Four current transistor temperature sensor and method

[0002] 1. Field of the Invention

[0003] This invention relates to the field of transistor temperature sensors, and particularly to methods of reducing measurement errors due to intrinsic base and emitter resistances in such sensors.

[0004] 2. Description of the Related Art

[0005] Numerous circuit devices, such as transistors, diodes and resistors, have operating characteristics that are temperature dependent. Because of their temperature dependencies, such devices are extensively used as temperature sensors. For example, germanium and silicon diodes can be operated at a constant forward-biased current, and the resulting forward-biased voltage measured to determine the temperature in accordance with the standard forward-biased diode equation:

[0006] V=kT/qln I/I_s

[0007] where V is the forward-biased voltage, k is Boltzmann's constant, q is the electron charge, T is the absolute temperature in degrees Kelvin, I is the forward-biased current and I_sis the diode's saturation current.

[0008] In practice, the measurement of temperature with a diode is subject to several inaccuracies. The precise voltage-temperature relationship depends upon the actual details of the junction, notably the doping densities on either side of the junction, the dopant profiles and the junction area, as well as secondary considerations such as bulk and surface defects in the material. These factors are difficult to quantify with certainty, and many of the parameters in the device equations (such as mobility) are themselves temperature-dependent. Other effects, such as conductivity modulation and series resistances, can also complicate the device's behavior.

[0009] Another approach employs two separate junctions which are fabricated on the same substrate, but which are operated at different current densities. This eliminates the effects of variations in doping levels and in the value of the bandgap voltage. The dual junction approach has been implemented with a pair of bipolar transistors whose emitter areas are in the ratio A. The difference in collector current densities gives rise to a difference in the base-emitter voltages for the two transistors. The relationship between the base-emitter voltage differential (ΔV_be) and the device temperature is given by the expression:

[0010] ΔV_be=kT/qln A

[0011] While this approach offers significant advantages over the single junction temperature measurement, it still has some limitations. There is a certain amount of tolerance in the transistor fabrication, which introduces an ambiguity into the emitter area ratio. Furthermore, the accuracy of the equation is reduced by ohmic resistances associated with the junction, specifically the base resistance r_band the emitter resistance r_e. The base and emitter resistances may be considered to include both the intrinsic resistances inherent in the device, and the resistances associated with connecting lines. Calibration of such a sensor is required for most applications, and the fact that at least a pair of junctions are required introduces the possibility that differential strain across the substrate could result in poor tracking of junction voltages with a consequent error in the small ΔV_bevoltage.

[0012] Another technique is described in U.S. Pat. No. 5,195,827 to Audy et al. Here, a single bipolar transistor is sequentially driven with three different currents, inducing three base-emitter voltages which are measured and used to calculate temperature. This approach also has significant shortcomings, however. Using three currents requires that the ratios between the currents be kept small, in order to avoid heating up the sensing transistor and thereby introducing error into the temperature measurement. Also, the calculations necessitated by a three-current approach are likely to require non-integer math, which can be difficult and/or impractical to implement.

[0013] A four current transistor temperature sensor and method are presented which overcome the problems noted above. The invention allows the use of large current ratios and simple temperature calculations, while still reducing or eliminating intrinsic base and emitter resistance errors.

[0014] A p-n junction, preferably the base-emitter junction of a bipolar transistor, is driven with four different currents in a predetermined sequence. Each of the four currents induces a respective base-emitter voltage, which is measured. The temperature of the transistor is calculated based on the values of the four driving currents and the four measured base-emitter voltages.

[0015] In a preferred embodiment, the four driving currents (I1, I2, I3 and I4) are arranged such that I1=n*I3, I2=n*I4, I1/I2=A and I3/I4=A, where A is a predetermined current ratio. In operation, I1 and I2 produce respective base-emitter voltages which are subtracted from each other to produce ΔV_be1, and I3 and I4 produce respective base-emitter voltages which are subtracted from each other to produce ΔV_be2. When so arranged, the difference between ΔV_be1and ΔV_be2is entirely due to the effect of series base and emitter resistances r_band r_e. The ΔV_be1−ΔV_be2value thus provides a correction factor which enables temperature measurement errors due to r_band r_eto be eliminated. This arrangement also allows the use of large currents ratios, and greatly simplifies the calculations required to determine temperature T.

[0016] FIG. 1 is a schematic diagram illustrating the basic principles of a transistor temperature sensor per the present invention.

