Monthly Archives: December 2018

news

Beckermus Technologies Announces Partnership with Engage Photonics

Beckermus Technologies was founded in 1998 and serves as an Excellence Center in the field of “Bare Die” assembly services for micro-electronic and micro-optical systems. Beckermus Technologies has been playing a vital role in the lifecycle of many products, for numerous companies worldwide, from early prototypes to mass production.

 

We are now glad to announce, that Beckermus Technologies have partnered with Engage Photonics, an Optics Design House, specializing in micro-optical systems. Engage Photonics has been selected based on our long-term experience working with its Founder and CEO, Costa Gurgov.

 

The partnership enables Beckermus Technologies to offer the complete product lifecycle:

 

  • Requirements study and Optical Architecture
  • Proof of Concept and prototyping
  • Optical design, system design
  • Component selection and optical characterization
  • Optomechanical design, tolerancing and integration
  • Manufacture, assembly and test.

 

The cooperation creates one-shop-stop full turnkey solution that not only includes design to production in-house process but also offers ongoing feedback between the design team and the assembly team.

Many old CPU chips and obsolete computer processors as background

CoreHW releases IP library with over 200 IP

CoreHW releases an IP library with over 200 silicon verified IP. Silicon proven solutions allow fast end-product development and implementation. The IP library gives a comprehensive look to company’s IP range and makes it easier and faster for customers to start their ASIC design process.

 

The IP listed in the library are developed with several Silicon Foundries for various process nodes, for example: 22nm FDX, 40 nm CMOS, 130nm RFSOI and 180nm CMOS. Silicon IP include solutions for example for imaging, Bluetooth and IoT, WiFi, among other various application areas.

 

CoreHW offers several licensing model options: white box or black box, single use – multi use. Company also offers support in IP integration and customization.

 

Examples of the IP solutions CoreHW offers:

  • CorePLL, High Performance 6GHz synthesizer, Synthesizer for LTE Carrier Aggregation systems
  • High Resolution 16b SAR ADC for imaging solutions
  • RF Front End for Bluetooth and IOT solutions
  • Receiver for WiFi Front-End
  • High Efficiency DCDC system for Power management ICs
  • Tunable regulators, Band Gap references, Operational amplifiers, comparators, Level shifters, and Digital control and Interface IP

 

Full IP library is available at http://www.corehw.com/IP

 

Requests and more information:

Tero Kainulainen

tero.kainulainen@corehw.com

 

 

About the company:

 

CoreHW is a fabless semiconductor company established in 2013 in Tampere, Finland. CoreHW develops full custom RF, Power Management-, Data Converter and High-Performance Analog IC and IP. CoreHW high performance IP are optimized for low power consumption and small die area. We support you in the IP integration, verification, productization and possible customization. Both black box and white box licensing models are available. We offer design services for customers worldwide, with over 20 years of experience gained from leading communications companies and research hubs. We are passionate about exploring the possibilities of integrated solutions and dedicated to our customers. We believe these are the reasons we are one of todays fastest growing IC companies. If somebody can do it, it is CoreHW.

 

 

computer cpu or central processor unit chip on mainboard.Technology background with computer processors CPU concept and blue circuit, board texture.

How Best-In-Class Companies Reduce ASIC Power Consumption – White Paper

Power, performance and area (PPA) are three core competencies on which an ASIC design is evaluated. In an ideal world, ASIC designers strive to improve the performance of their design in a minimum possible silicon area with lowest possible power dissipation. However, there exists a trade-off between these three key metrics which designers would need to gauge and choose a sweet spot which satisfies the requirements of the given application.

 

In the past 2 decades, with advent of mobile devices like cellphones, tablets, laptops, smart watches etc., power has taken the center stage in ASIC design. Low power design is critical from many standpoints- some obvious and some non-obvious ones:

 

  • Battery Life: Battery life is critical simply because no one likes to charge their devices frequently. Lower power dissipation would mean the devices can function for a longer duration without needed to be charged.

