Tuesday, February 10, 2009

Much Cheaper Than a New One: Custom Desktop Computer Rebuild
It had been a while since I had upgraded my main desktop machine. Been building my own custom desktops since the mid-1980's, starting with a PC-XT clone with an Intel 8088 processor, 640kB RAM (yes, kB) 30MB (yes, MB) hard drive, Hercules graphics, and a 14" monochrome monitor. The technology has changed substantially since that first machine, and my primary desktop has evolved through an Intel 286, AMD 386, Cyrix 486, AMD K6-2, AMD Duron (Applebred), AMD Athlon XP, and into the AMD Athlon 64 processors. Before this rebuild, my main desktop was running an AMD Athlon 64 Socket 939 4000+ and AGP 8x graphics. It's upper the limit of Socket 939 AMD processors, with AGP or PCIe graphics. By no means a slouch in speed (faster than most current mid-level machines), it was time to upgrade it to multi-core processor and PCIe graphics. Had some decision-making to do. There is some system design required when doing a custom build. One can't push any hardware combination together and expect it to work well. It might, but it might not. Learned that the hard way when I started building my 386 and 486 machines. Nearly all compatibility considerations are driven by the motherboard and its chipset. However, it shouldn't be the first choice. I choose the processor and desired chipset (Northbridge / Southbridge). Then I look for a motherboard with the chipset that supports the processor. Looked at both AMD and Intel processors. Intel holds the edge on speed, but that's at the bleeding edge of processor technology, and they're quite expensive. Back down slightly to the top end of AMD's processors, and they're less expensive than the equivalent Intel. In the process, I discovered the AMD quad-core processor clock speeds were still substantially lower than top end of AMD's dual-core. In addition, current operating systems and application software are still geared for the lowest common denominator in the majority of machines in the current "installed user base": 32-bit single-core processors. Even the default install for Microsoft's newest operating systems, Vista, is 32-bit. The latest budget machines are still shipping with 32-bit processors (check out what's inside many of the "loss leaders" in the computer store ads). I opted once again for another AMD processor, a Socket AM2 Athlon 64 X2 6400+ dual-core with 3.2 GHz core clock. One of the chipset makers I've used heavily in the past, Via Systems, basically stopped supporting AMD processors with the Socket 939 and 940 Northbridge/Southbridge chipsets. Disappointing as nearly all their chipsets were excellent. The only real choices for an AMD Socket AM2 or AM2+ are the AMD and nVidia chipsets. Which one picks also tends to lock in which graphics chipset will be used: ATi (owned by AMD now) or nVidia. It is possible to use an nVidia graphics card on an AMD chipset motherboard, and vice versa, as the PCIe socket it is plugged into, and the graphics interface behind it is an industry standard. However, it doesn't allow leveraging on enhancements that can be had if an ATi graphics card is used with an AMD chipset motherboard, or an nVidia card is used with an nVidia chipset. This is especially true if one wants to run multiple graphics cards in tandem (i.e. ATi CrossFire or nVidia SLI). I've used nVidia graphics for a long time (4000, 5000, 6000 and 7000 series) and have always been pleased. After looking at the nVidia and AMD Northbridge/Southbridge chipsets though, I settled on the AMD's 790FX and SB600 chipset. These are found on AM2+ motherboards which support dual, triple and quad-core processors. If I wanted to upgrade to a quad later (as they gain core clock speed), I could. It meant I'd be using ATi (Radeon) graphics. After reading some about ATi's Crossfire, I picked ATi's HD3870 with plans to use a pair of them in CrossFireX mode. The current bleeding edge is the HD4870, but they're expensive. Could have gone with HD4850, one notch down from the 4870's. It was pretty much a coin-toss until I found substantial rebates on a pair of HD3870's. Added up the power requirements for all this and realized quickly I'd need a new power supply; one that's bigger than the solid 500 Watt that's been used for several generations of motherboards. It wouldn't be enough for the processor and pair of graphics cards. The new hardware architecture was settled:
  • AMD Athlon 64 X2 6400+ Dual Core
  • AMD 790FX Northbridge and SB600 Southbridge
  • AM2+ socket motherboard with ATX form factor to fit case
  • 4 GB DDR2 800 MHz (PC6400) low latency RAM (4-4-4-15-T1 timing)
  • Two ATi HD3870 Radeon cards in CrossFireX
  • Sony DRU-865S SATA DVD/CD Burner (with Lightscribe)
  • 850 Watt Antec power supply
  • 300 GB SATA II hard drive (reused from current machine)
  • 3.5" micro-floppy drive (reused from current machine)
  • Flash card reader (reused from current machine)
  • Soundblaster Audigy SE audio (reused from current machine)
