Shock Testing, Round #1

One of the required ratings for our rugged VIUD is shock or acceleration and so we will need to build a test rig for this purpose. There are basically two ways of measuring the acceleration an object undergoes:

1. Install accelerometers on the object and measure the acceleration directly.
2. Take high-speed video of the object and indirectly measure the acceleration.

Our VIUD is far too small to install accelerometers on it which leaves us with the only other choice. 

High-Speed Video Camera

Using high-speed video necessitates buying a capable video camera but after much searching it seems that all of the dedicated high-speed cameras are far out of my price range of a few $100. Most vendors never list prices on their websites and instead say "Contact Us for Quote" which is just code for "really expensive", at least many $1000s if not $10,000s. 

Fortunately, after some further research I found out that Casio has a line of consumer cameras, the High Speed Exilim models, that support high speed video capture. Some models, including the ZR200, support frame rates up to 1000 fps at a resolution of 224x64 pixels: small in size but hopefully fast enough for our purposes. The price of around $220 is also nice although I hit a small issue when trying to purchase the camera: Casio doesn't sell their cameras at all in Canada. Fortunately I was able to eventually find a vendor in Japan that would sell me one (a lot of American vendors in general don't ship outside the USA).

An example video from one of the first few shock tests is below (prototype #3 at an impact velocity of 130 km/hr) :

The low resolution is immediately apparent but it is still usable for our purposes. Not surprisingly it takes pretty brightly lite environment to get a good video image at 1000 fps. The black & white pattern in the background is a scale (5 cm per large band) to give us the real world measurements within the video. More shock testing videos can be seen in our VIUD YouTube channel.

Shock Test Rig

The idea for the shock test rig is to make a custom air gun that fires the VIUD at a desired rate of speed at some target. The basic design of an air gun is pretty simple:

• Barrel for firing the projectile
• Valve to fire the gun
• Source of air under pressure

Unfortunately, my first naive attempt failed wonderfully: I simply hooked up an air compressor to a 1/2" ball valve attached to a 4 ft length of 1" pipe. Technically you could say it works if you considering "working" as barely being able to push the VIUD out of the barrel. The air compressor and its 1/4" hose simply can't supply the air fast enough into the barrel to give the projectile enough velocity.

The air gun design needs to be changed to hold a volume of compressed air close to the barrel with the largest possible pipe diameter connecting them (ideally the same size as the barrel if not larger). I first considered using a length of 2" or larger pipe but could not easily source the material locally. The next option considered was a small portable air tank. Initially this seemed like it would not work as all air tanks I looked at used a small 1/4" hose but on closer inspection the actual connection into the tank was a larger 1/2" NPT which is hopefully large enough to give our air gun some punch. I was also lucky enough to find a 5 gallon air tank on sale at my favourite store Princess Auto for only $25. I must confess to purchasing much more than just the air tank during my visit there (a very dangerous place to visit).

With a better design the assembly is relatively simple and uses all off-the-shelf parts available at any large hardware store. Total cost of everything to make the air gun was under $150 not counting the air compressor which I already had.

The Finished Air Gun for Shock Testing

The basic parts and features as shown in the above image are as follows:

• Barrel - 4 ft, 1" pipe with Copper pipe insert
• Firing Device - 1" Ball Valve 
• Air Chamber - 5 gallon tank with a 1/2" NPT connection
• Air Source - Electric air compressor 
• Control - Pressure gauge, regulator valve and shut off valve

I added a Copper pipe into the original barrel mainly to provide a smoother surface: the inside of the steel pipe was very rough and was scratching up the VIUD pretty badly during the first few runs. It also served to narrow the barrel's ID a little bit resulting in around 30% higher velocities at a given air pressure.

After a few trial runs it was obvious that the shock testing rig design works very well. Projectile velocity even at the lowest pressure setting of 15 psi is around 110 km/hr (60 m/hr) which is fast enough to be a little scary. This should be more than enough to reach the VIUD's terminal velocity of 55 m/s (200 km/hr, 120 m/hr) and beyond should we need it.

