Probably, many of you thought after the situation with Phobos-Grunt - what is so special about chips for space and why do they cost so much? Why can not put protection against cosmic radiation? What is the story of the arrest of people who have exported chips from the US to Russia? Where are all the polymers?

I will try to answer these questions in this article.

Disclaimer: Information obtained from open sources and may not be completely accurate. I personally do not work with military electronics, and who works - those articles can not write. I would be happy to add and correct the article.
')

Additional requirements for space and military circuits

In the first place - the increased requirements for reliability (both of the crystal and the case), resistance to vibration and overload, humidity, temperature range - much wider, because military equipment and -40C should work, and when heated to 100C.

Then - resistance to the damaging factors of a nuclear explosion - EMP , a large instantaneous dose of gamma / neutron radiation. Normal operation at the time of the explosion may not be possible, but at least the device should not be irreversibly damaged.

And finally - if the microcircuit is for space - the stability of parameters as the total radiation dose is slowly set and survival after encountering heavy charged particles of cosmic radiation (more on this later).

Why do the military like metal-ceramic corps?

It was a long time to find out , it seems that earlier (in Soviet times) the plastic did not withstand the tests on thermal cycling, was porous (i.e., was gaining moisture), and did not tolerate negative temperatures poorly.

And finally - this is an easy way to reduce the percentage of fakes, because on the market microcircuits in the metal-ceramic package not to buy.

But ceramics also have disadvantages - it is more expensive, less vibration resistance and, in general, the wire, on which the contact pads on the chip are connected to the microcircuit pins, can fall off from large accelerations (plastic wire is “supported” along the entire length by plastic).

About categories of microcircuits

In the west, chips are divided into categories of commercial, industrial, military, and space.

Commercial - the usual, most popular chips for home and office products, usually designed for the temperature range 0..75C.

Industrial / Military - the same conventional chips, but with additional testing, designed for a slightly wider temperature range (-40..125 for example) and optionally in a cermet package (chips that did not pass additional tests - can be sold as Commercial).

Space - radiation-resistant microcircuits for space applications; here, a cermet case is more likely the rule. On Military and especially Space microcircuits, there are significant restrictions on selling to sworn friends - you need to get special permissions, and if they sell them to us, then only for civilian equipment (for example, conditionally civil GLONASS).

In Russia, everything is divided somewhat differently: chips are sold with acceptance 1 (the so-called Quality Control Department - the technical control department when the plant itself tests the chips), Acceptance 5 (customer acceptance, in the case of the military - a military representative controls the tests) and acceptance 9 (when only the most qualified personnel is involved in the work - for space and nuclear power plants). The 5/9 acceptance in itself does not mean that the microcircuit is radiation-resistant - resistance to special factors is indicated in the (non-public) documentation on the microcircuit.

These additional tests, the ceramic case and small-scale production (when the cost of development is divided not by 1 million chips, but by 100) and lead to the fact that the military / space chip is at least 10 times more expensive than the civilian, and the maximum - maybe 100 ' $ 000 apiece.

However, not any microcircuit can be supplied to the Russian military equipment - there is a list of domestic (Belarusian circuits from Integral are also included in the “domestic”) electronic components that can be used to create the equipment, where everything is listed by name. If a plant creates a new microcircuit, then it cannot be used before being included in this list. For the overall development and assessment of how much the domestic industry produces, you can look at the list of 2010 here , and 2011 .

The use of imported microcircuits requires individual permission (with the appropriate formal bureaucracy that there are no domestic analogues, but as they appear, we will definitely use them).

How does the radiation on the chip

In the "pieces of particles" cosmic radiation consists of 90% of protons (i.e., Hydrogen ions), 7% of helium nuclei (alpha particles), ~ 1% heavier atoms and ~ 1% electrons. Well, the stars (including the sun), the galactic nuclei, the Milky Way - abundantly illuminate everything not only with visible light, but also with X-ray and gamma radiation. During flares in the sun - the radiation from the sun increases 1000-1'000'000 times, which can be a serious problem (both for people of the future and current spacecraft outside the earth’s magnetosphere).

There are no neutrons in cosmic radiation for an obvious reason - free neutrons have a half-life of 611 seconds and turn into protons. Even a neutron from the sun can not fly, except with absolutely relativistic speed. A small amount of neutrons arrives from the earth, but these are trifles.

