M-level diode junction temperature measurement method cancelling series and parallel parasitic influences

[0002] This Application claims the benefit of U.S. Provisional Application No. 60/246,311, filed Nov. 7, 2000, the disclosure of which is incorporated herein by reference.

[0003] Not applicable.

[0004] 1. Field of Invention

[0005] The present invention is related to temperature measurement techniques with semiconductor circuitry, a use of which may be to determine diode junction temperatures with greater accuracy than previously known. More specifically, the present invention relates to the method and apparatus of exciting levels of applied collector current I_Cto a diode, sensing changes in output characteristics such as base-emitter voltage (V_BE) values, and calculating diode junction temperature free of series and parallel parasitic effects from a comparison of applied input, sensed output and calculated constant values.

[0006] 2. Description of Related Art

[0007] The accurate determination of the temperature under which a diode is operating may serve vastly different purposes. Diodes, as with many semiconductors, exhibit general temperature dependence where at a given constant diode current, there will be an approximate 2 mV voltage drop across the diode for each 1 degree C. increase in operating temperature. Therefore diode performance is often scaled relative to operating temperatures.

[0008] The temperature dependence of a diode allows the diode to be used as a temperature measuring device in several applications, as well as a temperature compensating component in many other circuits. For instance, certain emitter-coupled logic (ECL) gates require a reference voltage in the center of the logic swing for all operating temperatures. The temperature dependence of diodes allows the construction of temperature-compensated bias supply circuits for ECL gates to better serve this purpose. However, a less productive result of this temperature dependence can produce unwanted values as well, such as reverse diode currents. Therefore to ensure proper design, it is often very important to know the operating temperature of operational diodes.

[0009] Prior art U.S. Pat. No. 5,195,827, issued Mar. 23, 1993 to Audy et al. sets forth a number of methods to measure diode temperature. One method involves the immersion of the device in a heated or chilled bath. Device input and output values are then measured and used to calibrate the device to the known temperature of the bath. However, as Audy et al. points out, this is a complex and time consuming method and accurate predictions at temperatures other than bath temperature are not guaranteed.

[0010] A second method involves the actual placement of a calibrated thermometer on the device. However, it is often difficult to achieve satisfactory contact between device and thermometer, and in addition, once contact is established temperature variations within the device substrate do not guarantee an accurate temperature reading from the point of contact. The prior art further states such temperature measurement techniques are subject to additional inaccuracies due to junction characteristics such as doping densities, junction area and surface defects. A solution addressing these factors has been the use of two separate junctions fabricated on the same substrate, operated at different current levels. Unfortunately this dual junction approach is subject to further inaccuracies due to series base and emitter resistance values present in the device.

[0011] As Audy et al. points out, many devices exhibit a defined temperature dependence which may be used to calculate the actual device temperature when several variables are known. Specifically, Audy et al. discloses a temperature measurement technique in which known input signals are applied to a p-n junction device, such as a bipolar transistor, and temperature dependent output signals corresponding to each input signal are measured. Since the operating characteristics of the device are known, a comparison of the excitation levels applied and the temperature dependent output values observed may be used to determine the operating temperature of the device. The applied excitation signal may be either base-emitter voltage V_BEor collector current I_C. If the applied excitation signal is base-emitter voltage V_BE, the temperature dependent output value measured is collector current I_C, and in a similar fashion, if the applied excitation signal is collector current, the temperature dependent output value measured is base-emitter voltage.

[0012] Audy et al. discloses a method which measures semiconductor substrate temperature by exciting and applying three levels of either base-emitter voltage V_BEor collector current I_Cto a p-n junction device and sensing the varying temperature dependent output signal values. The temperature may then be calculated from the known characteristics of the device through a comparison of the signals. The purpose of a three level excitation signal application is to eliminate parasitic base resistance r_band the series emitter resistance r_evalues present in conventional p-n junction devices.