 

  • Package and Cooling Costs: Package cost accounts for significant amount of the total ASIC cost. In order to dissipate the heat produced inside an ASIC, package is employed as a heat sink. A superior package would be able to dissipate the excess heat more effectively, however, it will cost more and therefore add to the overall ASIC cost. Similarly for ASICs used in big data centers need to use a cooling mechanism to dissipate excess heat. Keeping the power dissipation low would reflect in lower package and lower cooling costs.

 

  • Digital Noise Immunity: With increasing cell densities in modern nanometer processes, excessive switching in a concentrated region can result in spurious transitions referred to as glitches. By reducing the amount of power drawn from the power supply, designers strive to improve the noise margins of their devices.

 

  • Environmental Impact: Batteries used in mobile systems contains lithium and other toxic metals like mercury, cadmium and lead. These metals when contaminate the water supplies can cause many deleterious effects on the surrounding environment.

 

Figure 1: Increase in electronic waste generated over years. Image Courtesy: Washing State Dept. of Ecology

 

Today, ASIC designers use a plethora of techniques to reduce their power dissipation, and we would discuss them in detail. It would, however, be prudent to first talk about various type of power dissipations and how they trend with technology scaling.

 

Types of ASIC Power Dissipation

 

Dynamic Power

 

Dynamic power is dissipated in circuits when one of more transistors within the standard cell are switching from either high to low, or low to high. When the integrated circuit is working at “full throttle” which may be while running the most resource intensive application where in a significant or majority portion of the chip is switching states actively, dynamic power dissipation is usually the highest. It may be sub-divided into switching power and internal power.

 

Switching Power – The amount of switching power dissipated in a circuit is a function of switching frequency, the node capacitance and the operating voltage of the standard cells. This can be minimized by reducing the switching activity, with or without a reduction in the operating frequency, decreasing the node capacitance that’s been charged or discharged, and by reducing the operating voltage. Switching power is calculated as follows:

 

Where ß is the rate of switching, f is the operating frequency, c is the node capacitance and VDD is the operating voltage.

 

Figure 2: Switching power dissipated while charging and discharging of output capacitance

Internal Power – The component of dynamic power that is dissipated during circuit switching when both pull-up network and the pull-down network are ON intermittently is referred to as internal power. Internal power is directly dependent on the voltage and node capacitance, while indirectly dependent on the node transition. Slower transition times increase the internal power dissipated in the circuits.

 

Figure 3: Short circuit power dissipated when both PUN and PDN are ON.

 

 

Leakage Power

 

Leakage power is dissipated in circuits when none of the transistors within the standard cell are switching and the inputs are in a steady state. This component of power has become significant in the sub-micron technology processes. It is dependent on the ambient temperature, the threshold voltage of standard cells and the input combination of the standard cell during steady state. Leakage power increases exponentially with an increase in ambient temperature.

 

Figure 4: Gate leakage and Sub-Threshold leakage

 

The goal of designers is to reduce the overall power dissipation of their device- which means reducing the dynamic power during peak performance or by reducing redundant switching when the circuit is not active. Similarly, designers strive to reduce the leakage power of their device when the circuit is in a steady state. There are many techniques used to achieve this.

 

Techniques Employed to Save ASIC Power

 

Dynamic Frequency Scaling

 

Switching power has a direct dependence on the operating frequency. When operating in full-throttle mode, the operating frequency is typically very close to the specified frequency, however, this would mean that the chip is burning maximum dynamic power as well. When you are not operating in the full throttle mode, it is possible to scale down the frequency- with or without scaling the operating voltage. This helps reduce the switching power dissipated in the circuits. For example, many commercial CPUs today are able to change the operating frequency depending on the workload.

