting stage the frequencies at which the acoustic resonance occurs rapidly change due to the increase in gas pressure and the temperature of the lamp. Although this does not immediately appear to be a good means of detecting frequencies that have no visible fluctuation due to acoustic resonance, it has been found that the frequency and iteration sweep speeds can be selected such that the controller can be stopped at a frequency with a local minimum in the deflection of the arc reflex and trace this minimum until the end of the start. This has the advantage that when the starting period ends and the lamp has reached full light emission, an operating frequency has already been selected and the lamp can be operated without visible fluctuation from the end of the starting period. It has been found that this technique works well despite the lower possibilities of developing arc instabilities during startup due to lower gas pressures and temperatures. Favorably, the detection of deviations in the arc continues during steady-state operation in such a way that adjustments can be made in the operating frequency to account for small changes in filling pressure and gas and / or temperature caused, for example, for environmental effects. For control in the steady state, the speed and sweep of the frequency sweeps are selected in such a way as to avoid visible fluctuation. Random deviations in the arc can occur due to effects such as arc leaps at the electrode or concealments caused by the sudden entry into the arc flow of droplets of filler material found in the fixed lamp discharge base. If these random events are not taken into account, they would cause errors in the detection of standard deviations during the frequency sweep and cause operation at a new frequency corresponding to a false minimum in the standard deviation. In order to avoid this, in yet another embodiment, each frequency sweep is performed twice consecutively and operation is started at a new center frequency only if the difference between the respective set of deviations measured in the two iterations is found within a prescribed level. The lamp driver, or ballast, according to the invention includes a high frequency inverter for operating the HID lamp within a selected frequency range. The inverter is responsive to a control signal supplied by an arc stability controller that implements the selected steps of the method described above. Upon incorporation, the arc instability controller includes a microprocessor programmed with software to give the steps of the selected method. In contrast to the systems known here commercially available, the balasta employs universal operating principles suitable for operating HID lamps of different types, manufacturers and watajes despite the presence of acoustic resonance between these lamps over a fairly wide range of frequencies. In a favorable embodiment, the ballast includes a control device which ensures that the power supplied to the lamp remains substantially constant despite changes in the frequency of the inverter, and consequently the operating frequency of the lamp, during detection and balasta eludes it, particularly during the regime, in a manner substantially invisible to the user, this is substantially without visible fluctuation and substantially without visible changes in the quality and intensity of the light. In an incorporation of the ballast, the power control is obtained by controlling an overvoltage output of a boost converter that feeds the inverter. The gain of the inverter circuit, consequently the current of the lamp, depends on the inverting frequency. The voltage and current of the lamp are detected to obtain a power signal of the lamp that is compared with a reference power signal. The duty cycle of a boost switch found in the boost converter is controlled to adjust the overvoltage to compensate for changes in inverter gain. Because power control limits power, it also serves to prevent catastrophic failure of the lamp at the end of its life cycle. In yet another embodiment, a ballast includes a control to set the overvoltage during the ignition and start of the lamp, when the lamp presents only a very small charge to the reinforcement converter, to protect the elements of the circuit and the lamp. These and other objects, features and advantages of the invention will become apparent with reference to the drawings and the following detailed description, which are illustrative only, and not limiting. Figure 1 illustrates a displacement of a fluctuating discharge arc; Figure 2 illustrates a favorable sampling technique to measure the standard deviation in the conductance within a frequency window; Figure 3 is a material flow diagram of an open loop control to discover the minimum fluctuation in a selected frequency range and calculate the standard deviation s from the conductance samples, Gjk, in accordance with an embodiment; Figure 4 illustrates a simultaneous sampling of voltage and current to calculate the conductance; Figure 5a illustrates a global frequency sweep for detecting strong resonances; Figure 5b illustrates the location of strong resonances and the permissible operating windows. that are free from strong resonances; Figure 5c is a graph of the permissible operating windows in which strong resonances do not occur for several different 100 watt metal halide lamps A, B, C, D, E from different manufacturers; Figure 6a illustrates the deviation in the conductance representing weak resonances over a range of frequencies within the allowable window of Figure 5c; Figure 6b illustrates the global minimum for weak resonances, in the normal deviation of the conductance shown in Figure 6a; Figures 7a, 7b and 7c illustrate three control options for detecting the global minimum of standard deviations for weak resonances; Figure 8 illustrates the deviations measured around the operating frequency selected in steady-state operation; Figure 9 is a material flow diagram of a control loop for operating a high pressure discharge lamp; Figures 10a; 10b show the output of the routine of Figure 9 for a metal halide lamp of 100; Figure IA illustrates a situation where the standard deviation a has a broad minimum and a narrow minimum both less than the limit value sfl; Figure 11b illustrates the selection of the largest minimum as the new global minimum for the standardized data of the standard deviation; Figures 12a; 12b illustrate the representative conductivity responses for the arc deviations caused by arc flash and sodium flash, respectively; Figure 13 is a schematic block diagram of an HID lamp ballast with resonance detection and frequency control; Figure 14 is a schematic diagram of a power and surge control (Control D) of Figure 13; Figure 15 illustrates a pulse duration control of a control signal V, for the boost switch Ql to control the overvoltage and power supplied to the lamp; Figure 16 shows the circuit connections of the ICs incorporating various elements of Control A and Control D of Figure 13; Figure 17 is a block diagram of Control C of Figure 13 for Arc Instability Control; and Figure 18 is a schematic diagram of a half-bridge controller (Control B) of Figure 13. The Figure illustrates a discharge arc 1 between a pair of discharge electrodes 2 of an HID lamp in a vertical operating position. The arc tube is not shown for clarity. The arc is subject to fluctuation induced by acoustic resonance. The central position of the discharge arc 1 represents the position of the arc in an arbitrary time t in which it has a length L. The discharge arc that is in a compensated position l1 shown in dashed line represents the deviated position of the arc in a time t + 5t, from which it is seen that the arc has a greater length L +? l. The different lengths of the non-deviated and deflected arc do not cause the electrical parameters of the lamp, for example, voltage, current, conductance and impedance to have different values in each of these positions and is the main cause of the change of these electrical parameters. Therefore, acoustic resonance causes changes in the length of the arc and consequently changes in these electrical parameters. The detection of the changes in these parameters therefore gives the possibility of detecting the acoustic resonance electrically and of developing a scheme to evaluate in what frequencies the resonance and the deviation of the arc occur and to control the operating frequency of the lamp to make it operate at frequencies in which no visible fluctuation occurs. It should be noted that in certain situations satisfactory control can be obtained by detecting the deviations only in the voltage of the lamp V or in the current of the lamp I., using the impedance and, in particular, the conductance G = I / V have numerous advantages. Measurements of change in conductance or impedance will always have an equal or better signal-to-noise ratio than measurements of change in current or voltage alone. The signal-to-radio ratio is typically greater than 20dB in conductance or in impedance than in current or voltage alone. By measuring both voltage and current simultaneously to calculate their ratio with respect to impedance or conductance, noise contributions, for example from a power line, are substantially canceled out in these signals. If the voltage to the current is used by itself, the noise signals would remain. These advantages allow the detection of very small deviations in the arc which are not noticeable by the human eye and which in practice are not typically detectable using the voltage or current of the lamp alone. Additionally, the control based on the conductance or impedance is independent of the topology of the lamp driver circuit while the voltage-based or the sorptive control by itself depends on the topology. Finally, when both voltage and current are detected simultaneously to calculate the impedance or conductance, a