[0013] If only two excitation levels are applied, the temperature measurement will continue to be affected by parasitic base and series emitter resistance as shown in the dual junction approach, resulting in an incorrect measurement. To address this problem, Audy et al. discloses the use of a third excitation level which eliminates the effects of both r_band r_e. It is pointed out a three level excitation method works well with bipolar transistor devices, however when other p-n junction devices such as a conventional diode is measured, the accuracy of the three level excitation method may be degraded for various reasons. One factor leading to measurement degradation is the effects of parallel resistance and offset current effects present in devices such as diodes, which are not fully eliminated using a three level excitation method. Although the three level excitation method works well eliminating the effects of parasitic base resistance and series emitter resistance effects, so called series parasitic effects, the method fails to eliminate the effects of parallel resistances and offset currents, so called parallel parasitic effects. At full temperature this prior art method may read device temperature incorrectly by as much as several degrees C.

[0014] The multiple level excitation method taught by Audy et al. eliminates the inaccuracies of parasitic base resistance and emitter resistance through the use of three excitation levels of a single device, rather than a base-emitter voltage V_BEcomparison between two separate junctions as in the dual junction method. The use of three sequential excitation levels eliminates series resistances such as parasitic base and series emitter resistance but does not address parallel parasitic effects. The prior art method to determine p-n junction device temperature is based on the application of three or more collector currents and may begin with the application of collector currents I₁-I₄,

[0015] I=[I₁,I₂,I₃,I₄]

[0016] The temperature dependent output signal base-emitter voltage ΔV_BEis then measured as,

[0017] ΔV_BE=mV_T×ln[(I₃×I₂)/(I₁×I₄)]+R_S(I₃+I₂−I₁−I₄)

[0018] where R_Srepresents series parasitic resistance and ΔV_BErepresents the change in base-emitter voltage used to determine temperature. If collector currents I₁-I₄are chosen such that,

[0019] (I₃+I₂−I₁−I₄)=0

[0020] then,

[0021] ΔV_BE=mV_T×ln[(I₃×I₂)/(I₁×I₄)]

[0022] As seen in the prior art resulting equation, the change in base-emitter voltage value ΔV_BEmay be calculated free of series parasitic R_Seffects, however, current offset and parasitic parallel resistance are not taken into consideration and result in inaccuracies in ΔV_BEcalculations. Any temperature calculation based upon ΔV_BEwould also include these inaccuracies.

[0023] Therefore what is needed is a measurement technique to accurately determine device temperature with the ability to cancel out series parasitic effects, such as voltage offset and series resistance, in addition to parallel parasitic effects, such as leakage current and parallel resistance.

[0024] It is the object of the present invention to create a method and apparatus which may be used to determine diode junction temperatures with greater accuracy than previously known. The present invention achieves this goal by testing a device with four or more levels of applied collector current, sensing changes in output characteristics such as base-emitter voltage V_BE, cancelling or calculating the device leakage current value and cancelling the parallel leakage effects. The cancelled or calculated leakage current consists of both a constant leakage current, current offset, and an additional leakage current due to a parasitic parallel resistance. The p-n junction device temperature may then be measured by exciting four levels of applied collector current and sensing changes in output characteristics such as base-emitter voltage in which both parasitic series resistance and parallel leakage current errors are cancelled. Since the operating characteristics of a p-n junction device are temperature dependent, the excitation levels applied and output values observed may be used to determine the junction temperature of the device free of series and parallel parasitic effects.

[0025] The present invention is superior to prior multiple level excitation methods and apparatus in that four or more excitation levels are used to eliminate both series parasitic and parallel parasitic effects. The result is an M-Level excitation method, where M is greater than or equal to four, which produces a more accurate measurement of p-n junction devices than traditional three or four level excitation methods due to parallel parasitic effects. The prior art multiple level excitation methods only cancel series parasitic effects, such as series resistance and voltage offset values. Through the use of four or more excitation levels, the present invention determines the junction temperature of the device free of both series and parallel resistance effects, and voltage and current offset effects.

[0026] These and other objects, features and characteristics of the present invention will become more apparent to those skilled in the art from a study of the following detailed description in conjunction with the appended claims and drawings, all of which form a part of this specification. In the drawings:

[0027] FIG. 1 illustrates a diode circuit in which series and parallel resistance is modeled, as well as voltage and current offset;

[0028] FIG. 2 illustrates a simplified diode circuit in which only current offset is modeled;

[0029] FIG. 3 illustrates a simplified diode circuit in which only parallel resistance is modeled;

[0030] FIG. 4 illustrates a diode circuit in use in a test device in accordance with one embodiment of the present invention;

[0031] FIG. 5 illustrates typical results of the M-Level excitation method of the present invention compared with the prior art three and four level excitation methods; and

[0032] FIG. 6 is a flow chart of the current offset elimination method in accordance with one embodiment of the present invention.