Measuring Acceleration

With our shock testing rig and high-speed video camera we're all set with the exception of having to figure out exactly how to measure our VIUD's acceleration from the video. My first attempt was to simply save individual frames from the movie, paste them into Inkscape (a 2D drawing editor), measure the movement and somehow convert that into an acceleration. This was very tedious and didn't work well at all so it wasn't long before I was searching the Internet for an alternative.

It was a quick search as the first Google result led me to an article about measuring the acceleration of a jumping Arboreal Lizard and that quickly led me to Tracker, a free application used for measuring velocities and accelerations from videos. The interface is a little wonky (some things just stop working randomly) and the documentation is mostly the "figure it out yourself" type but it does exactly what I need and was pretty easy to pick up its basic usage.

Tracker: A Free Video Analysis and Modelling Tool for Physics Education

Shock Testing

The procedure for shock testing is pretty straight forward:

1. Choose an air pressure (higher pressure for higher exit speeds)
2. Choose a target type (wood, stone, etc...)
3. Record the impact using the high-speed camera
4. Analyse the video using Tracker to determine the speed and acceleration
5. Test the USB to ensure it still works (basic error checking and bit change tests)

Using one of the third prototypes for testing we have the following results so far:

Pressure	Target	Speed	Max Acceleration	Tests Passed	Notes
15 psi	Wood	80 km/hr	900-1000 g	Passed	Minor scratches from inside of barrel
15 psi	Wood	80 km/hr	1100 g	Passed
15 psi	Wood	80 km/hr	1200 g	Passed
20 psi	Wood	110 km/hr	1500 g	Passed
25 psi	Wood	130 km/hr	1800 g	Passed
15 psi	Wood	100 km/hr	1300-1500 g	Passed	Added internal Copper sleeve to barrel
15 psi	Stone	100 km/hr	1500 g	Passed
20 psi	Wood	125 km/hr	1900 g	Passed
25 psi	Wood	150 km/hr	2100 g	Passed
25 psi	Wood	145 km/hr	2000 g	Passed
30 psi	Wood	160 km/hr	2100 g	Passed
35 psi	Wood	180 km/hr	2500 g	Passed
40 psi	Wood	190 km/hr	1900 g	Passed	Broke the wood target!

So far the VIUD passes all shock tests with flying colours which is good news indeed. One surprising thing is the large accelerations involved. Even with video at 1000 fps the impact occurs in less than one frame which means that even these large accelerations may well be underestimates. This also makes it difficult or impossible to distinguish the difference between target materials. In theory a stone target should result in considerable more deceleration than a wood target but we would probably need a 10,000 fps or higher video camera to discern the difference.

Amazingly the last test broke all three layers of the 2" thick wood target. The first 1/2" MDF layer broke completely through likely due to the combined effect of all previous tests. The next two 3/4" Pine layers broke into multiple pieces along their grain. Looks like I'll need to make a sturdier target, probably out of multiple layers of MDF, hardwood and plywood.

Broken 2" Thick Wood Target After a Dozen Shock Tests (Side View on the Right)

An interesting note is that the few rugged USB competitors that do explicit shock ratings are only rated at 40-50 g which means our shock ratings exceed theirs by an amazing factor of 50!

An Interesting Question

The following question is one that I was wondering about recently:

At what height can the VIUD be dropped from and survive?

To answer this relatively simple question will take us down a variety of interesting paths which were not obvious at first. To consider how our USB drive may survive we need to first look at failure mechanisms from dropping it from a considerable height:

• Impact -- Drive fails when hitting the ground.
• Temperature -- Drive fails due to heating from travelling at high velocity in air (like a meteor).

We'll first consider the impact failure mode since if the VIUD can't survive the ground impact we won't have to worry about temperature effects at higher speeds.

Terminal Velocity

To consider whether our drive will survive impact with the ground when dropped from a given height we need to consider its terminal velocity. When an object is dropped in air it has two forces on it: the force of gravity down and the force of air in the opposite direction. The force of gravity on the object is essentially constant but as an object travels faster the the force of air resistance gets higher. At some point the force of air resistance equals the force of gravity and the object stops accelerating and reaches its terminal velocity.

