On the meeting of the Skolkovo and Rusnano in Palo Alto

Interesting information from the blog of a good man and a good specialist Yu.Panchula:


First of all, RUSNANO with Mikron Zelenograd bought FAB 90 nanometers. I understand that we are talking about this transaction with ST Microelectronics. Although I have also read recently an article in EE Times about a deal between IBM and Zelenograd Angstrom. All sorts of types of goats Latynina at this point would exclaim it’s not enough, you need to 15 nm, as in Intel’s labs. But the reality is that a significant portion of chips in the world (such as microcontrollers, even the new ones) made it to this technology, and if the Russian design centers need a more fancy technology, that is TSMC in Taiwan, Global Foundries in Dresden, etc. In addition, a heap of applications with extreme temperature conditions (such as the controller of automobile engines) few nanometers still does not fit. Yes, and these can make faby radiation hardened chip that is necessary for the satellites.

Updated faby — it certainly is good, their presence contributes to the fact that in Russia there were experts on different technologies. But it is much more interested in me said Dmitry Akhanova transactions with companies and Crocus SiTime, after which I appreciated the elegance business schemes implemented by his organization. In fact, Rusnano USA is in the states of key technologies (MRAM, MEMS clock), which lack a little bit before the widespread use of, and then invest in them a lot of money, on condition that develops technology company will open an office in Russia. In this case, the technology is in a state that 1) it spent decades of research and even money from DARPA; 2) all of the major problems associated with the introduction of technology has already been resolved, and 3) the technology does not dominate the market because it is not yet to have reached Hands mass producers involved support orders for previous technologies (flash memory, quartz clock). Russia bought this most important place in the future of the market, and Russian physicists and electronics engineers get interesting field of application of their talents.

By the way, Crocus, and SiTime up in our glorious city of Sunnyvale, which houses the office and MIPS Technologies, and my house of the Camellias. Total:

Crocus Technology
Now offices in Santa Clara and Grenoble
An article in EE Times

The technology: Memory MRAM has a real chance to become the new mainstream, universal memory technologies instead of DRAM and Flash, and even oust SRAM for use in the caches. Advantages and disadvantages of the existing mainstream:

SRAM — fast but expensive and does not store the values of nutrition privyklyuchennom

DRAM — much cheaper, but slower, and does not store the values when power is off

Flash — stores values when power is off, but slow

Article about MRAM in wikipedia

Magnetoresistive memory has a performance comparable to the type of memory SRAM, such as cell density, but less power than the memory type DRAM, it is faster and does not suffer degradation over time in comparison with flash memory. It is the combination of features that can make it a "universal memory" that can replace SRAM, DRAM and EEPROM and Flash. This explains the large number aimed at the development of its research.

Of course, at this point MRAM is not yet ready for widespread use. The huge demand for flash memory market compels manufacturers to aggressively introduce new processes. The latest factory, which manufactures such as flash memory chips of 16 GB firm Samsung, using 50nm process technology. On older production lines are manufactured memory chips DDR2 DRAM, which uses 90 nm manufacturing process the previous generation.

Magnetoresistive memory is still very much a work "in progress" and is made with outdated processes. As the demand for flash memory currently outstrips supply, then will not soon a company that decides to transfer one of its factories with the latest technological process for manufacturing magnetoresistive memory chips. But in this case, making the magnetoresistive memory at the moment plays a flash memory cell size, even using the same process.

Interestingly, one of the investors in Crocus-and (in addition to RUSNANO) — is the US-French fund Sofinnova, who many years ago and was my venture capitalist — cm 1 and 2. Small world!


SiTime Corporation
Synchronization Systems
According to the EE Time is one of the 10 most potentially interesting startups this (2013) year
The fastest growing semiconductor company in North America
The technology: Generators clock — one of the foundations of electronics. In this MEMS clock oscillators have a real chance to oust crystal oscillators.

Help from Wikipedia: Participation DARPA:

 http://en.wikipedia.org/w…illator # Commercialization 

The US Defense Advanced Research Projects Agency (DARPA) funded a wide range of MEMS research that provided the base technologies for the developments described above. In 2001 and 2002, DARPA launched the Nano Mechanical Array Signal Processors (NMASP) and Harsh Environment Robust Micromechanical Technology (HERMIT) programs to specifically develop MEMS high stability resonator and packaging technologies. This work was fruitful and advanced the technology to a level at which venture capital funded startups could develop commercial products. These startups included Discera [54] in 2001, SiTime [55] in 2004, Silicon Clocks in 2006, and Harmonic Devices in 2006.

The explanation for programmers who are not familiar with Hardware: Why do we need Oscillator / clock / clock generators, etc. ? Not only for time measurement or a fixed data transmission rate accurately. Clock generator — is one of the basic foundations of digital electronics, along with logic elements and flip-flops.

I once made a blogopost "Introduction to the Design harvera chips for those programmers who have never done this"

Repeat the main points. If the baud rate generator (or equivalent mechanism) was not, it would be difficult to make digital circuits, the values on the conclusions of which would depend not only on the current inputs, but also from a saved state. In other words, from simple logic gates such as AND, OR, NOT pretty easy to build a circuit that computes f (a, b, c) = a + b * c, but it’s hard to do anything iterative, such as the factorial.

Let me illustrate this thesis is an example:

Clarification of what is what in the picture:

Box to the left — an adder, a combinatorial design, which takes input two numbers and some time T outputs the sum of them. Prior to the expiration time T on the adder output will be some kind of undefined values, and after all the T ustakanilos.

Time T should be less than the time between clock pulses. Specifically, on the contrary. The designer sets the program-sintesayzeru the frequency with which he wants to dri
ve the design, and the program selects the appropriate adder (fast, big, slow, small, etc.) into this frequency. If this works out.

Box on the right — it’s a flip-flop (in Russian called "trigger"), such figovina that stores bits of information between the pulses of the clock signal. While every arithmetic adder ustakanivaetsya, gurgles and champing at this time flip-flop stores the value from the previous cycle clock / clock and gives it only as a read-only value for calculations.

In the end, all the combinatorial logic (adder) ustakanivaetsya, and finally comes to the rising edge of the clock signal, which allows the flip-flop to record the new values for the next cycle.

With modules, gates (combinatorial logic elements — "and," "or" "no", etc.) and flip-flops, in principle, possible to implement all the hardware. And any algorithm. And with a high degree of parallelism. You can add a hundred numbers for a nanosecond, if you really want and there’s a lot of gates and mudflats (though about Watts — a more subtle question.)

Note: Yes, I know that the above scheme can be "easier" to implement with two T-flip-flops, but I showed it that way, to illustrate the general principle of "combinatorial logic + flip-flops" (see picture from Wikipedia )

Summing up his post, I note that, after studying the issue I personally was pleasantly surprised by the finding of logic operations RUSNANO. Of course it can have all sorts of difficulties in the Russian territory, but the American part of the plan looks pretty clear and elegant.



Like this post? Please share to your friends: