WO2006037312A1 - Method for maintaining the temperature of an electrically heated sensor on load changes - Google Patents

Abstract

The invention relates to a method for the precise maintenance of the mean temperature of an electrically-heated sensor, whereby the sensor temperature remains between or very close to an upper temperature limit T<sub

Description

 description

Method for keeping constant the temperature of an electrically heated sensor during load changes.

Technical area

The invention relates to a method for keeping constant one and the same average temperature of an electrically heated sensor before and after a load change, wherein this temperature is maintained above the temperature of the medium surrounding the sensor.

State of the art

The constant maintenance of the temperature of a heated sensor is mainly used as a measuring method in anemometry. Constant-temperature anemometry (CTA) is a technique that has been used for many decades, mainly in fluid dynamics research, with short response times. Because of their frequent use in the academic environment, there are a large number of generally accessible scientific papers and investigations.

The basic structure of a type 1 CTA circuit (proportional method) is shown in FIG. 10 (according to a representation by Dantec Dynamics). In addition to Dantec Dynamics A / S, Skovlunde / Denmark, TSI Inc., St. Paul, Minn., Is one of the world leaders in constant temperature anemometers. Furthermore, CTA circuits of type 2 are known (switching method, see Fig. 11), in which a comparator ver¬ by the rapid on / off a fixed heating voltage ver¬ tries to keep the sensor temperature constant.

The CTA technology has received great general significance for about 10-15 years as "air genome meter" or "air mass meter" (LMM) in connection with internal combustion engines in automobiles, after the ever stricter emission standards only today still from electronic engine control units. can be met with metered fuel injection. However, these can only generate an optimum air-fuel mixture if the amount of air taken in per working cycle can be measured (see Auto & Elektronik 1/2002, p. The automotive supplier Bosch, for example, has LMMs according to the latest, conventional design in its product range.

The main criticism of the prior art is that the problem of the thermal inertia of the sensor could still not be solved satisfactorily: According to the currently valid understanding, the achievable response times are always proportional to the mass of the sensor.

The CTA technique and the hot-wire or Hitzfilm anemometry (with vanishingly small sensor mass) are therefore currently commonly understood as synonyms.

Initially, automotive LMMs actually had a heated, very thin platinum wire as the sensor element. Later, sensors were transferred to heat-film technology, which, however, still represent a notorious cause of operating disturbances due to their tiny dimensions (eg failure due to soot particles and, mainly in the case of type 2 circuits, failure due to control failure due to inadequacy) the load control range of this type of circuit upon contact of the sensor with a liquid, eg with a water droplet).

Similar problems relate to the hot wires of the research CTAs whose diameter of 5 microns or less is only a small fraction of the thickness of a human hair (for comparison: 60 to 100 microns) and not only very easily tear, but also careless with careless operation can burn through electrically.

Because of the high fragility of the hot wire sensors (and because of their high price), the use of the CTA technique is generally considered very expensive; In research, on the other hand, their use is nonetheless indispensable.

Although robust sensors, such as those required by industry, can in principle also be operated in state-of-the-art CTA mode (see Seydel, Kolahi, Rock - "Model-based flow detection using a current available on the market). mungswächters "in: Proceedings of the MessComp 1997, Wiesbaden, pp. 249-258), but only relatively small speed advantages are achieved compared to the extremely slow-acting constant current method used in industry (CCA). inter alia, strong vibrations can occur in the CTA mode) do not outweigh.

Presentation of the invention

In the new CTA method according to the invention, the mean sensor temperature can always be kept within a selectably small tolerance range by using suitably calculated heating voltages. This results in a clear improvement in the temperature stability compared to the known CTA method, in which a load change regularly either leads to strong vibrations or to a change in the average effectively kept constant temperature. As a direct result of this improvement, short response times can be achieved which are not dependent on the mass of the sensor to the extent that is usual, so that "hot-wire-typical" response times are possible even with significantly more robust sensors.

1) Basic formulas

Time-dependent cooling of the sensor (cooling law according to Newton), the sensor cools without supply of electrical power as follows:

(T _um = ambient temperature, T ₀ = initial temperature) Cooling time, which elapses between reaching an upper temperature T ₀ and reaching a lower temperature T _u :

t _k =

Time constant:

_ m - c ^{T = ~} ^ A < ³⁾

(m = mass of the sensor [in kg], c = heat capacity of the sensor [in Ws / (kg ^* K)], α = surface and temperature related heat dissipation [in W / (m ^{2 *} K)], A = effective area of the sensor in contact with the medium [in m ² ], time constant T [in s]).

Time heating process:

(T _e = final temperature in thermal equilibrium under the influence of a constant electric heating power)

Time during heating, which passes between leaving the ambient temperature T _um and reaching an upper temperature T ₀ :

2.1) Wie sich ein echtes Konstanttemperatur-Verfahren verhalten sollte

2.1) How a true constant temperature method should behave

The thermal inertia of a sensor always occurs only when its temperature changes. If the temperature is kept artificially constant by means of a control circuit, the reason why thermal inertia effects should be eliminated is also eliminated. Of course, no control circuit can make do without any slightest changes in the sensor temperature, but these changes could be amplified, for example, electronically, so that they would not be macroscopically noticeable on the sensor. A more or less strong heat loss of the sensor to the medium makes itself noticeable immediately on a tiny change in the sensor temperature, which, however, compensates the control circuit again in practically the same moment by a correct adaptation of the electrical heating power. The sensor temperature has therefore not changed, and there is no reason why thermal inertia effects should occur. Incidentally, no exorbitant temperature gradient between sensor and medium must prevail, because even at a slight excess temperature, the sensor and the control circuit reacts in principle immediately to any load change. Because even a slight over-temperature compared to the medium in principle completely sufficient and because the temperature of the sensor remains constant even under heavy load changes, you can use resistors with a very high temperature sensitivity (eg NTCs, PTCs) as sensors whose general temperature / resistance characteristics - line plays no further role (they are always operated only in the closest surrounding of a single point on their characteristic curve). Moreover, as said, there is no reason ever to have to operate it so hot that it could possibly be damaged. And again to come to speak of it: if no thermal inertia effects occur, then could be, very exaggeratedly said, but actually an electrically heated brick used as a fast-reacting sensor, as long as it is ensured that the very large amount of heat that he the medium gives off, can be replenished at any time by a sufficiently powerful power source. If one were to continue to roll the "brick" like a bread dough so that its active surface in contact with the medium would increase in proportion to its mass, then a) its heat loss would increase and at the same time b) its A larger active surface not only brings higher reaction rates, it also integrates over the entire surface, ie local microturbulences, individual liquid droplets, soot particles etc. would not significantly affect the measurement result. A distinction "for gases" or "for liquids" is also not necessary. You can move the sensor as often and quickly from a gaseous medium in a liquid medium and vice versa, without changing its temperature stability something.

The method according to the invention can be described as a true constant temperature method, since it comes very close to the above-described ideal image, as will be described in more detail below.

2.2) Why the previously known methods are not true constant temperature methods

Since the two terms "constant temperature method" and "constant temperature anemometry (CTA)" have been occupied for many decades, the question arises as to why the CTA methods known hitherto have only so few practical similarities to the one shown above Have a picture of a true CTA method.