The terminal velocity of an object is expressed in general as:

where:

V_t = Terminal Velocity in m/s

m = Object's mass in kg
g = Force of gravity in m/s²
ρ = Density of air in kg/m³
A = Projected area of the object in the direction of motion in m²
C_d = Drag coefficient

Most of these parameters should be straight forward with the exception of the drag coefficient. Depending on the shape of the object its drag will be more or less which in turns affects its terminal velocity. We can use the measured coefficients of known shapes to estimate what it will be for our VIUD:

• Long Cylinder = 0.82 (direction parallel with axis)
• Short Cylinder = 1.15 (direction parallel with axis)
• Cylinder = 1.17 (direction perpendicular with axis)
• Sphere = 0.47

Depending on how our rugged drive falls it could have a drag coefficient anywhere from 0.82 to 1.17. Since the coefficient is inversely proportional to the terminal velocity the smallest coefficient will result in the largest, or worst case, terminal velocity so we'll use C_d = 0.82 for the rest of the calculations.

Plugging in all our values (assuming worst case in everything):

m = 0.06 kg

g = 9.8 m/s²
ρ = 1.22 kg/m³
A =  0.00041 m²
C_d = 0.82
V_t = 54 m/s (195 km/hr, 121 mile/hr)

From this estimation we know that no matter how high we drop our USB drive it should hit the ground at a maximum of 54 m/s.

Terminal Velocity as a Function of Height

Our previous calculation was only interested in the VIUD's terminal velocity at ground level but for the next step we will need to know its terminal velocity at any given height above the ground. Since the density of air drops as we go higher the estimated terminal velocity will increase as higher altitudes.  

We first need to find the density of air as a function of height. We can use an analytic equation for the density of air within the troposphere, or roughly up to 15 km in height and there are also a variety of more complex air density calculations if you look for them. What we'll end up using is simply a table of measured air densities which also conveniently has the force of gravity which does decrease slightly within the first 100 km. More detailed tables are also available if you need finer altitude steps.

Using this air density table and the same VIUD parameters as before yields the following graph of the terminal velocity versus altitude:

We can assume for simplicity that above around 100 km the air density is so small that there is no effective terminal velocity.

Velocity of a Dropped Object as a Function of Height

The next step in the modelling is to figure out what the velocity profile of an object dropped at a specific height is. Assuming the object is dropped far from the surface of the earth there will be three basic areas to consider:

1. Space - No terminal velocity.
2. Upper Atmosphere - Terminal velocity begins to drop rapidly.
3. Lower Atmosphere - Object reaches a steady velocity of 54 m/s.

To determine acceleration in space we simply need to know the force due to gravity from Earth (we'll ignore all other astronomical bodies for simplicity):

where:

g_o = 9.8 m/s²

r_e = Mean radius of earth, 6371 km
h = Altitude above the mean radius of Earth in km

While the object is in space computing its velocity is relatively straightforward. The fact that the force of gravity increases as its approaches Earth makes it more complex but it can be solved with a simple numerical integration at regular intervals from the drop height to 100 km where the terminal velocity begins to take effect.

To figure out the net acceleration of the object once it hits the atmosphere we'll need to know the drag force which is where the prior terminal velocity equation was derived from:

Our net force on the falling object within the atmosphere is then expressed as:

F_net = mg_h - F_d

Since both the force of gravity and terminal velocity depends on height we'll keep doing a numerical integration solution. Graphing the velocity versus altitude for a couple of different drop heights yields the following velocity profiles:

Assuming we did everything correct in the solver the graphs appear to make sense. As the object falls in space it continuously picks up speed until it hits the atmosphere at around 80-100 km. From there its velocity begins to quickly decline as it hits more dense air. At around 10 km it reaches its final terminal velocity of 54 m/s falling at roughly a constant speed. 

