Diving Deep on the Zune: An Inside Look at the Hardware [Archive]

Darius Wey

12-05-2006, 03:00 PM

<img src="http://images.thoughtsmedia.com/zt/2006/wey-20060722-zune.gif" alt="" /> Last month, <a target="_blank" href="http://bunniestudios.com/blog">Andrew "bunnie" Huang</a> (well-known for his work on hacking the security architecture of Microsoft's first-generation Xbox) <a target="_blank" href="http://www.bunniestudios.com/wordpress/?p=131">revealed</a> the guts and glory of the Zune. He took time out this month to give us an inside look at the hardware, covering just about every component from head to toe. Read on for more! <div style="page-break-after: always;"> </div> Zune Thoughts: When we first saw the Zune, our initial impression was that it was quite thick. We thought there would be wasted space inside, or perhaps room in the chassis for a bigger hard drive. Based on your photos, however, it looks like they used up all of the space. How is the 30GB iPod able to be 50% thinner? What makes the Zune so much thicker from a components point of view? The battery, Wi-Fi chip, or something else? Andrew: First off, I haven't taken an iPod apart, so I can only draw comparative conclusions based on the few tear-downs of the iPod that I can find on the internet. I'm using <a target="_blank" href="http://www.appleinsider.com/image.php?i=ipodvideoteardown&id=1338&pe=1">this photo</a> of the iPod as my reference for comparison. One can see from the two photos that the key to the iPod's smaller size is its smaller display. Because the display has a small footprint, they can lay out the circuit board and the display side-by-side; the Zune, on the other hand, has to stack the display, circuit board, and hard drive, adding considerable thickness to the device. The Zune's display is sufficiently large that Toshiba probably would have a very difficult task of getting everything to fit into the small space that's not occupied by the display. They also probably could not have made it much smaller length or width-wise because the area not occupied by the display is used up by the UI components. Therefore, the "killer" component is the display. It's just one of those unfortunate things where the decision to make a little extra (0.5") display diagonal creates a cascade of other engineering problems that inflate the device in every direction: it was the proverbial straw that broke the camel's back. One thing Microsoft could have done to atone for the size impact of the display is to use a touchscreen-only interface, allowing them to drop the cluster of buttons and thereby reduce the length/width footprint. The design could feasibly do this, although clearly there are reliability and cost implications that may have been unacceptable. In the end, I would say the larger display wasn't worth it in this case, because if you're already watching video on a tiny display - and come on, 3" is still tiny - you might as well leverage the portability feature as much as possible, even if it means giving up a little display area to make a thinner and more portable device. After all, a portable device needs to be "backward compatible" with the existing infrastructure of pockets and fashion accessories, and there are certain breakpoints at which you just push yourself out of (or into) a market with that last millimeter of thickness, as the iPod nano aptly demonstrated. The other thing that struck me was the battery pack's size. The Zune battery is rated for 800mAh @ 3.7V; depending upon which website you trust, the 30GB iPod video's battery is around 400mAh @ 3.7V. The volume of a battery pack is pretty much related to the capacity of the battery, and the Zune requires twice the juice of an equivalent iPod for the same playtime... so the Zune battery pack is probably about twice the volume of the iPod's. Also, the plastic housing is pretty thick on the Zune. It has that dual-molded translucent plastic over colored plastic all-around, whereas the iPod has a metal backing. The metal case is a bit stiffer and thinner than the plastic case, so that saves a millimeter or so here and there. The problem is, of course, that if the Zune used a similar material system it'd just look too close to an iPod, and they had to distinguish themselves - being big is distinctive, I guess, like driving a Hummer around. They seem to have a thing about size - remember the original Xbox controller, and how huge it was? You also see a lot of big SUVs in Seattle. Maybe they just have big hands and deep pockets out there. I'm just kidding, I couldn't resist. Zune Thoughts: Looking at the components they used in the Zune, how would you grade the overall design? We've read comments stating that they used many discreet components, and the lack of integration made the Zune bigger and less battery efficient than it should have been. Is that a fair assessment? What would you change or improve upon? Andrew: The overall design is fairly average. I've seen better, and I've also seen worse. The number of smaller components used by the design is pretty much an artifact of their choice to use the iMX31; it has some unique properties that necessitate a lot of extra components. For example, the iMX31 features dynamic voltage frequency scaling (DVFS), where the processor literally overclocks itself based upon load demand. If it's not doing much, it runs in a low-frequency, low-voltage mode to extend battery life; when going all out, it overstresses the processor and trades off long-term operational stability for short-term performance gains. The aggressiveness of this is up to the system designer and their anticipated obsolescence