Since the answer to this question is the key to understanding the method described here, it should be given in advance: none of the known CTA methods is able to adapt the electrical heating power at the moment of a load change immediately correctly.

As a result, the properties listed in the first paragraph after the passage "... in virtually the same moment by a correct adaptation of the electrical heating power ..." for the previously known CTA methods are not attainable on principle explain follows:

In the case of the known proportional method (type 1), a sudden increase in the instantaneous heat loss is assumed. In this case, the sensor temperature drops only slightly in the first moments, whereupon the control circuit likewise reacts only with slight increases in the heating voltage. In fact, it should now respond with a massive, erratic increase in order to prevent a further drop in the sensor temperature under the new conditions. In the end The sensor temperature drops over a longer period of time before the (too slow) increasing heating current is finally able to reverse the lowering tendency. With the same inertia of the process, there is even an overshoot above the setpoint temperature, because the now too high heating current can only be lowered very slowly. In practice, these oscillations are therefore always (the CTA manufacturers also prescribe it!) Reduced by an incomplete compensation of the heat lost at the sensor, which of course has the unfortunate consequence that under different load situations inevitably different average Sen¬ semperaturemperaturen constant held (load change -> temperature change of the sensor -> thermal inertia).

In the known switching method (Type 2) is also assumed a sudden increase in the instantaneous heat loss. The process now automatically produces longer heating phases (ON) than cooling phases (OFF). If one considers mathematically the resulting mean sensor temperature which is kept constant, ie if one calculates the average sensor temperature by means of the integrals via the cooling and heating curves, then one realizes that the mean sensor temperature has in fact already fallen below the setpoint value. If the heat loss increases even further, it may happen that only a single, continuous heating phase is produced during which the sensor temperature drops unrecoverably (the method has failed). This happens, for example, regularly when a type 2 CTA designed for gases comes into contact with a liquid, for example with a water droplet. That this represents a very serious problem, for example, with air flow meters in the automotive industry, one sees in constant Lösungsver¬ search, such as in US Patent 6,752,014 Hitachi. If the setpoint temperature is very far above the medium temperature and if a high heating voltage is used, as is often the case, then cooling and heating curves in type 2 become approximately straight lines, so that the mathematical problem of the mean sensor temperature perspective. Regardless, however, the larger problem is the low load control range of the Type 2 method, which results from the principle of operation P = P _max * t _h / (t _h + t _k ): it is easy to see that neither the value P = O still the value P = P _max can be realized within the scope of a meaningful control. From this follows the existence of gray zones for still acceptable ratios of t _h / t _k to each other, and it follows again that the usable load control range must be limited. In practice, the effective load control range is in fact not sufficient, for example, to change the context from a gas to a flowing liquid or vice versa to be able to tolerate (it always comes to regulatory failure).

2.3) How the average sensor temperature can be kept constant at one and the same value before and after a load change

As has been explained in 2.2, in the prior art so far only the (load-dependent) constant maintenance of various mean sensor temperatures has been achieved. In other words, after each load change, a slightly different mean sensor temperature is set and kept constant (as a result of this thermal inertia problem). According to all research so far, it can be assumed that this circumstance is currently not generally known and that the phenomenon of "thermal inertia", as it still occurs in the known CTA methods, therefore generally refers back to the mass of the sensor In fact, however, it is the aforementioned methodological deficiencies of the known CTA methods which make the greatest contribution to the so-called thermal inertia of the sensor. In the special case of exactly 100.0% recirculation of the heat lost at the sensor before and after a load change theoretically keep one and the same average sensor temperature constant, but this theoretical possibility can hardly be realized in practice, because one would have to do this Yes, you can exactly set the exact amplification factor -> this is practically the case t is not possible If less than 100.0% of the heat loss is returned (in practice, it is usually about 80%), then one obtains the said load-dependent, different average sensor temperatures, while on the other hand one leads more than 100.0%. back, this results in a self-oscillation, in the course of which the mean sensor temperature constantly increases.

In the following it is shown, how already with the basic elements of the invention, ie with calculated heating voltages and with the use of a temperature window, ie with an upper temperature limit T ₀ and a lower temperature limit T _u , which are both above the medium temperature T _m , to a simple procedure that solves the task without any problem: It is a) a first, low heating voltage U _h o applied to the sensor, which is successively increased in successive steps by a voltage amount Δ \ J U _hn = U _hn -i + Δ \ J until ge the sensor has heated to a high sensor temperature T ₀ . Thereafter, b) the heating voltage according to U _hn = U _hn -i - ^ U lowered again until the sensor has cooled due to heat loss to a lower sensor temperature T _u , then c) the heating voltage according to U _hn = U _h ni + ^ U again increased until the sensor has heated up again to the high sensor temperature T ₀ , and ultimately the process with steps b) and c) is repeated constantly.

Fig. 9 is intended to illustrate the basic operation of this method. Shown is the behavior before and after a load change (in the left half of the drawing is a low heat loss, in the right half, the heat loss has increased). The bold curve represents the course of the sensor temperature, the thinly marked, triangular curve shows the course of the electrical power (P _e ι) converted at the sensor. It can be seen here: a) the average sensor temperature remains constant before and after the load change, b) this happens automatically and therefore requires no adjustment measures with an accuracy that is close to the impossible, c) the above method can be varied to a large extent without significantly affecting its character For example, in Fig. 9, U _hn = U _hn -i ± Δ \ J was not used, but P _hn = P _h π-i ± P, d) The term "computation" does not automatically imply the use of a digital microcontroller but can often also be realized with means of analogue technology (which incidentally was the case here, otherwise the smooth, triangular curve would have become rather a triangular-shaped stair curve).

Although this method can automatically solve the basic task of a true CTA, namely to maintain the mean sensor temperature which is actually kept constant under all load situations and thus already somewhat ahead of the known CTA method, it is in turn itself (eg by means of its known Type 1 similar tendency to oscillate, which still depends on the mass of the sensor) still far from a true CTA method, as will be the subject of the following Aus¬ guides. 2.4) Ideally, the average sensor temperature should be kept constant

Method based on Type 1: Assuming a constant loss of heat at the sensor, it is immediately clear that one would basically only have to apply the heating voltage which is exactly suitable for this purpose in order to permanently compensate for the heat loss. Under the effect of this matching heating voltage, the heating curve according to Eq. 4 lead to the end temperature T _{e of} the sensor would set exactly at the Solltempe¬ temperature. After a change in the instantaneous heat loss, it would be possible, in principle, to immediately calculate a new, exactly matching heating voltage from the first measurable consequences of the change, apply this and then have again achieved that the final temperature becomes identical to the setpoint temperature.

