Updated USB Specifications

Last week was spent updating the VIUD specifications on the new store site which includes a number of values related to the USB (read on for an explanation of each benchmark):

	16 GB	32 GB	64 GB	128 GB	256 GB
Usable Capacity	15,961,546,752 bytes	31,651,479,552 bytes	62,911,414,272 bytes	125,818,503,168 bytes	253,121,003,520 bytes
File System	FAT32	FAT32	exFAT	exFAT	exFAT
HD Tune Pro Benchmarks
Read	106.5 - 116.7 MB/s	114.2 - 127.9 MB/s	119.9 - 130.7 MB/s	181.2 - 193.3 MB/s	56.0 - 146.8 MB/s
Average Read	111.7 MB/s	121.9 MB/s	123.5 MB/s	187.7 MB/s	142.1 MB/s
Access Time	0.889 ms	0.973 ms	1.16 ms	0.747 ms	0.811 ms
Crystal Disk Mark Benchmarks
Read Seq Thread	121.6 MB/s	130.1 MB/s	133.3 MB/s	208.8 MB/s	149.2 MB/s
Read Rnd Thread	16.4 MB/s	14.3 MB/s	14.1 MB/s	6.03 MB/s	22.1 MB/s
Read Seq	119.8 MB/s	127.6 MB/s	131.6 MB/s	205.4 MB/s	146.6 MB/s
Read Rnd	15.0 MB/s	13.3 MB/s	13.6 MB/s	5.47 MB/s	19.2 MB/s
Write Seq Thread	24.2 MB/s	34.8 MB/s	38.2 MB/s	132.5 MB/s	94.1 MB/s
Write Rnd Thread	0.051 MB/s	0.097 MB/s	0.69 MB/s	0.224 MB/s	1.37 MB/s
Write Seq	22.4 MB/s	25.6 MB/s	17.8 MB/s	129.8 MB/s	22.6 MB/s
Write Rnd	0.054 MB/s	0.055 MB/s	0.81 MB/s	0.0048 MB/s	1.07 MB/s
USBFlashBench Benchmarks
Read	116.70 MB/s	124.93 MB/s	127.54 MB/s	206.48 MB/s	131.13 MB/s
Write	22.72 MB/s	29.40 MB/s	32.89 MB/s	133.83 MB/s	55.93 MB/s
Link	157023	157008	157022	157018	157011
4GB File Copy
Copy	2:38.3 (24.1 MB/s)	1:49.5 (34.8 MB/s)	1:42.1 (37.3 MB/s)	0:30.2 (126.3 MB/s)	0:44.8 (85.2 MB/s)
Read	0:33.4 (114.1 MB/s)	0:31.2 (122.1 MB/s)	0:30.6 (124.6 MB/s)	0:18.4 (206.9 MB/s)	0:28.8 (132.7 MB/s)

The VIUD's usable capacity is slightly less than the stated capacity which is typical of just about any mass storage device. Part of the overall device's drive space is taken up by the file allocation table which takes more space on larger drives. The other part, however, is just due to the USB board manufacturer using a slightly smaller size than stated. Perhaps this is due to a technical reason or, more likely, an economic one.

HD Tune Pro is a basic disk benchmarking and testing application. Unfortunately the write tests don't seem to work on newer operating systems but the read and error tests do give a nice visual display. I'd like to point out the low minimum read speed for the 256 GB VIUD which can be easily seen in the following graph:

VIUD 256 GB HDTunePro Read Benchmark

All other VIUD capacities have a roughly flat graph for the blue line but the 256 GB model has much slower read speeds at exactly 32 GB intervals for whatever reason.

CrystalDiskMark is another basic drive benchmarking tool that outputs several read and write tests:

• Seq Thread = Sequential read/write, 128 KB block size with multiple queues and threads
• Rnd Thread = Random read/write, 4 KB block size with multiple queues and threads
• Seq = Sequential read/write, 1 MB block size with single thread
• Rnd = Random read/write, 4KB block size with single thread

USBFlashBench is a simple tool that tests multiple block read/write sizes and then uploads the results to a website and creates nice graphs like:

USBFlashBench Results for the 256 GB VIUD

Finally, the 4GB file copy test is simply the time it takes to copy a 4 GB (4,000,000,000 bytes) file composed of random content to the VIUD and the time it takes to read it back. Due to caching we have to make sure to unplug the VIUD after the write test and then plug it back in to do the read test.

Product Shots!

Work on creating the VIUD store started last week and I'm already well on my way to getting it finished. I decided to go with Shopify as it seems to be a good sized Canadian online e-commerce system with the features I need.