much simpler sampling scheme than what is possible when sampling the voltage or current alone can be used. Although the detection of impedance and conductance share the above advantages, conductance detection is preferred for the reason that, before, during and immediately after switching on, the lamp current is zero or very small and the initial voltage of the lamp is very high. During these distances, the impedance R = V / I would be infinite or very high. Conversely, the conductance G = I / V is zero or small during these periods and can always be calculated. Using the conductance also makes the method / balasta less sensitive with respect to the lamp. For example, if a lamp is replaced or replaced with a lamp of a different type, V and I would change but G = V / I would remain in the same relative range. With conductance, this control can be considered even more universal, that is, it can be applied to a variety of lamps of different type, model and type. Accordingly, throughout the remainder of the detailed description, the incorporations will be described with reference to the detection of conductance, and those of ordinary skill in the art will appreciate that the impedance can be substituted in the following embodiments as long as the evaluation is avoided. in the first seconds after switching on the lamp. Additionally, those with ordinary skill will appreciate that, in cases where the signal strength of the voltage or current is high enough to obtain satisfactory control, sampling of the voltage or current alone may be used instead of the conductance. . The method and the ballast according to the invention employ several control phases, each of which depends on the sampling of the conductance in a plurality of frequencies and the calculation of a selected deviation in each frequency. Therefore, a favorable technique of sampling calculation and conductance deviation will be described before analyzing the control phases. Calculation of Conductivity (Sampling) The task is to calculate the conductance g (t) from the rapidly changing voltage and current. For a HID lamp that operates with a typical high frequency ballast with distorted sine waves, is the lamp voltage an alternating periodic function V (t) = V (t + p) with a period p = 2p1? . In case the frequency of the current ft and therefore the frequency of the voltage are above the audible range, that is, above 20 kHz, then p <; ? 50μs. By definition, the current is given by l (t) • = g (t) V (t). The conductance or conductance g (t) is only a positive function of slow change of time, almost constant during a properly chosen sample time T. At least, two samples are taken in a period p (to avoid detecting only small signals), and for a certain time T > p (of the order of the s) a total of N samples is taken. By definition, we will consider the values of G: m * - -. (4) where in general < f > denotes by definition the averaged value of the observed values f¿, i = l, 2 ... N: K 15 < f > - (5) N í The absolute values of the sums of the calculations of the conductivity are taken to avoid a sum equal to zero, and also to obtain a maximum value of the sum (an alternative is to take G = [< I * > / < V2 >] v '). The substitution of the definitions and the equation for the current leads to: while as the values g are assumed, they are almost constant during the sample time T. When the current and the voltage are sampled at the same time the two sums on V (are canceled.) This can be achieved very simply in practice with a board It is not necessary to sample a whole number of periods p, because the sums are always greater than the same samples V, to detect acoustic resonances, the general scheme is to operate the gas discharge lamp to a plurality of frequencies, calculate in each frequency the conductance a plurality of times taking a plurality of simultaneous sample of the voltage and current of the lamp, and for each frequency calculate the standard deviation of the conductances taken in that frequency. following equations: s, ("i '1,2 ...- V-; k - U ...- V, (7) N * (8) Vsy. | / Ys,., Vj l? L take each frequency f ^ a number of NIV x V0 samples (i = l a i = NIV; j = l a j-NG), a standard deviation sk is determined. All samples used to calculate sk are taken at the frequency fk. Then the next frequency fk is selected and repeated for sampling and calculation. The frequency with the smallest standard deviation is the frequency with the smallest arc deviation caused by acoustic resonance and the frequency (frequencies) with the highest standard deviation (the largest standard deviations) are the frequencies with the largest arc deviation. Figure 2 illustrates a scheme that will use the equations. A symmetric frequency sweep is performed on a frequency sequence (k = 0, 1, 2 ... 20). At each frequency fk, a sequence (j = 1, 2 ... NG) of conductivities G is determined by measuring for each conductivity G simultaneously selected samples 1 ^ and V ^. This is illustrated in Figure 2 for the frequency of k = 2. The frequency sweep is initiated at a central frequency fc, decreases (portion A) at a minimum frequency f, increases (portion B) at a maximum frequency f b, and then decreases again (portion C) at the center frequency f c. Portion B has the same set of frequencies as portions A and C combined. It is desirable to sample the conductance with the frequency increasing as well as decreasing due to the hysteresis in the bow deflection response. By performing the frequency sweep in this manner, the effects of hysteresis are eliminated. The frequency sweep section is the difference between fa and fb. The scanning speed is the speed at which the frequencies fk are modified. In equations 7 and 8, the indices i, j and k used in the current and voltage indicate that they are three-dimensional networks at first. For the implementation in a lamp ballast it will be advantageous to avoid having to provide memory to store these values. For the general case, the standard deviation is defined by: ,2 . (9) N l However, after taking the square root of the term after the addition in equation 9 and replacing equation (9a), where x «-l¿ £ ^ *, - (° a) Equation 9 becomes: Using equation 10 (where x¡ = Gk), the calculation of the standard deviation sk of i "I" ° 2 - ~ S * l - (~ S *? (10) Gk conductances sampled at a frequency of fk can be performed in software without having to store all currents and voltages for each sample of the network. Material flow diagram to achieve this while performing the frequency sweep shown in Figure 2 is shown in Figure 3. When using the conductance, the sampling scheme to determine standard deviations is greatly simplified in comparison with the use of the standard deviation in voltage or current alone This is illustrated in Figure 4, which shows a voltage waveform V and current I with two sample periods i and i + 1, representing the arrows the location and the amplitude of the samples taken In Figure 4, the samples of period i are taken at different points with respect to the phase of those of period i + 1. Therefore,? ¡lj or S¡V¡ of period i is not the same as in period i + 1, of t in the way that there would be an error in the calculation of normal deviation of the current or voltage alone. If only voltage V or current I is sampled, sampling must be triggered at the same instant in the voltage or voltage waveform so that the standard deviation is exact. This would entail additional detection and triggering devices in the balasta controller that used this technique. This shot would also introduce an error in the calculation of the standard deviation. When using the ? Conductance, voltage and voltage should only be detected simultaneously. The values of the current and voltage detected simultaneously are normalized in the I / V ratio that defines the conductance in such a way that the instant in which the samples are taken with respect to the waveform detected is irrelevant. This simplifies the sampling scheme and its implementation in the balasta. The deviation of the arc between the electrodes due to a force, induced for example by an acoustic resonance, can be described by a differential equation of second order in time. A typical time constant of ™ = 50 ms describes the time to achieve some deviation. This time was determined by deviating the arc of a 100 W metal halide lamp with an external electromagnetic force of known value and duration from a coil placed around the lamp. The equivalent force on the arc caused by an acoustic resonance can be described by F = F0sin (27r? Ft). With this force, the greatest response is observed at a frequency of? F = 3Hz. At higher force frequencies, the frequency response decreases by 40dB / decade. Depending on the strength of the force on the arch, the deviation will be large or small. In ballasts that use sinusoidal current to activate the lamp, some resonances can deflect the arc against the wall of the arc tube. Said resonance is defined herein as a strong resonance. All other resonances not capable of deflecting the arch against the wall are defined as weak resonances. In order to provide a control scheme to prevent a strong resonance from driving the arc towards the wall, the control must respond at a speed much greater than 50ms and apply a different operating frequency at which strong resonances do not occur in a much less than 50ms. When measuring the fluctuation due to weak resonances, a time of at least of the order of 150ms must be used. Strong resonances are important as the deflection of the arch towards the wall can cause the lamp to explode. Weak resonances are important in that they cause fluctuation of the arc that is very annoying for humans. Characteristics of the Lamp It is important to understand the starting of the lamp, that is, the first minutes after lighting the lamp, to create a control scheme for the lamp. This is illustrated in Table I below.

