By Jeff Miller
Product Marketing Manager
Internet of Things (IoT) designs mesh together several design domains in order to successfully develop a product that interfaces real-world activity to the internet. Individually, these design domains are challenging for today’s engineers. Bringing them all together to create an IoT product can place extreme pressure on design teams. Figure 1 shows the elements of a typical IoT device.
This IoT device contains a sensor and an actuator that interface to the Internet. The sensor signal is sent to an analog signal processing device in the form of an amplifier or a low-pass filter. The output connects to an A/D converter to digitize the signal.
That signal is sent to a digital logic block that contains a microcontroller or a microprocessor. Conversely, the actuator is controlled by an analog driver through a D/A converter. The sensor telemetry is sent, and control signals are received, by a radio module that uses a standard protocol, such as WiFi, Bluetooth, or ZigBee, or a custom protocol. The radio transmits data to the Cloud or through a smartphone or PC.
Each of these major IoT functional blocks can be assembled from off-the-shelf, discreet components. However, there is strong pressure to converge the component from Figure 1 into a smaller number of individual packaged devices.
Convergence improves the cost, size, performance, and power consumption of the IoT device. By creating a multifunctional chip, the part count can be reduced and design integration can be improved. Figure 2 shows two examples of convergence. A radio chip company adds a microcontroller and the A/D and D/A converter. A sensor company adds the analog signal processing and A/D converter.
The IoT Design Challenge
Convergence is the first clue to the fundamental challenge of IoT design. But let’s dig deeper by looking at a hypothetical IoT design for discussion purposes (Figure 3).
One tire pressure monitoring device is embedded in each of the tires of a race car. The tire pressure values are sent to an in-car base station that then sends the data to the Cloud. This data is available for the racing team to monitor. If the pressure is getting too low, the team is alerted and the driver is instructed to make a pit stop.
A MEMS pressure sensor constantly measures the air pressure for the tire. The analog signal from this sensor is amplified and converted to a digital signal. A digital interface sends the signal to the microcontroller for processing, which, in turn, sends the data to the radio. The in-car base station receives that data from the radio, and then uploads it to the Cloud. The racing team’s software interprets the data stream and presents a read-out of the tire pressure. A battery charges a supercapacitor that powers the microcontroller and the radio.
The tire pressure design points out the fundamental challenge to IoT design: the four design domains that Figure 4 shows all live together in the IoT device.
While design convergence can involve two or more design domains, the IoT challenge is even greater. IoT design requires that all four design domains are designed and work together, especially if they are going on the same die. Even if the components are targeting separate dies that will be bonded together, they still need to work together during the layout and verification process. In the tire pressure design, the A/D and amplifier are analog, the digital interface and microcontroller are digital, the radio is RF, and the pressure sensor is a MEMS device. The design team needs to capture a mixed analog and digital, RF, and MEMS design, lay out the chip, and perform both component and top-level simulation.
The Tanner Solution
Tanner provides a single, top-down design flow for IoT design, unifying the four design domains. Whether you are designing a single die or multiple die IoT device, you can use this design flow for creating and simulating this device:
- Capturing and simulating the design. S-Edit captures the design at multiple levels of abstraction for any given cell. You can represent a cell as a schematic, RTL, or SPICE, and then swap those descriptions in and out for simulation. T-Spice simulates SPICE and Verilog-A representations of the design, which is fully integrated with S-Edit. ModelSim simulates the digital, Verilog-D portions of your design.
- Simulating the mixed-signal design. S-Edit creates the Verilog-AMS netlist and passes it to T-Spice. T-Spice splits the netlist automatically to partition the design for analog simulation and for digital simulation in ModelSim, as Figure 5 shows.
Both simulators are invoked automatically, and, during simulation, the signal values are passed back and forth between the simulators. This means that regardless of the design implementation language, you just run the simulation from S-Edit and the design is automatically partitioned across the simulators. Then, you can view the results using the ModelSim or T-Spice waveform viewers.
• Laying out the design. Create the physical design using L-Edit, which allows you to create a full, custom layout of the IoT design. The parameterized layout library of common MEMS elements and true curve support facilitate MEMS design.
Implementing the MEMS Device
The MEMS component is key to determining device performance because of the associated packaging and fabrication process. In the pressure sensor in our example, pressure is exerted on a diaphragm above an etched cavity. The package must be deep enough to accommodate the cavity that Figure 6 shows. To characterize the sensor, you need to simulate the device with pressure exerted.
