Name that Ware, July 2009

August 4th, 2009

The Ware for July 2009 is shown below. Click on the image for a much larger version.


I thought I’d try something new for this competition. Above is a scan of a completely blank PCB. I don’t actually know what it’s from. I just bought it from a vendor in the Shenzhen mobile phone market who also didn’t know what it’s from — it was just used for shop decoration to attract customers. Since I actually don’t know what it’s for, I’ll have to rely on the depth of your comments to convince me. If nobody can convince me, then there would be no winner. On the other hand, it’s quite possible I could choose a winner who incorrectly identifies the ware, but that’s ok — this game has always been more about explanation and sharing thought processes, rather than about absolute accuracy.

I do think a blank PCB like this is fairly interesting to post for a number of reasons. First of all, it gives you a real appreciation for the sheer density of routing on a mobile phone motherboard. Second of all, this is a great example of an HDI-buildup, blind/buried via PCB. If you look at the large version of the scan, you can see the dimples where they used a laser, instead of a drill, to cut tiny vias into the outer layers of the circuit boards. There are almost no through-holes, and via-in-pad (typically forbidden when using through-hole vias) is very common in this design. Finally, this pair of PCBs is actually still in its manufacturing frame, which reveals a number of test structures that you never see (visible on the right hand side) that are used to verify that the PCB’s construction is correct.

You can also see in here how the two copies of the board are inverted across the horizontal mid-plane. I’m not 100% sure why they do this, but I suspect it has to do with balancing throughput on the SMT line. One side of the PCB is very dense with components and the other side isn’t. Since the board has components on both sides, you will have to flip the board over at least once to process each side. On high-throughput SMT lines, this means that you’ll have two (expensive) SMT pick/place (or chip shooting) machines back to back, one to handle the top, and the other to handle the bottom after flipping. If you have both boards top-side up, then the SMT machine handling the denser top side would run much longer than the one handling the sparser bottom side, and you would effectively be idling one machine for a relatively long period of time. This would increase the amount of capital you have to spend to scale your production line and require greater amortization for a given throughput, ultimately driving up product cost (or reducing your net profit). Mirroring the board about a central axis also allows you to use the exact same programming for the two back-to-back SMT machines (separated by a robotic flipping mechanism) so you only have to optimize one SMT program, although at the expense of possibly having to incorporate a greater mix of parts within a single SMT machine.

Winner, Name that Ware June 2009

August 4th, 2009

The Ware for June 2009 was a Hoya L-210-P2 laser power supply. It’s used to drive a high energy pulse laser at 532nm, and it’s used as part of a probe station setup that’s capable of cutting metal wires on integrated circuits. There’s one unfortunate part about using a 532nm laser to try and cut metal wires on a chip: with modern ICs pushing 45nm geometries, a 532nm laser is more like a sledgehammer and less of a scalpel for doing circuit edits.

At any rate, this Ware generated quite a comment thread–I didn’t think it would be so interesting or difficult to guess. Many people came close and a couple even guessed it. For being the first to guess correctly, as well as for the detailed and thoughtful comments, KE5FX is the winner of this competition. Very good sleuthing and great intuition; send me an email to claim a prize. Thanks to everyone who played!

As requested, here’s some links to larger photos of the ware: 1 2

Nokia Schematics!

July 26th, 2009

I was wandering around on a rainy Saturday afternoon in the mobile phone market in Shenzhen and I spied a stall keeper working on a phone motherboard. Unlike most of the other folks in the market, he was working from a set of schematics — that got my attention. I asked him where he got his schematics from and he kindly dispatched his young son to walk me over to the small tool shop on the other side of the market where, buried underneath a pile of single-use BGA SMT stencils, was a collection of mobile phone schematics for just about every phone made.

I don’t know about you, but getting my hands on schematics gets me really excited. This is like, the Ultimate Hardware Geek Pr0n. Sure, undressing a mobile phone and revealing its tender innards to my gaze — the sweet perfume of flux residue unleashed, curling into my nostrils — is one level of hardware voyeurism. But, getting the schematics for the phone and peering into its very circuit diagrams — that’s a whole new level, like tearing off the undergarments and ravaging the bosom of the phone. I was excited. I brought the manuals to the clerk and asked how much…wincing at the price I may have to pay to bring these prized morsels back to my hotel room. I breathed a sigh of relief when he asked for only 75 quai — a little over $10 US — for the whole three-book collection. I didn’t even haggle. I grabbed my booty and ran for the nearest taxi.