I've spent the better part of the last week taking and preparing product shots of all the various VIUD models for use in the store. There are a lot of good online guides for taking product shots but to summarize what I've used:

• Niken D90 camera with SB-600 flash
• Small light tent
• Piece of white paper over the flash to disperse flash and soften shadows
• A few extra lights to get better shadow and reflections (some "real" camera lights would be very nice but I've made do)
• About half-zoomed in on the lens
• Auto color balance on the camera
• Manual mode on the camera with a high aperture (F20) for a larger depth of field
• Adjust the exposure time and bias to get a good picture (white background without too much shadow)
• Adjust the picture's brightness, contrast, and saturation afterwards to get a realistic and good looking image

The shots turned out pretty good and I think they can speak for themselves!

Getting Closer...

So my summer vacation is unfortunately over and I'm getting back into the finishing stages of the VIUD preparation. All I can do before actually selling them is to do the coating of all the USB boards I currently have.

Through all the various testing I've done, mainly the shock testing, I've narrowed down a good combination of coatings and epoxies that seems to maximize the durability of the USB. I start with two layers of Silicon conformal coating.

Silicon Conformal Coating a USB

This is thin, flexible coating that gets into all the small nooks and crannies on the circuit board and even by itself seems to help a lot in minimizing damage caused by shock. Each coat requires at least a day to dry out.

Drying Out the Conformal Coating

Once the two conformal coatings are dry it is time for another two coatings of a thicker and more flexible insulating coating.

Coating a USB in Insulating Varnish

This gives the USB circuit board a considerably thicker and flexible coat to help absorb even more shock. It also helps prevent epoxy from leaking through the USB connector and making a considerable mess which is difficult and time consuming to clean up. Like the conformal coating each insulating coating takes a day to dry.

Drying Out the Insulating Coating

Next step is to just create an online store and actually start selling things! 

Week(s) in Review: 29 April 2016

In the past month or so I've finally managed to complete finishing the Titanium bases and have all bases laser engraved with the VIUD logo and serial number and they're looking great!

 I also experimented with custom logo/text and it worked out very well but the price point to have it done is unfortunately large enough to make it only worthwhile for larger quantities (say 10 or more).

Custom Logo and Text Example

There's not much left to do! I can conformal coat all the USB boards and then find, design and setup an online store. Unfortunately, it looks as though I may be away all summer which will put another delay in the VIUD progress but it won't be long now!

Week(s) in Review: 14 March 2016

Its been a busy month for me in other projects, particularly with the release of the Thieves Guild DLC for Elder Scrolls Online. 

Besides a little bit of label preparation the only other news to note is that the anodized cases have arrived and they look great!

Titanium VIUD (Left) and 4 Anodized Aluminum Colors

Next steps are to polish the Titanium bases so I can have the VIUD logo and serial number Laser engraved. After that we are ready to start actually selling things!

Week(s) in Review...16 February 2016

Its been a very busy week or two for me with both VIUD and non-VIUD related projects....

Aluminum Anodizing

I'm happy to say that the 150 Aluminum VIUD's are finally off to anodizing! The polishing of the bodies went well with no unforeseen issues. The preparation of the plugs took a little longer than expected. Originally I was looking into buying or making a broaching set for squaring the rounded inside corners of the plug. However, I decided to simply file the corners by grinding a small square file to be the exact width of the hole and making a simple handle to hold the file.  Filing the Aluminum plugs took some time but was relatively easy and we'll see if the same holds for the Titanium plugs. I wasn't going to polish the plugs but the filing created a sharp edge which was good to get rounded with a quick buffing.

Next step was cleaning everything. The buffing process caked a considerable amount of debris in all the nooks and crannies of the parts so I soaked everything for a day in Mineral Spirits and then brushed/wiped each part and left it to dry. 

For this production run I'm doing four basic colors for the Aluminum VIUDs:

• Black (already done)
• Red
• Blue
• Green

There are a large number of potential colors I could use but there is a large minimum charge for the anodizing so we'll start with these four colors and expand in another production run if there is demand for any more.

A Little More Impact Testing....

I did another run of impact testing with a new set of USB circuit boards and a modified epoxy filling technique to ensure everything works as expected and there are no premature failures. Fortunately, the VIUDs fared better than my new stone target which broke clean in half while testing the Titanium VIUD at speeds above 200 km/hr.

Another Broken Stone Target

Fortunately I think I'm done with impact testing for now as I don't know what I can actually use for a stone target that won't break on me...perhaps a really thick custom made concrete pad with rebar reinforcement.

