Next-gen thermal and fan management controllers: A Programmable Solution

by Jim Davis , TechOnline India - August 03, 2011

Here, we examine how the proper use of a programmable SoC solution can simplify advanced thermal management solutions, save BOM cost by integrating nearly every discrete used in such systems and enable new capabilities to greatly enhance reliability and marketability of the end products these solutions reside in

Today’s thermal management systems are composed of a variety of discrete components: MCUs for PWM-generation (multiple of these for large fan control systems), MCUs (either shared or dedicated) for temperature sensing and the host application processor (CPU, FPGA, ASICs etc.) that share fan speed management with the main application processing functions. 

For large end products like chassis-based communication systems, there are typically very complex thermal management solutions made up of many of these discrete components.  In much smaller form factor products, these solutions typically are still composed of multiple discrete solutions but are much simpler.  In this article, we’ll examine how the proper use of a programmable system-on-chip solution, such as Cypress’s PSoC technology, can simplify advanced thermal management solutions, save BOM cost by integrating nearly every discrete used in such systems and enable new capabilities to greatly enhance the reliability and marketability of the end products these solutions reside in.

 

Enabling Limitless Sensor Interfaces—Quantity and Variety

One typical challenge in high-power industrial or communications systems is to control the thermal environment in the application.  The first step to implementing these controls is to know the actual conditions—temperature sensing.  There are two options when measuring temperature—analog or digital temperature sensors. 

You have high and low-end versions of both, so whether you need to measure
down to +/- 0.1 degree Celsius or if the standard +/- 1 degree is good enough, then the real decision factors come down to size, distance and cost.  A diode or transistor is the cheapest form of analog temperature sensor you can implement, it’s also the smallest; however, the distance between the diode and the device measuring the voltage with an ADC, is the most important factor since we are talking about microvolt measurements. 

The most popular digital temperature sensor, by far, is the I2C-frontend temperature sensors — an integrated ADC, diode temperature sensor and I2C interface for extracting the temperature values.  Digital temperature sensors are great for long distance measurements but come at a price significantly more than a simple diode.  Then there are also thermocouples—great for ambient temperature sensing, thermistors, PWM-based digital temperature sensors and many other forms and types.

A system-on-chip device that incorporates programmable digital and analog functionality enables you to interface to any and all of these different forms of temperature sensors and with a large enough density device, you can also interface to the most, in terms of quantity, sensors than any discrete, fixed-function MCU device available.  This allows you, as the systems engineer or architect, to really focus on the functionality you need versus the devices that are out there that
might be able to support your needs at the lowest cost.  Additionally, when you remove the constraint typically imposed on thermal designs of the quantity of temperature sensors, you now can incorporate many more temperature sensor points in your application, better understand the thermal conditions and really optimize fan placement, speeds and algorithms to reduce end system costs, power consumption and acoustic noise through optimized fan speed control.
 
Delivering Unique Fan Control Capabilities

Fan control, for either a 3- or 4-wire fan, is basically implemented through a PWM interface where you adjust the duty cycle of the PWM period to modify the actual fan’s speed.  Typically systems with few fans, less than four for example, utilize MCUs with built-in PWM peripherals to control the independent speeds of those fans—and when the fans out number the available PWM peripherals, then a single PWM interface will support multiple fans.  This works fine and has been the de facto standard in fan control functions; however, this
limits the control and optimizations you can enable through independent
fan control. 

Additionally, to calculate the actual fan speed, each of the fans output a tachometer signal that has to be interfaced with a timer or counter to determine the RPM-speed of the fan.  While most applications do not necessarily care what
the exact RPM is of a given fan, this signal is extremely important in determining whether a fan stall or rotor lock failure has occurred or not.

Programmable logic-based solutions remove the constraints a typical MCU may provide with the ability to independently control more fans than any other discrete solution.  Additionally, with the ability to independently control and monitor each fan in a given system  you gain the ability to:

1) implement hardware/logic-based closed-loop speed control;

2) optimize each fans’ speed, thus acoustic noise and energy consumption, for exactly what the system needs to maintain a target temperature; and, lastly

3) to implement advanced predictive fan failure and fan-aging algorithms based on historic PWM duty-cycle and actual RPM speed analyses. 
 
Hardware or logic-based closed-loop speed control is essentially the ability to use the programmable logic to implement a PWM peripheral and multiplex the tachometer outputs of a fan into a central counter block that, in a coordinated effort with a firmware command, sets and maintains each fans’ duty-cycle. 

This function is typically implemented either in the firmware of the MCU controlling each of the fans or more commonly is simply not used and instead the design engineer of the end-application spends several man-weeks characterizing
the duty-cycles to fan-speeds in the application in order to only worry about the duty-cycle.  This characterization process is cumbersome and must be redone for every new SKU or end-application design due to each of the architectures’ unique form-factor.

