Archive for the ‘Ponderings’ Category

Made in China: Getting Started

Sunday, July 8th, 2007

I found setting up a supply chain in China to be enlightening, and I’d like to share more of my experiences with the readers of this blog.

Presenting this material is challenging; there is a lot of detail and its easy to get overwhelmed and lost in the vagaries of electronics manufacturing. In order to focus my writing and create a more succinct body of works, this series of posts will be thematic. In other words, instead of talking about the details of how pick and place machines work in China, I’ll post about a more abstract theme such as “automation” and just show sound bites to give the reader a flavor of what’s going on.

The theme of this first post is “Getting Started”.

In September of 2006, Chumby was just a team of about a half dozen people, and we had given away about 200 early prototype chumby devices at Tim O’Reilly’s FOO camp. The devices were well received by the attendees of the conference, so I got the go-ahead to build the Asian supply chain.

Before we went to China, we had a trusted vendor in the US give us their best quotation for the job, so we had a reference point to work downwards from.

In a nutshell, we called up a lot of our friends who had some experience in China, and we lined up a series of about a half dozen factory tours in China. We hit quite a variety of places on our tour, from specialty factories as small as 500 people to mega-factories with over 40,000 people.

Being open source helped a lot in terms of kick-starting the process, because first we had no fears about people stealing our design–we’re giving the design away–and there were no NDAs to sign when sharing critical information, like the bill of materials. I think this kind of openness actually gave us a better reception with the factories in China; they seemed more willing to open up to us because we were willing to open up to them. Also, there was no question in the factory’s mind that they were in a competitive situation. It was plain that anybody could and would quote and bid on the job (in fact, we received a few unsolicited quotations that were quite competitive), so it saved a round of huffing and puffing. And finally, it seems that despite handing out our BoM without an NDA, nobody in the general public is aware of the strategic improvements we rolled into the chumby hardware since the FOO prototype (then again, chumby has no serious competition at this point in time).

In November, Steve Tomlin and I went to China to do the factory visits, and some of my first impressions were documented in my post Adventures with the Venture Communist.

There is no substitute for going to China to tour the factory. Pictures are always deceiving, and you just can’t get a sense for the scale and quality of the factory without going to see the actual facility. In general, factories welcome you to take a tour–I wouldn’t work with one that didn’t allow me to come visit. However, most factories do appreciate a week prior notice before you come and visit, although as your relationship with the factory progresses things should become more open and transparent.

While almost every factory will “clean up” the day you come to visit, a sharp eye and the right questions can see through any quick veneers put in place. One place I always liked to visit was the QC room. I expect to see rows of well-maintained and well-worn binders with design documentation and QC standards, as well as “golden samples”. I would point at a couple of random binders and demand to view their contents, see the product’s golden sample, and verify that they knew what was going on in the binder–that it wasn’t just some random data they threw in there. Also, hard investments in equipment is a good sign: the best manufacturers I visited all had a couple rooms with sophisticated equipment for thermal, mechanical, and electrical limit testing, and of course operators were in the room actually using the equipment (I could definitely believe a Chinese manufacturer would buy a room of equipment just for show and not actually use it).

After reviewing several manufacturing options, we decided on one, which was PCH China Solutions. PCH itself owns only a few facilities, but it has a comprehensive network of trusted and validated vendors–primarily in China but also Europe and the US as well. Not surprisingly, the factories that PCH sub-contracts to were some of the best facilities we visited while we were in China.

PCH is actually headquartered out of Ireland, and as such, most of their staff engineers are Irish, so there was no langauge barrier. They are hard working, resourceful, well-trained, and as a bonus they always seemed to know the best places to find a pint no matter where we are. I had no idea China had so many Guiness taps.

In retrospect, this is definitely the way to go–I learned a lot about what it takes to do business in China by shadowing their engineers while I was out there. From 10,000 miles away it looks easy but it’s anything but easy working with even the best Chinese contract manufacturers. I’ll cover some of the “frustration stories” in another themed post later on.