[0033] The present invention provides an improved method and apparatus to measure p-n junction device temperature by testing a device with four or more levels of applied collector current I_C, sensing changes in output characteristics such as base-emitter voltage V_BE, calculating the device leakage current value and cancelling the parallel leakage effects. The calculated leakage current consists of both a constant leakage current, current offset, and an additional leakage current due to a parasitic parallel resistance. The p-n junction device temperature may then be measured by exciting four levels of applied collector current and sensing changes in output characteristics such as base-emitter voltage in which both parasitic series resistance and parallel leakage current errors are cancelled. Since the operating characteristics of a p-n junction device are temperature dependent, the excitation levels applied and output values observed may be used to determine the junction temperature of the device free of series and parallel parasitic effects.

[0034] In the preferred embodiment, a diode is shown as the p-n junction device and in this sense, acts as a traditional bipolar transistor with the base and collector connected. However, as pointed out, several parasitic resistance, voltage and current offset values will affect measurements and result in inaccuracies in the calculated temperature value. For instance, series resistance and voltage offset effects, considered series parasitic effects, introduce error into measured base-emitter voltage V_BEvalues as seen in the prior art examples. In addition, measurement errors also result from parallel resistance and current offset effects, considered parallel parasitic effects. The present invention, unlike the earlier prior art, cancels out parallel parasitic effects in addition to series parasitic effects. Therefore the present M-Level excitation method provides improved accuracy when measuring p-n junction device temperature, including diode devices, which respond poorly to the prior art methods.

[0035] The new method to determine diode junction temperature begins with the calculation of ΔV_BEin which the first component of parallel parasitic effects, current offset, is introduced for elimination.

[0036] $\Delta V_{\mathrm{BE}}={\mathrm{mV}}_T\times \ln \left(I_2/I_1\right)+R_s\left(I_2-I_1\right)+{\mathrm{mV}}_T\times \mathrm{ios}\left(1/I_2-1/I_1\right)$

[0037] In this equation ΔV_BErepresents the change in base-emitter voltage, I₁and I₂represent levels of applied excitation current, R_Srepresents series parasitic resistance and ios represents current offset. The current offset value ios has been added in the equation, representing ios error in ΔV_BEmeasurement. The parallel resistance R_Pvalue, representing the R_Perror in ΔV_BEmeasurement, is to be considered later. A four level excitation method results in the following base-emitter voltage ΔV_BEcalculation,

[0038] $\Delta V_{\mathrm{BE}}^1={\mathrm{mV}}_T\times \ln \left(I_3/I_1\right)+R_s\left(I_3-I_1\right)+{\mathrm{mV}}_T\times \mathrm{ios}\times A$ $\mathrm{where},A=\left[\left(1/I_3\right)-\left(1/I_1\right)\right]\Delta V_{\mathrm{BE}}^2={\mathrm{mV}}_T\times \ln \left(I_4/I_2\right)+R_s\left(I_4-I_2\right)+{\mathrm{mV}}_T\times \mathrm{ios}\times B$ $\mathrm{where},B=\left[\left(1/I_4\right)-\left(1/I_2\right)\right]$ $\Delta V_{\mathrm{BE}}=\Delta V_{\mathrm{BE}}^1-\Delta V_{\mathrm{BE}}^2\left(A/B\right)$

[0039] If currents are selected such that,

[0040] I₃−I₁−(A/B)(I₄−I₂)=I₃+I₂(A/B)−I₁−I₄(A/B)=0

[0041] then,

[0042] ΔV_BE=mV_T×ln(I₃/I₁)−mV_T(A/B)×ln(I₄/I₂)

[0043] The series parasitic resistance (R_S) and current offset ios values are cancelled. For example,