There are 2 charged particle belts around the earth - the so-called Van Allen radiation belts: at an altitude of ~ 4000 km from protons, and at an altitude of ~ 17 000 km from electrons. Particles there move in closed orbits, trapped by the earth’s magnetic field. There is also a Brazilian magnetic anomaly - where the internal radiation belt comes closer to the ground, up to a height of 200 km.



Electrons, gamma and x-rays
When gamma and X-ray radiation (including secondary, obtained due to the collision of electrons with the apparatus case) passes through the microchip - the charge in the gate dielectric of the transistors gradually accumulates, and the parameters of the transistors begin to slowly change - the threshold voltage of the transistors and leakage current. An ordinary civilian digital microcircuit after 5000 is happy can stop working normally (however, a person can stop working after 500-1000 is happy).

In addition, gamma and X-ray radiation causes all pn junctions inside the chip to work like small “solar cells” - and if radiation is usually not enough in space for it to greatly affect the operation of the chip, during a nuclear explosion the gamma and X-ray flux may already be enough to disrupt the operation of the chip due to the photoelectric effect.

Then - flash / EEPROM memory. Someone may still remember the old memory chips with UV erasure:



To reduce the cost, a version without a quartz window was also produced, which was considered once-programmable. But craftsmen still managed to erase it - X-rays. Exactly the same effect exists in space - radiation quickly erases data in eeprom / flash memory, therefore, FRAM / MRAM memory is actively explored for space applications (we have Integral and Angstrom doing this). The memory on burned-out and short-circuited jumpers — fuse and antifuse — is not erased from radiation, Micron understands this. In the west, however, they fly on a cheap near-civil eeprom, and in general they have no problems.

In the low orbit of 300-500km (where people fly), the annual dose may be 100 glad or less , respectively, even in 10 years the dialed dose will be tolerated by civilian chips. But in high orbits> 1000km the annual dose may be 10'000-20'000 glad, and the usual chips will gain a lethal dose in a matter of months.

Heavy charged particles (CHP) - protons, alpha particles and high-energy ions
This is the biggest problem of space electronics - TZCH have such high energy that they “punch” the microcircuit through (together with the satellite body), and leave behind them a “loop” of charge. At best, this can lead to a program error (0 to become 1 or vice versa - single-event upset, SEU), at worst - to cause a single-event latchup, SEL. In the case of a latched chip, the power is shorted to ground, the current can go very large, and lead to the combustion of the chip. If you have time to turn off the power and connect before burning - then everything will work as usual.

Perhaps that was what happened with Phobos-Grunt - according to the official version, non-radiation-resistant imported memory chips failed at the second orbit, and this is possible only because of the THF (according to the total radiation dose in a low orbit, a civilian chip could have work).

That snapping limits the use of conventional ground-based circuits in space with all sorts of software tricks to increase reliability.

There are several ways to deal with snapping:

1) Monitor the current consumption, and quickly distort the power
2) Use chips on a sapphire substrate (Silicon-on-sapphire, SOS, more generally, Silicon-on-insulator, SOI) - this eliminates the formation of bipolar parasitic transistors and, accordingly, latching. Software bugs however can still be. Silicon-on-sapphire plates are expensive, it is difficult to process them, and they have limited use in the civilian sector - accordingly, production is expensive.
3) Use the so-called triple-well process - it also greatly reduces the possibility of microcircuit snap-in due to additional isolation of transistors by pn-junction, but does not require any special plates or equipment and, accordingly, the production itself is much cheaper than silicon on sapphire.

Historically, in the USSR and Russia, they worked more on sapphire with silicon, and in the west they try to use ordinary silicon from triple-well as much as possible (to combine with commercial products and reduce the cost), but SOS / SOI is also made of necessity.

Neutrons + ¹⁰ B
Boron is used for doping silicon and in the form of borosilicate glass to isolate metal layers. The problem is that natural boron by 20% consists of Boron-10, which reacts very well with neutrons with the release of alpha particles right in the heart of the microchip. This led to errors in the operation of microchips, especially memory.

Neutrons are produced as secondary radiation, or come from the earth, as we remember in cosmic radiation, they are not.

¹⁰ B + n → [ ¹¹ B] → α + ⁷ Li + 2.31 MeV.

This is one of the problems that was solved by using only the ¹¹ B isotope for manufacturing microcircuits. Now neutrons pass through the microchip almost without difficulty, without causing errors. By the way, this property of boron is used for emergency shutdown of atomic reactors - boric acid is poured into it, enriched with ¹⁰ B isotope - alpha particles are not a problem there.