FOO Camp 07 and RNA Folding

Monday, July 2nd, 2007

I was at FOO camp last weekend and it was a blast. As usual, Tim brought together quite an interesting crowd of people. It was a pleasant surprise to see old friends from MIT, some whom I hadn’t seen in years. To date, four FOO alumni worked on the same robotics team (ORCA) building autonomous submarines at MIT back when we were all students there, and at least three students/ra’s of my MIT graduate advisor, Tom Knight, have also attended FOO. Of course, I got to meet some interesting new people, including a group of folks who have expertise and great interest in manufacturing in China (we had a little round table discussion about doing business in China and China’s economic role in the world). I also gave a little presentation about how chumbys are made in China, something which I will blog about in the next couple of days through a set of posts forthcoming (I have a lot of material to go through so it’s taking me a while to organize them and write them).

One FOO attendee who I was most fortunate to stumble upon was Christine Smolke. She gave a fascinating talk about the capabilities of RNA that really opened my mind. As many may be aware, the “central dogma” of biology is being rewritten, and RNA is starting to take a more active role in everything from heredity of genetic traits to catalysis of reactions. Recent findings have caused some hypotheses to be revisited, such as the “RNA world” hypothesis, which indicate that life may actually have started through self-replicating strands of RNA, instead of DNA.

The most interesting connection I made listening to her talk was with my experience looking at the protein folding problem. In a nutshell, protein folding is one of the “grand challenges” of computer science today, and the basic mission is to predict the 3-D structure of a protein given its amino acid sequence–in my opinion, one important part of the “uber-tool” for nanotechnology engineers that would create a catalyst for an arbitrary substrate (another application for protein folding is also to elucidate the structure of proteins that cannot be crystallized and are thus unsuitable for X-ray diffraction analysis).

Protein folding is hard. I mean, really hard. It’s one of the few computational problems that truly scare me. There are whole supercomputer projects devoted to the subject, from DE Shaw’s ambitious project to IBM’s Blue Gene series of machines, to Stanford’s Folding at Home distributed computing project. My facts are a couple years out of date but iirc, a typical goal for such a big project would be to fold one “small-ish” protein of about 50 to 100 amino acids in about a month–a reaction that happens in a cell on a timescale on the order of milliseconds. And, the problem doesn’t scale particularly well. The reasons why protein folding is hard are numerous, and most of them have to do with the enormous dynamic range of timescales required for the simulation, the very sensitive interactions that the numerous hydrophilic and hydrophobic amino acids have with the surrounding water, and the sheer number of particles involved. The simplifying assumptions made in even the most sophisticated simulations today are crude compared to the actual conditions in the cell. The way a protein folds depends upon the rate of sequence output, the temperature, pH conditions, presence of helper molecules, coordinating ions, and even post-folding sequence modifications–all things that challenge current computational models.

To illustrate the point, even the iconic double-helix of DNA is a direct result of its interaction with its surroundings. The double helix arises from the fact that the base pairs are “greasy” (hydrophobic) and they repel water, so they stick together…thus, a structure that might otherwise look like a straight ladder collapses in on itself to minimize the distance between the rungs, squeezing out the water, and in the process twisting the backbone into a double helix; the process also requires coordinating ions from the water to neutralize the concentration of charges brought on by the collapse into the double-helix. Before I learned about this I just took the twisting of DNA for granted…shows how little I know about the real mechanics of biochemistry, but boy, is it fascinating.

Christine’s talk on RNA got me thinking…RNA is nice, as it can function single-stranded, and is very pliable. It only has four base pairs, instead of the twenty basic amino acids found in proteins. The secondary structure of an RNA molecule is also predictable. And, RNA can be active on a variety of substrates. Granted, RNA may not be as effective, efficient, or as versatile as the more complex protein counterparts, but I can’t help but wonder if maybe a good baby-step would be to first try to solve the RNA folding problem. It’s only a hunch right now but it feels like RNA might be an easier beast to tame than proteins. And as a molecular tinkerer, I’d rather have a tool that creates less than optimal results but is available sooner, can iterate faster, and is more affordable, instead of a tool that gives ultimate results but also comes at enormous cost and effort. There are a lot of simple molecular problems that need solutions today, and perhaps from these learnings we can eventually develop smarter tools for the more complex problems.