The Aluminum VIUD managed to survive up to impact speeds of 230 km/hr on the stone target which seems to be the typical failure point for the modified epoxy technique. The Titanium VIUD broke a new record by surviving impact speeds up to 250 km/hr on the stone target which is rather incredible when you think about it let alone experience the violent event itself.

Cap Magnets

I've always been trying to figure the best way to secure the magnets in the cap. Initially I tried superglue which works reasonably well but is a little messy and tends to fail at low to moderate impacts.

I then tried denting the metal around the magnet on the inside using metal point punch. This works pretty well in Aluminum but in the much harder Titanium it is very difficult to hit the punch hard enough to secure the magnet but not so hard as to crack or break the magnet. This method also tends to fail at moderate impacts.

On the last set of impact tests I tried another method: covering the magnet with a small amount of epoxy (the same stuff I use to fill the base). I didn't think this would work very well as you can only fit about 0.05" (1.3 mm) of epoxy above the magnet and still leave room for the USB connector. Fortunately I was completely wrong and the epoxy didn't start cracking until very high impacts, far surpassing the other two methods I tried.

Shipping Things

With the help of my lovely assistant I also figured out the details on how the VIUDs are going to be shipped. Originally I was planning on just using small bubble wrap bags from Canada Post but my wife found a better idea of small cardboard boxes which ended up being a little bit cheaper and looks/works better as well. 

Next Week(s)...

While I'm waiting for the Aluminum parts to be anodized I can start finishing the parts for the 100 Titanium VIUDs. I'm also working on getting the details sorted out for the Laser marking.

Week in Review...30 January 2016

It has been a busy week or two in VIUD production, although I didn't get as much done as I wanted due to work that needed to get done on www.uesp.net.

More Impact Testing

About 100 more impact tests with the air cannon have been done to test a few different types and methods of epoxy fills in order to improve the VIUD's shock resistance (more on that in a future blog post...soon). I've managed to successfully test speeds over 300 km/h (200 miles/hour) with impact shocks nearing 5000 G which seem to be the limit of what can be accomplished with the current shock testing rig for a couple of reasons.

At lower speeds the air gun pressure was roughly linear with the output speed of the VIUD but exceeding 25 psi of pressure the output speed quickly flattens out (see the graph on the right). Since the theoretical terminal velocity of the VIUD is around 200 km/h (120 miles/hour) its going to be difficult to get it to go much faster than this. I may be able to get a little more speed out it by mounting the VIUD in a foam "bullet" to increase the firing efficiency but I don't know exactly how much that will get me.

At these high speeds the impact targets are also taking a good beating. The wood target is beat up pretty well but has survived 100 impacts and can probably take 100 more before needing to be replaced. The stone targets, however, have fared less well...

Stone Targets Destroyed by VIUD Impact Tests

Against the heavier and tougher Titanium VIUD at speeds above 200 km/h (120 miles/hour) a new stone target (left image above) broke after just 4 impact tests. I've since started using a thicker 2" Arkansas Limestone target which is a bit softer but seems to take the hits a bit better.

I also seem to have reached a measurement limit with the high-speed camera. The Casio Exilim EX-ZR200 was the cheapest camera I could find that could go up to 1000 fps. This is much faster than most cameras but this speed seems to limit the measured impact accelerations to under 5000 G. At this magnitude the actual impact occurs in well under one frame (<< 1 ms) so it is likely underestimating the actual impact force. I'd love to get a better camera that can do at least 10k fps or better but the ones I've looked at are quite pricey.

Aluminum Case Preparation

Before I can get the remainder of the Aluminum cases anodized there is a fair amount of preparation work needed. In truth, they could probably be anodized as they are and turn out OK but I'd prefer to take a little bit of time and make them better than just "OK".

First step was to make a simple jig that can hold the VIUD cap or base securely, quickly, and accurately in the lathe without damaging them. It was relatively straightforward to design and manufacture an Aluminum collet piece sized exactly to the outer diameter of the VIUD.

Simple Jig for Holding a VIUD Cap/Base in the Lathe

Using this jig I increased the size of the cosmetic chamfer on the ends of the cap and base from essentially 0.00" (it was barely what you could call a "broken" edge) to 0.01". While this seems like an insignificant modification it stops this edge from feeling sharp and will help the anodizing from chipping prematurely.

Next step is to polish all remaining Aluminum 150 caps and bases. The finished surface as they are is typical dull Aluminum. By polishing to a near mirror finish the resulting anodized surface will be much shinier and it also helps break any remaining sharp edges for a better feel as well as preventing the anodized surface from chipping.

