Hacker News Reader

Comment by taylorfinley | Hacker News Reader

Call me cynical, but the ceo of a major defense contractor saying it’s going to take years of testing and R&D to refresh a circuit board design from the late 80s is just a cancerous grifter piggy squealing at the trough. You could crowd source a replacement PCB in a week if they published the design spec. Sure, testing is required, but call this what it is. And don’t try to tell me we can’t share missle guidance tech from the 70s, even the model rocketry community has open source thrust vectoring.

View on HN

It'd be interesting to see what'd happen if the military just pivoted into having top consumer electronics companies manufacture weapons using best-available current technology, perhaps designing them from scratch. The "peace dividend" from the end of the Cold War was literally spent on the Internet - my wife's father used to make guidance systems for SLBMs in Silicon Valley, now his son and son-in-law both work for Google. The rapid progress in software, networking, and electronics over the last 30 years is because a lot of the top engineering minds who otherwise would've gone into defense are now going into tech.

This is also what happened in WW2 - the bulk of the U.S. war machine wasn't built by the shipyards and artillery vendors that had supplied WW1, but by new consumer giants like GM, Chrysler, Raytheon, Boeing, Hewlett Packard, Fairchild that pivoted from consumer markets to the war, or defense-specific startups like Grumman & Kaiser that were founded in the 10 years leading up to the war to capitalize on new technologies. WW1 technology was generally obsolete by WW2; the war was fought in a fundamentally different way, with different suppliers, because tech had advanced.

So what happened to USSR MIC talent pool? We poached them too?

A lot of them, yes - I work with a number of ethnically Russian engineers who probably would've gone into military hardware had they not emigrated to the U.S. In a company founded by someone whose parents were mathematicians who fled the then-moribund Soviet Union, no less. Bunch of my favorite computer games are written by programmers from former Soviet-bloc countries who now can make a lot more money writing projectile physics for computer games than guidance systems for actual projectiles.

Definitely not. It’s a total coincidence that the F-35 VTOL system is very similar to the innovative Soviet one used on the Yak-141 …

I think that happened before or around 1991 so is less relevant to the present discussion

You can already see it (somewhat) in the drones. It was already a thing in the middle of 2010s, now everyone is not only taking notes but already producing drones from OTC components. Seen those drone loght shows?

having top consumer electronics companies manufacture weapons

Do you want your weapons to come from China and Taiwan?

They already do:

https://archive.is/EgDT1

This has been a known problem for 10 years, but supply chains are hard, and oftentimes the alternative to buying from foreign companies is having a worse-performing product.

> if they published the design spec.

Apparently this is something the US military is doing with the F-22 replacement, NGAD. They will require ownership or a perpetual license or similar to the NGAD spec and blueprint, rather than allowing a defense contractor sole ownership of the design. Gives them a little leverage and ability to source multiple manufacturers, diversifying its supply line.

The US government has been requiring those rights for decades. They probably have them to the Stinger as well, they just assume that no one else can start a manufacturing line faster

That is a far cry from the DoD publishing the specs.

> They will require ownership or a perpetual license or similar to the NGAD spec and blueprint, rather than allowing a defense contractor sole ownership of the design.

This has been standard in all US government contracts for decades. As long as I can remember. You must give the US government an unlimited perpetual license.

You can't crowd source mil-spec ICs. Many have no commercial equivalent and even when they do they often use unusual packages that can't be sourced anywhere else.

I completely agree with this. Mil-spec and certified ICs are a completely different beast.

As we can see, they have to survive decades in storage, need an absurdly high pass rate, need to survive incredible abuse in battlefield situations AND survive who-knows-how-many G’s of acceleration (/ vibration) catching up to a target.

That sounds a lot like car grade electronics, TBH.

I suspect that there is some combination of military contractors forgetting how to move fast, and the military being so wrapped up in process instead of results at play here.