Ah, if only I had the time and the money…too many interesting things to do! I wonder if I had become a professor instead of a professional, if I would have had the priviledge to investigate such interesting diversions, or if I would simply be consumed by the tenure clock…

Shenzhen Diary, April 2007

Saturday, May 5th, 2007

I’d thought I’d write a brief (okay, it ended up much longer than I thought it would…) blog entry about some of my most recent experiences in China.

I’m currently out in Shenzhen working on bringing up the Chumby production line. This means everything from raw material approvals to programming, testing, QA, and certification for not only the Chumby hardware itself but also its accessory line. It’s a lot of work.

I’m very lucky to be working with a fabulous company out here which serves as my liason between vendors and contractors. I’ve been told explicitly by Chumby that I can’t mention the name of this company because it is a “strategic” advantage for Chumby at this point, but I also made it clear that eventually I will talk about this company publicly, after the Chumby hardware is launched and the magnitude of the strategic gain from radio silence is reduced. I’ve got some nice photos of Chumbys in various states on several production lines that I’d also like to share as well someday.

Shenzhen is quite an up and coming city. I’m staying in a serviced apartment on Xinwen road called the Frasier Futian. It’s cheaper than a hotel yet it has similar amenities (such as maid and laundry service and complementary breakfast), although the quality control of the details of the facilities isn’t quite the same as at a name-brand hotel. I’d recommend this as a place for anyone who needs an extended stay in Shenzhen. The location is good–several very nice restaurants in walking distance, a Starbucks and some convenience shops–and it’s about a 5-10 minute walk to the MTR (the subway system in Shenzhen–you can go all the way to Hong Kong on the subway from here). Most importantly, the internet works great here; I have no trouble using Skype and even video conferencing works well.

The most miserable two days of my life were also spent around here. I got food poisoning–I think it was from handling money (which is filthy dirty out here, it literally smells of human waste) and then eating my food, because I’ve been reasonably careful to avoid anything that’s uncooked here in Shenzhen. It was the worst food poisoning I’ve had in my life, it lasted five days until I finally broke down and went to a doctor in Hong Kong and got antibiotics for it (I’m told you never want to go to a doctor in Shenzhen if you can help it). The doctor gave me Ciproflaxin–which I have an allergy for–but the doctor was like, “do you get a rash around your mouth?” and I was like “no, only on my hands and feet”, and she was like, “well, take the Ciproflaxin then. It won’t kill you and you’ll need the strength of Cirpo to get rid of those bugs in your gut.” It worked; I got better. I did get that rash, but I also didn’t die. One really nice thing about the doctors offices in Hong Kong is that they also fill your prescription at the office, so I didn’t have to run around Hong Kong searching for a pharmacy. I was diagnosed and taking my first dose of antibiotics all within 15 minutes (and at a fairly reasonable price–about $125US for a weekend visitation, including the medications).

Anyways, while I had the food poisoning and before I could see the doctor, I had to go to a factory in Donguang at 8AM and I was up until the next morning at 3 AM debugging the manufacturing problems in the very first run of the Chumby circuit boards coming off the line. The fever, cramps, dehydration and constant trips to the toilet didn’t help progress. I spent the night in the factory dorms (where the workers also stay), which isn’t too bad of a place (actually a bit nicer than many of the dorms at MIT) but I ran out of toilet paper around 5 AM (I was making trips to the bathroom every hour) and I had to ehm–improvise. There is no concierge service in a factory dorm! The next morning I had to get up at 8 AM again and continue debugging. Fortunately, most of the significant problems were resolved and the root cause of all the remaining issues are at least understood and solvable in time for production (or so I hope!). There are more interesting stories around the root cause of some of these issues, but let’s save them for another day.