You could create a 3D model of the pressure sensor and then analyze the physical characteristics. But you need a 2D mask in order to fabricate the MEMS device. How do you derive the 2D mask from the 3D model? You follow the mask-forward flow that Figure 7 shows, which results in a successfully-fabricated sensor.
Start with 2D mask layout in L-Edit to create the device. Then, instruct L-Edit to automatically generate the 3D model from those masks in order to provide a simulation of the fabrication steps that occur at the Fab. Perform 3D analysis using your favorite finite element software and then iterate if you find any issues. Make the appropriate changes to the 2D mask layout and then repeat the flow.
Using this mask-forward design flow, you can converge on a working fabricated MEMS device because you are directly creating masks that will eventually be used for fabrication, rather than trying to work backwards from the 3D model.
The fundamental challenge of IoT design is working in four design domains: analog, digital, RF, and MEMS. The Tanner design flow is architected to seamlessly work in any of the design domains by employing an integrated design flow for design, simulation, layout, and verification.
A companion video to this whitepaper is available at: http://go.mentor.com/4ibuk
©2015 Mentor Graphics Corporation, all rights reserved.
By Dr. John Parry (CEng, CITP, MBCS, MIEEE) and Byron Blackmore
Why is PCB Thermal Design Important?
Many aspects of a printed circuit board’s (PCB’s) performance are determined during detailed design—for example, making a trace a specific length for timing reasons. Timing issues are also affected by temperature differences between components. Thermal issues with the PCB design are largely ‘locked in’ during the component (chip package) selection and layout phases. After this point, only remedial actions are possible if components are found to run too hot.
We advocate a top-down approach starting at the system or enclosure level  in order to understand the flow environment for the electronics, which is critical for air-cooled electronics. Assumptions made about the uniformity of the airflow in early design that subsequently proves unachievable can have a disastrous impact on the commercial viability of the product and meeting the market window.
Optimizing the Thermal Layout
The golden rule is to start early and start simple. The mechanical engineer responsible for the thermal integrity of the product should aim to provide as much useful feedback as possible to the electronic engineers to guide the design, about the thermal impact of their choices, especially during early design.
From the mechanical engineer’s perspective, at the PCB level, this entails helping with package selection and the best positioning of components to utilize system air flow for cooling. Inevitably, both layout and package selection are driven primarily by a combination of electronic performance and cost considerations, but the consequences of those choices on thermal performance should be made as clear as possible, as temperature and cooling also affect performance and cost.
1: Start Pre-Placement / Pre-Layout
There is a lot of work that can be done well before layout is completed within the electrical design flow. Indeed, any influence that thermal considerations have on the design need to be factored in before this point. A lot of work can be done with a simple representation of the enclosure  to provide information about the air flow profile over the board.
You can start by simply smearing the total board power over the total board surface. This will give you a temperature map that will give an indication of any hot regions that are caused by a mal-distributed air flow, and the enclosure-level air flow should be optimized ahead of the PCB design. For this, you can treat the board as a block with an isotropic conductivity of between 5Wm-1K-1 and 10Wm-1K-1. The results at this stage will be quite insensitive to the value chosen.
A word of caution – components inject heat locally into the board, so the heat flux density into the board below a component will be higher than the average for the board. As a result, the local board temperature will be higher than that predicted in the simulation, so at this stage, the board temperature should not be used to try to estimate component temperatures. To do that, the model needs to be refined.
If the board temperature at any point is close to the maximum component case temperature, then it is very likely that this limit will be exceeded once the component heat sources are represented discretely. This may be expected, for example, if one or more components are known to require a heatsink.
2: Get Component Power Guesstimates
For this reason, it’s very important to know a best guessed estimate (guesstimate) of the individual power budgets for the main heat dissipating components that will be used in the design, and the approximate size of those packages. This will allow you to describe these as footprint heat sources in your simulation, smearing the remainder of heat uniformly over the board surface.
Ahead of part research and selection, which happens at the start of the schematic capture phase of the project, the system architect will already have some idea of what key components will be required, what will need to be positioned close to what, and what the size of the components will be. He may be anticipating using some components that were selected for another product, or retain components from the previous generation product.
3: Use 3D Component Models before Package Selection
It’s important to try, despite the difficulty, to include some form of 3D component model in the simulation before the component selection is finalized. By feeding the thermal results back before this milestone is reached, it’s more likely that thermal performance will get considered as part of the package selection criteria. Some integrated circuits (ICs) are available in more than one package style, and not all package styles perform equally well from a thermal point of view. As a result, the need for a heatsink later on may be eliminated by appropriate package selection.
