Remote Thermal Sensing Diode Selection Guide
Introduction
This training is intended for designers who build systems that use thermal sensors with remote diodes; specifically, remote diodes that are discrete Bipolar Junction Transistors (BJTs).
The information is organized based on important criteria for selecting the remote sensing diode to use with Microchip's high accuracy, low-cost remote diode thermal sensors.
Microchip does produce temperature sensors that are designed to work specifically with CPU thermal diodes. Therefore, these discussions are about selecting an appropriate BJT as well as providing a list of acceptable BJTs, several are mentioned.
Throughout this training, the phrase “remote diode-connected transistor” refers to a discrete, diode-connected (Base-Collector junction shorted) BJT. Please ensure that you possess a working knowledge of temperature sensing that uses diode-connected transistors before proceeding.
Overview
This is a practical approach for selecting a remote diode-connected transistor to use with a thermal sensor, as illustrated in Figure 1.
Discussions of the semiconductor parameters of the transistor that affect the accuracy of temperature measurement are included here as the requisite feature of a remote thermal sensing diode.
A short table of qualified discrete 2N3904 Negative Positive Negative (NPN) transistors is provided here. It lists devices from other manufacturers that have been tested and met established standards of accuracy.
Diode Parameters
These three semiconductor parameters are the primary factors when considering diode-connected transistors in temperature sensing applications.
- Ideality Factor (η)
- Forward Current Gain (beta or hFE)
- Series Resistance (RS)
Ideality Factor (η)
The ideality factor is a parameter in the diode current-voltage relationship. It approaches a value of 1.0 when the carrier diffusion dominates the current flow and approaches 2.0 when the recombination current dominates the current flow. This term is a constant on any particular device, though it can vary among individual devices.
Temperature sensors are calibrated during the final test to provide accurate readings with a diode that has a typical ideality factor. For the purposes of this document, the typical ideality factor value is expressed as ηASSUMED and the ideality factor value of the user’s diode-connected transistor is expressed as ηREAL.
The temperature indicated by a temperature sensor will include an error from the real temperature, as defined by the equation in Equation 1. To use this equation, the temperature values must be converted to the Kelvin scale. The result will be incorrect if the values used to reflect are in Celsius or Fahrenheit scale.
Generally, a 2N3904 transistor is the preferred remote diode. Several samples of each of the transistors listed in Table 1 were evaluated and their ideality factor was determined to be ~1.004. Typically, the ideality factor is not stated in the datasheet for a transistor. While transistor devices (other than the ones cited here) could be used, to be confident of accurate operation, they should be qualified before use.
Qualification of these devices is performed by obtaining data, on the parameters described in this training, from the device manufacturer. Precision thermal equipment is required to measure the parameters. Contact your Microchip Field Applications Engineer for additional support.
Table 1: Typical Ideality Factor Values for 2N3904 Diode Connected Transistors
Manufacturer | Typical Ideality Factor |
---|---|
ROHM Semiconductor | 1.0038 |
Diodes® Incorporated | 1.0044 |
NXP® | 1.0049 |
STMicroelectronics | 1.0045 |
ON Semiconductor® | 1.0046 |
Chenmko CO., LTD. | 1.0040 |
Infineon® Technologies AG | 1.0044 |
Fairchild Semiconductor® | 1.0046 |
National Semiconductor | 1.0037 |
In Equation 1, the ideality factor value that the temperature sensor is calibrated for is ηASSUMED and the actual ideality factor value of the diode-connected transistor is ηREAL. In this equation, the temperature measurement error is not a constant offset but increases as TREAL, the temperature of the remote diode-connected transistor increases.
Figure 3 shows the temperature-measurement error that is induced solely from the differences between ηASSUMED and ηREAL. In this figure, ηASSUMED is 1.004, a typical ideality factor value for a 2N3904 NPN diode-connected transistor. Temperature sensors are typically calibrated in the range of the 2N3904 (1.004) because this is also very similar to the ideality factor of the majority of substrate diode-connected transistors that are found on CPUs and GPUs.
Figure 3 also shows why true two-terminal discrete diodes are not used in temperature sensing applications instead of three-terminal devices such as the 2N3904. Generally, a discrete two-terminal diode would perform equally well as a thermal diode in temperature sensing applications. However, characterization in the labs determined that discrete two-terminal diodes typically have an ideality factor much higher (1.2–1.5) than ηASSUMED of 1.004. This discrepancy between ηASSUMED and ηREAL would cause unacceptable temperature measurement errors at all temperatures.
Forward Current Gain (beta or hFE)
A typical temperature sensor forces two fixed currents (IF1 and IF2) into the thermal diode to measure temperature, as shown in Figure 4.
The temperature sensor measures the voltage, VBE, which is developed based on the collector current, not the emitter current.
The forward current gain (beta) of a transistor is not a constant over all operating conditions. It varies over temperature and as a function of IC. The variation in beta over temperature does not induce temperature measurement error. However, if the transistor has a large variation in beta as a function of IC, the temperature reading can be inaccurate due to beta-induced error.
If the value of beta is relatively constant over the range of forced emitter currents, then the ratio of IC2:IC1 remains equal to the ratio of the two forced emitter currents and induces no error. It only becomes a problem when the beta variation causes a mismatch between the IC2:IC1 ratio and the IE2:IE1 ratio.