[0017] FIG. 2 is a schematic diagram of a preferred current source arrangement for the present invention.

[0018] FIG. 3 is a schematic diagram showing a preferred current source implementation for the present invention.

[0019] FIG. 4 is a block/schematic diagram illustrating a temperature measurement system which employs the present invention.

[0020] A four current transistor temperature sensor per the present invention is shown in FIG. 1. A p-n junction 10 is employed as a temperature sensor. P-n junction 10 is preferably a bipolar transistor Q_s, but other bipolar devices, such as a junction or Schottky diode, could also be used.

[0021] As indicated in FIG. 1, bipolar transistor Q_shas an associated series base resistance r_band a series emitter resistance r_e. These resistances may be due to the intrinsic properties of the p-n junction itself, as well as lead and connection resistances. As noted above, these resistances often degrade the accuracy of prior art temperature sensors which employ a p-n junction as a sensing element.

[0022] The present sensor is arranged to provide four different currents through p-n junction 10, each of which induces a respective voltage across the junction. As shown in FIG. 1, a current source 12 is arranged to provide currents I1, I2, I3 and I4 to sensor transistor Q_sin a predetermined sequence, which induces voltages V_be1V_be2, V_be3and V_be4, respectively, across the transistor's base-emitter junction.

[0023] Temperature can be determined by calculating the difference between the base-emitter voltages induced by two different currents. By measuring V_be1with current I1 applied to transistor Q_s, and measuring V_be2with current I2 applied, the difference ΔV_be1between V_be1and V_be2is given by:

[0024] ΔV_be1=(kT/q)*ln[(I1/(1+(1/β)))/I_s]+I1*r_e+(I1/β)*r_b−(kT/q)*ln[(I2/(1+(1/β)))/I_s]−I2*r_e−(I2/β)*r_bvolts  (Eq. 1)

[0025] where k is Boltzmann's constant, q is the electron charge, T is the absolute temperature in degrees Kelvin, I_sis the transistor's saturation current, and β is the transistor's gain. This simplifies to:

[0026] ΔV_be1=kT/qln(I1/I2)+(I1−I2)*(r_e+(r_b/β))  (Eq. 2)

[0027] Similarly, V_be3and V_be4are measured with currents I3 and I4 applied, respectively, with the difference ΔV_be2between V_be3and V_be4given by:

[0028] ΔV_be2=kT/qln(I3/I4)+(I3−I4)*(r_e+(r_b/β))  (Eq. 3)

[0029] If r_b, r_eand β are known, either of equations 2 and 3 could be used to determine the temperature of transistor Q_s. With each expression requiring the application of only two currents, the I1/I2 or I3/I4 ratios can be larger than would be permissible with a three-current scheme, and still not unacceptably heat the transistor. The larger current ratios provide a larger ΔV_be, which tends to increase the accuracy of the measurement.

[0030] However, transistor parameters r_b, r_eand β can be difficult to ascertain, and may vary from transistor to transistor. These problems are overcome when the invention is arranged in accordance with the preferred embodiment shown in FIG. 2. Here, the ratios between currents I1 and I2, and between I3 and I4, are equal to a common value “A”. In addition, currents I1 and I2 are made equal to n*I3 and n*I4, respectively. When so arranged, the difference between ΔV_be1and ΔV_be2is entirely due to the effect of parasitic resistances r_band r_e. This arrangement results in an expression for temperature T as follows:

[0031] T=(q*ΔV_be)/[k*ln(I1/I2)]  (Eq. 4)

[0032] where ΔV_beis given by:

[0033] ΔV_be=ΔV_be1−[((n/(n−1))*(ΔV_be1−ΔV_be2)],  (Eq. 5)

[0034] and

[0035] ΔV_be=ΔV_be2−[(1/(n−1))*(ΔV_be1−ΔV_be2)],  (Eq. 6)

[0036] in which ΔV_be1=V_be1−V_be2and ΔV_be2=V_be3−V_be4. Thus, when I1−I4 have the relationships specified above, errors that would otherwise be present due to the series base and emitter resistances are eliminated.