We can do a quick check to make sure our model is outputting reasonable numbers. If we use the equation for the velocity of a falling object after a given distance, sqrt(2gd), and take the force of gravity at the midpoint of the 10,000 km drop height (3.1 m/s²) we find a velocity of 7700 m/s at 100 km which is reasonably close to the model's value of 8500 m/s.

Aerodynamic Heating

We're finally at a point we can find what we're looking for: the temperature of the object as it falls through the atmosphere at a high velocity, otherwise known as aerodynamic heating. While our VIUD is a relatively simple cylindrical shape it is still too complex for a simple modelling of aerodynamic heating. Realistically we would want to do some finite element modelling but this is beyond my current capabilities so we'll have to settle for some approximations.

All the simple aerodynamic heating models I've seen have assumed a thin plate travelling with the thin edge towards the direction of motion. This isn't too far from our VIUD but since it is thicker than a plate we could assume the heating would be more severe than that of a plate (how severe is difficult to guess).

where:

T_wad = Temperature of air just beside object, K

T_∞ = Temperature of air far from object, K
r = Recovery factor, estimate for air of 0.84-0.87
M = Speed of object in Machs
 = Heat capacity ratio, 1.40

We can get the normal air temperature as a function of altitude from the same data tables we used for air density. The heat capacity ratio doesn't change far from a value of 1.4 for air until you get into temperatures above 1000^oC which is irrelevant for this model. 

From this presentation we'll use the very simple approximation to find the temperature of the USB body: 

T_net = T_air + 0.5 (T_obj - T_air) + 0.22 (T_wad - T_air)

I'm unsure exactly how they derived this but I'm assuming it is only valid for a particular model and may not be accurate at all for our case. Unfortunately, all other methods I've investigated are extremely complex and a little beyond my limited understanding of heat dynamics (we'll look at alternate temperature models in another post). Since this model depends on temperature and speed of the object we'll do a numerical integration just like for our velocity calculation which yields a graph like:

Not surprisingly there are large temperature increases when the fast moving object first begins to hit the atmosphere. While it is difficult to tell from this graph the actual amount of time spent at these temperature is relatively small: for temperatures above 1000 K only 10 seconds are spent for the object dropped at 10000 km and 19 seconds for the object dropped at 1000 km. The object dropped from further up experiences a higher temperature spike but for less time as it is moving much faster (8 km/s compared to 4 km/s).

Since the melting point of Aluminum is only 933 K (660 °C, 1220 °F) we can guess that the objects dropped above 1000 km would have significant damage, assuming they survived at all. The object dropped at 100 km experiences a much gentler temperature curve only reaching a maximum of 410 K (137 °C, 280 °F) and only being above the boiling point of water for 10 seconds. This is far below what our USB drive has survived in temperature testing so we can be reasonably sure of its survival in this case.

Although, if we were to be optimistic for the higher altitude drops, at the high altitudes the air density is very low (a million times lower than that at sea level) and the amount time experienced at high temperatures so low (temperature rating for the VIUD at 2300 K  for 40 seconds) that a more accurate temperature model could reveal the VIUD can survive drops from higher altitudes.

Conclusion (Kind Of)

Although the accuracy of our temperature results is a little suspect and unverified we have at least a reasonable guess that dropping our rugged USB drive from up to 100 km will not destroy it from temperature affects. More research into an accurate temperature and heat model needs to be done to get more accurate results, assuming we want them.

References

• Aerodynamic Heating
• Aerodynamic Heating Simulation Paper
• Aerodynamic Heating Presentation from Queen's University
• Terminal Velocity
• Drag of Cylinders and Cones
• Basic Atmosphere Data Tables
• More Detailed Atmosphere Data Tables
• Falling Bodies in the Atmosphere
• Aerodynamic Drag
• Air Density Equations
• XKCD: Steak Drop

Designing the Cap

One of my pet peeves since USB flash drives first came about is loosing their caps. The first several flash drives I bought had separate caps with no way to store them when they were open. This inevitably led to the game of "try to remember where I put the cap" after using the drive. Later USB models either had a way to store the cap on the other end of the drive or no cap at all, like the slide-to-open SanDisk models.