date, and the quality of their testing program (due to the unique choice of voltages that the processor uses, it is conceivable that a bug in the software could burn out the processor given enough time, although I must emphasize that this is an unlikely scenario). Implementing DVFS requires a lot of extra components in the power supply system. Also, due to the 90nm technology used to implement the iMX31, it requires glue logic to convert voltages between older legacy standards, such as those used on the hard drive, and the lower-voltage standards supported by the iMX31. Finally, features like the FM radio are a little bit wart-like in the sense that it appears they introduced a whole additional stereo codec just for the FM radio: there actually is another codec hiding in the device that is unused for reasons I don't quite understand. I'm guessing that some product manager decided that a high-quality 16-bit digital experience was needed for FM audio, instead of the slightly lesser default codec that already existed in the design, but really, consider the source material: FM radio is not CD-quality audio to start with. So they introduced some extra components for what seems to be a game of specsmanship. One other competing theory is that they used the extra codec because they couldn't get sufficient noise isolation for the audio lines to pipe them around the circuit board, but man, I'd be disappointed if that was the reason. It's difficult to make quiet signals, but not impossible. What is interesting is that the iMX31 is a processor that is used in many cell phones, and the design parameters of the iMX31 is completely consistent with a very high-end, small and compact cell phone which would retail for a few hundred dollars, assuming no carrier subsidy. However, the Zune tries to pack this high-end processor into a low-cost device, which probably necessitated some trade-offs to be made that disfavor size over cost. As for the manufacturing technique, it's very typical for a Japanese-manufactured device. I don't know what it is, but Toshiba has a certain manufacturing methodology that you can identify just by looking at it. It's hard to say what I would change or improve upon, because I don't know what the internal design goals of the Zune are. However, the pattern of design embodied in the Zune is fairly typical of the other products you've seen like this coming out of Seattle - overpowered processors that help ease the task of running generic software - under the theory that it's all about the software, and hardware should grow to fit the software - and some warts that are the inevitable result of not owning and understanding the hardware design process. It's pretty clear that Microsoft loses some agency over its destiny when it subcontracts hardware design. If anything, the fact that the device is ODMed by Toshiba hurts the bottom line, because Toshiba isn't a charity doing this for free! Zune Thoughts: What about the CPU/DSP - did they make a good choice there in terms of functionality and performance? There has been speculation that one of the reasons the Zune is so thick is that it has more "under the hood" than an iPod and thus has room to grow into a more functional unit as they update the firmware. Would you agree with that assessment? Andrew: I don't particularly agree with the assessment that the Zune has more under the hood. Apple packs several processors in their iPod video, according to the tear-down I am <a target="_blank" href="http://www.appleinsider.com/article.php?id=1338">referencing</a>. The iPod has a dedicated audio decoder, and a Broadcom video chip that integrates both an 300MHz ARM11 plus a 150MHz "videocore" co-processor. The Zune uses a single 533MHz ARM11. Comparing just the two ARM11's against each other neglects the advantages that the performance-tuned co-processors provide, especially in a battery powered device. Lots of MHz is great for a general purpose machine, but the generality comes at the price of higher power and higher cost. If you know that all you are going to do is to decode video, why not shed logic paths you never use and optimize the system? To that extent, if you believe the literature on the Broadcom videocore feature, the ARM11 needs to run only housekeeping code at 10MHz while the 150MHz videocore churns away at H.264 playback. Thus, you get comparable performance to a 533MHz ARM11 doing it all at once, but with just a fraction of the cycles, therefore a fraction of the wattage. The Broadcom literature also claims that the videocore is fully programmable to support any codec, but I imagine that it requires some expertise and care to program - but that's not a big deal, because you only need to write it once. Given that the iPod's battery is half the size of the Zune's, you can see how this design trade-off benefits in terms of power and ultimately size. The philosophy of going with a big, huge general purpose core is very typical of the Microsoft view of the world. Their core focus is creating high-level code, and hardware is abstracted away through the magic of the compiler; this pattern was established in the PC space. As long as Intel and AMD make backward and cross-compatible instruction sets, Microsoft is happy. Intel and AMD's price wars have guaranteed Microsoft a plentiful supply of powerful, cheap processors. This is also why we have the multi-platform compatible DirectX, keeping NVIDIA and ATI in each other's backyards and thereby keeping graphics card prices down. Apple, on the other hand, has not been shy about digging into the hardware, as it played a hand in defining the PowerPC specs and