Component temperature, either in the form of a case temperature or junction temperature, depending on how the manufacturer has specified the component, is the key measure used to indicate whether the design is acceptable from a thermal perspective. At this stage, however, we can only get a rough estimate of component temperature.
In the absence of any other information, the simplest 3D component model that can be used is a conducting block. FloTHERM includes material properties that are tailored to give a case temperature prediction for different package styles.
For plastic components, a thermal conductivity of 5Wm-1K-1 to 10Wm-1K-1 is recommended  and 15Wm-1K-1 for any ceramic components. 5Wm-1K-1 will clearly give a worst-case figure for case temperature.
By representing the package body in three dimensions, the effect of the component on the local air flow and, correspondingly, any downstream components is taken into account. Large components can shield smaller, lower profile components from cooling air, and the wake formed behind a component is a region in which the same air gets recirculated, so any components in that region are likely to be hot.
One useful tip is to try to align any rectangular components so that their long side is parallel to the primary flow direction. This both reduces the overall pressure drop as the flow ‘sees’ less of an obstruction, and produces a smaller wake, minimizing the effect on downstream components.
4: Feed Back Thermal Results
At this stage, you can start to feed information about the PCB’s performance back to the PCB design team. Although the simulation is relatively coarse at this stage, the principle simulation results—the airflow distribution over the board and the resulting board temperature map—are very powerful tools that you can use to show what you have to work with in terms of available cooling air, and what that may mean in terms of component temperature.
It is worth emphasizing that these nominal component case temperature values are subject to change as they are based on:
- an assumed layout
- very rough power estimates
- uncertainty about package selection
- unknown layer stack-up and copper distribution within the PCB, and
- a preliminary heatsink size and design (if already known to be necessary)
It is nevertheless a useful start, providing both an understanding of the system performance and a model that can be refined as the design is elaborated. This model provides a very useful platform for investigating the effect of component placement on the temperature of a component and its neighbors, so adjustments can be made easily and the model rerun often in a matter of minutes, not hours.
The results will give some indication as to which components, if any, will be likely to need some form of heatsink, which can be investigated next. These are also the components that need to be refined in terms of their modelling, once more information is known about package selection, so this exercise helps prioritize where to invest effort in developing the thermal model.
5: Size Heatsinks Early
For any components that may be too hot, investigate how effectively the use of a heatsink brings down the component’s temperature. If the flow is mainly normal to one side of the package, a plate (or extruded) fin heatsink is likely to be most suitable. If not, then a pin fin heatsink should be considered.
FloTHERM and FloTHERM XT have a Heatsink SmartPart, which can be used to parametrically define heatsink geometry. Start by making the base of the heatsink the same size as the package and investigate different numbers of fins, fin height, and fin thickness. The aim is to see if the heatsink can simply be mounted on top of the package, or if a larger heatsink might be needed, which will require board real estate for the mechanical attachment, as this information needs to be fed back to the PCB design team as early as possible. If so, it is essential to select an existing heatsink that provides adequate cooling performance, or design a custom heatsink, before the board can be routed, as the mechanical attachment for the heatsink may affect component placement.
Heatsinks are essentially area-extending devices, which increase convective heat transfer to the air by providing a larger surface area for the air to pass over. Heatsinks are normally made of an aluminum alloy to allow the heat to spread effectively across the base and up the fins. The base itself acts as a heat spreader, and so helps to reduce the component temperature. Start by using short, widely spaced fins to minimize the resistance to the airflow and any wake caused by the heatsink, as this will impact the cooling of downstream components.
If this shows that the component can be cooled by a relatively small component-mounted heatsink, this activity can stop at this point, but will need to be revisited later.
When including a heatsink, it is essential to include the thermal resistance of the thermal interface material (TIM) between the package and the heatsink. The ultimate choice will depend on many things, but a standard thermal pad, having a thickness of around 0.2mm and a thermal conductivity of around 1.0 Wm-1K-1, will be conservative for early design use.
This article provides a basic overview of the key considerations in PCB Thermal Design. To learn 5 other important tips, please follow this link:
1. 12 Key Considerations in Enclosure Thermal Design…A High-Level ‘How To’ Guide, Mentor Graphics White Paper
2. Tony Kordyban (1994), Estimating the Influence of PCB and Component Thermal Conductivity on Component Temperatures in Natural Convection, Third International FloTHERM User Conference, September 1994, Guildford, U.K