As MEMS companies such as SiTime, Discera and Sand 9 are continuing to commercialize and develop their oscillator technologies, we connected and spoke with Dr. Hisham Haddara, CEO of Si-Ware Systems. In this detailed interview, Dr. Haddara provides an overview of MEMS based oscillator technologies including key performance metrics and considerations for the overall oscillator solution architecture. Dr. Haddara's company provides a unique perspective on the oscillator technology space in their role as ASIC design developers for Discera's MEMS resonator devices. Si-Ware has also developed an IC based "all-silicon" oscillator technology of its own.
MEMS Investor Journal: What are programmable clock oscillators and what is advantageous about them?
Dr. Hisham Haddara: Any electronic system such as hard drives, cameras, video cards and others will typically use several clock oscillators to provide timing information necessary for the proper operation of the system. The advantage of a programmable clock oscillator is the ability to customize the oscillator frequency after batch processing. Low price levels by processing large volumes of “virgin” chips can be reached while the frequency can still be customized according to customer specifications *after* manufacturing.
MEMS Investor Journal: For clock oscillators, what are the main performance measures? In simple terms, what is the meaning of each performance measure?
Dr. Hisham Haddara: The main performance measures for clock oscillators are frequency range and resolution, stability and tolerance, jitter, power consumption, temperature range and driving capability.
Frequency range and resolution
The larger the frequency range you can cover, the wider the market you can potentially serve with your low cost device. A fine resolution of the frequency steps (the increment) allows greater flexibility in selection of frequency by customer.
Stability and tolerance
Clock oscillators produce a specified frequency. Any drift of such frequency over temperature is undesirable as it may negatively impact the performance of the system. Frequency tolerance is the amount of drift that occurs over temperature. It is usually measured in parts per million (ppm) and the lower it is the better. This is a key indicator of the timing devices. In our work with Discera, for example, we ensure a low frequency tolerance is achieved through an effective temperature compensation technique that corrects for any drifts in the MEMS resonator.
Jitter
The output of a clock chip is a square wave that has a specific period (inverse of the frequency). Jitter measures how stable this period is from cycle to cycle. A low jitter signal will have consistent time periods between every two consecutive edges, while a jittery signal will have periods that vary randomly from cycle to cycle around the nominal value. Jitter can be thought of as noise in the time base of the system. This may lead to degradation in the signal quality. In extreme cases, jitter may lead to clock cycle slipping in microprocessors leading to erroneous calculations.
Power consumption
Reduced power consumption is always a goal since it has a direct impact on battery life for battery operated devices.
Temperature range
The temperature range over which the part maintains proper operation, meets specs and is confined to the specified stability tolerance.
Driving capability
The maximum capacitive load the chip can drive. The larger the driving capability, the more circuits the clock oscillator can drive without the need for additional buffering. This simplifies the customer’s system design and avoids unwanted buffer delays.
MEMS Investor Journal: You recently announced that you are collaborating with Discera, a company that has developed MEMS resonator technology. What is your role in the project with thme? What is Si-Ware providing and what is Discera providing for the overall solution?
Dr. Hisham Haddara: The overall product is a programmable clock oscillator based on Discera’s MEMS resonator. It consists of the MEMS resonator (which is a passive device) and an electronic ASIC that provides all the active circuitry and intelligence.
The overall programmable clock oscillator is Discera's product. They provide the MEMS die and outsource the ASIC design to Si-Ware. However, they handle the ASIC fabrication, packaging of the overall solution and production testing. Si-Ware’s part is to design the ASIC, according to Discera’s requirements, and provide heavy support throughout the production stages.
What Si-Ware brings to the solution is an extensive expertise in analog and mixed-signal design which is gained from our work with many other clients in this technology area. The know-how is crucial for creating a functional MEMS resonator and extracting the best possible performance out of it.
Another important advantage of Si-Ware's capabilities is our broad experience in a variety of MEMS devices such as resonators, inertial sensors and optical MEMS. A thorough understanding of modeling and calibration of MEMS devices is key to the success of any MEMS based module. Finally, we support our customer heavily in all stages of the product lifecycle including bench testing and characterization, proper production flow as well as problem debugging and solving.
MEMS Investor Journal: Who are the top 3-4 competitors for your clock oscillator product based on Discera’s MEMS resonator technology?
Dr. Hisham Haddara: Discera would be in a better position to answer this question, as they have a better understanding of their competitors. Si-Ware’s contribution was meeting the requirements of Discera with the best performance and lowest power consumption possible. This product achieves a very wide temperature range, very wide frequency range (150MHz) and a high driving capability. The competitiveness of the overall product will depend greatly on the MEMS resonator itself.
MEMS Investor Journal: What is the oscillator’s power consumption and how does that compare to a quartz-based timing devices? How does it compare with other MEMS based devices such as those from SiTime?
Dr. Hisham Haddara: Our solution's power consumption is around 6.2 mA at 20MHz, with no load. Quartz is generally of lower power due to the better quality factor (often called the Q factor) of the quartz resonator. However, the advantage of silicon based oscillators is in price, due to the economics of batch processing, as well as reliability.
The power consumption of a MEMS based oscillator will depend greatly on the quality of the resonator itself and the frequency it operates at; in other words, the higher the frequency, the better the performance and less power is needed to generate the programmable output.
MEMS Investor Journal: You also mention that that the working temperature range is extended? What is the standard temperature range and what is it for your solution? How are you able to extend the temperature range?
