3
\$\begingroup\$

According to this

article, even though it's 30 years old, ECL is more radiation resistant than normal CMOS chips by a factor of 1000. It is also less susceptible to DPA side-channel attacks. So in particular, are ECL chips used on military hardware/spacecraft current?

\$\endgroup\$
  • \$\begingroup\$ Interesting question. I would doubt it, given the more recent advances in making standard CMOS radiation resistant via redundancy, ECC, and careful layout, but perhaps in some specialized applications where radiation is expected to be extremely high? \$\endgroup\$ – Nate Strickland Mar 13 at 23:14
  • \$\begingroup\$ Sometimes newer/higher quality versions of older technologies like this are used in hardened electronics, especially those designed to withstant extreme, weaponized pulses in military use, although I know nothing of this tech in particular. I remember reading at the time (roughly 2007) that in some countries/jurisdictions it may be illegal to design/own EMP hardened electronics of some types for civilian purposes. The necessary methods and materials will change a lot depending on use case, such as how big and what kind of pulse must be withstood and available space. \$\endgroup\$ – K H Mar 13 at 23:42
  • \$\begingroup\$ ECL certainly has a high radiation resistance, but unfortunately is only used in SSI and MSI chip applications. If you were limited t building a logic solution with ECL it would be expensive (integration level, cost, power). Newer CMOS/FET based logic can be radiation resistant too. This may help you: www-physics.lbl.gov/~spieler/radiation_effects/rad_tutor.pdf \$\endgroup\$ – Jack Creasey Mar 13 at 23:46
5
\$\begingroup\$

The killer for CMOS built on bulk silicon is the current that radiation produces in junctions that are supposed to be providing isolation. One way around this is to build CMOS on top of an insulating substrate, such as sapphire.

The lowly RCA 1802 "COSMAC" processor was used in a number early satellite projects precisely because it was available in a SoS (silicon-on-sapphire) process that was inherently radiation-resistant.

\$\endgroup\$
  • \$\begingroup\$ Old 's CMOS was high impedance >300 Ohms and now is < 10% of this. yet smaller junctions can be damaged more easily by cosmic particles. My 1st computer by HP also used SoS in 1978.which gave higher speeds. \$\endgroup\$ – Sunnyskyguy EE75 Mar 14 at 3:09
1
\$\begingroup\$

You need to be very careful when you talk about "radiation". The article you linked is talking specifically about cosmic rays, which are single nuclei that have extremely high energy. When one of these nuclei passes through silicon it frees many electrons, which may then be collected by any existing electric field and result in unwanted current. It is generally true that bipolar technologies are less susceptible to cosmic rays. On the other hand, it's pretty difficult to build a modern computer platform using bipolar technology.

However, there are several other important kinds of radiation. Space environments are also subject to low-intensity, long-term radiation effects from high energy electrons. The magnitude of the radiation depends strongly on the spacecraft's orbit or flight path. Electronics used in nuclear reactors, nuclear medicine, or nuclear weapons may be exposed to gamma rays, x-rays, and/or neutrons. Each of these different kinds of "radiation" may produce different unwanted effects in electronics. There is also a problem with transient dose effects in some circumstances, but I won't go into that here.

Sometimes bulk CMOS is the right choice for radiation-hardened electronics, sometimes SOI CMOS is better, and sometimes bipolar technologies are better. If you really want to know the current state-of-the-art in unclassified technology, I recommend the IEEE Transactions in Nuclear Science, particularly the annual proceedings of the Nuclear and Space Radiation Effects Conference.

\$\endgroup\$
-2
\$\begingroup\$

Your assumption that CML is best for Rad hardening and current signatures for side-channel attacks. This is based on old information which is outdated.

TI's recent report contradicts your assumptions on Rad hardened technology indicates MOSFET now are more robust than BJT's to displacement damage dose (DDD) levels. https://www.ti.com/seclit/ml/sgzy002/sgzy002.pdf

Although ECL/CML are still immune to the side-channel attacks as the current signature is nearly constant non-existent with differential currents. Differential mode switched currents are still common communication channels today, as they improve immunity to EMI as well.

The real improvements in lower silicon impurities are what has made CMOS more robust as well as lower impedance. (25 Ohms in 3.6V) 50 Ohms on 5.5V families.) the impurities are excited to cause partial discharge (PD) and latchup. I have frequently discussed how PD works in this forum for damage, but in this case it is semiconductors.

===================================================

(Opinion) > I have not seen any Rad hardened ECL/CML in recent decades.

but in 1985 Honeywell's foundries had Rad Hardened ECL VLSI with 3500 Gates in BGA packages.

enter image description here

Rad. Hardened CMOS seems to be the preferred choice.

enter image description here

"The HMXADC9225 will not latch up due to any of the above radiation exposure conditions when applied under recommended operating conditions. Fabrication with the SIMOX substrate material provides oxide isolation between adjacent PMOS and NMOS transistors and eliminates any potential SCR latchup structures. Sufficient transistor body tie connections to the p- and n-channel substrates are made to ensure no source/drain snapback occurs."

\$\endgroup\$
  • 1
    \$\begingroup\$ This would be good news, if it was not so terribly outdated, by several generations. We now have oscilloscopes that can sample at 100GHZ real time, so I know the military has stuff at least that good. \$\endgroup\$ – Sparky256 Mar 14 at 1:59
  • \$\begingroup\$ Do you know of any Rad hardened 100GHz DSO? And why does that comment make this irrelevant? My opinion was they do not supply CML as CMOS has improved. since CD4xxx in 1985 @Sparky256 I don't see your logic \$\endgroup\$ – Sunnyskyguy EE75 Mar 14 at 2:47
  • \$\begingroup\$ Since we get the military hand-me-downs, they likely do have a rad hardened 100GHZ DSO, but it is not public news. Military electronics and software are redacted until they do not exist. I did not downvote you, so others have same or stronger opinions. \$\endgroup\$ – Sparky256 Mar 14 at 2:52
  • 1
    \$\begingroup\$ People without opinions who downvote are clueless Here is TI's Rad Hard brochure ti.com/lit/sg/slyt532f/slyt532f.pdf \$\endgroup\$ – Sunnyskyguy EE75 Mar 14 at 2:54
  • 1
    \$\begingroup\$ THose who downvote without valid comments may lack the competence to do so. \$\endgroup\$ – Sunnyskyguy EE75 Mar 14 at 3:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.