[0044] I=[2, 10 100, 20]μA

[0045] $\begin{array}{c}\Delta V_{\mathrm{BE}}^1=m{V}_T\times \ln \left(100/2\right)+R_S\left(100-2\right)+m{V}_T\times \\ \mathrm{ios}\left[\left(1/100\right)-\left(1/2\right)\right]\\ =m{V}_T\times \ln \left(100/2\right)+R_S\left(98\right)+m{V}_T\times \mathrm{ios}\left(-0.490\right)\end{array}$ $\begin{array}{c}\Delta V_{\mathrm{BE}}^2=m{V}_T\times \ln \left(20/10\right)+R_S\left(20-10\right)+m\mathrm{VT}\times \\ \mathrm{ios}\left[\left(1/20\right)-\left(1/10\right)\right]\\ =m{V}_T\times \ln \left(20/10\right)+R_S\left(10\right)+m{V}_T\times \mathrm{ios}\left(-0.05\right)\end{array}$ $\begin{array}{c}\Delta V_{\mathrm{BE}}=\Delta V_{\mathrm{BE}}^1-\Delta V_{\mathrm{BE}}^2\left(0.490/0.05\right)\\ =m{V}_T\times \ln \left(100/2\right)-m{V}_T\times 9.8\left[\ln \left(20/10\right)\right]+\\ R_S\left(98-9.8\times 10\right)+m{V}_T\times \mathrm{ios}\left(-0.490+0.05\times 9.8\right)\\ =m{V}_T\times \ln \left(100/2\right)-m{V}_T\times 9.8\ln \left(20/10\right)\end{array}$

[0046] The influences of series resistance R_Sand current offset ios have been cancelled. However, a comparison with the prior art four level excitation method base-emitter voltage ΔV_BEcalculation not considering current offset shows there is no common solution for both the new method and the prior art method unless the ΔV_BEof the measurement is zero.

[0047] $\begin{array}{c}\underset{\_}{\text{PRIOR ART}}\\ \Delta V_{\mathrm{BE}}=m{V}_T\times \ln \left[\left(I_3\times I_2\right)/\left(I_1\times I_4\right)\right]\end{array}\begin{array}{c}\underset{\_}{\text{M-LEVEL CONSIDERING CURRENT OFFSET}}\\ \Delta V_{\mathrm{BE}}=m{V}_T\times \ln \left(I_3/I_1\right)-m{V}_T\times 9.8\ln \left(I_4/I_2\right)\end{array}$

[0048] Adding an additional current level increases degrees of freedom by 1 allowing measurements to be optimized. If five levels of excitation are used, such that collector currents I₁-I₅are applied,

[0049] I=[I₁, I₅, I₂, I₃, I₄]

[0050] The temperature dependent output values ΔV_BEmay then be calculated as,

[0051] $\Delta V_{\mathrm{BE}}^1={\mathrm{mV}}_T\times \ln \left[\left(I_3\times I_2\right)/\left(I_1\times I_4\right)\right]+R_s\left(I_3+I_2-I_1-I_4\right)+{\mathrm{mV}}_T\times \mathrm{ios}\times A\mathrm{where},A=\left(1/I_3+1/I_2-1/I_1-1/I_4\right)$ $\Delta V_{\mathrm{BE}}^2={\mathrm{mV}}_T\times \ln \left[\left(I_5\times I_5\right)/\left(I_1\times I_2\right)\right]+R_s\left(I_5+I_5-I_1-I_2\right)+{\mathrm{mV}}_T\times \mathrm{ios}\times B\mathrm{where},B=\left(2/I_5-1/I_1-1/I_2\right)$ $\begin{array}{c}\Delta V_{\mathrm{BE}}=\Delta V_{\mathrm{BE}}^1-\Delta V_{\mathrm{BE}}^2\left(A/B\right)\\ =m{V}_T\times \ln \left[\left(I_3\times I_2\right)/\left(I_1\times I_4\right)\right]+R_S[I_3+I_2-I_1-\\ I_4-\left(A/B\right)\left(\left(2\times I_5\right)-I_1-I_2\right)]+m{V}_T\times \mathrm{ios}[A-(B\times \left(A/B\right)]\end{array}$

[0052] If currents are selected such that,

[0053] I₃+I₂−I₁−I₄−A/B((2×I₅)−I₁−I₂)=0

[0054] And if,

[0055] I₅={square root over (I

[0056] then,

[0057] ΔV_BE=mV_T×ln[(I₃×I₂)/(I₁×I₄)]

[0058] If the following five excitation levels are applied as collector currents,

[0059] I=[I₁, I₅, I₂, I₃, I₄]=[1, 5, 25, 50, 97.76]μA

[0060] The temperature dependent output values ΔV_BEare determined as, $\begin{array}{c}\Delta V_{\mathrm{BE}}^1=m{V}_T\times \ln \left[\left(50\times 25\right)/\left(1\times 97.76\right)\right]+R_S(50+25-1-\\ 97.76)+m{V}_T\times \mathrm{ios}\left(1/50+1/25-1/1-1/97.76\right)\\ =m{V}_T\times \ln \left(12.787\right)+R_S\left(-23.70\right)+m{V}_T\times \mathrm{ios}\left(-0.9502\right)\end{array}$ $\Delta V_{\mathrm{BE}}^2=m{V}_T\times \ln \left[5\times 5/\left(1\times 25\right)\right]+R_S\left(-16\right)+m{V}_T\times \mathrm{ios}\left(-0.6400\right)$ $\begin{array}{c}\Delta V_{\mathrm{BE}}=\Delta V_{\mathrm{BE}}^1-\Delta V_{\mathrm{BE}}^2\left(1.485\right)\\ =m{V}_T\times \ln \left(12.787\right)\end{array}$

[0061] FIG. 1 introduces the modeled parasitic effects eliminated through the use of the M-Level temperature measurement technique. In FIG. 1, an illustrative circuit 100 is shown which includes series parasitic resistance 106, series voltage offset 108, parallel parasitic resistance 110, parallel current offset 112 and the affected device, diode 114. The diode 114, including a cathode 118 and an anode 116, acts as a traditional bipolar transistor with the base and collector connected. The anode 116 of diode 114 is electrically coupled to the resistance 106 via the voltage offset 108 completing the series effects, and the current offset 112 and the parallel resistance 110 are each electrically coupled in parallel to the diode 114 completing the parallel effects.

[0062] FIG. 2 introduces the simplified model for the elimination of current offset effects in temperature measurement through the M-Level temperature measurement technique. FIG. 2 shows a simplified circuit 200 with current offset only and includes a diode 206 and current offset 204. The diode 206 includes a cathode 210 and an anode 208, and the current offset 204 is electrically coupled in parallel. In FIG. 2, if a prior art four levels of excitation are used such that collector currents I₁-I₄do not meet current ios cancelling conditions previously discussed, ios is not cancelled, and we can see the effects of ios to the Vbe measurements. As per prior art, apply a four level current to the forward biased diode 206.

[0063] I=[I₁, I₂, I₃, I₄]=[2, 10, 92, 100]μA

[0064] The temperature dependent output values ΔV_BE(voltage across anode 208 and cathode 210) for each excitation level are,

[0065] V_BE=[V_BE¹, V_BE², V_BE³, V_BE⁴]

[0066] ΔV_BE^3,1=V_BE³−V_BE¹

[0067] ΔV_BE^3,1=mV_T×ln[(92+ios)/(2+ios)]  equation 1

[0068] ΔV_BE^4,2=V_BE⁴−V_BE²

[0069] ΔV_BE^4,2=mV_T×ln[(100+ios)/(10+ios)]  equation 2

[0070] ΔV_BE=ΔV_BE^3,1−ΔV_BE^4,2

[0071] ΔV_BE=mV_T[ln((92+ios)/(2+ios))−ln((100+ios)/(10+ios))]

[0072] ΔV_BE=mV_T[ln(92/2)×((1+ios/92)/(1+ios/2))−ln(100(1+ios/100)/10(1+ios/10))]

[0073] For current offset 204 (ios) in the range of tens of nA,

[0074] 1>>ios/2>ios/10>ios/92>ios/100

[0075] The temperature dependent output value ΔV_BE(voltage across anode 208 and cathode 210) may then be measured as,

[0076] ΔV_BE=mV_T[ln(92/2)−ln(100/10)]+mV_T[ios/92−ios/2−ios/100+ios/10]

[0077] V_BE=mV_Tln[(92×10)/(2×100)]+mV_T[−ios/2+ios/10+ios/92−ios/100]

[0078] ΔV_BE=ΔV_BE⁰+dΔV_BE  equation 3

[0079] where,

[0080] ΔV_BE⁰=mV_T×ln[(92×10)/(2×100)]  equation 3a

[0081] dΔV_BE=−mV_T(ios/2−ios/10−ios/92+ios/100)  equation 3b

[0082] The calculated base-emitter voltage ΔV_BE⁰is the pure measurement without the current offset ios effects. The measurement error due to current offset 204 (ios) is dΔV_BE. The current offset 204 (ios) only influences the ΔV_BEmeasurement significantly at low currents, e.g. 2 and 10 μA. For higher currents, the influence can be ignored, e.g. ios/92 or ios/100. This is a very important concept to the extent that this method cancels out parallel resistance R_Peffects as well. As will be seen in the following equations, the parallel resistance R_Pacts like a constant current offset 204 when four levels of excitation are used.

[0083] If,

[0084] ΔV_BE=ΔV_BE⁰+dΔV_BE

[0085] ΔV_BE=mV_T×ln[(92×10)/(2×100)]−mV_T(ios/2−ios/10)

[0086] ΔV_BE=mV_T×ln[(92×10)/(2×100)]−mV_T[(4×ios)/10]  equation 4

[0087] If for example, there is a leakage current offset ios=15 nA, using the prior art method, $\begin{array}{c}\Delta V_{\mathrm{BE}}=m{V}_T\times \ln \left[\left(92\times 10\right)/\left(2\times 100\right)\right]-\\ m{V}_T\left[\left(4\times -0.015\mathrm{\mu A}\right)/10\mathrm{\mu A}\right]\\ =m{V}_T\times 1.526-m{V}_T\times 0.006\end{array}$  ΔT=303(K)×(0.006/1.526)=1.2° C.

[0088] A temperature of 1.2° C. higher is read. If we add two more levels of excitation to the device, I₅,I₆=(4, 8) μA, then,

[0089] dV_BE¹=mV_T×ln(4/2)+mV_T[(ios/4)−(ios/2)]

[0090] dV_BE¹=mV_T×ln(4/2)+mV_T(−ios/4)  equation 5

[0091] dV_BE²=mV_T×ln(8/4)+mV_T[(ios/8)−(ios/4)]  equation 6

[0092] dV_BE²=mV_T×ln(8/4)+mV_T(−ios/8)

[0093] ΔdV_BE=dV_BE¹−dV_BE²

[0094] ΔdV_BE=mV_T(−ios/4)+mV_T(ios/8)

[0095] ΔdV_BE=mV_T(−ios/8)  equation 7

[0096] Recall that equation 4 produced,

[0097] dΔV_BE=−mV_T[(4×ios)/10]

[0098] dΔV_BE=−mV_T[(ios/8)×((8×4)/10)]

[0099] dΔV_BE=3.2ΔdV_BE  equation 8

[0100] Therefore, if collector currents I₁,I₅,I₆=(2, 4, 8) μA are used to measure ΔdV_BEand equation 8 is used to cancel out the measurement error dΔV_BEin equation 4, we obtain the true ΔV_BEmeasurement without the influence of current offset 204 (ios). A simplified flowchart in FIG. 6 illustrates the steps taken.in the new method of current offset effect elimination.

[0101] In FIG. 6, two Parts are shown, Part 1 in which the current offset error is determined, and Part 2 in which the calculated current offset error is eliminated from the temperature determination. In Part 1 of FIG. 6, the excitation levels shown at 602 and described as I₁-I₆, are applied to the device 618 at 604, resulting in sequential output signals (V_BE) 606. The sequential output signals are used to calculate the current offset error dΔV_BEat 608.

[0102] In Part 2 of FIG. 6, excitation levels shown at 610 and described as I₁-I₄, are applied to the device 618 at 612, resulting once again in sequential output signals 614. These excitation levels shown at 610 are not required to be four new additional excitation levels. The excitation levels may be I₁-I₄of the M-Level excitation levels shown at 602 and included in I₁-I₆. In Part 2 however, the measurement error value dΔV_BEcalculated in Part 1 is cancelled from ΔV_BEcalculations. The temperature calculation in 616 is then determined from a comparison of the sequential output signals 614 and the sequential test signals 610, free of current offset effects.

[0103] We may extend the concept for parallel resistance R_P, which creates an additional leakage current dependent on Vbe. FIG. 3 introduces the simplified model of the elimination of parallel parasitic resistance R_Pin temperature measurement through a similar use of the M-Level temperature measurement technique. FIG. 3 shows a simplified circuit 300 with parallel parasitic resistance only and includes a diode 304 and resistance 302 which is electrically coupled in parallel.