Comparing Original (Left) and Polished (Right) Aluminum VIUD Bodies

Next Week...

I've just started polishing the bodies so that will be my first task next week followed by finishing the VIUD plugs and then cleaning everything. There are more impact tests I'd like to do as well if I have time. The next week will be very busy in UESP related work but stay tuned for more progress reports in a week or two!

Production Starts!

While I've been silent on the VIUD front for the past few months it has been mostly due to waiting for the first 300 VIUD cases (200 Aluminum, 100 Titanium) to finish production. They took a long time to get here but they finally arrived and are looking great!

What 300 VIUD Cases Look Like

Titanium VIUD

Anodized Black Aluminum VIUD

I'd love to say that I'll begin selling VIUDs in the near future but there are just a "few" tasks left before that can happen:

• Polish and clean the 150 non-anodized Aluminum.
• Anodize the 150 Aluminum sets 3 different colours.
• Laser engrave everything with a logo and serial number.
• Order all the various parts required for assembly.
• Finish testing and finalizing the epoxy type and assembly procedure.
• Create a few tools and jigs to help with assembly.
• Install 600 magnets.
• Create a VIUD shopping site. I'll probably use a service that offers pre-built stores but there is still a good amount of customization and setup to do.
• Create a simple assembly and non-conformance tracking database.
• Get shipping procedures and costs finalized.
• Figure out a sales price for the VIUDs.
• Start selling!

Sounds exhausting and while my summer is usually busier than average I hope to be able to finish everything in this list and start selling a few VIUDs this fall. Stay tuned!

The Finished Product!

While it took much longer than I had anticipated I'm happy to say I can hold in my hand the world's toughest USB drive!

The Completed Revision 5 VIUD Prototypes

The completed VIUDs are even better looking than the individual parts and work great. I've been carrying two in my pocket for the last two months and while they are definitely on the heavy side for a USB drive they are not excessively so. There's a bit of chipping on the corners/edges of the Aluminum model which is expected and probably mostly due to the Titanium model being next to it. The Titanium model still looks flawless.

The magnetic cap works great and is strong enough to hold the VIUD to the fridge door or other metal surface as designed. I haven't tested the lanyard attachment but my initial concern of the edges being too sharp were fortunately unfounded. I've only done a quick 3000 psi pressure test so far but both test units passed easily.

Assembly

Adding the Internal Epoxy

Assembly of the 15 prototypes went well but did reveal a few minor design and assembly flaws:

• The anodized threads were just barely too tight and required manual retapping. I'll specify a looser thread tolerance for the internal threads to compensate.
• The slots for the magnets were just barely too small and had to be made a bit larger. Slightly larger OD dimension for the slots are needed depending on how shock testing goes. I fear if I make it too loose the cap magnet will come free too easily.
• Weighing the amount of epoxy in the base worked great but the magnet caused havoc with the scale. Adding a few non-magnetic spacers fixed this easily.
• The scale used has an annoying tendency to turn off in the middle of deploying the epoxy into the base. Hopefully there's a way to disable the auto-power off feature or a different scale will be needed.
• The amount of epoxy needs to be reduced slightly as when it is cooked in the oven it expands and overflows. The overflow is actually good in a sense as it means there are no air pockets in the base but it requires more time consuming clean up afterwards.
• Oven baking to cure the epoxy worked just fine. The magnets help to keep the base secure the metal pan.

Baking at 90°C (200°F) to Cure the Epoxy

What Next? 

There will have be a small break in the VIUD development this summer due to my personal schedule...who knew getting married would keep you so busy! Sometime in the fall I'll get back into more testing of the finished revision 5 VIUD prototypes to see exactly what sort of punishment they can take.

Assuming no significant design change is required (I doubt it but you never know) I hope to be able to get a KickStarter going for the initial VIUD production run later this fall. Stay tuned to the blog for the eventual announcement.

Prototype #4: The Last One?

After a long overdue wait I finally have the pieces of prototype #4 in my hands, ready for assembly, and they are looking great!

Fresh from the Shop and Looking Good!