TABLE I Prior to ignition, the pressure p in a typical 100 Watt metal halide lamp is equal to approximately 0.3 bar. During start-up, the operating pressure increases to the steady-state operating pressure of approximately 15-20 bar, typically within 120-200 seconds after ignition. The damping in filler gas, that is, the resistance to inducing an acoustic resonance, and therefore the change in the position of the arc, is inversely proportional to the pressure, so that the damping decreases by a factor of approximately 50. during startup. Consequently, the intensity of the acoustic resonance and the arc instabilities increase. During the first 30 seconds after switching on, acoustic resonance does not occur because it remains essentially a low-pressure discharge lamp. Therefore, detection of resonance and fluctuation through conductance measurement can not feasibly occur in this period. During the post-additional start, between t0 to «30 s and te8tacJonar) or to« 120s, the gas content, and therefore resonant frequencies, which change rapidly due to the higher temperature of the TL lamp, the filling pressure of the lamp, the speed of sound (a) and the entry of metal halides in the flow of the arc. The location of the resonant frequencies does not. it will stabilize until the lamp has reached the steady state at testac. { onaHo. If you want a ballast that could operate HID lamps through a wide range of < *, nominal wattages (20-400W, for example), the detection of deviation in the conductance could be used to locate, for any HID lamp, an operating frequency that would be free from visible fluctuation induced by acoustic resonance. In practice, said ballast is not plastic in commercial terms because the voltage and the current that passes through the electronic devices 0 _. determines the cost; in such a way that each lamp is put into operation by a ballast designed for this maximum power. It is enough, and it would be a great improvement in the technique, if a high frequency ballast could operate lamps that had arc tubes in a similar way (for example, from different manufacturers) and a reduced range of wattage. A favorable voltage incorporation is based on the previous selection of a relatively narrow operating frequency window in which strong resonances (which cause the arc to deviate against the arc tube wall) do not occur for the intended range of lamps desired. After lighting the lamp, the method and balasta that implement this method work within the previously selected window to concentrate on a frequency in which no visible fluctuation occurs for weak resonances for the specific lamp that is being controlled. Additionally, because the environmental conditions or other factors can modify the acoustic resonance nodes of the lamp, the method and the ballast continue the detection and monitoring along the t * 5 operation of the lamp to prevent the lamp from wandering. in acoustic resonance. The above techniques will explain in general terms with reference to Figures 5-7. These figures illustrate the detection of strong resonances and resonances weak during both start-up and during the stationary regime. After this, a specific control algorithm that performs a favorable incorporation using these techniques will be analyzed. Strong Resonance Detection: Pre-selection of Windows The frequencies f (mn at which acoustic resonance occurs are a plurality of nodes provided by Equation 3. These frequencies can be confirmed through experiments by varying the operating frequency, or central, fc over a wide range fa a fb (see Figure 5a) in a plurality of frequencies fk and calculating the standard deviations of the conductances by taking a plurality of simultaneous samples of the voltage and current of the lamp at each frequency fk, as discussed above with respect to Equations 7, 8, 10 and Figure 2. Starting of this exploration, the frequencies at which strong resonances occur will become apparent because they will have the highest standard deviation, and can also be observed visually. These frequencies are shown as points in Figure 5b. In order to avoid damage to the arc tube, strong resonances must be detected by an iterative procedure, as illustrated in Figure 5b. Due to the damping, and the second-order deflection response of the arc, the arc deviation will be the lowest at high scanning speeds and greater at low scanning speeds. If a low scanning speed is initially used, there is a danger that the arc will deviate against the wall of the arc tube for a sufficient time to damage it. Therefore, the first frequency scan should use a high scanning speed of approximately 1000 kHz / s and will reveal frequency windows in which the strongest resonances are observed without damaging the arc tube. These frequencies should be avoided in further explorations. The frequency sweeps must be repeated at successively lower scanning speeds of, for example, 100 kHz / and 10 kHz / s. This will reveal the frequencies at which the strongest resonances occur. In Figure 5b a line indicates the frequency range for each scanning speed of the frequency sweep over which the center frequency fe varies, all with its respective sweep speed. Frequency windows, having a section of a few thousand hertz, will also be observed, in which strong resonances do not occur and are illustrated in Figure 5b, on the line called R, as solid lines. These windows are the frequencies at which the lamp should be operated to avoid strong resonances. This procedure is used either in the ballast of the lamp, or by the designer of the ballast when previously selecting a window of wide frequency to the lamps of which it is intended to control with the ballast. If done by the designer of the ballast, this procedure can be repeated for each lamp that is operated with the ballast. Then a common window can be selected in which each of the lamps that it is desired to operate by the same lamp ballast without the presence of strong resonances. In Figure 5c, a graph of the permissible operating windows for several metal halide lamps of 100 W A-E, each from a different manufacturer, is shown as solid lines. The common allowable operating window of this graph is at 20-25 kHz. It is not necessary that a previously selected window is completely free of strong resonances. The danger associated with strong resonances can be avoided by using a larger sample or the technique analyzed under the heading "wall deviations", for example. However, the previously selected range must have some, even if they are quite narrow, stable regions. Weak Resonance Detection Once a strong resonance free sale is selected, the lamps should be operated only within this previously selected window. Again, the general scheme is to vary the frequencies within the selected window to detect those frequencies that do not occur at weak resonances that would cause visible fluctuation. Preferably, the frequency sweep and sampling technique shown in Figure 2 are used to vary the center frequency between f and fr as shown in Figure 6 (a), which represents, respectively, the limits of the previously selected window of Figure 5c, for example f = 20 kHz and fr = 25 kHz in this specific case of metal halide lamps of 100W. The objective is to find the global minimum of the standard deviation of the conductivity Gp, inside the window f a fr, as shown in Figure 6 (b). Several control options are possible. In a first option illustrated in Figure 7a, the lamp is turned on at time = t ,, being operated at a central frequency halfway between f and fr and allowed to perform the heating at steady state, starting at testßc! onar-0, where the location and intensities of the resonant frequencies are relatively stable. The frequency sweep and the sampling technique of Figure 2 is then carried out starting at tßßtac, "" ,, ,, ", and the center frequency is adjusted to the frequency where the standard deviation is the smallest, which corresponds to the global minimum M shown in Figure 6b. Because the frequency does not vary during the start-up period, there is only a small chance that the selected center frequency fc is at a resonant frequency such that visible jitter will not occur during start-up. However, when the center frequency varies after reaching the steady state, visible fluctuation will occur during full light emission. Although the fluctuation could only last a few iterations of the frequency sweep, a period of one minute or the like, may nevertheless be disturbing to the user of the lamp. Another option (illustrated in Figure 7b) has to start the frequency sweep and the sampling technique during start-up, for example at a delay time t0 of about 30 seconds after turning on, again over the entire width of the window previously selected, fa fr. This has the disadvantage that visible fluctuation will occur both during start-up and when steady-state operation has begun. During start-up, the resonant frequencies and therefore the global minimum are all changing rapidly as the lamp heats up rapidly. When a large-section frequency sweep is used, the global minimum changes faster than the time it takes to drive each frequency sweep and sample iteration. The result is that with large sections, the global minimum can not be set until the stationary regime starts at testacjonar5o. This option, although still useful, is less attractive than the first option because the visible fluctuation will occur both during start-up and the start of steady-state operation. A third and more favorable option (Figure 7c) is to start the frequency sweep and the sampling technique