TX2 surviving extreme mechanical force

Hey guys

I’m working on a product utilizing a TX2 that will be subject to sudden shocks between 100-300 G’s, and am trying to think of a way to keep it from breaking. Anyone have any ideas?


Is weight an issue? Is physical volume an issue? What kind of temperature range is the environment (what kind of cooling requirements are there)?

Should’ve mentioned all this. Thanks for asking.

Weight is an issue - less is better (it will be carried for extensive periods of time). Temp range is outdoors, all seasons. It will be in an enclosed case, but the case will be exposed at times to direct sunlight, and it will be doing real-time convolutional neural network processing, so it will be running at full-power when in use. So I imagine it will need a fan or two.

What carrier board is being used? Is it the dev kit? Which connectors are used? Will the dedicated camera socket be used? What buttons need direct access? Will the Jetson be operating while the shocks are hitting it?

EDIT: Forgot to ask, will the force be from random directions, or primarily in one direction? More information on what’s causing the shocks might help.

Not sure on carrier board; looking at the Orbitty right now for prototyping, but am researching custom design. For the initial prototype, just a small hdmi screen, USB camera, and a few hardware buttons will be connected. In the longrun, the dedicated camera socket is likely to be used. The Jetson will be operating during shocks. Shocks are predictable and one-directional, caused by gun recoil forces (following the same plane as the board itself, e.g. it will be pushed longways).

The following is a bit of “out of the box” thinking, and may not be practical by itself, but gives you an idea of what to consider (it’s the kind of challenge which is entertaining to think about).

Are you familiar with conformal coating? This is typically used to seal against moisture and is more or less a soft rubbery coating which is non-conductive. For all-weather use this is almost mandatory, and for liquid immersion cooling it is a good idea (some coolants are mostly harmless, but you never know till the project has been tested for long term). Typically you would conformal coat everywhere which does not need a direct connection, e.g., you wouldn’t coat the heat transfer locations (such as top of CPU or a TX2 module heat transfer plate, nor connectors).

If you conformal coat then you can do liquid immersion cooling. If you suspend the Jetson via springs at the usual mounting points, and then use liquid immersion you could get rid of the cooling fan and have what is essentially a hydraulic shock mounting mechanism. The mechanism as a whole could be further mounted inside some sort of foam rubber style mount (essentially a hammock or foam rubber bag) or even recursively add more spring mounts to the outside of the original container and immerse that too inside of a separate liquid chamber. Immersion coolants may also have the effect of an antifreeze, thus providing protection against colder temperatures (though if this is to operate at altitude decreased pressure might require hermetic sealing).

In materials science part of the mechanical strength behavior is given a rating called modulus of elasticity. This is essentially a rating of how much the material flexes under stress. Materials with less elasticity tend to be more brittle versus materials with more elasticity being spring-like. You will find that within a given thickness of overall material that many layers of smaller thickness survive better than a single thicker layer. You will also find that in many uses having a modulus of elasticity which goes from less flexible to more flexible at a uniform rate will increase survivability beyond that of a uniform elasticity.

If shocks are sufficiently intense you might consider many containers-within-containers where the suspension is stiffer in the outer container and softer in the inner container. Liquids are nice because the material occupying the center does not require a seal the way pneumatic suspension does. You can perhaps control how “stiff” the suspension is by how much clearance your inner volume has from the sides of the outer container.

I suggest the liquid immersion in part because the fan is one of the weakest links in reliability. I’m not sure which immersion liquid you would use, but since the TX2 does not produce much heat even at full performance you probably won’t need a “great” coolant, you’d just need one which is reasonably inert, reasonably ok at cooling, and works at your required temperature range. You might even consider something like ordinary unflavored gelatin since it can be formed around a container in any shape you want, or dental impression materials (such as alginate). Here is an example of a liquid coolant:

Sometimes mineral oil is used.

Despite liquids having significant weight you are not necessarily looking for large volumes. Any kind of mechanical shell which is immersed in another liquid container can be extremely thin and not require much strength…a very thin aluminum shell (almost as thin as a foil) would do the job and also supply some EMI shielding so long as the liquid has a more or less uniform pressure between inside and outside.

If your screen has any vibration its life will be in doubt even with some extreme steps. Sealing connectors and cables and anything going to the board could be difficult. Many of the screens out there use what is essentially a lot of fingers held in place only by friction.

I don’t know if any of the carriers you are considering are IPC class 3 (basically mil-spec), but you might ask what IPC class the carrier you are looking at is manufactured under (class 2 is good/“siginifcantly above average” quality, class 1 is junk, class 3 is extreme quality). If that kind of shock is to be survived long term you will need quality…the lower the quality the more your mounting mechanism will need to compensate. I’m going to guess that most of what you find is class 2 which will probably work, but if you see class 3 and need this to last a long time, go for that. I spent time as a class 2 inspector, and can tell you class 2 is mostly as good as class 3 unless there is a harsh environment.

First off, thank you so much - this is tremendously helpful information, and has given me a lot of direction for further research. Admittedly I didn’t notice the reply until now because I missed the email notifying me of your response - I guess I should learn to check the forum proactively.