Prototype #4 has a number of design changes from the previous prototypes originating from a variety of tests, feedback, and research:

• Magnetic base and cap (it works very well...not sure how I overlooked this design initially)
• Slightly longer for a combination of aesthetics and to hold long-style USB boards if I ever need to
• Hole in the cap for a lanyard attachment
• Slightly simpler interior design for cheaper production (at a small cost of strength)
• Very high strength materials (Aluminum 7068-T651 and Titanium Grade 5)
• Hard anodized Aluminum (dark red)
• Laser engraved logo and serial number

Close-up Showing the Laser Engraving Details

The production itself was basically flawless with only a few minor issues encountered:

• Thread in base was 0.05" deeper than specified (makes assembly a little trickier)
• Threads in the Aluminum base were a bit tight (re-tapped manually after anodizing)

Other than the smaller quantities (10 Aluminum and 5 Titanium) the production process was identical to that of the (hopefully) eventual larger order. This minimizes any surprises and gives me a good idea on the base material cost and time required for assembly.

All that is left now is the final assembly of the USB boards and epoxying before I can get back to destructively testing a few units for fun!

Prototype #4 Production: How to waste 4 months....

I haven't updated this blog in a few months as I was waiting for a small production run of prototype #4 to be completed. I only expected this to take a few weeks but it ended up being a annoyingly painful 4-5 month process.

The Quotes

I originally tried to get quotes from 5 different machine shops from (names have been removed to protect the guilty):

Shop 1	Shop 2	Shop 3	Shop 4	Shop 5
High-end CNC shop Used for prototype #3 Al = $70/set for x100 Ti = $130/set for x50	Advertised as a large, high-volume CNC shop Did not respond to several quote requests and inquiries	US company via internet Has simple CAD software for instant quotes Al = $25/set for x100  Ti = $80/set for x50 Many production issues	Local machine shop Took 2 weeks to respond to initial quote request Could not source desired Al alloy Al = $160/set for x100 Ti = $350/set for x50	Local machine shop Did not source or work with desired materials

I don't understand why I never got a response back from shop #2 despite them looking like a large and reputable shop. I guess they're doing so much business they don't need to reply to quote requests although a simple "We cannot do this order for you at the moment" type reply would be nice.

I tried shops #4 & 5 as they are close in the area as me but both had issues with my material choices. I understand not wanting to work with certain materials, like Ti Grade 5, but I'd expect a machine shop to be able to source materials better than an amateur like myself can. ...or perhaps I'm overestimating how commonly Al 7068-T651 is used. Even with cheaper materials shop #4's quote is more than double that of shop #1 and 4-6 times that of shop #3.

I've used shop #1 for a large number of orders in my previous job as well as for making two pieces of the VIUD prototype #3. They produce excellent quality parts and are very easy to work with. I've always considered them as an "expensive" shop although after this episode I've warmed up to their prices...you get what you pay for after all.

An Unfortunate Choice

Given the wide range of quotes I went with the cheapest one from shop #3....unfortunately I choose poorly. Shop #3 has a slightly different operating practice than most other machine shops. It supplies a very simple custom CAD program that you use to create designs and submit them for instant quote requests. While their CAD application is very limited compared to SolidWorks or AutoCad it works well enough and being able to get instant quotes for simple parts is very convenient. Their quotes for the parts were also far cheaper than anything else I could find by over a factor two.

While shop #3 appears to have a lot going for it they had multiple issues actually delivering them. Trying to contact an actual person is difficult and usually you end up talking with someone who can't, or doesn't, actually answer your questions. Their lead time was longer than most shops at one month but this was extended twice to two and three months without explanation.

Oh....but it gets so much worse. After finally actually completing my order of 15 pieces after 3 months there were three serious quality control issues:

1. Knurling was the incorrect type (straight instead of diamond)
2. The internal thread in was made at the incorrect depth



Worst Knurl Ever

3. The depth of cut in the threaded plug was wrong.

I could probably live with the incorrect, and poorly made, knurling but the other two issues were a deal breaker. I know that mistakes happen in production but these parts are so simple and the mistakes so basic I'm at a loss to explain how they might of occurred. Call me crazy but I expect a competent shop to have the basic ability to make things according to the drawing, especially trivial parts like this. Coincidentally they messed up on each of three parts.

I asked, and the agreed to remake everything, but after another few weeks and several more manufacturing issues I was told they "couldn't find anyone to make it" and I had to cancel the order. The only nice thing I can say is that it didn't cost me anything...except 3 months of time.

The End Result

So +3 months after placing the order for prototype #4 I had nothing to show for it. At this point it is close to Christmas which scratches off another couple of weeks before I can source and place an order with shop #1. I finally get the 15 pieces in my hands some 5 months after the initial order with shop #3.

This whole process highlights a simple wisdom I've learned over the years:

Supplier relationship is often more important than price

Shop #1 may not have had the best price but they are incredibly easy to work with and produce excellent quality work (the prototype #4 pieces look great). I'm looking forward to using them in the future for larger production quantities of the VIUD.