You know it’s a good teardown when [Michel] starts off by saying to not ask him where exactly he got the guidance section of an FGM-148 Javelin from. This shoulder-launched anti-tank guided missile (ATGM) is a true marvel of engineering that has shown its chops during recent world events. As a fire-and-forget type guided missile it is designed to use the internal IR tracker to maintain a constant lock on the target, using its guidance system to stay exactly on track.
Initially designed in 1989 and introduced into service in 1996, it has all the ceramic-and-gold styling which one would expect from a military avionics package from the era. Tasked with processing the information from the IR sensor, and continuously adjusting the fins to keep it on course, the two sandwiched, 3 mm thick PCBs that form the main section of the guidance computer are complemented by what looks like a milled aluminium section which holds a sensor and a number of opamps, all retained within the carbon-fiber shell of the missile.
In the video [Michel] looks at the main components, finding datasheets for many commercially available parts, with the date codes on the parts confirming that it’s a late 80s to early 90s version, using presumably a TMS34010 as the main CPU on the DSP board for its additional graphics-related instructions. Even though current production FGM-148s are likely to use far more modern parts, this is a fun look at what was high-end military gear in the late 1980s and early 1990s.
Thanks to [Claus] for the tip.
I wonder how small you could fit all this today? Maybe an Infiray module and any fast CPU like ESP32 or STM32.
ASICs most likely. Gotta keep price reasonable for a throwaway.
ASICs absolutely cannot be used due to their ease of reverse engineering and the subsequent countermeasures that would crop up to the specific targeting algorithm being implemented in hardware. FPGAs are very coinvent to use because the bitstream can be ephemeral and not even stored on the projectile as well as updated for new target profiles.
ASIC doesn’t mean the device can’t have firmware. This board seems to be full of custom chips. But the small volumes likely make FPGAs way more cost effective.
This is true, but the amount of planning and red team review that would go into ensuring that the portions of fixed logic used can’t inform adversaries of what the firmware is doing would be immense.
To be clear, I’m talking about using ASIC for guidance, not ancillary functions.
This is an anti-tank weapon and will likely be captured launcher and all during warfare.
Ephemeral or not that bitstream will be captured.
People are probably overplaying the sophistication of the targeting it likely has 3 modes.
1. Fly directly at the target.
2. Attempt a top down attack.
3. Aim to hit it’s front or side perpendicularly (probably least important).
Tanks aren’t that fast compared to a missile they may as well be stationary.
For the most part it’ll be about sticking to the precalculated route determined at launch so the processing on the missile is only concerned with that.
I have no idea what the specs of something like this would be but do think the above accurately describes the project needs.
Two modes – top attack and direct attack. There is no ‘fly around to the side’ attack available for any missile.
And the missile tracks the target and course corrects in real time, not flying a precomputed path
At 4 km it can take almost 30 seconds to reach the target – plenty of time for a tank to move enough to turn a kill into a clean miss.
Good to know @Agammamon, to be clear I’m literally making this up but trying to think logically.
Regarding the 30s flight time, of course, and there will be wind and there will be motor problems with the rocket etc.
You will always need course correction. We all agree that the missile has a seeker.
In this domain of fast moving rocket and slow moving tank the targetting algorithim could assume the tank to be stationary with a drift in the missile course. I don’t know what missiles really do, I’m guessing what I’d make the device do.
Again I don’t imagine anyone is trying to hit a fast moving tank (more than say a few miles per hour). Tanks probably are most vulnerable when parked in a street surrounded by buildings. Not flying down the highway at 50 mph.
Regarding the third attack mode, again I totally made it up. I imagined the most complex attack mode I could that is the warhead would turn to hit it’s target perpendicular, mind you this is not “flying around” this is a turning even if only in the air before the warhead is triggered.
program the FPGA from the launcher
program the launcher from a satellite downlink
a bit more fragile but near impossible to capture bitstream by traditional means.