A solution that incorporates programmable logic can implement multiple PWM and Counter peripherals enabling the embedded engineer to implement a single, dedicated PWM for every fan in their system while sharing a common Counter function.  With the right programmable logic in place, a smart tachometer function in addition to the simple Counter utility can take a desired RPM speed and
self-regulate the PWM duty cycles for each of the PWM functions to maintain the desired speed.  With this type of implementation, additional features, that have never before been possible with the standard MCU firmware control approaches, such as high accuracy fan speed control to minimize acoustic noise and energy
consumption as well as predictive fan failure and fan-aging algorithms can easily be implemented.  

Fans are notoriously inaccurate and maintain, typically, +/- 10% or more error with respect to the RPM speed given a specific PWM duty cycle.  With a closed loop, hardware-controlled system, designers can easily obtain and maintain 1% accurate RPM speeds without high latencies, which if implemented in firmware, would result in acoustically unpleasant fan oscillations.  Additionally, fans are also relatively slow to respond to PWM duty cycle adjustments. 

To minimize the amount of under- and over-shoots a fast hardware-controlled system may introduce, which if not addressed would also lead to acoustically unpleasant oscillations, a damping factor should also be used.  For example, Cypress released an intelligent fan control component and application note for use with their PSoC devices and PSoC Creator software, which implements the
hardware-controlled fan management system discussed above.  In this example, an engineer customizing their fan controller solution simply applies their system-level parameters such as the damping factor, tolerance, type of control, etc. within this component that, in turn, configures the appropriate programmable logic and firmware API calls (refer to Figure 1 below).
 

                           

                            

 

 Figure 1 - Intelligent Fan Control Component

 

 

With a programmable system-on-chip which integrates programmable logic alongside an MCU, the ability to implement predictive failure detection algorithms is much simplified.  While running the closed-loop or hardware-controlled implementation of the fan controller the system knows what the desired fan speeds as well as the required duty cycles to achieve those speeds over time. 

With the right algorithm running in the background on the MCU, you can monitor
the duty-cycle trend over time to detect an increasing or decreasing duty-cycle to achieve the same RPM speed.  An increasing duty-cycle trend is an early indication of a mechanical fan defect that is resulting in the need to increase the power to the fan to maintain the same RPM speed.  A decreasing duty-cycle trend is an early indication of a clog in the air filter for the fan or something
else impacting the air flow of the fan resulting in less air resistance and less air flow and thus less energy or power required to spin the fan to obtain the same RPM speed.  These features can only be implemented in systems that take advantage of a closed-loop control mechanism otherwise the combination of the duty-cycles and RPM speeds over time would not be available for this type of analyses.
 


Achieving Full Thermal Management Solutions

With a programmable solution like Cypress’s PSoC programmable system-on-chip, you can now combine the temperature sensing and unique fan control capabilities described above all into a single device.  With the combination of a rich analog peripheral set, programmable logic and an integrated MCU the integration of these features, including the advanced predictive fan failure algorithms, can easily be integrated into a single device.  Additionally, further system-level benefits can be realized by off-loading the main host application
processor from the typical thermal management duties such as the higher-level temperature sensor aggregation and fan speed control algorithms.  

With an integrated solution, engineering a full thermal management system can be greatly simplified for the largest and most complex of applications.  For example, a software tool can be constructed to ease the design of a full thermal management solution by abstracting the low-level details of the analog or digital temperature sensor interfaces, the programmable logic control of the fans as well as the variety of advanced temperature sensor aggregation, fan speed control and thermal zone control algorithms.  For example, in Figure 2 below, a software tool properly engineered could present these types of parameters and settings to simplify the thermal management design.

 

                                        

 

Figure 2 - PSoC Creator Thermal Management Component

 

Using this type of solution can greatly offload the main host application processor enabling that device to focus more on the important features of the end system such as traffic management in a network switch or data throughput in a server or storage application.  Furthermore, with the integration of the many discrete devices typically used to implement the full set of thermal management features, a great deal of cost savings and board-space savings can be realized.  Whether
it’s for the cost savings or the performance and reliability capabilities, a programmable system-on-chip architecture enables the next generation thermal management solutions.

 
About the Author

Jim Davis is Global Marketing Manager for wireless products at Cypress Semiconductor Corp., based in San Jose, California. He joined Cypress in January 2008 and prior to that served for 8 years in the United States Air Force as a communications officer. He has a bachelor’s degree in computer science from United States Air Force Academy and a master’s degree in software engineering from the University of Maryland.
 
If you liked this article, go to the Industrial Control DesignLine home page for the latest in industrial control design, technology, trends, products and news. Also, get a weekly highlights update delivered directly to your inbox by signing up for our weekly industrial control newsletter here.
 

 

About Author

Comments

blog comments powered by Disqus