In the end, all I can say is what doesn’t kill you makes you stronger…and I also carry around a bottle of hand disinfectant now for after when I handle money. Surprisingly, it was about the hardest thing to find in the drugstores out here. It was buried in with the urgent care first aid area, oddly enough, next to the alcohol used for disinfecting wounds. I think the store clerks didn’t understand that this product is for preventative use, and not for urgent care use.

I also had a little adventure finding a fedex out here. Normally, fedexes aren’t hard to find, but there was a three-day vacation called the “5-1” (for May First). It’s a holidy declared by the communist party to celebrate the efforts of the laborers. Fedex was closed for three days and I urgently had to send a package. So, I hopped on a boat to Macau because I knew the casinos would be open on May 2nd and I couldn’t imagine a high-end casino not having a business center with a Fedex. And indeed, the Wynn Macau had a fedex with english speaking staff that happily processed my package on time!

Macau was also a very interesting place. There is a lot of hype about Macau being the next Las Vegas, with casinos bigger and more fantastic than Vegas. Macau casinos were certainly opulent and impressive, with a great many gimmicks and displays, but I think the hype is a little bit too much. I’ve been to Vegas many times, and currently, I think Vegas still trumps Macau. At least, the food in the Vegas casinos is better, and the clientele is more fun. Let me tell you, gamblers in Macau are hard core. In Vegas, you see people at least smiling and having fun at most tables. In Macau, even the simplest of dice games had chain-smoking men staring intently at the table. Nobody drinks alcohol–they all have coffee or tea, to keep their minds sharp. You can feel the intensity–the casinos are eeriely quiet compared to the din of the Vegas casino. I’m not that into gambling, so I guess it wasn’t for me.

I thought the local food in Macau was quite good. The Portuguese egg tart is tasty–it’s a variant of the Chinese dim sum called “dan gau” (simply put, an egg tart). While the Chinese version tastes quite eggy, the Portuguese version uses a type of flan for the interior, so it’s mildly sweet and has a nice texture to it. I also ate some local Portguese food, which was very nice. It’s very similar to a Japanese curry, I thought, although with less curry and a slightly more watery sauce. You can really taste the flavor of the meat through the sauce–I had the oxtail and chicken. I have no idea what it’s called, I just asked the waiter what was popular. Incidentally, I wasn’t as impressed with the high-end casino food.

Speaking of food, I also went to the most fantastic noodle shop in Hong Kong today (I had to run down there to shop for fabric samples for the Chumby–they have a great textiles market in Sham Sui Po). It’s the Crystal Jade–apparently a chain restaurant–I went to the one in the Harbour City Ocean Center. They have hand-pulled noodles there and oh man they were so good. The Harbor City Ocean Center is near the Tim Sha Tsui MTR stop, just stop into any hotel near the MTR station and ask the concierge for directions.

This post has grown much longer than I thought it would be, so I’ll stop the story telling now and leave you with these two pictures of street signs that I thought were just…interesting.

Obviously, the no bikes sign isn’t obeyed. Makes you wonder about the other ones.

Scritch scritch–huh?

Friday, March 9th, 2007

I was poking at the CIA World Fact Book, which is a fascinating resource in itself, and noticed this table of “current account balance” by country. It’s a big table, so I won’t quote it here, but to understand my puzzlement, take a look at the top ten names, and then look at the bottom ten names…and then notice the magnitude of each balance. One of these kids is not like the other…

Is this something to be concerned about? Warren Buffet seems to think so.

For contrast, try out this table of GDP by country. This is closer to what I expected for the top 10 and bottom 10.

On a possibly related note, I was trying to import some Korean DRAM today, and it was stopped in customs, where I discovered that the US is charging a 40%(!) tariff on DRAM chips imported from certain Korean manufacturers. I guess this partially explains why the Idaho-based Micron seems to dominate the inventory in distributors in the US, and how they can charge a 2-3x price markup over what I can get overseas. Then again, it’s somewhat comforting to know that some of my DRAM comes from the same state that my french fries and potato chips do.