I’ve read and re-read the response; this is so helpful and I have so many follow-up questions, but I’m also really tired at the moment and hesitate to ask too much because I should probably do it on a fresh mind. But two things come to me immediately:

  1. If I use liquid, how much would the resistance of an inner shell against the liquid on recoil potentially absorb shock? I know there are many variables - what liquid is in use, distance of travel, etc. But in general is this an effective solution (no pun intended) under any circumstances?

  2. With modulus of elasticity, can you provide an example of something that varies vs. is fixed? I get the idea that it can vary, but I’m not clear what that might look like in practice. I’ll also google this more and try to do my own research (I’m not a materials guy, so this will be fun to read).

Thanks again!

The trick of liquid suspension is not that it reduces shock. The trick is that the shock is a perfect uniform force across the surface of the entire assembly, and so long as pressure can be survived for that instant the board and components won’t be bent. Imagine that if you mounted a PCB via four mount points on each corner…the force could easily crack the board. So you would be tempted to add another mount point in the center. Add enough force, and this won’t be enough…you’d add more support points. It is a bit like calculus…you are approaching the limit as shock increases but support points become nearly infinite…there is no lever in a perfect suspension. The springs are just to return to center after being disturbed.

The actual amount of motion depends on the edges. If there is no clearance between the board edge and the walls you have a cylinder which is as rigid as if it were solid steel. If you increase the gap the flow around the edges increases the freedom of movement of the board within the fluid. This spreads the force over time. But remember you probably don’t want to do all of the work in a single liquid suspension…the liquid suspension chamber itself can be mounted via more conventional means. Having a thin suspension layer implies no more fan is required and you probably don’t need a lot of liquid. More would of course give you more room for error. It’s probably something you need to experiment with.

One more thing I just thought of…surface tension and viscosity matter. Less surface tension against the outer cylinder and less viscosity implies a better result for a given gap between your inner hardware and cylinder walls.

Modulus of elasticity is actually used extensively in dentistry. The dentin of a tooth has a certain amount of flex, and the enamel on top has almost no flex. If the tooth were just dentin it wouldn’t survive abrasion…if the enamel did not have a slightly elastic dentin backing it up its life would be far shorter due to cracking. This is almost the same issue with circuit boards…you’ll find an extra stiffening layer added to some of the more complicated boards (the TX1 and TX2 module in particular use a metal plate not just for heat transfer…this is partially a mechanical stiffener). The shock of repeatedly surviving gun recoil could exceed 300 Gs…some components, even when supported uniformly, might fail (I’m thinking of electrolytic capacitors and crystals…the components internally are different densities and produce stress on their mounting points), but if the entire structure is uniformly supported your chances of components not failing are maximized. A second form of dealing with shock by spreading this over time works in tandem with the liquid becoming an infinitely fine-spaced mounting system (which is why you might want to suspend the first container inside of a second container…you’re buying the quality of spreading the change in force over more time).

You will find plastics (often used in dentistry) have different abrasion resistance and flexibility. There is a tendency to use such materials in layers starting with the least flexible material being the outer surface, and the layers going inward being more flexible. Materials used for restoration have part of their life determined by choice of materials, e.g., porcelain fused to a hard gold alloy will have a better seal and mechanical strength improvement over pure porcelain.

If you’ve ever been into metallurgy (or simply watched some television shows) you’ll see that a higher quality knife or sword uses a softer core on the inside, and the outer layers are quenched and hardened…this inner core has a higher modulus of elasticity, the outer layer is hard and resistant to dulling, but would be brittle and snap if not layered on top of something slightly more flexible. This could be extended to multiple layers where outer layers are thinner and harder and inner layers become softer…more layers with gradual changes in modulus of elasticity will give you a nearly indestructible metal for something like a sword. This is accomplished by either change in material composition or via change in metal grain structure. Plastic polymers change based on how much linkage there is between chains (this is also why some plastics crack more with age versus other plastics…it’s a case of cross-linking continuing very slowly after it appears the polymer has finished setting…it never really finishes setting).

Reading your posts has become the most interesting part of my day these last few days. Thank you for your insights and perspectives. I remember watching a video on how clay was used to add curvature to Japanese swords by protecting one side from heating while leaving the other exposed. While they’re going for other effects as well, I understand your explanation of modulus of elasticity (and never appreciated the importance of variability in material give until now).

I’ll do more research on liquid immersion cooling and suspension. I’m going to think more about the problem in general, but may be back if it’s alright. Thank you again!

Yes, the clay not only reduces cooling rate (and thus curvature), it also changes the metal grain structure, which in turn means more elasticity. I suppose you could say that the shock from a hard metal hitting a hard metal is dissipated over more time when backed by the more elastic (softer) metal.

Your goal in the Jetson surviving extreme force would be first to prevent flexing (the liquid suspension does this), and to dissipate the shock over time. The gap between the carrier board and the cylinder wall is equivalent to the very hard metal being backed by a somewhat hardened metal, and the second suspension of this chamber with a more elastic mounting point is what would complete reduction in shock sufficient for the board to survive. The greatest difficulty is probably in doing this with low weight and keeping the liquid suspension itself exposed to air or other cooling environment.

Like a submarine, I suspect a Jetson can survive a greater crushing force than it can a non-uniform flexing/lever type force.