We now turn to a couple of interesting myths:

And let's wrap the satellite in radiation protection, and put civilian chips

With a grin, nature looks at toy particle accelerators of beastmen — at the Large Hadron Collider, they (or rather, would have) achieved miserable energies at 7 TeV for protons, and 574 TeV for lead ions. And with galactic cosmic rays, particles with an energy of 3 * 10 ²⁰ eV sometimes fly to us, i.e. 3000000 TeV. Where do such particles come from? this is above the theoretical limit of the energy of the Graisen – Zatsepin – Kuzmin cosmic particles . In human-readable units, this is about 50J, i.e. in one elementary particle energy is like that of a small-caliber sports pistol.

When such a particle collides, for example, with an atom of lead in radiation protection, it simply breaks it to shreds. The shards will also have gigantic energy, and will also rip apart everything in their path. In the end - the thicker the protection of the heavy elements - the more fragments and secondary radiation we will receive. Lead can only be greatly attenuated by the relatively soft radiation of terrestrial nuclear reactors.

A high-energy gamma-radiation has a similar effect — it is also capable of breaking heavy atoms to shreds due to a photonuclear reaction .

Finally, let's take a look at the design of the x-ray tube:


Electrons from the cathode fly toward the anode of a heavy metal, and when colliding with it, X-rays are generated due to bremsstrahlung . When the electron of cosmic radiation arrives to our ship, then our radiation protection will turn into such a natural x-ray tube, next to our delicate microcircuits.

Because of all these problems, radiation protection from heavy elements, like on earth, is not used in space. Protection is used mostly consisting of aluminum, hydrogen (from various polyethylenes, etc.), because it can only be broken up into subatomic particles - and this is much more complicated, and such protection generates less secondary radiation.

But in any case, there is no protection against TZCH, moreover - the more protection - the more secondary radiation from high-energy particles, the optimal thickness is about 2-3 mm of Aluminum. The hardest thing is to have a combination of hydrogen protection and slightly heavier elements (so-called Graded-Z ) - but this is not much better than pure hydrogen protection. In general, cosmic radiation can be reduced by about 10 times, and that’s all.

Another myth is that modern technical processes are less radiation-resistant.

The chance to get an error in a particular transistor is proportional to its volume, and it quickly decreases with decreasing technology (since transistors become not only smaller in area, but also thinner). In addition, an anomalous increase in radiation resistance with modern thickness of gate dielectrics (3 nm and less) is noted.

In general, modern resistant technical processes (65 nm and less) routinely produce microcircuits that withstand a radiation dose of 1 million rad, which exceeds all reasonable requirements for durability. Resistance to snaps and software errors - achieved through triple-well and special architectural solutions.

About soft-bugs (single-event upset)

Those. when because of the TZCH we had a distortion of the memory contents or the logic worked incorrectly.

It remains to fight this only with architectural methods - majority logic (when we connect 3 copies of each block we need at some distance from each other - then 2 correct answers "overload" one wrong, using more error-tolerant memory cells (out of 10 transistors, instead of the usual 6), using error correction codes in memory, cache and registers, and many others.

But it is impossible to get rid of errors completely - we can get lucky and the TZCH (or rather a whole fan of secondary particles) will pass exactly along the chip, and almost 5% of the chip may work with an error ... Here we need a highly reliable system of several independent computers, and the correct one programming.

How to develop space and military chips

From the previous article, we already know that microchips do not grow on trees, it is long and expensive to develop them. This fully applies to military and space circuits. The situation here, however, is aggravated by low volume - and on its own initiative it becomes extremely difficult for the plant to develop something: it is conditional to spend $ 1 million on development, and customers need only 10 chips. How much do they need to sell? $ 100,000? 200'000 $?

Therefore, the state finances OCD for the development of the necessary chips for the industry, and these OCDs are dark. For an example, you can look at the list of OCD of one Integral (by the way, there are already some small FPGAs). This is how the domestic ARM appeared - Milander, performing OCD, bought a license for the Cortex-M3, made a microcontroller for the military and produced it in the right quantity, and then released it in a civilian version (and plastic case) at a competitive price.

Of course, not everything can be worked out with reasonable costs. One of the sore spots is large FPGAs. The FPGA chip itself is not difficult to develop, but the synthesis software can be very complex. In such cases, it may be advantageous to purchase imported chips in the form of plates with a large margin, to test and package them. Probably that is how domestic FPGA 55764 and 55763 appeared - which are software compatible with Altera but have a different pinout.