at t0 (see point "A"), but with a first section much smaller than the width of the window f and fr. In comparison with a typical section of several kHz for the section f a fr in the previous points, the first section of f a fr in this third _ option is approximately 0.1 kHz. It is likely that the instantaneous global minimum is not found during the first iterations. However, each iteration will result in the site of the closest center frequency in the instantaneous frequency in which the global minimum or a local minimum currently exists. This will probably be found at either end of the narrow section of the frequency sweep. After several iterations (see point "B"), the frequency sweep and the sampling loop are set to the instantaneous global minimum. After each successive iteration, the new central frequency closely matches the instantaneous global minimum. Therefore, this procedure quickly locates the instantaneous global minimum after several iterations, and tracks the instantaneous global minimum to the steady state. This has the advantage that fluctuation can occur only during the first iterations near the start of the start in cases where the lamp has a low light emission. Thereafter, the procedure will have already found the instantaneous global minimum, and because the section is small and close to the minimum, fluctuation will not occur near the end of the start and during the steady state. This control option is the most favored because the detection of the resonant frequency during startup is almost imperceptible to the human eye. Stationary Rate Detection Once the overall minimum and the corresponding preferred operating frequency are determined, it is desirable to continue the verification of whether the desired operating frequency should be adjusted during the steady state. If the temperature of the lamp changes due to changes in the ambient temperature, the speed of the sound will change with a corresponding change in the resonant frequencies. Therefore, such changes must be detected and the frequency of operation continuously adjusted. This is achieved by dynamically repeating the frequency sweep and the sampling loop previously described with a second section fu a fv (Figure 8). The second section is preferably smaller than or equal to the first section used during startup. The second section should be selected in such a way that the lamp does not fluctuate visibly during this steady state detection process, and typically around 0.1 kHz. Operating Routine Figure 9 shows a material flow diagram for a routine to operate an HID lamp in accordance with the third control option (Figure 7c). The algorithm finds the minimum fluctuation and the corresponding operating frequency fmJn in a previously selected power frequency sale (fco ± Section0 / 2). The range of power frequencies is previously selected based on experimental investigations as described above or through the initial step of the lamp controller, to discover a window of resonance-free and strong operating frequencies for the lamps that are desired to be controlled by the lamp driver. At the start of the program, the center frequency fc is initialized at a design frequency of balasta fc0. Then a high voltage is applied to the lamp to turn it on. After a predetermined time t, the current of the lamp I is sampled to determine whether the lamp has in fact been turned on. The lamp current is below a value I0, the lamp has not been turned on and the ignition is attempted again. When the current I is greater than I ", the lamp has been turned on and the time t and a counter-variable, Nfl, are initialized to O. Then the lamp is allowed to start at the operating frequency fc0 until t is greater than one. time t0 previously selected near the start of the start. The time t0 is selected to be long enough such that the pressure and temperature of the lamp are high enough for acoustic resonance to occur. From time t0, a plurality of frequency sweeps of a previously selected first section and a scanning speed are performed through the Open Loop slave routine (shown in Figure 3) while the lamp is still in the start stage, which it is prior to the moment in which the lamp has reached the steady state in the testacJonar1o time, for example at 120s. The inputs of the Open Loop program in the center frequency fc, the frequency section "Section", the number NF of the different frequencies sampled, the number NG of the conductivity samples taken in each frequency other than NF, the number NIV of samples I and V taken in each sample G, k, and the time ífF to obtain the NG samples in each frequency. These variables are illustrated in Figure 2. The outputs are the smallest standard deviation sm (n of the conductance and the corresponding frequency fmjn.) During the start-up period, the frequencies at which acoustic resonance and fluctuation do not occur are changing according to the gas pressure and the temperature change During this control stage i, the center frequency is continuously updated in the direction in which the resonance-free frequency is moving, therefore, once the lamp has reached the speed stationary yt >; testac) onar < 0, the central frequency fc usually has already reached the optimum value where weak resonances and visible fluctuation do not occur. As long as time t is less than testationary, the routine remains in branch "A" of Figure 9. Once the test frequency is reached, the central operating frequency fc is adjusted to f-, ¡n, which as illustrated in Figure 6b, is the global minimum of the steady state. The routine follows branch "B" if the standard deviation smin to f ^ is less than sfl, which has a value corresponding to that in which the visible fluctuation occurs. If sm- is greater than sfl, the visible fluctuation in the lamp in fmin is still occurring and the global minimum has not been found. The frequency sweep is carried out again through the branch "C", but with a greater section equal to the width of the window fa fr (Section = Section0) and a greater number of frequencies (NF = 100) to find the minimum global in which visible fluctuation does not occur. If, the minimum in which the visible fluctuation does not occur is found, the program enters a steady-state detection mode, with the narrowest section, through the branch "B". If visible fluctuation occurs after the first iteration, the global scan is repeated until a free fluctuation window is found. If the number of iterations Nfl of branch C exceeds a pre-set number, for example 10, the lamp goes out. The following are typical parameters for operating the above routine for a 100 W metal halide lamp: NF = 20, NG = 20, NIV = 20, «Stf = 50 ms, fc0 = 23.5 kHz, Section0 = 3 kHz, Section = 0.1 kHz, t, = - 10 ms, I0 = 0.1 A, t0 = 30s, G. It should be noted that the value sfl is a conservative threshold value for the standard deviation of the conductance at which the fluctuation will be observed by the Humans. This can be estimated (see reference W.F. Schreiber, Fundamentals of Electronic Imaging Systems, Springer-Verlag, Berlin (1991) pp. 14-16) and can be confirmed by adjusting its value until no visible fluctuation occurs. The advantage of using the conductance is that the signal ratio with respect to noise and sensitivity are high enough so that the deflection response of the arc can be evaluated at levels below that which can be WO 9ß / 20S7ß PCT / IB95 / 01106 detected by the human eye. Therefore, during the steady state, fc can be continuously tracked by sweeping the frequency and induced arc deviations that result can be found at levels that are commensurable but not visible to the human eye. Figures 10a, 10b are graphs of the minimum frequency fmin and omn, respectively, compared to the NF for a 100W metal halide lamp that is operated in accordance with the material flow diagram of Figure 9 (which it works in accordance with the control option of Figure 7 (c)) with the values of the parameters listed above. After only about 6 iterations, the omfn was below 0.005, the level at which the fluctuation is visible. From then on, sm} n remained very low at an approximate level of visible fluctuation of 1/10. Visible fluctuation occurred only during the first iterations, which occurred during the early part of the start-up when the lamp was still at a low light emission level. Once the lamp reached the steady state, branch B (Figure 9) continued to monitor and adjust fc without visible fluctuation. Alternative Selection Criteria In the previous incorporations, the frequency was selected as those with the least standard deviation. Other criteria can also be applied to select a new center frequency after each iteration of the Open Loop routine. For example, a wide range that surrounds the operating frequency is preferred, because the operation in a wide minimum would be more stable than in a very narrow minimum. In Figure 11, the global minimum is denoted as "AA" and has a narrower range than the preferred position denoted by "BB". In both sites, s is less than Ofi t And no visible fluctuation occurs. Because the deviation is sufficiently low at the widest minimum such that no jitter occurs, the widest minimum can be selected as the center frequency. This can be selected by adding the results of sk of a number 2n of successive frequencies fk (from k-n to k + n) around each minimum smaller than sk and obtaining a new series of s, unifying the data. The most standardized data are shown in Figure 11 (b). The new center frequency is equal to the frequency allowed with the smallest standard deviation s. The selection of the new center frequency is therefore based on the standard deviation in the conductance, and the overall minimum detection. Wall Deviations It is desirable to turn off the lamp when the discharge arc touches and remains in the arc tube wall. Therefore, for the relevant lamp type, a s ^ must be determined for the s_ that corresponds to a deviation that is large enough to cause the arc to touch the wall. During start-up, a wide sweep is performed if s >; smá ?. If in the sweep of frequency holds for the sm) n that asmin < smáx 'e- * arc will not be on the wall and the minimum can be selected. However, if < xsmfn, then the arc will remain on the wall and the lamp should go off. The constant a is selected to provide the desired statistical reliability, and is typically found between two (2) and four (4). Other Arc Inestabilities Arc deviations may occur for reasons other than acoustic resonance, for example, randomly, due to arc jumps and electrode or sodium flashes. Figures 12a and 12b show representative conductivity responses for these two events. In Figure 12a and 12b the curve does not indicate the conductivity values when random events do not occur. Curves 2 indicate the conductivity values when an arc jump occurs (Figure 12a) respectively when a sodium flash occurs (Figure 12b). The presence of such random events will cause a false change in the central operating frequency if the deviations caused by this are not ignored. These deviations from random events can be distinguished from the acoustic resonance deviations by repeating each measurement and checking whether the same result is obtained within a desired statistical confidence level. If the results are not within the desired level of reliability, the measurements should be repeated again before adjusting the center frequency. For example, a 5-frequency sweep can be performed twice, each time calculating a standard deviation of all the conductivity samples taken through the entire frequency section. If the s of the second sweep (s2) is within a desired reliability level of the s of the first sweep (sx), that is, i9. or "= ßo" 1, then a random event has not occurred and a new operating frequency can be selected. If s2 > ßs1, then a random event has occurred and the frequency of operation should not be modified, ß is typically chosen between two (2) and four (4). Those with skill Those skilled in the art will appreciate that many other tests can be used, for example comparing the average, the minimum, or "i the maximum of the standard deviation of the conductivity samples measured at each frequency between the two iterations 20 Lamp Bulb Figure 13 is a schematic block diagram of an HID lamp ballast, or controller, in accordance with the invention to operate a HID lamp at high frequency and to detect and prevent operation, to frequencies that cause acoustic resonance / arc instabilities. The following embodiment illustrates a certain balasta topology and reveals specific numerical parameters selected for a specific application, in particular 100W metal halide lamps. This embodiment serves as an illustration of one of the many possible implementations of the ballast using the method described above for the detection of arc instability and for the overvoltage of the power control of the lamp as previously analyzed. Accordingly, those of ordinary skill in the art will appreciate that the following embodiment is only illustrative, and not limiting, and the disclosed operating principles can be used in many different balasta topologies, with different operating parameters. The ballast includes a DC source 10, a boost converter 20, a rectangular signal inverter 30 of Direct Current - Alternating Current and a lighter 40. Elements 10-40 turn on the lamp and provide a high frequency, substantially current enusoidal CA to the lamp 50. After the lamp is turned on, the Control C controls the operating frequency of the inverter 30 to prevent arc instability / acoustic resonance of the HID lamp in accordance with the method described above. The Control D controls the reinforcement circuit 20 to limit the overvoltage during the lamp's lighting phase (because the lamp only presents a small charge during this time) to prevent excess voltage from being applied to the lamp and to the lamp. elements of the circuit. Control D also controls the fuzzy circuit 20 to maintain constant power to the lamp despite the changes made by Control C in the operating frequency of the inverter to avoid acoustic resonance. Control A operates the boost converter 20 at the boost frequency determined by the Control D while Control B operates the inverter 30 at the inverting frequency determined by Control C. The DC source 10 includes a pair of terminals input 1, 2 to receive a normal AC power line voltage of 110-120V. A rectifier consisting of diodes Dl-D4 provides a full-wave rectified DC voltage of approximately 160 V through the rails RL1, RL2. The CD source 10 may also include an EMI filter 5 to isolate the power lines from the interference generated by the lamp controller. The boost converter 20 reinforces and controls the DC voltage through the rails RL1, RL2 at a level such that a selected power is provided to the HID lamp 50 through the inverter circuit 30. The boost converter typically reinforces the voltage from 160V to approximately 380V. The boost converter also provides power factor correction. The converter 20 includes an inductor Ll having one end connected to the cathodes of the diodes DI and D2 and its other end connected to the anode of the diode D5. A switch Ql is connected between the CD tracks RL2 and RL1 at a junction between the inductor Ll and the diode D5. Switch Ql is a mosfet and includes a body diode BD1 and a parasitic capacitance called Cdsl. The control grid of the switch Ql is connected to the control A, which 0 provides a periodic voltage signal for controlling the switching frequency and the duty cycle of the switch Ql in a manner that was described in more detail. The duty cycle and the switching frequency control the current flow through the inductor Ll in such a way that in conjunction with the capacitors C5 and C5 of the inverter circuit 30 the voltage on the CD rails RLl, RL2 through capacitors C4 and C5 remains constant at the desired level. Capacitors C4 and C5 act as an energy storage element to provide constant power to the lamp, or even when the line voltage crosses zero. The inverter circuit 30 is a semiconductor DC-AC inverter powered by voltage with switches of the series Q2, Q3 connected through the lanes CD RL1, RL2. Switches Q2, Q3 are mosfets. The source of the switches • *! 5 Q2 is connected to the rail RL1, the power socket of the switch Q2 is connected to the source of the switch Q3 and the power socket of the switch Q3 is connected to the rail RL2. The capacitor Cds2 and the diode BD2 are the parasitic capacitance and the body diode, respectively, of the switch Q2. The capacitor Cds3 and the diode BD3 are similarly the parasitic capacitance and the body diode of the mosfet switch. The control bars of switches Q2 and Q3 are connected to control B, which will be described in more detail: The output of the half-bridge inverter, which appears through points MI, M2, is a rectangular wave signal generally high frequency as will be familiar to those skilled in the art.A LCC network of capacitors C6, C7 and inductor L2 are series connected between a midpoint MI between switches Ql and Q2 and a midpoint M2 between half-bridge capacitors C4 and C5.The HID lamp 50 is connected in parallel with the capacitor C6.The LCC network provides a function that shapes the wave and limits the current, or balancing, to provide a substantially sinusoidal lamp current to the HID 50 lamp from The inverter output is present through the midpoints M1-M2 The LCC network also functions as a lighter to turn on the lamp at the time of the initial application of power to the controller. LCC is set to a third harmonic of the inverter's initial operating frequency to provide a high start voltage of approximately 2500 V selected for this specific application. Starting the lamp in the third harmonic has the advantage of reducing the initial ignition current taken from the boost converter compared to a conventional start in the first harmonic. When the inverter starts to work, an ignition voltage is caused by the resonance of the LCC network and the third harmonic of the inverter output at the initial operating frequency. After the lamp starts, the impedance of the lamp is much smaller than that of capacitor C7, so that the waveform and current limitation is controlled mainly by the LC network of C6 and L2. Therefore, the lighter uses the variation of the impedance of the lamp (from approximately 1M before the ignition to approximately 100 in steady state) to change the gain of the circuit including the LCC network and the lamp of a suitable gain to provide a voltage Ignition at a lower gain suitable to operate the lamp. Instead of the LCC network shown, other lighters can be used, for example the well-known pulse lighters or other resonant lighters, for example an LC or an LLCC network that employ the ignition in the first or third harmonic. Additionally, in the LCC network shown, an active frequency change can be used whereby the operating frequency of the inverter is adjusted to an initial frequency (usually higher than the steady-state frequency) to cause the ignition voltage and then would adjust to a second different frequency for lamp operation after power-up. Control B controls the switching