Method based on Type 2: In a cyclic "ON / OFF" method, the average sensor temperature can only be kept constant at the setpoint T _so n when the cooling curve between an upper limit T ₀ (= T _so u + ZT ) and an un¬ direct limit T ₁₁ (which also _always n above the highest possible medium temperature lie¬ gen needs, T _u = T - ZiT) and the heating curve between the same limits complement each other In other words, the average sensor temperature of the heating curve. (Equation 4) must always be greater by the exact same amount ("mountain"), by which the mean sensor temperature of the cooling curve (GI.1, "valley") always remains slightly below the setpoint temperature for physical / mathematical reasons These requirements can then be satisfied, and only then, if no fixed maximum voltage is applied in the heating phase, but a calculated voltage which leads to a final temperature of exactly T _e = 2 * T _so n _-Tr would, if you let it rest permanently. For real change to 2 * T _so -T n _to occur while never because it is in the nature of a style similar to the Type 2 method is that the heating voltage is already switched off when reaching T ₀ again. But the section of this special heating curve lying between T _u and T ₀ is the exact counterpart to the course of the preceding cooling curve between T ₀ and T ₁₁ . In this case too, after a change in the instantaneous heat loss, it would be possible in principle to calculate a new, exactly matching heating voltage from the first measurable consequences of the change, to apply this and then to ensure that the mean sensor temperature remains identical to the setpoint temperature. Summary: assuming that the currently exactly required heating voltage could be determined, optimum heating curve characteristics in the sense of a true CTA process can be achieved both for Type 1 and Type 2 processes. In a true CTA method, the knowledge of the instantaneous heat loss is also a concomitant of the scheme, which holds itself with this constant knowledge even in ideal operation, and is not, as in the known CTA method, only the end result of a previously relatively blind expired Rege ¬ ment.

2.5) How to find the exact required heating voltage

2.5.1) Execution of a measuring cycle

First, the temperature of the sensor must be at the upper limit T ₀ . Now the supply of electric heating power is reduced to a minimum. There is a drop in the sensor temperature according to Eq. 1. Now the time is measured ^, which passes until the sensor temperature has dropped from the upper limit T ₀ to the lower limit T _u .

You can t | _If necessary, also with a sequence of many, at short intervals Δ \ aufein¬ other temperature measurements determine. The time t _k is measured as the time nz \ t, where n is the number of temperature measurements that occurred between the receipt of the measured value T ₀ and the receipt of the measured value T ₁₁ , wherein as a rule between the two marginal measured values, between which T ₀ or T _u is interpolated must be. The circuit of Fig. 7 uses this measuring method. However, a direct measurement of the time t _k (as in the circuit of FIG. 6) is always preferable, since time measurements are among the most technically feasible measurements ever.

In a method based on type 2, the upper limit T _{0 is} regularly reached, whereupon the next measuring cycle is triggered. With a reference to the type 1, the upper limit T _{0 is reached} only as a result of a load change, the measurement cycle is then triggered. The load change in type 1 leads first to reach the lower one Limit T _u , the sensor temperature is forcibly and briefly brought to the upper limit T ₀ with a high heating voltage, followed by the measuring cycle followed.

2.5.2) Formation of targets for the heating voltage to be applied

Here it is determined to which result the heating voltage to be applied should result.

In a method of type 1, one would demand from the heating voltage to be applied that, if possible, it should set a final temperature of the sensor which is identical to the setpoint temperature. However, since the final temperature of the sensor can not be set immediately and therefore it could not be determined until relatively late whether the desired result has occurred, it is better to form basically only one target for the equivalent type 2: here it is known that the ideal heating voltage would lead to a final temperature T _e = 2 * T _so n -T _um , if you let it rest permanently. At the same time, this ideal heating curve also has the easily measurable property that, starting from the lower limit T ₁₁ , the upper limit T _{0 is reached} in exactly the same time t _h as when t _k in the case of cooling said measuring cycle was measured ge.

The target value is thus t _h / t _k = 1. If an applied heating voltage causes the temperature of the sensor to increase from the lower limit T _u to the upper limit T ₀ in the time t _h = t _k , then it must now be exactly required heating voltage U _h2 (for a switching process based on type 2).

It is then also possible to indicate immediately the exactly required heating voltage U _M for a method according to type 1, since both voltages are related to one another in the simplest manner:

U - ^ (6) Incidentally, even with the less stringent target value t _h / t _k = const, in many cases it is possible to arrive at a passable control, except that then a slightly different setpoint than the corresponding nominal value is kept constant.

In principle, one could also describe the real CTA process described here as a synthesis between the known processes of type 1 and type 2, but this would lead too far here. It should therefore be said that in a true CTA method there is no exclusive reference to only one of the two known basic types, but (as is obvious from equation 6) elements of both basic types can be used in free combination. For example, For example, the voltage required by a Type 1 method can be determined by a very short interlude (a few cycles) of a Type 2 method, and then, for an indefinite time (until the next load change), only the Type 1 process characteristics become effective again are. Another example of synthesis is the use of the variable heating voltages of type 1 (which in principle can only change mathematically continuously there) and which are switched on in the case of a true type 2 CTA (that is, mathematically discontinuous). Therefore, instead of "true CTA" one could also say, for example, "synthetic CTA".

2.5.3) Formation of an estimate of the heating voltage to be applied

A first heating voltage U _{h is} estimated (0 <U _h <= U _max ) and applied. In general, one will use a well-founded estimate based on the measured time t _k (see 2.5.5.2), but in principle one could make a completely free choice.

2.5.4) Quality assessment of the applied heating voltage

The following statements apply to a method based on type 2: in view of the application of the heating voltage (the sensor temperature is then currently at the lower limit T _u ) one begins with the measurement of the heating time t _h , which passes until the upper limit T _{0 is} reached. There are exactly three possible cases that can occur: 1) the estimated value exactly matches the currently exactly required heating voltage, 2) the estimated value is too high and 3) the estimated value is too low.

Case 1) presents no problems, the upper limit T ₀ is reached exactly in t _h = t _k , thus one has obtained the desired control and can initiate a new measurement cycle. The quality assessment consists of the quotient t _h / t _k (or what is basically exactly the same, from the quotient t _k / t _h ) and has the optimum value 1.

Case 2) increases the sensor temperature too much, so that the upper limit T _{0 is reached} too early. The quality assessment (t _h / t _k <1) is poor, but the estimate may be considered process neutral (ie, not harmful to the process) since the upper limit has been reached and a new measurement cycle can be initiated.

Case 3) does not increase the sensor temperature sufficiently fast. If the sensor temperature rises and the upper limit T _{0 is reached} within a still tolerated time t _h = f * t _k (f = tolerance factor, eg f = 1, 15), the estimated value is given an unfavorable quality assessment (t _h / t _k > 1), but evaluated as process neutral, since a new measurement cycle can be initiated without further action.