One of the design requirements for our VIUD is some way to prevent the cap from becoming easily lost. There are a few ways in general of accomplishing this:

• No cap at all (slide-to-open or flip-to-open)
• Cap storage at other end
• Lanyard attachment

The first option is not applicable for our design due to other requirements, notably the need to be waterproof. The last option is useful to have but is not a complete solution...it doesn't work if you don't have the USB drive on a lanyard. With only one option left we now have to figure a way of storing the cap on the other end of the drive when it is in use.

Cap Storage

The initial design from essentially the very start was to have the cap physically fit onto the other end of the VIUD by way of a properly sized o-ring (see below image).

Prototype #2 Showing the Cap On and Off

By carefully tuning the size of the groove and the o-ring size and hardness the cap snaps on and off easily and doesn't fall off by itself. While I particularly like this design it does have a few drawbacks. One is that the exposed o-ring is vulnerable to damage, particularly at high temperatures like in flame tests. Not a huge deal though, especially considering the o-ring is easily replaced and only costs a few cents.

The other, more significant, drawback is that it lengthens the body by around 3/8". The USB flash drive is too large to fit inside the part where the cap is stored and while it doesn't make it that much longer it is one factor of many that contributes the overall length of the device.

Another thing to consider is the long term reliability of this cap storage method. While the second and third prototypes have survived one year of use very well I did notice the cap storage became a little looser over time likely due to the o-ring wearing out.

Magnetic Cap

I was all set on using the prior design for cap storage until someone innocently suggested trying a magnetic cap storage. The idea is simple enough: a small magnet inside each end of the VIUD which permits the cap to magically "stick" to the end of the drive when open. Using a magnetic cap storage would eliminate all the draw backs of the previous method in addition to letting us make the overall design look more symmetric and good looking.

There are a few questions to answer before accepting this new concept:

• What size magnets do we need to use?
• How close to each other do they need to be?
• Will the magnets affect the USB flash drive?

The first two questions were quickly answered by creating a few test pieces and doing some stress modelling. Ideally we'd like the magnets close to the surface but need to consider the strength of the body when under stress. Modelling and testing revealed that a distance of 0.07" would be acceptable. With 3/8" diameter rare-earth magnets their pull force is more than sufficient with a 0.14" separation and the body wall is still strong enough at this thickness.

On a whim I tried attaching the test magnet pieces to my fridge but they were just barely not strong enough to support the entire weight of the VIUD. Increasing the magnet size to 1/2" diameter made the magnet more than strong enough to hold the VIUD to a metal surface which is a neat feature.

In order to facilitate the magnetic cap attaching to the end of the drive a small recess was made in the cap that fits nicely to the chamfer in the other end of the body (see below image). 

Test Pieces for the New Cap Design With Internal Magnets

As for how the magnets may affect the USB flash drive: in theory they shouldn't affect the flash memory at all however I did test it. I placed two large rare-earth magnets at each end of a USB for one month. Testing before and after revealed no changes as expected. The only change was that USB connector became very slightly magnetized.

Temperature Issues

One thing I completely forgot about with regards to the magnets was the effect of temperature on them. Magnetic materials have a property called the Curie Temperature which is the temperature at which the material completely loses it permanent magnetism. A material's magnetism is actually gradually lost as you increase the temperature towards this point so in order to pass all our temperature ratings we will need to take a closer look at this effect.

The material used for the magnets are a Neodymium-Iron-Boron compound, the more common and cheaper form of rare-earth magnets. Neodymium magnets have a Curie temperature of 320 °C (610 °F) which seems high but they will actually begin to lose their magnetism at much lower temperatures depending on their size and strength. It is actually a rather complex thing to figure out but a small Neodymium magnet like the one we're using has a rated working temperature of only 80 °C (175 °F). If you exceed this temperature to around 100 °C (212 °F) the magnet's strength is reduced by up to 60%.

Now technically these temperatures are fine as it is unlikely that the VIUD would normally be exposed to temperatures high or long enough to affect the magnet significantly. Even if the temperatures were exceeded the worst case is the magnets losing some of their strength...hardly a major failure. However, it would be nice for our temperature ratings to be able to achieve higher temperatures for longer than permitted from the Neodymium magnets alone.