[0104] Define a leakage current through R_pthe parallel resistance 302 as ilk,

[0105] I_lk=V_BE/R_p

[0106] Recalling equation 3b, the leakage current influences the ΔV_BEby dΔV_BEand thus influences the temperature measurement.

[0107] dΔV_BE=m×V_T×Σ(i_lk/I)

[0108] dΔV_BE=m×(V_T/R_P)×Σ(V_BE/I)

[0109] As the excitation current I increases, V_BEincreases logarithmically. The contribution of dΔV_BEsharply drops at large levels of excitation current. The largest contribution of dΔV_BEis at low excitation current levels. Because the V_BEdifferences are small compared to the individual V_BEmeasurements, ilk can be treated as a constant leakage current offset without significantly impacting accuracy. Take a real measurement as an example. In FIG. 3, if a M-Level excitation method is used such that collector currents I₁-I₆are applied to the forward biased diode 304,

[0110] I=[I₁, I₂, I₃, I₄, I₅, I₆]=[2, 4, 8, 10, 92, 100]uA

[0111] The temperature dependent output values V_BE(voltage across anode 306 and cathode 308) for each excitation level are,

[0112] V_BE=[557.707, 576.4766, 595.2266, 601.2661, 662.1276, 664.4616]mV

[0113] V_BE/557.707=[1, 1.0337, 1.0673, 1.0781, 1.1872, 1.1914]

[0114] This data shows that an increase in excitation current by a factor of 50 causes V_BEto increase by a factor of only 1.1914.

[0115] The contribution of dΔV_BEwithout treating the parallel leakage current offset as a constant is,

[0116] I=[2, 10, 92, 100]μA

[0117] V_BE=[557.71, 601.27, 662.13, 664.46]mV

[0118] i_lk=V_BE/R_P

[0119] dΔV_BE¹=m×V_T×[(i_lk1/I₁)−(i_lk2/I₂)−(i_lk3/I₃)+(i_lk4/I₄)]

[0120] dΔV_BE¹=m×V_T/R_P×[(V_BE¹/I₁)−(V_BE²/I₂)−(V_BE³/I₃)+(V_BE⁴/I₄)]

[0121] dΔV_BE¹=m×V_T/R_P×218.17 kΩ

[0122] The contribution of dΔV_BEif the parallel leakage current offset is assumed to be constant is:

[0123] Assume, i_lk=V_BE¹/R_Pfor all levels of excitation.

[0124] dΔV_BE²=m×V_T×i_lk×[(1/I₁)−(1/I₂)−(1/I₃)+(1/I₄)]

[0125] dΔV_BE²=m×V_T/R_P×V_BE¹×[(1/I₁)−(1/I₂)−(1/I₃)+(1/I₄)]

[0126] dΔV_BE²=m×V_T/R_P×222.60 kΩ

[0127] This result shows that dΔV_BE²is only 2% different from dΔV_BE¹. Therefore, treating the parallel leakage current as a constant current offset cancels on the order of 98% of the ilk error. In an earlier discussion we found that a 15 nA current offset would contribute 1.2° C. of error to a temperature measurement. Thus, using the M-Level method to compensate for parasitic parallel resistance leaves a residual error of only about 0.03° C.

[0128] Since both the parasitic series influences and parallel influences are cancelled, the temperature measurement derived from base-emitter voltage V_BEvalues does not include the inaccuracies of the prior art models. The new M-Level excitation method results in a more accurate measurement. In FIG. 4, a device is shown in diagram 400 to illustrate these test results. A Multiple Level current source is provided at 404, electrically coupled to the anode 416 of diode 414 via pin 422. By pulling pin 420 of the test device 406 below ground and conducting a current 410, a leakage current is created in diode 414 while simultaneously measuring V_BEvalues. As can be seen in FIG. 5, at a constant device temperature, if current value 410, Irext (FIG. 5), increases creating larger and larger leakage currents, the prior art four level excitation method results in larger and larger inaccuracies in temperature measurements. The new multiple level excitation method maintains an accurate reading, illustrating the corrective effects of the new method.