Dr. Hisham Haddara: A standard commercial temperature range is from 0-70°C. An extended range is from -20 to 70°C, and an industrial range is from -40 to 85°C. Our temperature range is from -40 to 105°C.
To achieve such a wide temperature range, several factors are important. A robust design of the analog circuitry with ample margins needs to be ensured. We add calibration knobs to sensitive circuitry and adjust their behavior at temperature extremes. A sophisticated calibration routine dials these knobs in a way that ensures proper operation. Finally, accurate temperature compensation hardware is necessary to compensate for the temperature dependence of the MEMS element.
MEMS Investor Journal: What are the signal inputs and outputs for this oscillator from Discera?
Dr. Hisham Haddara: The ASIC and the MEMS resonator are housed in the same package and the interconnections between them are internal to the product. The ASIC connects to the outside world via only 4 pins: VDD and Ground which deliver the power, the Enable pin that turns the chip ON or OFF, and Output that transmits the chip output clock signal. This ensures compatibility with the existing footprint of the widespread quartz clocks.
One important feature of this product is the ability to communicate, calibrate and program the chip via only those four pins. Si-Ware’s technology enables this by multiplexing those functions on the same 4 pins of the chip, assigning multiple uses for each pin. This is visible only during calibration phase.
MEMS Investor Journal: When and how are the chips calibrated? Does each chip have to be calibrated individually? How does this calibration step affect the cost?
Dr. Hisham Haddara: Calibration is done on the production line according to a certain calibration routine designed to optimize the performance of the each individual chip to the frequency requested by the user. The frequency along with the calibration settings are then programmed on to the chip.
Obviously, calibration does increase cost, but this increase will be a function of the calibration routine, the parallel processing that can be built in the production line and whether this calibration is done at room temperature only or at several temperatures.
MEMS Investor Journal: Based on the oscillator performance measures that we discussed above, what are some examples of low, mid and high end applications?
Dr. Hisham Haddara: Among the two most important metrics are the frequency stability and jitter. Low end applications (including printers and some automotive applications) demand frequency stability as high as 1% and a period jitter of 10’s of pico seconds (ps). Medium end applications such as digital still cameras, hard disk drives, PCIExpress and USB3 demand a frequency stability of 50 to 100 ppm (parts per million) and a period jitter of less than 3 ps. High end applications such as SONET, wireless and GPS would generally require frequency drift of less than 25 ppm over the extended temperature range and a jitter of less than 0.5 ps.
Power consumption depends primarily on whether the clock is for a portable (battery operated) device or not. Portable devices would generally require a clock generator part to be less than 2mA.
MEMS Investor Journal: You mentioned earlier that MEMS based oscillators offer a cost advantage over quartz. What is the specific cost comparison of quartz vs. MEMS based oscillators? You also mention that reliability is better for MEMS based oscillators versus quartz. How is reliability typically measured and what is the numerical comparison based on that criteria?
Dr. Hisham Haddara: One important way of measuring reliability is the shock impact and vibration that a part withstands without being broken. These are measured in g-forces (where ‘g’ corresponds to one gravitational force, 2g would be double the gravitational force, etc). MEMS-based clocks can withstand a shock impact of thousands of g, which is typically x10 higher than quartz. An even more reliable technology from that regard are all-silicon clocks (no MEMS or quartz resonator), which is a technology in which Si-Ware has significant advancements.
For the applications where MEMS based clocks compete, the cost will be generally lower than quartz by a significant percentage, provided that a good yield is achieved. It is important to note that prices depend on the specifications and applications; also both quartz and MEMS based oscillator providers are continually pushing price down. This is also an area where all silicon clocks (IC clocks) offer a clear advantage, since the resonating element is on the chip and does not add any extra cost.
*********************************************
Dr. Hisham Haddara is the founder of Si-Ware Systems (SWS), a leading fabless semiconductor company. Prior to founding SWS in December 2003, he was with MEMSCAP from May 2000 as Executive Vice President and President of MEMSCAP Egypt. During that time, he established and managed an R&D center focusing on RF IC design and MEMS component design for optical applications. In 2002, he became President for Asia and the Middle East overseeing business and development activities in the region.
Dr. Haddara was co-founder and first general manager of ANACAD Egypt in 1994. In the same year, ANACAD was acquired by Mentor Graphics Corporation (MGC), where Dr. Haddara has worked till May 2000. During the period 1994-1997, Dr. Haddara managed the Mentor Graphics Egypt facility growing its development and business activities successfully. In 1998, Dr. Haddara was appointed chief scientist in the analog and mixed-signal simulation division where he started an initiative for analog design re-use.
Dr. Haddara obtained his B.Sc. and M.Sc. degrees in Electrical Engineering both with highest honors from Ain Shams University in 1980 and 1983 respectively. He received his Ph.D. degree in Microelectronics from the National Polytechnic Institute, Grenoble, France in 1988. His research focused on modeling and characterization of hot carrier reliability in deep sub-micron MOS transistors. His Ph.D. research work won the CNRS Ph.D. thesis award in 1988, awarded by the French National Scientific Research Center (CNRS).
Copyright 2011 MEMS Investor Journal, Inc.
Thanks Investor Stuff, it is an honor to SiWare and to Egypt all. Dr. Hisham has been effectively searching much in new technologies in Microelectronics since long before. He deserves this much. And, I can suggest that Nano Science and Nano Technology can serve much in this field. Perhaps it differs.
Posted by: Amr Morsi | March 25, 2011 at 11:18 PM