There are two solutions to the temperature issue: high temperature Neodymium and Samarium-Cobalt magnets. There are types of Neodymium rare-earth magnets that have a higher working temperature rating of 150 °C (302 °F) which should be good enough for our purposes. Even better are another type of rare-earth magnets, Samarium-Cobalt, which are rated up to 300 °C (570 °F). Which one we use mostly comes down to whether we can find a supplier for them at a reasonable cost.

Lanyard Attachment

While I wasn't going to have this option initially it would be good to add something to attach the VIUD to a key ring or lanyard like almost all the other rugged USB competitors have. Assuming the lanyard attachment will be on the cap there are a few possible methods of making the hole for it:

Choosen Lanyard Attachment Design

1. Diagonal hole
2. Two perpendicular holes
3. A hole sideways with extra material removed

The first two are the simplest from a manufacturing standpoint but the main drawback is that they interfere with the magnetic cap storage. I ended up deciding on the last option and while it is slightly more complex (and thus expensive) it doesn't interfere with the cap storage and it looks pretty good.

Materials

The choice of material for our ultimate rugged USB drive is probably the second most important choice, surpassed only by the design itself. When looking for a material there are a few criterias we're interested in:

• Strong
• Light
• Cheap/easy to source
• Easy to machine
• Hard
• Low thermal conductivity

You may note that several of these are in conflict with each other. For example, "easy to machine" and "hard" are basically exact opposites as are "strong/light" and "cheap".

Wood

I only mention wood as a material since I am an experienced wood worker and love working with it. A great looking USB drive could be made out of wood but for the purpose of a rugged/indestructible drive it is a poor choice. The yield strength of wood is a complex subject but for a rough estimate it is more than ten times weaker than even a low strength Aluminum alloy (yield strength of ~5k psi).

Brass

I used Brass for the first prototype and while it is cheap and easy to machine it is also heavy (8 g/cc) and weak (yield strength of ~20 kpsi) and thus not a good candidate for our design. 

Aluminum

Aluminum seems like a good choice of a material although we'll have to look more closely at which of many alloys in particular. In general, Aluminum is light (2.7 g/cc), easy to machine, and can have yield strengths approaching 100 kpsi in some alloys. We'll look at a few of the most common alloys, particularly those with high strengths.

6063

This is a common low cost and readily available alloy but is low strength with a yield strength around 20 kpsi.  It is great for quickly testing a prototype design and I've gone through a few feet of 1" diameter 6063 rod over the design process for the VIUD including the second prototype. 

Al 7075-T6

7075  is the most common of the so called "aircraft Aluminum" alloys developed for its higher strengths. It is around twice as expensive than 6063 but for our VIUD it still only works out to around $4 of raw material per unit. 

The "T6" designation is the "Temper" of the alloy which are standards of how the metal is prepared and is a very important designation of the material. For example a 7075 non-tempered alloy has a yield strength of only 15 kpsi while 7075-T6 is near 80 kpsi. A temper of T6 indicates that the Aluminum is solution heat treated and artificially aged.

Al 7068-T6511

One of the stronger aircraft Aluminums with yield strengths approaching 100 kpsi for the T6511 prepared alloy. Price for 7068 is around twice that of 7075: getting higher but hopefully still acceptable. T6511 is a combination of "T6"  solution heat treated and artificially aged)  and "T511" (cooled from hot working and artificially aged at elevated temperatures with minor straightening after stretching).

Titanium

Titanium is also an obvious choice when creating something that needs to withstand a lot of abuse. It is heavier than Aluminum at 4.5 g/cc making it about half as heavy as Brass and Steel. Where Titanium excels is its strength with certain alloys reaching a yield strength of 140 kpsi. The downside, however, is that due to its hardness, especially with the higher strength alloys, it is difficult to machine which leads to higher production costs.