The first step, before disconnecting anything from the existing machine, was backing up its hard drive completely. I used an external USB hard drive, and cloned the hard drive inside the machine to the external drive. These things are handy. Following drive upgrades, I've been putting the old drive into an external USB case. The contents of the 300GB drive fit onto the 250GB external USB drive with room to spare. The new machine will have a different processor, different supporting chipset on the motherboard, and equally different graphics. That requires completely different operating system drivers. The cleanest, most trouble-free rebuild requires completely reinstalling the operating system. I've tried simply booting up with the existing hard drive without reinstalling the operating system in the past. Doesn't work very well, and requires more time and effort to get the new machine working right compared to backing up the drive, wiping it clean, doing a complete operating system install, reinstalling application software, and then copying legacy data to be retained from the backup. A clone of the old drive ensures everything is saved, and can be retrieved. The only things to be reused in this project are the computer case, hard drive, floppy drive, flash card reader, and sound card. The case was originally bought to house an AMD K6-2 (Socket 7) system on an ATX form factor motherboard running Windows 95. It's solidly built out of heavy gauge steel; not a cheapie and worth reusing. Buying a new one with its materials and build quality, even without a power supply, would cost well over $100. The next step, after backing up the hard drive completely, was gutting it completely, down to just the case itself. Only one problem with reusing this case though. It's a horizontal desktop box which will be a little crowded after all the new hardware is installed, and it wasn't made to dissipate the heat generated by current higher-end processors, motherboards and graphics. Although there's an air intake fan on the front, the only exhaust is out through the power supply in the back. That sufficed until now, but it's insufficient for this upgrade. Pushing the very hot air generated by current high-end processors and graphics card(s) through a power supply would be hard on it. It needs some exhaust fans on the back, and the flat panel on the back, just above the rectangular hole for the ATX motherboard connectors, is the only viable location for mounting cooling fans in the rear of the case. It's the only unused area on the case back. Some measurements of the panel shows a pair of 60mm fans can be mounted next to each other. The task now is drilling the holes for the fans, and that requires a bi-metal hole saw. Turns out I've got one just the right size. The same size hole saw used for drilling holes in metal doors for door handles will make a hole just the right size for 60mm cooling fans. The fan mounting hole centers are marked first, and lines drawn between opposing centers locate the center point for the hole saw's pilot drill. Drilling sheet metal, even if it's steel, can deform it, especially when the drill begins to break through to the other side. A piece of scrap wood is clamped to the area of the case being drilled out to provide support. The other technique is not putting too much pressure on the drill, or too high a drill speed, using just enough to let the drill's cutting edges do the work. When it does break through, there's not as much metal left for the cutting edges to grab onto and twist. It also reduces the amount of burr left behind that must be removed after drilling the hole. The holes have been finished and the burrs removed. The remnants of marks made to locate hole centers can still be seen, but removing those doesn't matter much. It's on the back of the case that won't be seen when the computer is in use. I used a fine-tip Sharpie "permanent" marker. Some alcohol or similar solvent would easily remove the marks if desired. The fans have been mounted taking care to route their power cables so they don't get pinched, and won't interfere with the rest of the computer's components when they're installed in the case. The pair of fans selected are variable RPM to minimize fan noise to the level necessary for cooling. They have heat sensors on them, starting at a low RPM at room temperature, that increases fan RPM as heat from the air passing through them increases. These types of fans do the job for exhuast air; not recommended for intake fans (the air would always be cool). A look from the back with the new fans. I had considered adding a couple of grilles over the fans to keep things (like fingers) from coming into contact with spinning fan blades. The motor mounting in the fan box provides sufficient protection for the fan blades on these particular fans. They'll be OK without grilles over them. Note the horizontal line marked above the fans, and the small hole just above the fan on the right. The top case cover has a flange on the back, and the hole is for one of the five screws that attaches the case cover to the case. The fans had to be mounted below this line and hole as the case cover screw protrudes into the case when its cover is installed, and the fan mounting screws would interfere with the cover flange if it extended over the top of them. The fan mounting screws are nearly flush, but not quite. A vacuum cleaner and its hose attachment were used to completely clean all the sawdust, and more important, the metal filings out of the case. I removed