I guess the question is, would customers of finished products like the chumby pay extra just because it contains US-made DRAM.

I’m guessing not.

A Physics Puzzler

Friday, February 23rd, 2007

I was eating dinner and sharing a couple bottles of wine with my friend Mike Fitzmorris last night, and he posed me this two-part question to which neither of us are quite sure about the answer:

Part one: Imagine an iron bar in free-space (no gravity, perfect vacuum, perfect darkness). If you impart a force such that it spins about the short axis, does it ever stop spinning?

It seems that with nothing around to apply a counter-force to the bar, it should spin forever, cosmological and bizarre quantum effects aside.

Part two: Imagine the same iron bar under the same conditions, but magnetized. If you impart a force such that it spins about the short axis, does it ever stop spinning?

The answer to this part seems to be that it should stop spinning eventually, but I can’t explain exactly why. The spinning magnetized iron bar would generate a fluctuating magnetic field, which means that it would be emitting some kind of remotely detectable signal, which, by definition, means that energy is being lost and therefore the bar should eventually stop spinning…

But exactly what is applying the counter-force to cause the bar to stop spinning? The bar is in a perfect free-space, so the electromagnetic waves will never interact with anything…The best explanation I could come up with is that the bar’s self-magnetic field is “catching up” with its emitted magnetic signature, and this interaction induces a counter-current in the metal which would cause an opposing magnetic field that would result in a net braking effect. I’m not very happy with this explanation, however.

Anybody know the correct answer to this question, or have a good explanation for why it might be so? There must be a simple answer and I’m just not seeing it.

[some edits and an addendum here to clarify this post]

A lot of interesting views in the comment area, thanks everyone for the thoughts!

Let me rephrase the thought experiment, because I think my poor phrasing has lead to many people to point toward flaws in the problem statement rather than the problem itself.

Suppose we think now of the spinning magnet in free space as a flywheel, and I couple energy out of the flywheel via a large coil. The changing magnetic field of this flywheel would cause electrons to move in the coil, and clearly since I am coupling power out of the flywheel via the coil’s capture of the changing magnetic field, the flywheel must slow down, otherwise energy is not conserved. In other words, the flywheel’s kinetic energy is converted into electrical energy by this mechanism.

If I remove this coupling coil, no energy is being directly removed from the system, so you might think, okay, the magnetic flywheel should not spin down ever since there are no mechanical frictional effects.

However, now imagine that I create an enormous superconducting coil that spans a diameter of 10 light seconds away from the flywheel. As I start spinning the flywheel, the changing magnetic field flows away at the speed of light. If the argument is that no energy is lost into this magnetic field, then the flywheel will not slow down at all. However, 10 seconds later, this magnetic field hits the coil, and all this energy is captured in the coil…but where did this energy come from?

Clearly, at time = 10 seconds, the flywheel can’t possibly begin instantaneously spinning down; the information about the presence of the coil would have to propagate its way back to the flywheel first, otherwise we have created a mechanism for transferring information at a rate faster than the speed of light. So for the “flywheel does not slow down” hypothesis to be valid, for a net 10 seconds, the system mysteriously has “extra” energy–I took energy out via the coil, yet the flywheel maintained its original rate of spinning and all its kinetic energy for the 10 seconds that it took for the information about the existence of the coil to make it back to the magnetic flywheel.

Therefore, in the absence of any device to couple energy out of the system (other than radiation to free space), the flywheel must slow down with time, otherwise I can spontaneously create energy. Therefore, there must be a completely local mechanism within the flywheel that causes the energy to be debited at the instant that it radiates from the flywheel. This is true whether the bar spins near the speed of light, or if it spins very slowly as any changing magnetic field can compel an electron to move and therefore transfer energy to another part of the system.

It is this local energy-debiting mechanism that I am trying to grasp. Any explanation also has to work if I replace the iron bar with a non-conductive magnet, like a ceramic or organic magnet. That’s a flaw in my counter-eddy current explanation that I proposed above.

Here is a quick sketch that helps illustrate the question.