Another thing to note about Titanium is its lower thermal conductivity of 16.4 W/m-K compared to Aluminum's 173 W/m-K. This means that using Titanium for our VIUD design should render it much more resistant to temperature. Titanium is also much harder than even anodized Aluminum which further increases its usefulness as a"rugged" material.

There are a variety of alloys but we'll only look at the two most common ones, Grade 2 and Grade 5 (6Al-4V).

Ti Grade 2 

This is the softer and more common Titanium alloy with a yield strength of the annealed alloy near 50 kpsi. This makes it less than ideal as it is weaker than the 7075 Aluminum despite being more expensive to manufacture, although surprisingly probably not by much. The hardness of Al 7068 is B90 on the Rockwell scale while Grade 2 Titanium is around B98, larger but less than I had originally assumed.

Ti Grade 5 (Ti-6Al-4V)

This is the much stronger alloy with yield strengths nearing 130 kpsi for the annealed version and approaching 170 kpsi for surface treated alloys. The price for this strength is its machining difficulty and high material cost, about four times that of 7068-T6511 making the material cost per unit around $25.

Review

The following table summarizes the important material properties we've been discussing:

Material	Density g/cc	Yield Strength psi	Heat Conductivity W/m-K	Hardness Rockwell
Aluminum 6063	2.8	20,000	200	B30
Aluminum 7075-T6	2.8	65,000	200	B80
Aluminum 7068-T6511	2.8	95,000	200	B90
Brass 360	8.5	20,000	110	B25-B80
Titanium Grade 2 Annealed	4.5	40,000	16	B98
Titanium Grade 5 Annealed	4.5	120,000	7	C36
Wood, Maple	0.7	6,000	0.7	N/A

The obvious choice of material is the Aluminum 7068-T6511 with 7075-T6 being a close runner up. Grade 5 Titanium is great but will end up being very expensive to use and perhaps is better used for a "special edition" version with a higher price.

Resources

Some of the online resources I found useful for material property references are listed below:

• Matweb Material Reference Library
• Aluminum alloy material reference
• Titanium alloy material reference
• Titanium Grades 

Temperature Testing, Round One

So far we haven't done any temperature testing other than the failed attempt with the first prototype and a propane torch which melted the USB connector's plastic. Our investigation into being the toughest USB drive in the world showed that a good temperature rating is one thing we are currently lacking.

We will perform some initial temperature testing with the one of the third prototypes which has been destined for eventual destructive testing, although we will try to be careful and not destroy it so we can get as much data from it as possible.

Failure Modes

Before going ahead with the testing it may be useful to review exactly how our VIUD may fail at high temperatures:

• Plastic in USB connector melts
• USB flash drive fails
• O-rings Melt
• Internal epoxy melts/fails
• Body melts

The primary failure modes we're interested in would be the first two. The plastic melting in particular is the most likely failure as, depending on the exact type of plastic, it occurs at relatively low temperatures of 150-315 °C (300-600 °F)  although Wikipedia states a melting temperature for Polyethylene as low as 105 °C (220 °F).

It is difficult to determine at what temperature the flash memory will fail at although it appears to be a gradual process: the higher the temperature the more likely bits will flip by themselves. There is an interesting paper that shows results for accelerated temperature testing in the range of 170-250 °C (340-480 °F) and gives the data retention for their device at 105 °C (220 °F) as 10 years. We can assume from this that the failure temperature for the flash memory is also relatively high and not likely to occur before the connector plastic melts.

Similarly the last three failure mechanisms occur at temperatures above that of the plastic melting point. If Viton o-rings are used they are rated to 200 °C (390 °F) and probably won't completely fail somewhere above that. The epoxy used in this prototype is only rated to 125 °C (260 °F) but again won't completely fail somewhere beyond this temperature and the actual epoxy that is going to used is rated much higher (300 °C, 570 °F). The melting of Aluminum (660 °C, 1220 °F) and Titanium (1670 °C, 3030 °F)  is so high as to be mostly irrelevant.