Equation 3 shows the error induced from the non-constant value of beta at the two currents. βF1 represents the beta of the transistor at the current value IF1 while βF2 represents the beta at the current value IF2. N represents the fixed ratio of the two forced (IE1 and IE2) currents. If beta is constant over the range of the two currents (βF1 = βF2), then there is no temperature measurement error induced because of beta variation.
Figure 6 presents a plot of allowable beta variation over the sensor’s sourced current range (10 – 400 μA) to be able to still maintain at least one-degree accuracy at 70°C. The beta of the transistor must reside between the two lines in the plot, over the extremes of the current range of the temperature sensor, in order to maintain 1°C accuracy with the selected diode-connected transistor. The x-axis represents the beta of the diode-connected transistor at IF1 while the y-axis is for the beta at IF2. It varies over the sensor’s sourced current range.
Figure 7 shows typical values of transistor beta for a limited sample of these devices. These devices were characterized in Microchip characterization labs. This data should not be used as a guaranteed value for the specific transistor, only a typical representation for the limited quantity tested by Microchip.
The conclusion to draw from the previous figures and Table 2 is that for the set of 2N3904 transistors tested by Microchip, the beta was consistently high and flat. The measured value of beta easily resides inside the two lines of the above figure, over the entire temperature sensor’s sourced current range.
Table 2 quantifies the error induced from beta variation using the 2N3904s that were tested. As demonstrated through the tested devices, beta variation has a very small effect on temperature measurement accuracy.
Table 2: Temperature Error Due to 2N3904 Beta Variation at 70 °C
Manufacturer | Temperature Error (°C) |
---|---|
ROHM Semiconductor | +0.07 |
Diodes Incorporated | +0.00 |
NXP | +0.04 |
STMicroelectronics | +0.03 |
ON Semiconductor | +0.01 |
Chenmko CO., LTD. | +0.15 |
Infineon Technologies AG | +0.03 |
Fairchild Semiconductor | +0.00 |
National Semiconductor | +0.00 |
Series Resistance (RS)
Series Resistance (RS) is another parameter that affects temperature measurement accuracy. It causes the temperature sensor to report the temperature higher than the actual temperature of the thermal diode. The relationship between temperature offset and RS is displayed in the equation shown in Figure 8.
The temperature error induced by RS is a constant offset for all temperatures. When using a typical Microchip temperature sensor, the magnitudes of IF2 and IF1 induce approximately +0.67°C error per Ω of series resistance. For different 2N3904 devices characterized by Microchip, the RS was found to be less than 1Ω. This does not include the RS due to PCB traces connecting the sensor and remote diode; this only represents the RS found in the characterized 2N3904 devices.
Table 3 quantifies some typical values of RS found for a sample of different 2N3904 devices. This value of RS for the set of 2N3904s tested was found to have a positive temperature coefficient and typically increased by approximately 5% per +10°C increase.
- Table 3 should not be used as a guideline for offsetting the temperature reported by a Microchip temperature sensor. Microchip temperature sensors are typically calibrated using a 2N3904 diode-connected transistor which already compensates for this RS error term.
- Table 3 is a reference to help thermal designers understand the possible effects of non-idealities in temperature measurement.
Table 3: Typical Values of Series Resistance for 2N3904 Diode Connected Transistors
Manufacturer | Series Resistance (RS) @70°C |
---|---|
ROHM Semiconductor | 0.68 |
Diodes Incorporated | 0.65 |
NXP | 0.72 |
STMicroelectronics | 0.58 |
ON Semiconductor | 0.90 |
Chenmko CO., LTD. | 0.73 |
Infineon Technologies AG | 0.57 |
Fairchild Semiconductor | 0.60 |
National Semiconductor | 0.51 |
Tested Diode List
Table 4 lists a limited selection of 2N3904 NPN transistors that have been characterized as meeting the specifications to obtain 1°C accurate measurements.
Table 4: Tested Diodes for Temperature Sensing Applications
Manufacturer | Model Number |
---|---|
ROHM Semiconductor | UMT3904 |
Diodes Incorporated | MMBT3904-7 |
NXP | MMBT3904 |
STMicroelectronics | MMBT3904 |
ON Semiconductor | MMBT3904LT1 |
Chenmko CO., LTD. | MMBT3904 |
Infineon Technologies AG | SMBT3904E6327 |
Fairchild Semiconductor | MMBT3904FSCT |
National Semiconductor | MMBT3904N623 |
Conclusion
In conclusion, while differences were seen between the various manufacturer’s versions of 2N3904 BJTs, the results when using them with Microchip temperature sensors were very consistent. For all typical 2N3904 devices tested, the temperature never varied more than ±0.2°C from the true temperature. The 2N3904 devices listed in Table 4 (or any BJT/diode with equivalent parameters) will yield accurate temperature measurement results when used with Microchip temperature sensors. Microchip supplies a family of temperature sensors for many applications. Several special functions, such as resistance error correction and ideality configuration are available. In addition, some devices are designed to work specifically with CPU thermal diodes. Please consult your Microchip representative or visit the Microchip website for additional information.