This weapon actually seems quite remarkable in it’s design and engineering well modern variants anyway.
https://www.youtube.com/watch?v=SUdHzKRiBX8
@hjf
Satellite data links could be subject to jamming and ASAT warfare. Cryptography is the more likely use of protecting the bitstreams; think Stargate DHD
missile alone is $80k, cost is not a factor here
Maybe not for the government, but for the manufacturer trying to maximize how much of that $80k goes to their pockets it does. As long as reliability isn’t hindered enough that anyone notices they probably don’t care
For USG defense contracts to manufacture munitions, the profit margin is usually stipulated, and any theoretical savings between processors in volume manufacture will accrue to the DoD. The customer will absolutely be made aware of and demand validation of any cost/benefit tradeoff in processor selection before they sign off on it.
I don’t think taping out an ASIC for such a low volume device makes sense.
infiray does make some nice microbolometers that would be used for this purpose but they are well north of 10,000USD. I don’t think an ESP32 or STM32 have the performance to implement the targeting algorithms that were being used even 20 years ago, and the algorithms used today are significantly more sophisticated.
I think you are wrong about that ESP32 or STM32 performance thing, it’s quite surprising what you can achieve with simple means.
The IR though would have to have a real high update rate, and that tends to cost.
I don’t doubt its possible to implement an algorithm that can guide a missile of similar velocity in the most basic sense on ESP32/STM32, but that you cannot implement the algorithms being used today let along 20 years ago on them. The CPUs are simply too slow, they will not reach the frequency response of the FPGA based system; also since they’d be overloaded just doing basic guidance, counter EW subroutines could not be implemented.
Like Paul says, a CPU is not going to be up to the task no matter. It’s a CPU vs. GPU type of situation here for the image processing. An FPGA is simply just the right tool for doing the tasks of the DSP-like algorithms involved in the target identification.
FPGA.
Given that it could be done on two boards of late-80s chips the actual gate count has to be pretty tiny by contemporary standards; but I’d be curious how much the timing requirements would mess with your options:
You’d certainly have options; it’s not like we’ve stopped having high speed machine vision needs; but it would exclude a lot of the “who cares if running motion detection and object recognition on the 30fps surveillance feed is running 750ms behind real time?” options that you can do in software on relatively feeble CPUs; presumably in favor of contemporary DSPs or software running on CPUs or GPUs sufficiently fast that they can just brute force their way to very low latency.
Navair Spike.
5.4 lbs, 1 lb warhead, 2 mile range.
Passive Thermal imaging and semi active laser homing.
They used ASIcs there.
Incoming ITAR violation notice in 3.. 2… 1…
No one cares, this is europe.
I found this video a little bit disturbing.
I’ve seen earliers military missiles internals and they had a very different design : several circular-shaped pcb, all densely populated and closely stacked. Obviousely the axial balance was taken into account. Most ic were also not referenced.
On the video the missile seems to be too big for the electronic, with a lot of wasted room that could be used for a bigger warhead / bigger propelant room and therefore a higher range.
Also amazed by the guy who “bought” some recent military stuff and put a tear-down on yt… who can do that ?
The missiles internals I saw were not accessible in any way other by naked eyes and nobody could touch it.
Sorry but this video looks like misinformation.
Believe what you will, but he mentioned how hard it was to come by and that it simply can’t just be “bought” (though _everything_ can be bought if you have enough money, so it shouldn’t come as a shock), he has a long running history of disassembling military hardware, he doesn’t live in an authoritarian state that arrest people who does “stuff the government might not like”, and this is not “recent” hardware (developer in the early 90’s).
I’m pretty sure that by the time anything appears on YouTube, foreign entities have already long before-hand procured, copied, placed into development a duplicate.
It hasn’t been that long ago that borderline military secrets were published in trade magazines and then hand-delivered to the same foreign entities.
Hell you can buy GP-5 gas masks off Etsy. For costumes, of course.