 

  1. Basic Mode- 400 MHz
  2. Moderate Usage- 800 MHz
  3. Active Multimedia Mode- 1.2 GHz

 

Dynamic Voltage and Frequency Scaling (DVFS)

 

Since voltage has a quadratic dependence on the switching power, it is most beneficial to scale down the voltage accompanied with scaling down the operating frequency to reduce switching power. This technique is referred to as Dynamic Voltage and Frequency Scaling. When you operate at a lower voltage, the devices get slower. In order to meet timing, one needs to operate at a lower frequency. There are many complexities that designers need to deal with in order to implement DVFS.

 

By employing DVFS, designers might be able to get away without employing an active cooling mechanism (like use of fans) to reduce the heat generated by the electronic components. DVFS is not restricted to mobile technology arena. Many other applications like automotive SoCs, desktops, servers etc. also employ dynamic voltage and frequency scaling.

 

Clock Gating

 

Clock signal has the highest toggling frequency of all signals in the design, with a toggle rate of 2.0 because it toggles twice per clock cycle. Consequently, dynamic power dissipated in the clock network accounts for 60—70% of the total switching power dissipated in the circuits. In addition to switching power dissipated in the clock tree network, sequential devices like flip-flops also dissipate a significant amount of internal power. It is therefore prudent to gate the clock when a certain portion of the chip is not active to save on the switching power. This is accomplished using a bunch of architecturally placed clock gates referred to as clock gating integrated cells (CGICs).

 

Figure 5: Clock Gating Integrated Cell

 

Clock gating is perhaps the oldest and the easiest method to save on dynamic power. Modern ASICs use many levels of clock gating to control the granularity of the design that is clock gated.

 

 

Power Gating

 

While clock gating helps reduce the switching power, power gating is useful to reduce the leakage power. As mentioned earlier, leakage power is dissipated during steady state where a small amount of leakage current flows through the transistor substrate to the ground. Power gating cuts off the voltage supply to the standard cells, and therefore limits the leakage power. This can be thought of as similar to standby mode in modern CPUs where a majority portion of the CPU goes to sleep, and only basic housekeeping and critical operations are kept awake to save on the battery life.

 

Figure 6: ASIC sub-divided into multiple power domains

 

There’s some penalty that designers need to pay when they employ power gating.

 

  1. Implementation complexity: With power domains, the design needs to be split into one or more power domains, with the power grid of each domain being fragmented. Ensuring robustness of fragmented power grids is an ever-growing challenge for physical design engineers.
  2. Low Power Verification: For any signal handshaking across power domains, designers need to carefully assess the need for level-shifters and/or isolation cells. Thus, this creates an extra overhead to ensure verification of all signals crossing power domains.

 

 

Downsizing Standard Cells

 

Bigger cells dissipate higher switching power. For all the cells along a timing path with positive slack, it might be possible to opportunistically downsize the cells to recover dynamic power. This is usually accomplished using modern EDA tools, and brings about an incremental change while operating in the ECO (Engineering Change Order) mode. The savings achieved by downsizing would be nowhere close to any of the savings achieved by using one or more architectural methods discussed above.

 

VT Swapping

 

Many flavors of CMOS and FINFET devices are manufactured which differ in their threshold voltage.

 

  1. High VT devices (HVT) – Least leakage, slowest.
  2. Standard VT devices (SVT) – Moderate leakage, moderate speed.
  3. Low VT devices (LVT) – Highest leakage, fastest.

 

Similar to downsizing, it is also possible to swap the cells with lower threshold voltage to cells with higher threshold voltage to save leakage power, but only in an ECO mode.

 

Summary

 

During the 90s during the desktop PC era, most of the efforts were concentrated on extracting the best performance out of the design. However, in past decade mobile devices like cellphones, tablets, smart watches and fitness tracking devices have shifted the momentum towards power. Low power is a key for design success. Ever increasing cell integration densities and increasing leakage with shrinking technology has made power optimization one of the most challenging problems that designers have to deal with. The architectural and design techniques discussed in this post are dispensable for designers to help meet their power targets.