Procedure

The procedure for temperature testing will be relatively simple:

1. Let test object return to room temperature.
2. Let test apparatus achieve desired temperature (if applicable).
3. Subject test object to the desired temperature for a set amount of time.
4. Cool object after test in water for at least 5 minutes.
5. Examine object for any obvious signs of damage.
6. Test USB for any new file blocks with errors.
7. Test USB contents for any changes in data.
8. Repeat test at slightly higher temperature and/or longer period as desired.

The test for bad blocks is done using HD Tune Pro which is a basic drive benchmark and tester. The test for changing bytes during a test is simply done by creating random files that fill the USB and using a MD5 sum to ensure their contents are identical before and after the test.

Test Results

The following table summarizes all the temperature test results so far:

Test Type	Temperature	Length of Time	Error Test	Bit Change Test	Notes
Oven	90 °C (200 °F)	5 min	Passed	Passed
Oven	120 °C (250 °F)	5 min	Passed	Passed
Oven	150 °C (300 °F)	5 min	Passed	Passed
Oven	180 °C (350 °F)	5 min	Passed	Passed
Oven	200 °C (400 °F)	5 min	Passed	Passed
Oven	230 °C (450 °F)	5 min	Passed	Passed
Oven	260 °C (500 °F)	5 min	Passed	Passed
Oven	290 °C (550 °F)	5 min	Passed	Passed	Case exterior exceeds 100 °C
Oven	290 °C (550 °F)	6 min	Passed	Passed
Oven	290 °C (550 °F)	7 min	Passed	Passed
Oven	290 °C (550 °F)	8 min	Passed	Passed
Propane Torch	1800 °C (3300 °F)	10 secs	Passed	Passed
Propane Torch	1800 °C (3300 °F)	20 secs	Passed	Passed
Propane Torch	1800 °C (3300 °F)	30 secs	Passed	Passed	Case exterior exceeds 100 °C
Propane Torch	1800 °C (3300 °F)	40 secs	Passed	Passed	Minor damage to external o-ring
Propane Torch	1800 °C (3300 °F)	2 min	Failed	Failed	Prototype #1, USB plastic melted
Oven	200 °C (400 °F)	15 min	Passed	Passed
Oven	290 °C (550 °F)	9 min	Passed	Passed
Oven	290 °C (550 °F)	10 min	Passed	Passed
Boiling Water	100 °C (212 °F)	60 min	Passed	Passed	Turned Aluminum black
Wood Fire	600 °C (1100 °F)	30 sec	Passed	Passed	Case exterior exceeds 100 °C
Wood Fire	600 °C (1100 °F)	45 sec	Passed	Passed	Damage to external o-ring
Wood Fire	600 °C (1100 °F)	60 secs	Passed	Passed	Damage to external o-ring

One interesting thing to note is that the boiling water test turned the drive a black color with tints of gold that actually looks rather nice (resembles a black anodized Aluminum). I'm assuming it is something in my water (a very hard water) that turned it this color.

Surprise...it's Black Now!

Comparing the Competition

Most USB drives don't have explicit temperature ratings but of those that do the LaCie Xtremkey has the highest:

• Operating temp.: 5 to 35°C (41 to 95°F)
• Non-operating temp.: -20 to 60°C (-4 to 140°F)
• Fire test: 30 sec. fire exposure
• Heat test: 200°C (392°F) / 3 min

Since we easily beat both of the listed tests by a considerably margin it is safe to say our VIUD would be the toughest USB drive available for temperature ratings.

Tentative Ratings

We still have to test our final design although I wouldn't expect too different results with the possible exception of the Titanium model performing a little better. We can still estimate our temperature ratings based on the tests so far:

• Operating Temperature: ? to 60 °C (? to 140 °F)
• Storage Temperature: ? to 90 °C (? to 200 °F)
• Air Storage: 290 °C (550 °F) for 5 minutes
• Boiling Water: 100 °C (212 °F) Indefinitely
• Wood Fire: 1 minute exposure (600 °C, 1100 °F)
• Propane Torch: 40 seconds exposure (1800 °C, 3300 °F)

The operating and storage temperatures may be changed by the specifications of the USB flash memory used. Next up will be testing the cold side of the temperature scale which should be interesting to see how low we can go.