Alright. Don’t take my double entendre too literally; it’s just fun to write that way. Anyways, on to the manuals…

As you can see, these are sold as service manuals.

A sample index entry.

Each phone has a scan of the circuit board that is annotated to call out the position and function of all the components. This helps with the repair process.

A detail of the schematic for the Nokia N95’s RF section.

The same publisher of these schematics also offers a wide library of schematics, including those for Samsung, Sony Ericsson, Motorola, and Chinese-local “copycat” phones (like the iPhone clone in the lower left hand corner).

It’s hard to say if these manuals are legitimate (in the sense that Nokia deliberately released the schematics for use as a repair guide), or if they were stolen and republished as a “repair guide”. I doubt, however, that they were reverse-engineered out of the phone, because the schematics contain references to codenames and lingo that would not be embedded directly in the circuitboard. A bit of searching around on Google reveals that these manuals are available for downloads in bits and pieces on various websites, but they all seem to derive from a Chinese origin, and not from Nokia directly.

When I give talks about Open Hardware, I emphasize that it’s fundamentally impossible to keep hardware closed, because the source is the same as the product. Schematics can be derived out of a circuit board layout — a completely legitimate activity under US law. There are shops in China that will pull out a netlist in a couple days for under $1k, and services in the US that will take a bit longer and cost you a lot more. Significantly, there is a whole lot more that goes into building hardware than a mere schematic design: it takes me days to capture a schematic, but it takes me months to get it into mass production. Thus, I believe that publishing a schematic makes the product more serviceable and more useful, yet has little negative effect on your competitiveness in the marketplace. This is a key difference between software and hardware, as the time between writing software and publishing it to “production” can be as short as a few seconds for web-based services.

From my personal perspective, having the schematics is handy for a number of purposes. Aside from satisfying a general curiosity about the phone’s structure, it’s interesting to see the details on how certain sub-circuits are implemented. For example, when I learned electronics at MIT, they never directly taught me how to do EMI mitigation, or ESD protection. While I know the theory behind it, the implementation is trade-craft know how; these are subjects where experience trumps knowledge. Therefore, seeing Nokia’s take on it expands my understanding of the subject.

The schematics are also very useful because of Nokia’s buying power. Picking the cheapest part for a mass-produced hardware design is a tricky exercise; when you leave the realm of buying a few hundred or thousand pieces at a time and move into really high volumes, often times the price of the part has less to do with its design features and more to do with its physical dimensions and who is buying a lot of it. If someone like Nokia is buying millions of a certain part a year, the supply of this part is very stable, lead times are shorter (usually), and the price goes down. So, if you’re a small company and you want to build something cheap, you want to pick parts out of the Nokia supply chain because you indirectly enjoy the benefits of Nokia’s buying power. Thus, these schematics are a good starting point for sourcing cheap parts for production.

Update: got the books for Sony-Ericsson, Motorola, and MediaTek-chipset phones. Found only one Samsung book and it was very out of date. From my observations, the books left on display in these small shops are remainders — typically a little older, perhaps a little less popular. I had to scour the market to find a few of the more recent books, which probably means they are out there but I just don’t know the right person yet. Interestingly, the MediaTek-chipset series of books included some “how-to” manuals on using the chipset to build your own phone. This may be part of the “open” repository of hardware knowledge I previously mentioned in my post about the Shanzhai phone-copiers-turned-innovators.

Eclipse (as good as it got)

July 21st, 2009

The eclipse got down to a little sliver on one side, but I couldn’t get a good photo of it. This was the last time the sun passed behind a cloud that allowed for my camera to take a good photo, about 10 minutes before peak occlusion.

Eclipse (so far)

July 21st, 2009

The eclipse so far…got lucky and a cloud passed over it so I could use my digital camera directly instead of my pinhole rig.