If, on the other hand, the sensor temperature does not reach the upper limit T ₀ in the tolerated time, then the time measurement is aborted (t _h / t _k »1), the estimated value is deemed to be potentially harmful to the process (ie there is a risk that the Sensor¬ temperature so that can not be kept within the temperature window, in particular at too low heating voltages) and the temperature rise to T ₀ is enforced by the application of a high or maximum heating voltage. The same happens when the sensor temperature reaches the lower limit T _u during the measurement of t _h or can not exceed this limit already from the start (more precisely: if the sensor temperature has not yet exceeded the limit T _u after elapse of a selectable minimum measurement time t _w ). The potentially harmful process value is thus converted into a process-neutral value (case 2) by this mandatory measure. Thereafter, a new measuring cycle is initiated. You certainly can not do all the ways of performing a quality assessment in detail. If, for example, the temporal, not disturbed by a load change course of the cooling curve with a sequence of n successive Tempera¬ turmessungen in the respective distance .DELTA.t and has stored as a curve T «(t) in a memory, then one could as a target for example, the heating curve T _H , _S oiι (t) = (T _u + T _o ) -Tκ (t) calculate. Now one does not necessarily have to use the quotient t _h / t _k for quality assessment, but can define a more general quality value Q which obtains an optimal score (= 1) if the temperatures T _H measured at each time t> 0 during heating (t) lie in a selected To¬ leranzbereich the width 2ε to the setpoint temperatures (TH, _SO I _I O) - ε <THOO ^< TH, _S oiι (t) + f). If ε has been designed relatively generously, the quality value can also be determined finer by the quotient t _h / t _k . Again, there are again the above-mentioned 3 possible cases. Case 1 has just been described, in case 2 T _H (t) deviates upwards from the setpoint temperatures (gets a bad score, Q <1), and case 3, where Tπ (t) deviates downwards from the setpoint temperatures. This case can lead to Q> 1 and to Q »1 (if the sensor temperature reaches T ₁₁ despite heating or does not even exceed it). One then also immediately breaks off and forces the temperature rise to T ₀ , as described above.

For the convergence methods described in 2.5.5, it is irrelevant in what way the quality assessment was obtained, it is only crucial that a quality assessment of the estimated values is carried out at all, and that monitoring of the heating curve for the purpose of neutralizing potential procedurally harmful estimates.

During the neutralization of a potentially harmful process value, ie in the forced heating to T ₀ , a high and therefore generally not adapted to the just exactly required heating voltage heating voltage must be applied during the required short time. So you do not, for example, in a simple load change within a gas u. For example, with the highest possible performance (which would suffice for fast-flowing liquids), it is quite obvious and sensible to preselect the instrument for gases or liquids. Assuming that there is still a gas preset to gases in the medium during the measurement, one will first use one for gases designed high voltage, which would normally bring the sensor temperature in a short time to T ₀ . The resulting heating is, as mutatis mutandis already be¬ written monitored. If the heating is not possible as planned, it can be assumed that a change of the medium from gaseous to liquid, whereupon only the actual maximum available heating power is used. The default setting can be made manually, done automatically (by evaluating the heating voltages used in the previous average) or ignored again on a case-by-case basis (eg with drop counters). Since high-frequency medium changes rarely occur in practice, said presetting would be a gain in terms of the achievable temperature stability under normal load changes. However, heavy load changes (in particular medium change from gaseous to liquid) would still be tol- erated, whereby the control would only be busy for one clock cycle longer.

2.5.5) Iterative, convergent improvement of estimates

In the first decision for an estimated value one could still use a more or less arbitrarily chosen value, but already in the second decision, the quality assessment of the first decision is available as additional information, so that the estimated value can now be improved in a targeted manner. In principle, the improvement of the estimated value always amounts to choosing the heating voltage higher than before, if a ratio t _h / t _k > 1 was obtained or to choose it lower (at t _h / t _k <1).

With each new estimated value, which follows a measuring cycle, the exactly required heating voltage is always approached better, in the final state the estimated value and exactly required heating voltage are identical.

It is obvious that repeated changes in the instantaneous heat loss at the sensor (load change) should not occur in a shorter time in the interest of a sensor temperature which is to be kept as constant as possible on average, than the method for producing the final state requires.

It is therefore preferable to use methods with the highest possible convergence rate: 2.5.5.1) Example of a very fast convergent method

1) Uh _Q = 2 °. Uh _max = Uh _max

2 _. a ₎ Uh _n = Uh _n _ _λ - 2-. Uh _max ; ± <1;n> Q h

2 _.b) Uh _n = Uh _n _ _{ι +} 2- ^» . Uh _max ; f * ≥ l; n> 0

2 _.C) Uh _n = Uh _n ^; l - ε <^ <l + εh

This is basically the interval halving method known from computer science or the database technology, which, however, had to be safeguarded by means of the surveillance measures mentioned in 2.5.4 (to protect against potentially harmful operating estimates). If during the monitoring a neutralization measure has occurred, the automatically following measuring cycle is first awaited, the measurement result t _{k of which is} no longer taken into account, and thereafter a planned increase in tension (2.b) or step 1 (if the method had already begun or had reached step 2.c). As soon as the heating voltage no longer changes (2.c, the tolerance value e is freely selectable, but should on the Auflösungsver¬ like the heating voltage generation oriented, ie small e at high resolution), can be waited for the occurrence of a new load change. If a load change occurs, which necessitates an increase in the heating voltage, the method starts again with n = 0 from step 1. If the load change requires a reduction in the heating voltage, then n = 0 can be continued from step 1, with but reasonably should no longer use the maximum heating voltage Uh _max in step 1, but better than the last used heating voltage Uh _n .

This process converges so quickly that it could already advantageously replace the currently industrially used electronic flow monitors, which operate on the constant current principle. 4 shows the time profile of the sensor temperature (thin sawtooth curve) and that of the heating voltages (bold rectangle curve) in a very rapidly converging process. Strictly speaking, in this example, not the heating voltages themselves are halved, but the electrical powers, but the principle remains the same. At first, the sensor temperature is still below the temperature window, which is why it is brought to the upper limit T ₀ with a maximum heating power. Switch off the heating power, decrease the temperature to T _u and start with step 1 (Fig. 4 no. 1): apply maximum heating power, measure the heating time t _h , turn off the heating power when T _{0 is} reached, measure the cooling time t _k , quotient t _h / t _k and because of t _h / t _k <1 select the continuation with step 2.a (FIG. 4 no. 2): Reduction of the applied heating power by a halving step. Here, however, after the time t _{k has} elapsed from FIG. 4 No. 1 plus a certain tolerance time, it has been established that the sensor temperature has still not reached the limit T ₀ -> forcibly increase the heating power to maximum, bring the sensor temperature to T ₀ , switch off wait for the heating power, drop to T _u , then continue with step 2.b (increase the heating power by a further halving step, Fig. 4 No. 3). Switch off the heating power, cool down mes¬ sen. Quotient t _h / t _k , finding that again t _h / t _k <1, that is to say in FIG. 4 no. 4, lowering the heating power by a further halving step, etc. Already from FIG. 4 no. 6 ff , the ideal control occurs with t _h / t _k = 1, which is of course no longer changed thereafter until the next load change (step 2.c).

2.5.5.2) A maximum fast convergent method

An extremely high convergence speed is provided by the following method, in which t _k -based estimates are used and in which, as a rule, even the first estimate corresponds almost exactly to the exact required heating voltage. As a result, this process can compete with the current research CTAs in terms of performance, but in contrast to these, they work with much more robust sensors.

The prerequisite for this method is the knowledge of the "temperature window heat" W _f , ie the heat energy required to increase the temperature of the sensor from T _u to T _0. Conversely, the sensor returns exactly this amount of heat to the medium when its temperature reduced from T ₀ to T ₁₁ . From each (at least short-term) stable load situation W _f can be easily determined, for example by applying the first-mentioned interval halving method. As soon as there is a stable control with the equality of the times t _k and t _h , it is always ensured that the mean sensor temperature is just the value T _so n. Since one directly knows the heating voltage used during a heating phase (it was deliberately applied in the selected height), one can calculate from this the electrical power P _h , which is proportionately attributable to the sensor and which was consequently delivered to the medium.

