Publication date: 13 April 2010
Clocks provide the electronic heartbeat for a diverse variety of consumer electronics, communications and computing applications. These timing components provide critical reference frequencies to processors, FPGAs, ASICs, DSPs, ADCs, DACs, memory and physical layer devices. The combination of frequencies required in a given application can change dramatically based on component selection of the other ICs in the system, since each processor, FPGA, etc. can have unique frequency requirements. In addition, timing requirements can vary considerably from application to application based on performance, processing and line rate requirements.
To satisfy this diverse range of requirements, the timing IC industry has responded by developing application-specific clocks. Each device is highly customized to meet the frequency, jitter, phase and skew requirements for a particular application. These ICs are simple, pin-controlled devices and require no firmware or configuration via a host processor. There are two significant limitations with this traditional approach.
First, if the application’s frequency, jitter or other performance requirements change, a new application-specific clock must be designed, manufactured, qualified and tested. Sample lead times for clocks customized using dedicated wafer masks can approach 8 weeks or more, which can be an issue in some time-sensitive designs.
Second and more importantly, application-specific clocks have traditionally only been available for high volume applications. As a result, mid and low volume designs have resorted to using a combination of clocks, crystal oscillators (XO), and crystals to complete their timing architectures. The resulting solutions are not optimized in terms of cost, power or real estate.
To address this market need, several timing suppliers have introduced programmable clocks that can be in-circuit programmable via an I2C interface. The microprocessor interface makes it possible for hardware designers to customize the clock for their given application. This flexibility comes at a price, however. Jitter performance can vary significantly across frequency configurations. Also there are no guarantees the clock can generate all timing references with exact (0 ppm) frequency synthesis error, forcing the hardware designer to continue using XOs in their design.
Perhaps the most significant limitation of traditional programmable clock solutions is that they require I2C-based firmware development. In many applications, an I2C interface is either not available or not desirable. If the system has a single clock IC that is used as the master reference for the design, a chicken-and-egg scenario can potentially develop. The clock needs to generate the right frequency for the processor at start-up, but in order to operate it must be programmed by the processor first! Often a XO must be used as the startup reference to solve this dilemma, increasing design complexity.
Field-programmable clocks are available that can be programmed on a socketed board using configuration software prior to installing the device in the end application. The process flow for procurement of field-programmed clocks is shown in Figure 1. This approach is less than ideal because it is not scalable. Prototype quantities can be easily managed but it is more difficult to support such a solution in volume production. This solution also requires a custom programming board, vendor-specific configuration software and forces customers to manage raw stock, device programming and finished goods stocking levels. Devices are not marked with a custom top mark, increasing the risk that a programmed device ends up on the wrong board.
A new type of programmable clock generator is available that addresses the deficiencies with traditional solutions. New products like Silicon Laboratories’ Si5355 Any-Rate 1-200 MHz Clock Generator can be completely customized using a web-based utility available at www.silabs.com/ClockBuilder. The process flow for web-customizable clocks is shown in Figure 2. Using a simple point-and-click graphical user interface, a user can quickly configure the device input (crystal versus clock), input frequency and output frequencies. The utility assigns a unique orderable part number for each custom configuration. Samples can be ordered directly via the website and are available in as little as two weeks after order placement.
In order to support web-based device configuration, a highly flexible device architecture is required. The Si5355 has four banks of clock outputs, with each bank comprised of two CMOS output clocks. Each bank can generate any frequency from 1 to 200 MHz, such that a single device can produce 8 outputs at 4 unique, non-integer related frequencies. All combinations of output frequencies are guaranteed to have 0 ppm frequency synthesis error. All Si5355 devices are pin-controlled, eliminating the need for I2C firmware development and simplifying device startup. The device has consistently low jitter across all input/output frequency combinations (50 ps peak-peak period jitter), providing significantly lower jitter than traditional programmable clocks. Also, up to three unique frequency configurations can be customized per device, with external control pins available to select the active frequency plan. This enables one part number to replace three SKUs with a single device, simplifying inventory management.
The most significant benefit provided by web-customized programmable clocks is that low and mid volume based applications can finally enjoy the benefits of having fully customized, application-specific clocks. All frequencies can be synthesized from a single IC with 0 ppm frequency synthesis error, making it possible to replace multiple discrete clocks and oscillators with a single IC. All device customization, including custom part number assignment, device programming, custom part marking, and logistics can be handled by the supplier, freeing up hardware designers and procurement to focus on higher value-add activities.
The advent of web-customizable clocks represents a significant step forward for the semiconductor industry. ICs are manufactured in ultra-modern, multi-billion dollar wafer fabs and are designed with state-of-the-art design tools, yet little has changed in terms of IC sales processes since the invention of the integrated circuit. By transitioning business processes like device customization and sample procurement to the web, the IC industry will be able to reap the same benefits that successful e-tailers like Amazon, eBay, and Apple now enjoy. Mass customization, improved customer service and shorter delivery times are all unique benefits that can be realized by this transition.
James Wilson, Sr. Director of Marketing, Timing Products, Silicon Laboratories Inc.
James Wilson serves as a Director of Marketing for Silicon Laboratories’ timing products. Mr. Wilson joined Silicon Laboratories in 2002 as a product manager focusing on optical networking solutions. Previously, he was with Freescale where he worked with the networking and communications systems group. Mr. Wilson holds a bachelor’s in mechanical engineering and a master’s in business administration from the University of Texas at Austin.