The missile almost certainly came from Ukraine. The US and other NATO countries have been sending anti-tank missiles over by the container load and pulling them all from older stockpiles. The White House and war hawks in Congress have resisted all calls for audits or accountability on these deliveries of military hardware so no one is tracking where these missiles go once they’re unloaded. Corruption in Ukrainian government and military circles also has a long and well-documented history.
In short – the dates line up, we’ve sent tens of thousands of them, no one is checking where they’re ending up and Ukrainian officials are famously corrupt.
Javelins were famously deployed in UA even before 2022, so if you genuinely think “audits” later would have prevented this teardown you’re frankly dumb. Also, all signs indicate munitions are tracked, just that the tracking is classified. GRU interns not being able to track them off public internet doesn’t make it “no one knowing it”.
Looks to me more like a discarded part from a missile upgrade. The rest of the missile was kept with new electronics.
This is a late 1980s/early 1990s version of the guidance computer. Any active ATGMs from the 90s would have been used up during the conflicts the US was involved in since 1996. As said by cplamb, this is most like part of an avionics upgrade. There’s no way in hell that the US or anyone else using Javelin ATGMs today still has versions with 1989-era electronics in it.
Your other accusations are equally unfounded/irrational considering the available evidence.
War is a hell of a thing, and being as RT’s apparently preferred entity in the invasion of UA initiated it, the most efficient way forward is for that entity to stop pursuing it. At that point, the costs and stakes for any friction and inefficiencies in UA’s government would be enormously reduced, and that entity could tend to its own affairs.
>> missile seems to be too big for the electronics, with a lot of wasted room that could be used…
Ironically, there *has* to be a lot of wasted room, because there needs to be a big void in front of the shaped-charge warhead for the metal jet to form.
The explosive itself is shaped like an open cone, lined with a ductile metal. When the warhead is detonated from the rear, the explosive front racing forward extrudes this metal into a hypersonic jet that does the actual penetration. think of slapping your hand down on a ketchup packet… but with way more death.
Though the jet is very powerful once formed, the actual formation process is easy to disrupt, say with reactive armor.
Hence if you look at the internals of a typical anti-tank missile, you see a lot of weird stuff. Up in the front where you’d expect lots of bang, there’s only a little warhead, which sets off any reactive armor and deflects the shockwave away. Then there’s a lot of empty space till you get to the big warhead, which goes off at the same time. In front of the main warhead there’s that nice big empty void where the penetrating jet can form in peace.
ITAR?
exactly
Why would the laws of the United States apply to people in not America???
They shouldn’t, although it would be nice if they had their own such rules which they may have, I don’t know.
Anyway, after President Oatmeal Brain’s (what a national embarrassment) catastrophic, bungled, total withdrawal from the stupid, multi-trillion dollar mistake called “The US longest war in history in Afghanistan” rather than leaving behind just 2.5k Air Force, Army Aviation troops and contractors at Kandahar Airbase to continue to provide air support for Afghan forces, a far more useful, but likely perpetual, use of troops to make up for the error of being there in the first place than the “why in the hell are we still there over 75 years after WWII” US NATO forces in Europe, our adversaries probably have all kinds of our military technology to dissect (or use) from the items we abandoned there.
I don’t think it’s a security feature. It’s obvious to me that they used FPGA to make the CI/CD pipeline more efficient. So far the primary users of the Javelin platform seem to be very happy but the secondary, tank-based users just don’t seem to be delighted by the UX. No doubt the product owner will push the devs to improve this before the next demo so it can be included in an upcoming release.
A TMS34010 CPU?
So an $80,000, state-of-the-art anti-tank missile uses the same chip as Mortal Kombat and NBA Jam arcade machines…
There’s a little AI in their thinking our world is a video game.
If there’s an OTS chip that can be ‘militarised’ (e.g. tested to Mil-Spec norms), then why not use it?
Someone run doom on it already