Under the conditions of an ideal control with t _h / t _k = 1, the relation holds

W _f = P _h - t _k = P _h - _h t (J)

Since W _{f is} a constant, Eq. 7 no more and no less than that one can calculate immediately from the measurement of a cooling time t _k , which heating power must be applied to the sensor in order to continue to keep the control in ideal operation:

W

P ₁ = f

(8th)

From the knowledge of P _h , the exact voltage required to drop across the sensor can be used for Type 1 methods

and after type 2

always calculated directly (R = resistance of the sensor at T _so n). In a real circuit, the calculation of the heating voltage to be applied must naturally take place taking into account all other resistors involved. For the circuit according to FIG. 6, for example, if the sum of R 1 + R ₂ + R ₃ relative to R ₄ + R _{5 is} so great that its contribution can be neglected

The thus-calculated estimation values always produce a ratio t _h / t _k = 1 and thereby cause the temperature of the sensor in the center is exactly T _so iι.

Changes in the medium temperature and load changes (including medium change from gaseous to liquid and vice versa) can not change this principle.

In practice, there may also be load changes that do not build up over a few control cycles, but occur so abruptly that they can completely devalue even the best estimate. Therefore, the quality assessment according to 2.5.4 is indispensable, mainly because of the associated safeguarding of the heating-up times with the monitoring / neutralization measures described there. In contrast to the interval halving method (2.5.5.1) is erhalte¬ ne (from the automatically at neutralizing action following measurement cycle) cooling time t _k, however, analyzed here to the regulation immediately to adapt correctly to the new load situation.

Over time, there may also be a slight deviation of the average ratio t _h / t _k from the ideal value 1. If you do not want to simply redefine W _f or you may (at the moment there are too many load changes per unit of time), the value of W _{f can} also be increased or decreased automatically until the average deviation has been eliminated. Thus, the method can independently adapt to factors on which W _{f can} still depend (for example, partial soiling of the sensor, aging) and compensate for this. In the rare case that the CTA encounters a high frequency of load changes immediately after switching on and thus has no opportunity to measure the current value of W _f itself, you can also use the W _f value from a non-volatile memory (eg EEPROM, eg position of a trimmer potentiometer, etc.), in which the value has been stored ex works or also, later, as the last independently measured value.

FIG. 5 shows the time profile of the sensor temperature within the temperature window before and after the occurrence of a very abrupt load change in the case of an extremely high temperature! convergent method (here type 2). First, there is a relatively low instantaneous heat loss, so that the cooling time t _k i is relatively large. After the lower limit T _{u has been} reached and the measurement of t _{k1 has} thus been ended, an estimated value for the heating voltage to be applied is calculated, as described above. This heating voltage is applied and would normally lead to a heating curve, as has already occurred at t _M (quotient t _h / t _k = 1). But now comes the sudden load change (strong increase of the instantaneous heat loss), which ensures that the upper limit T ₀ with the currently applied heating voltage can not be achieved in the maximum tolerated time t _max = f * t _k1 . Not only that, it even happens that the "wrong" limit (ie, t _u ) is reached during heating, so the heating voltage now immediately goes to maximum, which is seen by the steep rise in the sensor temperature T _{0 is} reached, the heating voltage is reduced to almost zero and a measurement of the cooling time t _{k2 is} made This result is already determined by the new value of the instantaneous heat loss, so that practically the first calculated estimated value is again able to To produce a heating curve which leads to the equality of the times t _k2 and t _h3 , this equality also arises again and again for all following times t _k and t _h .

If the course of the heating voltages is also included, the characteristic image of the new method (here based on type 2) is shown: FIG. 1 shows a situation with a relatively stable instantaneous heat loss, which is applied in a stable manner, while FIG 2 shows the situation under the conditions of a stable, relatively high heat loss. It can be seen that the control obviously aims at keeping the quotient t _h / t _k always at the value 1 and that it activates heating voltages at different levels for this purpose. Fig. 3 shows the new method based on the type 1, here before and after a load change. First, there is a permanent heating voltage U ₁ , which compensates the instantaneous heat loss at the sensor so precisely that a final temperature has been set, which corresponds to the setpoint T _so n. The load change (here an increase in the instantaneous heat loss) occurs at the moment when the sensor temperature begins to drop. As soon as the lower limit T _{u has been} reached, a maximum heating voltage U _max brings the sensor temperature in a short time to the upper limit T ₀ , then Ab¬ lowering the heating voltage to a value close to zero (U _k ) and measuring the Abkühl¬ time t _k , which is already determined by the newly occurred value of the instantaneous heat loss. Now, as in the above-mentioned methods, the calculation of an estimated value for the voltage to be applied. This differs from the result of a procedure according to type 2 only by the division by the root of 2 (see equation 6). The voltage (U ₂ ) obtained in this way is applied permanently, whereupon the end value of the sensor temperature normally returns to the desired value T _SO . If a load change causes a reduction of the same instead of increasing the instantaneous heat loss, the sensor temperature will be triggered at the upper limit T ₀ . One then dispenses with the heating step with U _max and initiates the measuring cycle directly to determine t _k .

2.6) From momentary heat loss to heat dissipation capability of the medium

Also, the CTA method described here initially allows nothing else than the determination of the instantaneous heat loss at the sensor. This loss of heat depends not only on the height of the temperature difference between the sensor and the medium, but also on the heat dissipation capability of the surrounding medium, the latter, in turn, being dependent on various material properties and last but not least on the flow state in which the medium is located relative to the sensor how much sensor surface is in contact with the medium, etc. In a solid contacted with the sensor, the heat dissipation capability is even influenced by the magnitude of the contact pressure, and even that does not mention all the known influencing factors. The heat dissipation capability is therefore a relatively complex physical quantity, in the formation of which very different other physical, geometric and chemical influencing factors are involved. Conversely, however, it is possible that precisely these other variables can be measured by measuring the heat loss itself. As a rule, the desired quantities are obtained from the measurement of the heat loss, if one has determined individual, likewise involved physical variables (eg temperature, pressure, immersion depth of the sensor, etc.) by their simultaneous, separate measurement or they already knows or keeps constant for other reasons.

In any case, the measurement of the instantaneous heat loss alone is sufficient in exceptional cases for a useful application. Even the simple task of flow detection requires knowledge of the current heat dissipation capability of the medium; the momentary heat loss is simply not enough (except if the medium is practically always at the same temperature).

For applications which are primarily concerned with the heat dissipation capability of the medium, the medium temperature (T _um ) must be additionally measured, for example in a known manner with a second, conventional temperature sensor or, at least as well, with a second CTA according to the inventive method in which all parameters except the temperature of the medium are kept constant.

Then, immediately after completion of a measurement cycle (out of a regulation that is in ideal operation, as described above), the current time constant T of the sensor can be calculated from the measured time t _k (see equation 2):

The values of the upper temperature _limit T ₀ and the lower temperature _limit T ₁₁ are known in advance, T _um and t _k were measured. The time constant of the sensor is not a constant in the usual sense, but, inter alia, a practically inertia-free function of the current heat transfer from the sensor to the medium (compare equation 3):

If, for example, the heat transfer is doubled while the medium temperature remains the same, and as a result the cooling time t | _< exactly halved, then this alone is attributable to the change of the factor, if the contact area A was kept constant. The mass m and the heat capacity c of the sensor are generally not subject to change (aside from the case of damage to the sensor).