frequency and pulse width of switches Q2 and Q3 in a well-known manner to provide substantially rectangular wave AC inverter voltage across the midpoints MI and M2 at frequencies within a range of suitable frequencies to operate the lamps you want to control using the ballast. In this instance, the range is between approximately 20 kHz and 25 kHz, which is the previously selected operating window, to avoid strong acoustic resonances / arc instabilities for the metal halide lamps of 100 W A-E shown in Figure 5c. In particular, Control B is responsible for a control signal of frequency from Control C of Arc Instability to operate the half-bridge at the frequency designated by Control C. The circuit and operation of Control D and Control A will now be described in greater detail with reference to Figures 14 and 15. Control D includes circuits for detecting the lamp voltage and lamp current to the VL and IL sites shown in Figure 13. The lamp voltage is detected in the voltage sensing circuit 60 as shown in Figure 14 which includes a voltage divider including resistors Rll and R12 and the capacitor Cll. The function of capacitor Cll is to isolate the DC component between the detected lamp voltage and the VL point and the ground. The diode Dll has an anode connected between the resistors Rll and R12 and its cathode connected to a side of the resistor R13. The other side of resistor R13 is connected to ground. The capacitor C12 and the zener diode D12 is connected to the cathode of the Dll diode. The diodes Dll and D12 form a half-wave rectifier to provide a DC voltage VL in the resistor R13 representative of the detected lamp voltage VL. The current of the lamp is detected through the current detection circuit 70 which includes the current transformer T. Connected in parallel with the transformer T is the resistor R13, the capacitor C13 and the resistor R15. The anode of the diode D13 is connected to one side of the current transformer T while its cathode is connected to the resistor R15. The output of the current detector circuit 70 is a DC voltage V, which is linearly proportional to the lamp current ILa? Rp. The Control D additionally includes a multiplier 61 that multiplies the signal VL representing the voltage of the lamp representing the signal V, the current of the lamp to obtain the signal V that represents the power of the lamp. The diode D14 and the resistor R19 are series connected between the output of the multiplier 61 and the inverting input of the error amplifier 65. The non-inverting input of the error amplifier 65 receives a reference signal V f which indicates the desired operating power for the HID 50 or Vref lamp that indicates the upper limit of the overvoltage. The resistor R20 is connected between a junction between the resistor R19 and the inverting input of the error amplifier 65, and the ground. The resistor R21 is connected between the inverting input and the output of the error amplifier 65. The comparator 67 receives the output of the error amplifier at its non-inverting input and a sawtooth wave output of the oscillator 63 at its inverting input. The comparator 71 compares the signal V, received at its inverting input with a signal Vr. { received in his positive input. The diode D15 has its cathode connected to the output of the comparator 71 and its anode connected to the anode of the diode D6. The cathode of the diode D16 is connected to one side of the resistor R18, the other side of which is connected to the inverting input of the error amplifier 65. A voltage divider including the resistors R16, R17, is connected between the anodes of the diodes D15 and D16. One end of resistor R16 is connected to the DC rail, or bus, RL1 at the site illustrated in Figure 13. A voltage V3 representing the overvoltage is thereafter present at the midpoint between the resistors R16, R17. During the lighting of the lamp 50, the Control Circuit D operates in an overvoltage control mode. During this time, the lamp is not yet lit and has a high impedance. As a result, the load on the reinforcement converter 20 is light and the tension in the CD rails RL1, RL2 will increase significantly without other measures being taken. After the ignition, the impedance of the lamp decreases, and the current taken by the lamp increases, until the lamp reaches the steady state. Limits on overvoltage during ignition and steady state are required to prevent catastrophic operating conditions, including those close to the end of lamp life. Control D detects the overvoltage Vbus and supplies it to the control A which adjusts the duration of the impulse of the reinforcing switch Ql to maintain the overvoltage at a predetermined voltage during the ignition phase. Because the lamp has a high impedance, the current passing through the lamp has a small value which is detected by the current transformer T and then rectified through the half-wave rectifier D13. The voltage D13 V ,, which is linearly proportional to the current of the lamp, is almost zero because there is little current from the lamp during ignition. In addition, the output of the multiplier 61 is smaller than the reference voltage Vref which results in a blocking of the diode D14, deactivating the power control loop during this period. The voltage V, is compared with the voltage Vr | through comparator 71. When V is less than VH, as is the case during ignition, the output of comparator 71 is a high voltage Va. As a consequence, the diode D15 is derived in reverse and the voltage control loop is activated. Therefore, D16 turns on or off depending on the value of the overvoltage. When the voltage control loop circuit initiates the ignition of a lamp, when the lamp has a very small charge, the overvoltage increases rapidly and is detected through the voltage divider of the resistors R16 and R17. While the detected voltage V3 is less than the reference voltage Vref at start-up, the D16 diode remains off. The output of the error amplifier, in the initial circuit condition before driving the diode D16, is given by «/ The voltage Vo is compared to the sawtooth wave generated by the oscillator 63 to obtain the pulse duration control for the boost switch Ql, thus controlling the energy stored in the electrolytic capacitors C4, C5 (Figure 13). By design, the working ratio of the reinforcement stage has a maximum value adjusted to approximately 0.48. Once the detected voltage of V3 reaches the reference Vrβ, the diode D16 starts to drive. The detected overvoltage is fed to the inverting terminal of the error amplifier 65 through the resistor R8. The output voltage V0 of the error amplifier 65 is given by where V0 is the forward voltage drop across diode D16. The voltage V0 and the output voltage of the oscillator 63 are fed to the comparator 67 to obtain control of the pulse duration of the reinforcing switch Ql. Figure 15 illustrates the voltage V0, the sawtooth wave of the oscillator 63, and the output wave of the comparator 67, which is the voltage of Vgs1 of the grid source controlling the reinforcing switch Ql. The smaller the voltage V0, the smaller the pulse duration of the control signal Vgsl which controls the switch Ql. As a result, the pulse duration will decrease when the detected overvoltage V3 increases. Based on the operation of the boost converter, the overvoltage will be reduced and will remain within a preferred range of, for example, 450V. After the HID lamp 50 is turned on, the control circuit D is switched to a power control mode to control the power to the lamp. Without additional measures, the power applied to the lamp will change when Control C changes the operating frequency of the lamp to control the arc instability due to the gain of the resonant LCC network or any other ignition topology, and therefore the power supplied to the lamp, varies with the inverter output frequency. Generally speaking, in the power control mode, the current and voltage of the lamp are detected and multiplied to obtain the total power of the lamp. The power of the lamp is compared to a power signal of reference in order to change the work ratio of the "1 reinforcement switch Ql to regulate the overvoltage through RL1, RL2 which results in an adjustment of the power supplied to the lamp. of voltage V, is greater than the reference voltage v "t. The comparator 71 emits a low voltage V2 which leads to the conduction of the diode D15. As a result, the overvoltage Vb (ls is set, deactivating the overvoltage control loop, and the power control loop is activated. of tension of \ l \ nmpprn VL and corrected Jn Iftm nrrt V, PP feed multiplier 61 to obtain a power signal V of the lamps, which is compared with the reference power V f to control the duration of impulse of the reinforcement switch. If you change the switching frequency of the half-bridge inverter, the power supplied from the lamp will increase or decrease because the value of the voltage gain for the lighter is different for different inverter operating frequencies. ^ A decrease in the power transferred to the lamp will be assumed as an example to illustrate the operation of the power control loop. The power of the detected lamp V from the multiplier 61 is supplied to the inverting terminal of the error amplifier 65 through the resistor R19 and the diode D14 and then compared with the reference power V f. In the power control mode, the output voltage V0 of the error amplifier 65 is given by Therefore, V0 will increase when the detected power of V decreases. When the detected lamp power V decreases, V0 increases and the pulse duration of V increases.

When the pulse duration increases, the boost switch remains on for a longer period of time within each cycle, thereby increasing the overvoltage through the rails RL1, RL2. As a result of the increase of the overvoltage buS ', the power supplied to the lamp increases. This negative feedback maintains the power of the lamp equal to the set reference power i by the signal V f. On the other hand, if the power of the lamp increases due to a change in the half-bridge switching frequency by Control C, the detected power Vp becomes greater, leading to a decrease in the pulse duration of the driving signal V, for the commutation switch Ql. The reinforcement converter will then be smaller than the overvoltage of the rails RL1, RL2 until the detected power equals the reference power value Vref. Certain HID lamps tend to use more power near the end of their life than their nominal wattage, which can lead to catastrophic lamp failure * - if this higher power is supplied through the ballast. Because the power control mode limits the power supplied to the lamp, it serves to prevent catastrophic failures. Rather, the lamp will simply stop functioning when the power corresponding to the signal V f is insufficient to equal the greater power required by the lamp at the end of its life.

From the previous analysis, it can be seen that the D control has two control modes, which are the voltage and power control modes. An objective of the voltage control mode is to set the overvoltage during the lighting of the lamp and before operation in mode. The overvoltage control can also be used to prevent the catastrophic life of the lamp, when the tendency of the ballast is to provide excessive power to the lamp due to the impedance variation with the life of the lamp. the lamp. The objective of the power control mode is to supply a constant power to the lamp even with changes in the switching frequency and line voltage. Figure 16 is a circuit diagram of the 15 controls A and D of Figure 13. The components are the same as those discussed in Figure 14 and bear the same reference numerals. The multiplier 61 of Figure 14 is incorporated in an IC of 14 connectors (model AD534 available from Analog Device Corp.) VL is the input in connector 1 and 20 V, it is the input in connector 6. Oscillator 63, the error amplifier 65 and comparator 67 are incorporated into a 16-connector high-speed PWM controller 80 (model UC 3825 available from Unitrode Corp.). The UC 3825 is optimized * »For power supplies of high-frequency switched mode 25 frequency and directly controls the switching of the mosfet Ql. Source SQ1 of switch Ql is connected to connectors 9, 10 and 12 and grounded. The grid GQ1 of the switch Ql is connected to the connector 14 through the resistor R22. In Figures 16 and 18, connector connections not specifically described herein are shown solely for the sake of completeness and are not necessarily for the understanding of the invention. The manufacturer of the IC specifies these connections. As shown in Figure 17, Control C of the Arc Instability Control includes two A / D converters 90, 95 for converting the detected lamp voltage VL and the lamp current V. to a respective digital signal. The signals VL and V are derived from the circuits 60, 70 of Figure 14. The control C further includes a microprocessor 100 which implements the program of Figure 9 in the software to control the operating frequency of the half-bridge inverter. 30. The microprocessor 100 receives the voltage and current of the detected lamp from the dual A / D converters and outputs a digital signal that is converted into analog signal f5 by the A / D converter 105. The signal fg represents the frequency output Snapshot of the software program when performing any of the operating modes in accordance with the previously described method. When implementing the program of Figure 9 in software in the microprocessor 100, Control C includes mechanisms for performing each of the methods described hereinabove with respect to the detection of frequencies in which acoustic resonance / arc instability occurs and to change the frequency of the inverter to a new central operating frequency . Thus, for example, with respect to the frequency sweep of Figure 2, Control C includes mechanisms for varying the operating frequency through a plurality of different operating frequencies, for measuring a plurality of samples of an electrical parameter of the lamp in each of the plurality of distinct operating frequencies, and mechanisms for calculating, at each plurality of operating frequencies, the standard deviation of the samples taken at each frequency, and mechanisms for evaluating, for example, detecting the smallest standard deviation , to select a new center frequency. The selection of control C components is based on the desired accuracy and sampling rates needed to effectively implement the program in Figure 9 and depends on the type of the lamp, its physical and chemical properties, the power of the the lamp and its dimensions. In one implementation, 20 was the minimum number of samples selected so that the current and voltage of the lamp obtained a measurement of the standard deviation at a frequency. Samples were taken at 100 individual frequencies during each frequency sweep, such that a total of 100 by 20, or 2000 samples are taken in each frequency sweep. The sampling rate should be slow enough to allow the arc to react but fast enough to prevent large bow movements. An adequate speed range is between 50 msec and 100 msec for metal halide lamps of 100 W. The minimum accuracy of the A / D and D / A converters 90, 95, 105 and the microprocessor 100 was selected in 0.005 which it is satisfied by 8-bit devices (8-bits = 1/256 = 0.004). In the implementation shown in Figures 13, 14 the A / D converters are an 8-bit, high-speed ADC 0820 model available from Linear Technologies. The D / A converter was also 8-bit; NE 5018 model available from Signetics. The microprocessor was an 8-bit 16Mhz microprocessor, model No. 87C550 with 4K EPROM available from Philips Semiconductors. The software program of Figure 9 was programmed in the Asse bler language. These devices provide a sampling range of 10 kHz to 20 kHz for the current and voltage of the lamp with a sampling rate between 50 msec and 100 msec, which was found sufficient to detect and effectively avoid the acoustic resonance of the lamps. 100W metal halide. Control B uses well-known techniques to drive the half-bridge inverter with the exception that it drives the half-bridge inverter at a frequency indicated by control C instead of at a constant frequency as is usually done. Figure 18 illustrates a circuit that implements Control B. The circuit includes IC 110 and 120. IC 110 is a high-speed 16-pin controller (model UC3825 available from Unitrode) that serves to receive the frequency which designates the output signal fg from Control C and provides control signals for the input of IC 120. The IC 120 is a 14-pin, high-voltage, high-speed MOS gate driver (IR model 2110 available from International Rectifier) . The IR 2110 has high-side and low-side reference output channels and is especially useful for half-bridge applications because it has internally adjusted dead time control. The IC 120 provides the grid / source drive signals to the mosfet and Q2, Q3 switches. The output signal of Control C enters connectors 5 and 6, which serve to control the variation of the duty cycle. The output of IC 110 is taken from connectors 11 and 14 and placed in connectors 12 and 10, respectively, of IR 2110 (IC 120). The grid GQ2 of the switch Ql (Figure 3B) is connected to the connector 1 through the resistor R31. The diode D31 is connected in parallel with the resistor R31 with its cathode connected to the connector 1 and serves to prevent a negative cesgo in this connector. Similarly, the grid GQ3 of the switch Q3 is connected to the connector 7 through the parallel arrangement of the resistor R32 and the diode D32, which fulfills the same function as the resistor / diode network connected to the connector 1. The source SQ2 of switch Q2 and source SQ3 of switch Q3 are connected to connectors 2 and 5, respectively. The microprocessor can be programmed to include different combinations of the steps described herein and / or to use one described option instead of another. For example, the controller can select an operating frequency that occurs at a wide minimum instead of a narrower, but smaller minimum, or implement the Open Loop routine at different points during start-up and steady-state, such as it was analyzed, for example, with respect to Figures 5 (a) - 5 (c). Moreover, the controller / ballast could be implemented in the form of a modular system including the controller one or more pieces each performing selected functions and emitting signals to other modules. In this context, for example, the modular system could include a ballast with a controller designed to collaborate with a series of control devices, each of which performs a selected combination of the functions described herein. From the above description, it can be seen that the Applicants have discovered certain steps that are universally applicable to gas discharge lamps for detecting and preventing arc instabilities and which are particularly useful for detecting / preventing arc instabilities due to acoustic resonance. These steps can be implemented in many balasta topologies, including single stage ballasts (such as those disclosed in U.S. Application Serial Number 08 / 197,530, filed February 10, 1994) and two-stage ballasts, and as already discussed, can use different lighters. The control is applicable for all (or a large) range of frequencies, basically from O Hz to several MHz and without perceived upper limit. The limitations refer to the speed of sampling and processing (which will be overcome by faster processors) and more importantly by the construction of the balasta (hardware). With a highly constant inverter gain over a wide range of frequencies, a wide frequency sweep rate can be implemented and will generally eliminate the need for the described prior selection of a wide frequency window to the designer. This will allow an even greater universality of control; for example, allowing a universal controller module to be connected to the ballasts optimized for the different wattage ranges. The method makes the balasta quite insensitive to variations in the dimensions of the lamp (in production lines), chemical changes throughout the life of the lamp, and changes in the characteristics of the lamp during its life. The method allows the use of high frequency electronic ballasts to operate HID lamps and prevents the destruction of the lamp and a catastrophic end of life. This control makes the ballast quite immune to the lamp because it is intelligent, universal, general and flexible.