In general, now the Russian electronics industry can develop and produce any military and space microelectronics (especially after the purchase of Micron equipment in 2007 and 2011), but for this, someone must order and finance this development taking into account the development and manufacturing period in several years old. Or directly, or through the state OCD. So if you hear in the interview of some manager the words “Here, the poor backward domestic industry doesn’t make us the right chips” it should be understood as “I’m too lazy to finance or get funding to create all the necessary chips”.

But of course, a lot of equipment made 5-15 years ago is built on imported key components - this is the result of our lost 90s, when everything was very sad in microelectronics (however, like everywhere else at that time). Forced use of imported components in the 90s and early 2000s is certainly bad and dangerous, but the choice was simple, either we are importing, or not at all. In recent years, it seems they have come to correct the situation with domestic military electronics, and it will be more and more difficult to find excuses for using imported components.

And we must remember that war is won first and foremost on the economic front. The one who spends resources more efficiently wins. Therefore, it is difficult to blame the defense industry for the fact that we do not develop “for ourselves” absolutely everything that the entire Western world develops together — everywhere compromises are needed.

There are similar problems with military electronics in the west - military chips are also expensive there because of the low volume (for example, the RAD750 is $ 200,000), and the recent scandal about the massive supply of fake chips for military equipment was not from a good life.

About bookmarks

Very often we hear about the "bookmarks" - the magic button, which can turn off imported chips. Of course, everything is not so simple - the electron is still protected from external radio signals, and the signal still needs to be managed to be sent.

But what is possible is to reduce the reliability of the chips supplied to us. As you know, reliability - already 10 years as a result of a compromise with speed and heat dissipation. And the ways to increase and decrease reliability are very well studied: it is enough, for example, not to add 1% copper to aluminum compounds, or to anneal a microcircuit not in deuterium, but in hydrogen - and the service life is reduced by 10 times. Whether it will detect testing is another question.

In addition, the use of imported components in key systems is a dependency that can be costly (and is already costly, because you have to buy such components). Well, buying ICs abroad - we help foreign enterprises to solve their problems with small-scale production.

Some danger exists in the manufacture of microcircuits in domestic factories, if masks are made abroad - not only can theoretically be copied and studied, masks can be modified - companies like Chipworks are quite capable of it (for example, reliability can be reduced by disrupting majority logic or damage to the structure of error correction). It will be very difficult to detect such modifications - I'm not sure that the ready-made masks thoroughly compare with their electronic original.

What is the story of the arrest of people selling chips from the US to Russia?

By itself, exporting from the USA even military-space class microcircuits is not a problem - this can be done quite legally with the passage of a corresponding bureaucracy. The problem is the provision of fraudulent end-use documents to avoid unnecessary difficulties in obtaining the necessary permits.

The list of microcircuits (p 20, probably incomplete, at least a couple of items are missing at the beginning of the list) caused everyone to be perplexed - there were no space ones, from the coolest - EV10AQ190CTPY - Quad 10-bit 1.25 Gsps ADC.

But the most important thing in this story is that all these comrades and firms were monitored from the very beginning - all correspondence, conversations, and so on. Accordingly, we read the previous section of the article on “bookmarks” and a possible decrease in reliability.

Discusses the topic here (for registered users only).

Summary

The use of civilian microcircuits in space is limited by the snapping effect, and perhaps at best in low orbits. In high orbits and in deep space, special radiation-resistant microcircuits are needed, since there we are deprived of the earth’s magnetic field protection, and the meter of lead will not save cosmic radiation from high-energy particles.

It is unrealistic to make “bookmarks” with remote disconnection in imported chips, but it is quite possible to reduce reliability and service life.

After the dark decade of the 90s, in recent years, relatively complex domestic microchips finally began to appear - microcontrollers, FPGAs (small ones of their own, large ones from imported plates with their own packaging and testing), processors (Comdivories, Elbrus, MCST R500, Milandrovskie ARMs). Work is underway on the conditional "breakthrough" technologies for military purposes (rad-resistant FRAM).

So if the end of the world does not happen this year, ever less military and space technology will come out with “Made in Taiwan” microcircuits and, less often, automatic interplanetary stations will surf the ocean.

I will be glad to hear about errors and additions - they will definitely be needed here.