Since the heat transfer from the sensor to the medium (at constant contact area A and after the influence of the medium temperature has been excluded) can only change if a simultaneous and same-direction change in the heat dissipation capability of the medium occurs, a direct proportionality exists between them ¬ the sizes.

Consequently, only the reciprocal 1 / τ of the calculated time constant (equation 12) must be formed in order to obtain a display proportional to the actual, temperature-compensated value of the heat-dissipation capability of the medium.

Furthermore, if the heat dissipation capability of the medium remains constant, the display of 1 / τ can be used to measure changes in the contact area A. In connection with robust sensors, which can be thought of as arbitrarily shaped (for example, also as a long wire), this results in possible applications for the measurement of immersion depths, water levels, fill levels and the like.

2.7 Circuit examples

2.7.1) Circuit according to FIG. 6

While in the known CTA methods the Wheatstone bridge circuit with four resistors is used (see Fig. 10 and Fig. 11), this circuit has an atypical bridge with 5 resistors, wherein R2 for the formation of two resistors. fined, exact switching or temperature limits ensures. The switching limits are sym metrically to the setpoint temperature of the sensor, and the value of R2 determines how far apart the boundaries. The circuit generates a comparator signal ("too hot") when the upper temperature limit is exceeded and another comparator signal ("too cold") when it falls below the lower temperature limit.

It is essential that the exact time can be measured, which elapses between the reaching of the one and the reaching of the other temperature limit. This time information about the phase duration (heating phase duration t _h and cooling phase duration t _k ) can be obtained here from the time offset between the switching of the comparators measured by the microcontroller (.mu.C).

To generate different voltages, the microcontroller controls a D / A converter (DAC). The usually relatively high-impedance output voltage of the D / A converter supplies after amplification and impedance conversion (OP1) the loadable variable heating voltages (eg U _f1 during the heating phases and U _k during the cooling phases) for the measuring bridge.

In general, it is additionally the ambient temperature T _by more than a second temperature sensor in a conventional manner to measure (not shown in FIG. 6 eingezeich¬ net).

2.7.2) Circuit according to FIG. 7

In this circuit example, there is no longer any classical measuring bridge, and no comparators are required. Rather, the voltage drop U ₃ over R ₅ is simply measured with the aid of a fast A / D converter (ADC). Since the variable heating voltage Ut ₁ or the voltage U _k used for cooling is known at any time because the microcontroller has generated it itself with the aid of the D / A converter (DAC), the value of Sensor resistance R ₄ , which is proportional to the temperature of the sensor (because the value of R _{4 is} quasi fixed to a point of the temperature characteristic of the sensor and the possible here in question minimal fluctuations of R ₄ therefore always be approximated by a linear function), simply according to the microcontroller

be calculated. Thus, the evaluation of the time profile of the sensor temperature (for example, by virtue of the virtual simulation of a temperature window) and the determination of the matching heating voltages could be carried out completely by software. Again, the ambient temperature (fluid temperature) T is generally additionally measured _by a second temperature sensor in a conventional manner (not shown in Fig. 7 shown). It would be possible here purely by software to change the setpoint temperature during operation, z. B. to maintain a constant temperature difference to Mediumstempera¬ tur. Of course, the same can be achieved with the circuit according to FIG. 6, if R ₁ or R _{3 are made} variable there (eg via a digital potentiometer controlled by the microcontroller).

2.8 Determination methods for detecting whether a foreign measuring device uses the erfindungsge¬ Permitted method

With the exception of certain thermostats, because of the indispensable contact with the medium, the sensor is that element of a measuring device which, in principle, can never be completely concealed or cast in. Under certain circumstances, the electrical supply lines of the sensor can be directly accessible, but need not. If one approaches the electrical supply lines (possibly nevertheless), one can simply connect an oscilloscope and see directly whether, after a load change, variably high, convergently connected heating voltages are used (compare FIGS. 1, 2, 4) , 5). Otherwise it is possible to bring a (small) temperature sensor into good thermal contact with the sensor of the device to be examined and draw the necessary conclusions from the thus measured temperature curve at the sensor (eg regulation keeps the ratio t _h / t _k = 1 (or t _h / t _k = c) constant (FIGS. 1 + 2, 4, 5), carrying out a measuring cycle during load changes with a permanently calculated estimated value calculated therefrom (FIG. 3) or a derived starting value for a simpler method (eg according to FIG. 9)). It should also be pointed out that the term "heating voltages" naturally also includes (alternating, PWM) voltages whose effective values can produce the objectively identical thermal effects as direct voltages DC voltages, this should ultimately include more exotic types of indirect heating of the sensor, such as by laser fall.

2.9 Examples of new developments using a true CTA process

It would go far too far to enumerate the large number of scientific and technical applications in which the known hot-wire anemometry has hitherto been used for a very wide variety of measuring tasks. A true CTA method is suitable in principle for exactly the same applications, but also brings a robust practicality, a tolerance to media changes and the ability to integrate the measured value over a larger sensor surface as essential new gen¬ properties of the CTA technology into play so that many, far beyond the previously known, new applications are possible. The following examples can therefore only give you a first impression:

a) For the industrial monitoring of fluid flows (also for paste-like media) electronic flow monitors can be built, which operate in contrast to the previously used CCA-based devices virtually maintenance-free and react compared to these by several hundred or a thousand times faster.

b) Handheld instruments for the rapid thermal classification of gases, liquids and surfaces for the purpose of a rough chemical analysis.

(By immersing a sensor in the quiescent medium, for example, one obtains a first characteristic value (which incorporates the heat dissipation capability of the quiescent medium.) By rapidly moving the sensor, a second characteristic value is obtained (in which the heat dissipation capability of, relative to the sensor Of course, knowing only these two parameters, it may not be possible to uniquely identify any chemical substance, but it may well be be assigned to the characteristic ratio of the measured values to each other clearly a particular substance from a small preselection of substances of interest, especially if one can take into account results from other measurement methods (eg the temperature) with. For example, it would be possible to differentiate between different petroleum types quickly and on the spot, without first having to commission a chemical analysis).

c) Applications of the new methods in robotics and mechanics where e.g. a gripping arm is equipped with a flat sensor. The instantaneous change in heat dissipation capability upon contact with different media or surfaces could easily be detected by the change in heat loss, and would e.g. enable a robot to distinguish whether it is z. B. a piece of metal, a piece of plastic or just into space (air) has seized. Last but not least, the measured heat dissipation capability of solids increases with the pressure applied. Both would be useful additional haptic information for robots.

d) electronic drop counters, e.g. in medicine, pharmacy or chemistry (constant medium change between gaseous and liquid). The correct dosage of medicaments (for example in the case of infusions) or in titrations (chemistry) could thus be electronically monitored or automated in a simple manner.

e) devices for the rapid measurement of mass flows, e.g. for measuring the intake air amount in internal combustion engines or e.g. for Verbrauchser¬ mediation in the industrial application of compressed air, industrial gases, liquids and other fluids. The heat loss of the flowing media increases in gases approximately proportional to the root from the flow velocity of the medium (at constant pressure) and also proportional to the pressure (at a different pressure). With some correction calculations (calibration), therefore, the mass flow and (with additional measurement of the pressure) also the real volume flow can be determined.

