How to Build a Temperature Sensing System Using Silego GreenPAK

In this project, we will test how to build a temperature sensing system using the Silego GreenPAK.

The purpose of this lab is to test the usability of GreenPAK devices in temperature sensing applications, specifically for CPU processors with a temperature domain of 25°C to 100°C. The temperature sensing system uses a remote temperature sensor 2N3904 (diode connected) and control logic implemented in the GreenPAK device. The goal is to achieve an internal temperature accuracy of +/- 1°C using the SLG46537 chip.

Figure 1. Temperature sensor application circuit implementing dual current method

theory

Two-stream method

The dual current method is a voltage subtraction method that uses two current sources, a diode, capacitors, and FET switches (see Figure 1). It aims to amplify the voltage difference across the diode caused by two different currents I_L and I_H.

The forward voltage of a diode varies based on the electric field induced by the current across the diode. By subtracting the forward voltage at two known currents, we can eliminate common variables such as offset and saturation current. The forward bias current ID of a diode is related to the saturation current Is and the forward voltage VD by the following formula:

ID= IseVD/ηVT

Where η is the ideality factor of the diode and VT = kT/q. k is the Boltzmann constant and q is the electron charge constant. For this example, I_H will be the bias current I1 and I_L will be the bias current I2. Therefore, their forward bias current equations are:

I1= IseV1/ηVT

I2= IseV2/ηVT

By dividing the two forward bias current equations, we eliminate the saturation current and reduce the input current variable to a scalar factor N.

I1⁄I2= N = e(V1-V2)/ηVT

If N is known, then by measuring the difference between V1 and V2 we can calculate the temperature T.

ln(N) = (V1-V2)q/(ηkT)

T = (V1-V2)q/(ηkln(N))

Ideality factor η

While k and q are constants and do not vary from part to part, the ideality factor η does vary between values ​​of 1 and 2. The closer this value is to 1, the more carrier diffusion dominates. The closer this value is to 2, the more recombination is involved. The higher this value, the more errors there are in temperature measurements, since temperature highly influences recombination rather than diffusion. [1]

When choosing a remote temperature diode, remote diode connected transistors (BJTs with the base-collector junction shorted) are better choices than rectifier diodes because their ideal values ​​are specified. Almost all BJT transistors have an ideality factor close to 1.

Other important parameters include the forward current gain β and the series resistance Rs. The forward current gain varies with temperature and collector current, and the series resistance shows a constant shift at all temperatures. It is recommended to choose a device whose β varies between the two currents I_H and I_L.

Experimental Circuit

The purpose of the experiment was to measure the general temperature deviation by testing several points: 40°C, 60°C and 80°C.

Figure 2 shows the block diagram of the application circuit used in the experiments.

Figure 2. GPAK block diagram for temperature sensing application circuit

Resistors R3 and R4 provide the currents I_H and I_L in Figure 1. The switches NMOS and PMOS are located inside the GreenPAK (PIN 13 and PIN 15). The internal GreenPAK design is shown in Figure 3. The update time of this sensor is 10ms.

Figure 3. Internal GPAK design

When NMOS is pulled low and PMOS is floating, capacitor C4 is charged by I_L. When PMOS is pulled high and NMOS is floating, C4 is charged by I_H. The control logic of the system is programmed to provide dead time (between NMOS and PMOS on time) to obtain the voltage difference on C4.

This difference goes into the IN+ input of the op amp (Silego's SLG88103). The voltage waveform across C4 is shown in Figure 4.

Figure 4. Voltage waveform across C4

Pin 7 of the GPAK device is connected to a voltage divider with a variable resistor (trimmer) that adjusts the switching voltage of the analog comparator ACMP0 in the internal GPAK design. The switching voltage is the moment when the voltage on PIN 7 reaches the voltage on PIN 6 (op amp output voltage). The output voltage of the op amp (and therefore the switching voltage of ACMP0) is different at different temperatures.