[0037] The value of n is preferably 2. When n=2, the expressions for ΔV_beare simplified as follows:

[0038] ΔV_be=ΔV_be1−[2*(ΔV_be1−ΔV_be2)],  (Eq. 7)

[0039] and

[0040] ΔV_be=V_be2−(ΔV_be1−ΔV_be2).  (Eq. 8)

[0041] Making n equal to two is preferred because, assuming the voltage measurements are converted to digital values, a multiplication by two is easily accomplished by calculating ΔV_be1−ΔVbe2and performing a left-shift on the result.

[0042] A preferred embodiment of current source 12 is illustrated in FIG. 3. Here, current source 12 comprises a current mirror having an input transistor Q_inand two output transistors Q_out1and Q_out2; current ratio A is established by making the emitter area (A*X) of Q_out1A times the emitter area of Q_out2(X) A control signal CONTROL establishes the current in Q_in. In operation, CONTROL is set to a first value, causing Q_out1and Q_out2to output currents I1 and I2. A switch connects first I1, then I2, to junction 10, and a voltage measuring means 20 measures the resulting base-emitter voltages V_be1and V_be2. The CONTROL signal is then changed such that the current in Q_inis halved, which also halves the current in Q_out1(I3) and Q_out2(I4). These currents are connected to junction 10 in sequence, and V_be3and V_be4measured. With I1−I2 and V_be1−V_be4known, ΔV_be1and ΔV_be2are calculated and provided to equation 7 or 8 to determine ΔV_be1which is then used by equation 4 to produce T. Operation is the same if current source 12 is arranged such that n is a value other than two, except that equation 5 or 6 must be used to calculate ΔV_be.

[0043] A system for determining and storing the value ΔV_be1ΔV_be2needed in equations 5 and 6 is shown in FIG. 4. Voltage measuring means 20 is implemented with a signal conditioning circuit 22 and an analog-to-digital converter (ADC) 24. Circuit 22 samples consecutive V_bevalues (e.g., V_be1and V_be2, or V_be3and V_be4) and provides the differences (i.e., ΔV_be1and ΔV_be2) to ADC 24, which converts the ΔV_bevalues to digital form. The ADC output is provided to a processor 26 which includes a first register 28, an offset register 30, and a subtractor 32.

[0044] When CONTROL is asserted (i.e., goes high), currents I3 and I4 are successively applied to junction 10. This produces voltages V_be3and V_be4, and signal conditioning circuit 22 calculates V_be3−V_be4=ΔV_be2, ADC 24 converts this to a digital value, which is stored in register 28 when the conversion is complete (as indicated by the ADC's generation of the “eoc” (end-of-convert) signal). CONTROL is then deasserted, the above sequence repeats for currents I1 and I2, and a digital representation of ΔV_be1appears at the output of the ADC. The output of register 28 (ΔV_be2) is subtracted from the ADC output (ΔV_be1) by subtractor 32, with the result stored in offset register 30. This value (ΔV_be2−ΔV_be1) can then be used in equation 5, 6, 7 or 8 to produce ΔV_be, which is used in equation 4 to calculate T. Note that, if n=2, the ΔV_be1−ΔV_be2value stored in offset register 30 can be doubled by left-shifting the data one bit; this can be useful when using equation 7 to calculate ΔV_be.

[0045] A controller (not shown) controls the system's operating sequence, by, for example, providing the CONTROL signal and controlling the switching between currents within current source 12. The controller function may be handled by processor 26, or may be a separate circuit block.

[0046] Signal conditioning circuit 22 might comprise, for example, a switched capacitor integrator which samples two base-emitter voltages (e.g., V_be1and V_be2) and integrates the difference to produce the ΔV_bevalue (e.g., ΔV_be1) provided to processor 26.

[0047] Note that the implementation shown in FIG. 4 and discussed above is merely exemplary. Many other signal conditioning circuit designs might be employed, and other topologies could be used to determine ΔV_be1−ΔV_be2For example, rather than use a signal conditioning circuit to calculate the difference between base-emitter voltages, each base-emitter voltage might be converted to a digital value, and the digital values subtracted as necessary to determine ΔV_be1and ΔV_be2. However, use of an analog signal conditioning circuit to calculate ΔV_be1and ΔV_be2is preferred, as this approach allows the base-emitter voltages to be amplified to a level sufficient for the ADC to resolve, thereby obtaining a degree of measurement resolution that would be otherwise be difficult to achieve.

[0048] While particular embodiments of the invention have been shown and described, numerous variations and alternate embodiments will occur to those skilled in the art. Accordingly, it is intended that the invention be limited only in terms of the appended claims.