f) High-speed dosing devices, eg for liquid adhesives, toothpaste or coffee powder, to name only three arbitrary examples that would normally lead to the immediate destruction of the sensor in classical hot wire anemometers and which, with sufficiently robust measuring instruments according to the CCA principle, would previously have been unthinkable for reasons of speed.

g) medical applications; the ability of the skin to dissipate heat changes as a result of perspiration, hormone, poison and medication effects, different opening widths of the blood vessels, tissue changes, diseases, fever etc. Possible applications: pulse measuring devices, clinical thermometers, polygraphs, tissue examinations on the patient and in the laboratory and much more more.

h) Further medical applications: Respiration monitoring by measuring the air flow, e.g. by a clip attached to the nose wing sensor. Advantages: the temperature of the sensor must be kept just above body temperature, and the sensor can detect the true air flow (the bigger, the better), without being irritated by local air movements in the room. Dittos can e.g. In pulmonary function tests, due to the high time resolution of the method, oscillations can also be detected, which can result from illness-related narrowing of the respiratory tract.

i) Mechanical length measurements, e.g. by an elastically deformable sensor. The more the sensor extends in length, the more heat transfer surface comes into effect in relation to the volume of the sensor -> increase in heat loss means length change (with conditions otherwise kept constant).

j) Measurement of the humidity (hygrometer). The more water vapor in the air (more generally: in one gas) is contained, the higher the heat dissipation capability of the gas or gas mixture in general.

k) measurements of the moisture content of various substances and objects. Example: at the same temperature, a dry sponge differs from a damp sponge, among other things, in particular also due to the different heat dissipation capability.

I) Applications for (fast) temperature measurement. Under otherwise constant conditions, the heat loss at the sensor changes only as a result of changes in the Temperature T _around the surrounding medium. This temperature T are _to be measured advantageous ge because the procedure requires only a slightly higher temperature sensor than the maximum occurring medium temperature and may therefore very sensitive to temperature changes of the medium.

m) applications for determining the heat dissipation capability, in which two CTA are used in combination according to the new method, one is eg in contact with a flowing fluid and the other, even in contact with the fluid, but flow-protected, to measure T _to is used.

n) Applications as a thermostat, e.g. For example, a small quartz watch quartz could be brought into firm thermal contact with the sensor, so that to a certain extent it would itself become part of the sensor whose temperature is kept constant. This would result in an increase in the accuracy of the accuracy with respect to quartz, which are exposed to changing temperatures with little effort. Prior to the advent of atomic clocks quartz crystals kept constant temperature served as official time standards, so that with the solution proposed here cheap and yet excellent timepieces could be installed in devices that should not depend on the local reception of Zeit¬ time transmitters.

o) As a last example, the use of a CTA circuit according to the invention (Fig. 8 No. 1) for the rapid measurement of water levels in separated media with unterschied¬ Licher heat dissipation capability, eg. As water (Fig. 8 No. 3) and air (Fig. 8 No. 2), ge shows are. In principle, a simple, insulated wire can serve as the sensor (FIG. 8, No. 6).

Let h _max (Fig 8 No. 4) be the height of the vessel and at the same time half the total length of the wire (because of the double wire guide). The height to which the liquid wets the wire is h _F (Fig. 8 No. 5).

The height of the air column is h _L = h _max - h _F. With the same wire length, the heat is lost by the factor f more heat than in air. The heat loss is proportional to the filling level, as shown below: For t \ = h _max (no liquid in the container) one obtains a certain heat loss W _min . For h _F = h _max (container completely filled with the liquid) results in a heat loss W _max = f * W _min .

For the heat loss as a function of the filling height, the following relationship results:

and resolved according to the filling level as a function of the measured heat loss W, which in the case of the thermal equilibrium can be equated with the directly known average electrical power at the sensor:

respectively.

K (W) = C ₁ - (W - C ₂ ) (16)

with - - TT - TT = ^c i = ^const and W _min = C ₂ = const.

W ^ - ⁽ J - I)

If direct disturbing influences are to be excluded by possible turbulent proper motions of the liquid, it is recommended to surround the measuring wire with a tube which is open at the bottom and at the top and encloses the measuring wire over its entire length.

In summary, the following results:

In a method for precisely maintaining the average temperature of an electrically heated sensor, the sensor temperature also remains at load changes between or in the immediate vicinity of an upper temperature limit T ₀ and a lower limit T _u . If the distance between the limits is chosen to be small, short response times can also be achieved with robust sensors.

For heating, variable-height heating voltages are used, which are estimates whose quality of results is assessed in relation to the targets. As a result of the quality assessment, iterative, convergent improvements of the estimates occur until the estimated value and exactly required heating voltage are identical. A distinction is made between the process-neutral and process-harmful estimates, the latter being converted into process-neutral estimates by means of acute countermeasures before they can have a negative effect on the temperature constancy.

Short description of the drawing

In the drawing show:

1 shows the time profile of the sensor temperature [T / K] and the heating voltage [U _h / V] with a low heat loss according to the invention,

2 curves similar to those in FIG. 1, with a higher heat loss, according to the invention,

3 shows curves similar to those in FIG. 1 or 2, with a sudden increase in the heat dissipation capability of the sensor, according to the invention, FIG. 4 shows the time profile of the sensor temperature (thin sawtooth curve) and that of the heating voltages (bold rectangular curve) in a very rapidly converging method according to the invention, FIG.

5 shows the time profile of the sensor temperature with a sudden increase in the heat loss in an extremely fast converging method according to the invention,

6 shows a circuit used in the method according to the invention, in which a sensor with NTC characteristic is used,

7 shows another circuit used in the method according to the invention, in which a sensor with a fast A / D converter is used.

8 shows a representation of a sensor in the form of an insulated wire in the use of the method according to the invention for the rapid measurement of liquid levels,

9 shows the time profile of the sensor temperature (bold curve) with the associated curve of the electrical power converted at the sensor (thin triangular curve), in the left half of the picture before a load change (low heat loss) and in the right half after a load change (higher heat loss),

10 is a prior art CTA circuit of a first type;

Fig. 11 is a known in the art CTA circuit of a second type.