Claims (21)

1. NOVELTY OF THE INVENTION Having described the foregoing invention, the content of the following CLAIMS is claimed as property 1. A method for igniting and operating a high pressure discharge lamp, comprising the steps of: a. operating the high pressure discharge lamp at a plurality of operating frequency; b. detecting and taking a plurality of samples of an electric lamp parameter in each of the plurality of operating frequencies; c. calculate a deviation, in each of the frequencies, of the samples taken in each frequency; and d. Evaluate the calculated deviations and select a lamp operating frequency based on the evaluation.
2. 2. A method according to claim 1, wherein the detected lamp parameter is at least one chosen from a series comprising the lamp voltage, the lamp current, the lamp conductance and the impedance of the lamp. the lamp.
3. 3. A method according to claim 1 or 2, wherein the calculated deviation is the standard deviation.
4. 4. A method according to claim 1, 2 or 3, wherein the deviations are evaluated to determine the minor deviation and the operating frequency is selected as the frequency corresponding to the smallest deviation. A method according to any of claims 1 to 4, wherein the high pressure discharge lamp is operated at the plurality of frequencies by performing a frequency sweep around a central frequency, the sweep including a portion with a set of frequencies where frequencies are increasing and a portion that includes the same set of frequencies where the frequencies are decreasing. 6. A method according to any of claims 1 to 5, wherein steps (a) - (d) are carried out during a step of starting the operation of the lamp, between the ignition of the high-pressure discharge lamp and the operation in steady state. 7. A method according to any of claims 1 to 6, wherein steps (a) - (d) are carried out both during the lamp start-up stage and during steady-state operation of the discharge lamp of the lamp. high pressure, and wherein the frequencies at which the lamp is operated during the starting stage have a first section, and the frequencies at which the lamp is operated after having reached the steady state have a second smaller section or equal to the first section. 8. A method according to claim 7, wherein the second section of the frequencies during steady-state operation is selected such that the high-pressure discharge lamp has no visible fluctuation for a human being. 9. A method according to any of claims 1 to 8, wherein the ignition and operation of the lamp is turned off, after a predetermined number of iterations of steps (a) - (d), the minor deviation detected is above a prescribed level. A method according to any of claims 1 to 9, wherein steps (a) - (b) are carried out twice consecutively and steps (c) - (d) are carried out only if the difference between The respective sets of samples measured in the two iterations are within a prescribed reliability level. 11. A method according to claim 10, wherein a respective standard deviation s and s2 is calculated for the samples of the first and second iterations and the prescribed level is s2 < ß sl t where ß is a constant selected for the desired statistical reliability level. 12. A lamp ballast for starting and operating a high pressure discharge lamp according to the method as claimed in claim 1, characterized in that the ballast comprises: ballast mechanisms for operating the gas discharge lamp during a first range of operating frequencies; and control mechanisms for controlling the frequency of operation of the ballast mechanism, including the control mechanism: mechanisms for varying the frequency of operation throa plurality of different operating frequencies within the first range of operating frequencies; mechanisms for measuring a plurality of samples of an electric parameter of the lamp at each frequency of the plurality of operating frequencies; mechanisms for calculating, at each frequency of the plurality of operating frequencies, the deviation in the plurality of samples taken at each frequency; and mechanisms to evaluate the deviations calculated to adjust the operating frequency of the ballast mechanism based on the evaluation. 13. A lamp ballast according to claim 12, wherein the mechanism for evaluating the deviation calculates the standard deviation of the samples taken at each frequency of the plurality of frequencies. A lamp ballast according to claim 12 or 13, wherein the parameter of the electric lamp measured by the measuring mechanism is at least one chosen from the series comprising the voltage of the lamp, the current of the lamp , the conductance of the lamp and the impedance of the lamp. 1
5. 5. A lamp ballast according to claim 12, 13 or 14, wherein the electrical parameter measured by the measuring mechanism is the conductance of the lamp. 1
6. 6. A lamp ballast according to claim 12, 13, 14 or 15, wherein the mechanism for varying the frequency includes mechanisms for effecting a frequency sweep around a central frequency, the sweep including a portion with a set of frequencies where frequencies are increasing and a portion that includes the same set of frequencies where frequencies are decreasing. 1
7. 7. A lamp ballast according to claim 12, 13, 14, 15 or 16, wherein the evaluation mechanism determines the frequency at which the deviation of the electric parameter samples is the lowest and adjusts the operating frequency from the balasta mechanism to the frequency with the minor deviation. 1
8. 8. A lamp ballast according to claim 12, 13, 14, 15, 16 or 17, wherein the control mechanism operates during a step of starting the operation of the lamp, between the lighting of the lamp and the operation in steady state. 1
9. 9. A lamp ballast according to any of claims 12 to 18, wherein the control mechanism operates both during the start-up phase and during the steady-state operation of the lamp, and wherein the frequencies at which it operates the lamp during the start-up stage have a first section, at the frequencies at which the lamp is operated after having reached the steady state have a second section less than or equal to the first section. 20. A lamp ballast according to any of claims 12 to 20, wherein the second section during steady-state operation is selected in such a way that the lamp has no visible fluctuation to a human being. 21. A lamp ballast according to any of claims 12 to 20, wherein the lamp controller includes mechanisms for extinguishing the lamp if the detected minor deviation is above a prescribed level.