While the previously discussed techniques help produce an initial Printed Circuit Board (PCB) layout, it is important to verify your design functions as specified. This section includes methods for measuring the response of individual components and of a PCB. With this information, it is possible to make intelligent design tweaks.
There are many ways to measure temperature. We could use thermocouples, RTDs, thermistors, diodes, ICs, or thermal imagers (infrared cameras) to measure the temperature.
The figure below shows a circuit based on the MCP9700 IC temperature sensor. Because all of the components draw very little current, their effect on PCB temperature will be minimal. There is enough filtering and gain to make VOUT easy to interpret. This circuit can be built on a very small board of its own, which can be easily placed on top of the PCB of interest.
- AN990, "Analog Sensor Conditioning Circuits – An Overview" DS00990A
- AN684, "Single Supply Temperature Sensing with Thermocouples" DS00684A
- AN679, "Temperature Sensing Technologies" DS00679
- MCP9700/MCP9700A/MCP9701/MCP9701A, "Low-Power Linear Active Thermistor ICs"
The MCP9700 outputs a voltage of about 500 mV plus 10.0 mV/°C times the board temperature (TPCB, in °C). The amplifier provides a gain of 10 V/V centered on 500 mV (when VDD = 5.0 V), giving:
Since the MCP9700 outputs a voltage proportional to temperature, VOUT needs to be sampled by an Analog-to-Digital Converter (ADC) that uses an absolute voltage reference. The absolute accuracy of this circuit does not support our application, so it is important to calibrate the errors. Leave the PCB in a powered-off state (except for the temperature sensor) for several minutes. Measure VOUT at each point, with adequate averaging. The changes in VOUT from the calibration value represents the change in TPCB from the no power condition.
To measure thermal gradients, simply measure the temperature at several points on the PCB. The gradient is then the change in temperature divided by the distance between points. More points give better resolution on the gradient, but reduce the accuracy of the numerical derivative.
Package Thermal Resistance
The way to estimate the temperature (T in °C) of a component is to multiply its dissipated power (P in W) by the package thermal resistance (𝜃JA in °C/W). This helps establish temperature maximum points.
To measure 𝜃JA, when it is not given in a datasheet, place the temperature sensor at the IC (usually, a thermocouple between the package and the PCB). Insert a small resistor in the supply to measure the supply current when it is on (IDD in A). Measure the change in temperature (∆T in °C) between the off and on conditions, supply voltage (VDD in V) and IDD. Then,
The easiest way to measure thermoelectric voltages is to thermally imbalance a difference amplifier circuit. The thermoelectric voltages have a polarity that adds (instead of canceling), as shown in the figure below. The differential input voltage is zero and the resistors are larger to emphasize the thermoelectric voltages. The large resistor on the right of the layout can generate heat, causing a horizontal temperature gradient at the resistors RG and RF. The gain (G) is set high to make the measurements more accurate. The thermoelectric voltage (VTHx) across one resistor is:
We can also place a short across one component, of a matched pair, with a copper trace on the PCB. The figure below shows a non-inverting amplifier layout that shorts RN (with a copper trace) to unbalance the thermoelectric voltages. It also connects the two inputs together and uses larger resistors to simplify measurements (VOUT = VDD/2, ideally). The short is easily removed from the PCB.
With the unbalance, we now have the thermoelectric voltage:
Troubleshooting Tips and Tricks
Using a strip chart to track the change in critical DC voltages over time helps locate the physical source of the errors. It can show how large the change is between two different thermal conditions (e.g., on and off) and it also shows the time constants of these shifts. They can be roughly divided into the following three categories:
- Time constant « 1 s, within the component (e.g., thermal crosstalk within an op-amp)
- Time constant ≈ 1 s, single component (e.g., in an eight-lead SOIC package)
- Time constant » 1 s, PCB and its environment
To quickly and easily change the temperature at one location on a PCB, do the following. Use a clean drinking straw to blow air at the location (component) of interest. Use a piece of paper to re-direct the airflow away from other nearby components. When troubleshooting, the paper can be used to divide a PCB area in half to help locate the problematic component. This approach does not give exact numbers, but can be used to quickly find problem components on a PCB.
You can use a heat sink (with an electrically insulating heat sink compound) to reduce the temperature difference between two critical points on your PCB. The greater the area covered at both ends of the heat sink, the quicker and better this thermal short will work.