has changed processor architectures three times now in the evolution of the Macintosh. This attention to the hardware has earned Apple the distinction of building very svelte, tight but more expensive hardware, which in my opinion remains a major brand differentiator for them in several key markets, particularly outside of Microsoft's core enterprise/business market. It is interesting to see how Microsoft's hardware offerings are at price parity with Apple's offerings when you remove the multi-source hardware market forces that Microsoft normally enjoys in the PC space. Zune Thoughts: Right now, the only video support on the device itself is WMV. Does the hardware support anything else from a codec level, or is that all in the software? Microsoft didn't pull an Apple, did they, where the hardware supports WMA, but Apple disabled it? Andrew: Well, in the case of the Zune, the heart of the iMX31 is a powerful ARM11 CPU doing all the number crunching. There is no codec-specific hardware that you might find on less powerful processors. I have personally seen the same iMX31 processor decode H.264/AAC video quite well, so there's no reason why it couldn't do more formats. It's just a matter of what codec software the designers decided to put on it. Since it seems most of the software is stored on the 30GB hard disk, there certainly is no storage limitation for adding more codecs. However, one obvious problem of adding a codec like H.264 or AAC to a hardware product is that you have to pay the MPEG4 licensing board a non-trivial licensing fee. So, even if the hardware could decode any video, it may not make business sense to pay a licensing fee to third parties, especially if you have your own codec solution, like Microsoft has. It depends on what the company's market goals are. One purpose of introducing a hardware platform is to create a barrier to competition, so quite frequently manufacturers will limit the compatibility of their devices to achieve this goal. Clearly, this is at odds with creating a convenient product to use that supports every format under the sun. Of course, Microsoft has no choice but to believe that its software technologies are superior and will win the day. If they didn't believe this, then what else would they be good at? Zune Thoughts: Hard drive size matters more to early adopters than most people, so there are some (like ourselves) wishing there was a bigger option than 30GB. It seems that you can get a 40GB in there, but a 60GB or 80GB driver would be too thick - you'd have to change the back chassis. That leads us to believe that the Zune team didn't want to release a higher-capacity Zune until the 60GB and 80GB drives are as thin as the 30GB drives. Does that theory make sense to you, or could they have fit a bigger drive in there? Andrew: I think it wasn't size, but cost. The Zune packs in Wi-Fi and an FM radio tuner; the extra silicon cost of the higher-end processor is probably cancelled out against the extra silicon cost of the iPod's multiple co-processors. Well, actually, if you dig into it a lot, the Zune probably pays a hair more because of some arcane trade-offs related to the use of the iMX31's companion chip that implements DVFS control, which has more features than a swiss army knife but the Zune only uses a fraction of those features. The companion chip was designed for economical integration into a cell phone, not a media player. At any rate, the Zune probably carries an additional $20-$30 of parts cost alone over the iPod video, which is killer when your margins are already squeezed. Remember, Toshiba designed and manufactured the Zune, and Microsoft adds the software - and Toshiba is not a charity. It stands to reason that Apple would have a better cost structure over Microsoft due to its higher volume and deeper involvement hardware manufacturing. Since extra capacity means extra $$$, and Microsoft is already on the edge, I think they traded off capacity for price. As for how this thing fits into the case, the 30GB drive is 5mm thick, and the 80GB drive is 8mm thick. The metal shell for the drive housing has about 7.5mm of clearance built into it, so you can probably shimmy in that last 0.5mm with a little elbow grease. Zune Thoughts: Thanks for taking the time to dive deep with us on the Zune hardware - it was a learning experience. Andrew: No problem! It's a hobby of mine. <a target="_blank" href="http://bunniestudios.com/blog">Andrew "bunnie" Huang</a> is a nocturnal hacker and the hardware lead for Chumby; his responsibilities include the architecture, design and production of Chumby's electronics, as well as writing drivers for and maintaining the Linux kernel on Chumby. Wielding a PhD in EE garnered from MIT in 2002, he has completed several major projects, ranging from hacking the Xbox (and writing the eponymous book), to designing the world's first fully-integrated photonic-silicon chips running at 10 Gbps with Luxtera, Inc., to building some of the first prototype hardware for silicon nanowire device research with Caltech. bunnie has also participated in the design of 802.11b/Bluetooth transceivers (with Mobilian), graphics chips (with SGI), digital cinema CODECs (with Qualcomm), and autonomous robotic submarines (with MIT ORCA/AUVSI). He is also responsible for the un-design of many security systems, with an appetite for the challenge of digesting silicon-based hardware security. bunnie is also a contributing writer for MAKE magazine and a member of their technical advisory board. bunnie studios LLC is bunnie's private banner under which he performs the bulk of his security and reverse engineering work.