Claims

Claims: 1) A method for keeping constant one and the same average temperature of an electrically heated sensor before and after a load change, wherein this temperature is maintained above the temperature of the medium surrounding the sensor, a) that, starting lowered from a high sensor temperature T ₀ which is above the temperature of the surrounding medium T _around and on which the sensor has been heated or is located, the filament voltage to a low value U _nied so that the sensor temperature due to heat loss to a low sensor temperature T ab¬ sinks _11, which is also _to still above the temperature T, then the cooling time T I _<gemes¬ sen is, b) that the sensor is then heated up again to the sensor temperature T ₀ and the heating time t _h required for it is measured, c) that then, d) if it is established that the heating time t _{h is} greater than or would become greater than the previous cooling time t _k applied by a tolerance value F ₁ , the sensor is reheated to T ₀ with a new, increased heating voltage and the method with step a) is restarted, or c2) if the heating time t _{h is} equal to or less than the cooling time t _{k applied} with the tolerance value Fi, the quotient Q is determined from the heating time t _h and the cooling time t _k , and c3) if the quotient Q lies within a number range between a lower limit Qu and an upper limit Q ₀ , c3.1) either the sensor continually alternately heated and allowed to cool is gelas¬ sen, in each case heated with the heating time t corresponding _h Heizspan¬ voltage U _h and cooled at each of the filament voltage U _n I_D according to a), c3.2) or the heating time t _h corresponding filament voltage U _h by the square root of 2 divided and so resulting filament voltage U _hiPerm is continually used for heating the Sen¬ sors, and the method step a) is begun again when the sensor temperature reaches the value T ₀ or T _u , the attainment of T _u followed by heating with an increased heating voltage to T ₀ , before starting with step a), or c4) if the quotient Q determined in accordance with c2) is not between the limits Q _u and Q ₀ or no value has been determined for Q due to the case d), in an iterative convergence method alternating heating and cooling cycles to the sensor temperatures T ₀ and T _u be repeated until a quotient Q results from the respective heating and cooling time, which is between the limits Q _u and Q ₀ , wherein the convergence through the use of higher heating voltages than before (after Q> Q ₀ and after Case d)) or by the use of lower heating voltages than before (after Q <Q _u ) is brought about, and then c4.1), then the process is continued with the steps c3.1) or c3.2). 2) Method according to claim 1, characterized in that F ₁ > = 1 is selected. 3) Method according to claim 1 or 2, characterized in that as a set value for Q, the value 1 and that

is selected, where T ₀ as target temperature T _so n

and T _{u are defined} as Ts ₀ H -ZiT ₂ . 4) Method according to one of the preceding claims, characterized in that the heating voltages are formed as follows with a rapid convergence method (interval halving method): 4.a) Uh ₀ = 2 ° • Uh _^ = U \ max 4. _b ) Uh _n = Uh ^ - I- "- Uh _{n ^} ; Q <Q _U ; n> 0 4.c) Uh _n = Uh "_, + 2" • Uh max Q ≥ Qo n> 0 4.6) Uh _n = C ^ ₁ ; ß ₍ <ß <Q ₀ 5) Method according to claim 4, characterized in that after steps d) and a) with a scheduled increase in voltage (4.c) or with step 4.a) (if the method had previously been settled and step 4.d) achieved was continued), then, as soon as the heating voltage does not change (4.d), at Eintre¬ th a new load change, if it leads to the case c1) or Q> = Q ₀ , the Ver go with n = 0 is again started from step 4.a), and if the new load change leads to Q <Qu, the method is also continued with n = 0 from step 4.a). 6) Method according to claim 4 and 5, characterized in that in the continuation with n = 0 from step 4.a) in the case Q <Q _u not the maximum heating voltage Uh _max in step 4.a) is used, but the last used heating voltage Uh _n . 7) Method according to one of the preceding claims, characterized in that from the heating voltage U _n derived quantities are bisected in half, in particular the heating power P _n = U _n ² / R _so n (R _SO ιι is the resistance of the sensor at T _SO ιι ) And that the actual heating voltage to be applied is calculated back from the interval halved. 8) Method according to one of the preceding claims, characterized in that, if in the convergence method according to c4) a control with the property Q = 1 has been prepared, then: W _f = P _h -t _k = P _h -t _h = const, wherein P _{h is} the heating power used during a heating phase, which is ent¬ on the sensor, and then the cooling time t _k is measured again and then the Heiz¬ power that must be supplied to the sensor is calculated as follows: W P - f and from this the heating voltage to be applied to the sensor

is calculated. 9) Method according to claim 8, characterized in that the long-term, average constancy of the ratio Q = 1 is monitored and as soon as there is a deviation from the value of 1, the value of W _f is increased or decreased so long, until wie¬ the average value 1 is present. 10) Method according to claim 8 or 9, characterized in that the value of W _{f is} determined once under the conditions of constant heat loss and stored in a non-volatile memory (eg EEPROM, eg position of a Trimmerpotentiome¬ ters) on which a measuring device can immediately access in a future performance of the proceedings. 11) Method according to one of claims 8 to 10, characterized in that, if the measuring device has made a new measurement of W _f , the new value overwrites the original setting in the non-volatile memory. 12) Method according to one of claims 4 to 11, characterized in that the frequency f of the control oscillation / = ^ι 2 -t _t 2 -t " is output or further processed as a variable proportional to the measure of the instantaneous heat loss. 13) Method according to one of the preceding claims, characterized in that an additional measurement of the temperature T is made _by the surrounding medium. 14) Method according to one of the preceding claims, characterized in that the current time constant of the sensor or its reciprocal

τ t _k is calculated, which (with the size of the contact surface of the sensor with the medium kept constant) as a temperature-compensated, proportional to the current value of the Wärmeableitungsfä¬ ability of the medium measured value is displayed. 15) Method according to claim 14, characterized in that the determined value of 1 / τ is displayed as a measure of the size of the actual contact surface of the sensor with the medium, the heat dissipation capability of the medium being considered constant. 16) Method according to one of the preceding claims, characterized in that the quotient Q = t _h / t _{k is} replaced by a quality value Q, whose formation is carried out over a series of temperature measurements, in which the temporal Abkühlungs¬ course T _κ (t ) of the sensor as a result of temperature measurements with associated Zeitin¬ formation and stored and from a setpoint curve according to T _H , _S oiι (t) = (T _u + To) -Tχ (t) is calculated for reheating and at the Q the difference between the measured temperature profile and the course of the setpoint curve is derived. If, at a time t> 0, it is determined during heating that the current sensor temperature is above T _H, then (t) ⁺ f, then Q is assigned a value <1, if T _{H is} exceeded, then t) -f receives Q assigned a value> 1, where ε can be chosen freely. If the current sensor temperature within the tolerance width 2ε follows each time point of the setpoint curve T _H , _SO ιι (t), then Q becomes 1. As soon as Q is assigned a value> 1, heating takes place with increased heating voltage to the sensor temperature T. ₀ and the initiation of a new cooling phase. The value of Q then proceeds to step c3). 17) Method according to claim 16, characterized in that the time duration t _{k of} a cooling Ab and the temperature of the surrounding medium T _is measured by that Equation tk is solved for T and a setpoint curve T _H , soiι (t)

for the heating process according to T T _{H, solι} (t) T

e for times 0 <= t <= t _k , where T _e = T ₀ + T _u -T _is used. 18) Method according to claim 17, characterized in that for the final temperature T _e the setpoint temperature T _so n is used and the setpoint curve Tπ.soiiCt) for times 0 <= t <= f * t _{k is} calculated, the factor f so is selected that the deviation of the temperature TH, soiι (f ^* tk) of the target temperature T _so n just below a previously selected value f. With this setpoint curve, the heating course over the quality value Q is assessed as in claim 16. Once the difference between the actual temperature of the sensor and the target temperature T n _so within a tolerance width of 2ε to T is n _so that heating voltage last found remains permanently applied. If this tolerance range is left as a result of a load change or if T ₀ or T _{11 is} reached for the same reason, the achievement of T ₁₁ or the falling below T _so n -ε is followed by heating with an increased heating voltage to T ₀ (ditto if T _so n + ε is exceeded, but not when T _{0 is reached} ), then step a) is started again.