Therefore, in the experiment, the resistance of P1 is adjusted until the switching moment occurs. Based on the measured resistance of potentiometer P1, the switching voltage (Vref) of ACMP0 can be calculated.

result

The GPAK-induced error (temperature measurement accuracy) was measured by testing the output of multiple devices (three GPAK devices). The test results of the three sensor circuits with only one different component (GPAK5 device) are recorded in Tables 1, 2, and 3, where:

The temperature inside the oven is measured using thermocouples.

The resistance of P1 was measured using a Fluke multimeter.

The Vref value is calculated based on the resistance value of P1.

Table 1. GPAK5_1

Table 2. GPAK5_2

Table 3. GPAK5_3

The results are further summarized using the graph in Figure 2 .

Figure 2. Graph of measurements from three sensors using different GPAK devices

Figure 2 shows that sensors 2 and 3 give overlapping linear plots, while sensor 1 gives a line that is very close to the linear trend line presented by sensor 1.

The slope of the curves for sensors 2 and 3, and the trend line for sensor 1, is 1.7mV/1°C between T=40°C and T=80°C. The maximum difference between the measurements of the three GPAK devices is 2mV. This means that the GPAK introduces a part-to-part error of approximately 1°C.

Error Source

Analog Comparator

The GreenPAK design shows that part-to-part variation is most likely to occur at the Analog Comparator component (ACMP0). The ACMP0 settings in this application design are as follows:

IN+ Source: pin6

IN-: extension. Reference voltage (pin7)

IN+Gain=1

Hysteresis: Disabled.

The ACMP's offset voltage (and thus the switching voltage) varies with temperature and supply voltage. The experimental conclusions are based on measurements of three GPAK samples - too few for statistical analysis.

As a typical representation, we can use Silego's ACMP Offset Voltage char data, which is based on a set of 35 components tested at temperature and voltage. The component ACMP offset results measured at room temperature with Ext.Vref=600mV and 5V supply voltage are shown in Table 4.

Table 4

Table 4. ACMP offset test at Ext.Vref=600mV_AutoPWR, 35 components, room temperature

Silego's Voffset calculation is based on the following equation:

VOffset = Maximum(|Vref,Ext-Vih| , |Vref,Ext-Vil|)

Data accuracy +/- 0.2mV

Test results based on 35 components show that the ACMP offset can be as high as 4.366 mV, which can introduce up to 2.5°C of error.

2N3904 Components

The transistor parameters that affect temperature measurement accuracy (remote accuracy) are the forward current gain β(I) and the series resistance Rs. According to Microchip Labs data for the 2N3904, these parameters have little effect on temperature measurement accuracy over the entire sensor source current range (4.5-920uA). [1]

Figure 3 shows typical β values ​​for transistors tested by Microchip and is representative of the limited number of 2N3904 transistors available. [1] Equation 3 gives the temperature error due to the shown β variation of approximately 0.02°C at 80°C. Similarly, using the Rs data for Microchip's 2N3904 group (approximately 0.7 ohms) and Equation 4 would yield a temperature error since Rs is approximately 0.8°C. [1]

Figure 3. Typical beta values ​​for a 2N3904 transistor at 23°C (Microchip Labs data)

Operational Amplifier Considerations

The SLG88103 voltage offset is typically 0.35mV (up to 2.4mV for VCM close to VSS (ground)), plus typically 0.16mV (T=80°C) to add the offset drift with temperature. The typical values ​​introduce a small error from the op amp for this application. However, considering the maximum offset value, the SLG88103 op amp may introduce an error of more than 1°C.

in conclusion

The output gain of the temperature sensing system used in the experiment is 1.7 (mV/°C).

Test results based on three samples of GreenPAK devices indicate that implementing GreenPAK alone in a temperature sensing system introduces approximately 1°C of error.

The key characteristics of the sensor are based on experimental results where the temperature domain is applied to the remote sensor while the internal temperature of the system with the GreenPAK is maintained at room temperature during the experiments. To view the application note and design files for this project, click here.

GreenPAK Key Features

Temperature range (25°C - 100°C)

accuracy

±1°C remote temperature accuracy (accuracy of 2N3904 sensor)

±1.5°C Internal Temperature Accuracy

1.7V-5.5V supply voltage

10ms update time

1.7 (mV/°C) output gain

29μA Quiescent Current