Acceleration isn't the biggest problem on such light devices, vibrations are. Those boards are exposed to high frequency vibrations that could detach or crack components because the pcb resonates inflecing rapidly back and forth.

A stinger missile has a rocket motor who's back blast can easily cause serious injuries or death to a grown adult man; every manual for shoulder launch munitions calls to make sure that the backblast area of a launcher is clear of friendlies.

There's 1.02kg of HTA high explosive to launch shrapnel with. The M67 fragmentation grenade carries 0.182 kg of high explosive, kill radius of 5 meters, injury radius of 15 meters.

Would you want, "Move fast and break things," development attitude with an electronically detonated bomb less then an inch from your head?

There’s a difference between being able to iterate quickly and move fast, and the “move fast and break things” ethos.

The P-51 Mustang went from conception to flying prototype in 179 days by a company that had never built a fighter before. Fighters aren’t famous for being lower stakes than bazookas.

I know that complexity doesn’t scale linearly, but we can use the Dragon vs Starliner development to make the point that legacy military contractors have forgotten how to move with purpose. The Starliner capsule has been in development for longer than the entire Apollo program, is years overdue, and still has no scheduled crew test flight. The dragon, meanwhile, made by a company that famously moves fast, has successfully, and safely, been flying crewed missions for years. That seems pretty high stakes to me…

Yes you can. Google is literally doing it now, and it's free! https://developers.google.com/silicon

On the cost and time scale of military hardware, it's not out of the question for them to make new copies of the old asics.

Most of those ics can be replaced with a single FPGA today.

Also, the DoD actually owns and operates their own semiconductor foundry in Sacramento that is specifically and deliberately setup to manufacture old and obsolete ics needed by the military.

So yes, there are challenges to sourcing them, you can't get them from digikey, but there is a whole infrastructure built up to keep supplying and or replace these things with fpgas or whatever

Edited - typo

My understanding is that it's the fabrication process that's different for mil-spec ICs, not the architecture. The chips have to survive intense acceleration, intense heat, transport in dust-laden inhospitable climates, years of storage afterwards, and then function with close to 100% reliability afterwards. These are all conditions that you would not subject your data server or iPhone to.

My understanding is that the fabrication of the dies are exactly the same, the actual silicon.

It's usually the encapsulation, enclosure and testing that's different. Those are really not big deals, it's not trivial, and they can't be interchanged with consumer product, but it's not as if you are making a fundamental different thing.

There are many mil spec fpgas available

> intense acceleration, intense heat, transport in dust-laden inhospitable climates, years of storage afterwards

What of these points can not survive an iPhone motherboard without battery and display?

The kind of vibration produced while strapped to a rocket motor is completely different that what iPhones normally experience.

Most people underestimate how violent that environment can be.

Soviets solved that by pouring epoxy over.

We did too, apollo was all hand wire wrapped and then epoxy potted

Or you accept slightly lower standards and get them for a fraction of the price in a fraction of the time.

And than a Stinger fails while being fired from, say, a F-35. And it fails catastrophically, resulting in the loss of the aircraft and, worst case the pilot. Which can happen, a stinger is tube filled woth rocket fuel carryong a warhead designed to take down planes.

I can already see the HN comments: Pentagon and defence conteactors to stupid to conduct simple quality control. The software obviously sucked, but what do you expect from government or old industry... Let's just say that Raytheon has a pretty strong incentive to be able to scale up Stonger production, and despite claims of the contrary, they are pretty good at doing so.

> And than a Stinger fails while being fired from, say, a F-35. And it fails catastrophically, resulting in the loss of the aircraft and, worst case the pilot.

I'm guessing that Stinger missiles, just like any other machine, have a failure rate above 0% already.

If we can produce them more quickly, efficiently, etc., but that failure rate increased that might be worth it if we get to produce more missiles.

Or maybe the new missiles are cheap enough that targets we previously couldn't cost effectively engage can be added to the menu.

Even in war, "good enough" is still important.

We can just as easily lose that F35 because we didn't have any missiles to launch, so it didn't even take off.

Stinger is always going to fail when dropped from F-35 since it is a man portable missile.

https://en.wikipedia.org/wiki/AIM-92_Stinger

On the other hand, the F-35 doesn't use them.

There is a variant called ATAS (air-to-air stinger) that can be fired from helicopters.

That's mostly a function of packaging and QC, no? The temperature range I can see from a quick search isn't much different than the strict automotive range of -50 to 150C, which standard process nodes can hit.

How much of that is owing to anything special about the IC vs manufacturing technology of the day? Legacy chip designs built on older nodes are probably more robust just due to physically bigger wires and components. Wikipedia lists a 1987 node size as 800nm[0].

[0] https://en.wikipedia.org/wiki/Semiconductor_device_fabricati...

There are more mundane issues like lead-free solder. RoHS isn't US law and the military will not use lead-free parts because of reliability issues. Most DIL and QFP parts are sold with unformed leads because vibration requirements generally require higher standoff from the board customized for each application so the leads flex less and can be supported from below with elastoemric compounds. Good luck procuring such things without some significant cash and the right contacts.

I'm confused, you seem to be conflating pcbs and parts. The ics are soldered to the board with the lead, or lead free solder. The lead or lead free solder thing only applies to securing the parts to the board, the pcba assembly process, not the chips themselves. Am I correct?

Makes sense on the lead height thing, the more you know.

Leads that come pre-tinned have to have a compatible solder on them. When you buy components now, the suppliers declare whether they're RoHS (lead-free) or not. You can't easily buy eutectic (lead bearing) tinned parts anymore because it's easier to have the global supply chain conform to the EU regulations. Note that military parts often come with gold plated leads which you can use your choice of solder on. You can't easily buy such things anymore from commercial suppliers.

Thank you!!

Not quite, the leads are (usually) coated with something. If it's pure tin it can grow whiskers from the top or even from under the solder connection, even if you use a leaded solder to attach it to the board. That's why even DIP parts come in leaded or lead-free versions.

There are packages which come with the solder, like BGA, which can (usually?) safely be reballed to use leaded.

Thank you!!!

couldn't the whole PCB covered in some resin to prevent whisker formation?

Metal is stronger than plastic

I can think of a few good reasons why one would not want to crowd source sensitive military hardware.

We basically do.

95,000 people worked on thr F-22 program. That's a crowd.

https://en.m.wikipedia.org/wiki/Lockheed_Martin_F-22_Raptor

The MCI can move fast if there's an incentive. Take the paveway gbu-28, 10 weeks from design to production use in war. Then there's WW2, and the B21.

It's the most American thing in the world to bid as high as possible at times like these.

Can too. In fact, I know someone who used to do a lot of teardown of parts for F14's.

Grumman at one time was making money hand over fist sourcing COTS microprocessors, acid washing off the identifying marks, flipping them over, then remounting them to final assemblies, and charging the Government a 2500% markup per unit until the deception was found out.

Frankly, I think the Government should 100% own the specs, and let anyone who wants to take a shot at building them cheaper do so.

Alao, in case of the stinger, ypu talk about compatability issies woth a myriad of other systems, oldy really old and new. After all, it can be fired as aanpad, from air vehicles, ground vehicles and from sea. The existing design is guaranteed to work with all platforms it can currently be fired from. Changing the design means loosing that guarantee, and would require a tremendous amount of testing.

So, there is a kernel of truth in saying it will expensive to recreate production tech to build current spec stingers. Also true that this kernel of truth is great opportunity to get more funding.

Why not?

I have no idea how publishing the exact specifications for a weapons system, and the presumptive flaws that can be exploited by countermeasures, would be a good thing.

The biggest challenge I see is the qualification process. If you want drop-in replacement electronics how do you guarantee that the functionality at the pin is an exact replacement for the original electronics?

Reply on news.ycombinator.com