and put back the drives in the drive cage while doing this. In retrospect I should have removed the CD/DVD burner, hard drive and 3.5" micro-floppy drive before drilling the holes and mounting the fans. I still use a 3.5" floppy disks occasionally. Booting from a floppy is required to flash a BIOS update on nearly every motherboard made. Now to deal with the motherboard. After looking at various AM2+ boards with 790FX/SB600 chipsets, reading some reviews on them, and comparing prices, I picked an MSI K9A2 Platinum that includes USB, Firewire and 1000BASE-T Ethernet on the motherboard. Even though it can handle up to four graphics cards in CrossFireX, I only plan to use two.Most motherboards come with plenty of cables for connecting peripheral devices (e.g. disk drives). This one is no exception. Also included is a backplate for installation in the computer case to provide the proper cutouts for all the rear connectors on the motherboard. The original ATX form factor had a fairly standard layout for its connectors, but that didn't last long. Centronics parallel and RS-232C serial ports were supplanted by USB, Firewire, Ethernet, and embedded audio connectors. Now, eSATA ports are becomming common (this motherboard has a pair of them).The next step is putting the backplate into the case. All the ones I've encountered with motherboards snap into the rectangular hole for them. Occasionally the motherboard makers will use one plate for several different motherboards and some tabs on them must be tweaked or completely removed for ports and connectors on the motherboard that wouldn't need to be for a different model number motherboard. The next photo shows the backplate installed into the large rectangular hole in the back of the case. The bottom halves of the two cooling fans can be seen just above it. Before handling the processor, motherboard, or any of the peripheral cards, there's an essential device called an ESD wrist strap that needs to be put on and the cord attached to a ground point on the case (can be clipped to any convenient, unpainted, exposed metal inside or around the case. While many folks get away without using one, the bare processor is an extremely ESD sensitive device. It can be easily damaged by ESD. An ounce of prevention is worth a pound of cure (and the cost of a new processor). When doing a rebuild that includes a new motherboard, I've found it easier to install the processor into its socket on the motherboard before installing the motherboard into the case. Doing it after the motherboard is in the case necessitates reaching down inside the case which isn't impossible, but is more difficult, especially with lining up the pins on the processor with the holes in the ZIF socket on the motherboard. This processor is the "retail" version that includes a heatsink and cooling fan for it. The amount of heat generated by this 135 Watt processor can be seen from the size of the and cooling fan and the heatsink it's mounted to, complete with copper pipes to facilitate heat transfer. It dwarfs the processor IC and is one of the reasons I added additional exhaust fans to the case. ZIF is an acronym for "Zero Insertion Force." ZIF sockets unlock to allow inserting an IC, cable, or connector with near zero "force" or friction, and then lock again after the device is seated. Processor ZIF sockets on motherboards have a locking arm along the side of the socket that's lifted to unlock it. The square socket takes a "Pin Grid Array." This processor has a grid of 940 small pins on its underside. (The four long "card edge" sockets to the left of the square processor socket are for RAM.) After the processor is carefully dropped into its socket, it literally drops into place when the pins are lined up correctly, the socket is locked using the locking arm and contacts inside the socket grip each pin tightly. The large black frame around the processor socket is for the heatsink, which attaches to and uses the frame to press the surface of the heatsink firmly onto the top of the processor. Next, the case is prepared for the motherboard. ATX form factor mounting holes are fairly standard. There are two or three variations, and some motherboards do not use all nine mounting points. Consequently, cases use removable brass studs and have threaded holes at all the possible mounting points for them. With this motherboard, I needed to move a couple of them as its mounting points were a different variant than the old motherboard. The studs are hexagonal which allows using a hex nut driver to remove and install them easily. Care must be taken to avoid over-torquing them as it can strip their threads, or worse yet, snap the stud off of its threaded end (requires extracting it from the hole to replace it with a new one; not easy to do). One of the cool tools I found some years ago is this one. Mounting points on the motherboard are surrounded closely by components and connectors, some of which are tall. This tool can grip a Phillips screw by the slots in its head, and it allows getting the screw started into its hole. If, perchance, the screw is dropped, it typically falls down into someplace that's hard to reach (one of many corollaries to Murphy's Law). There's a small magnet on the other end of this tool that allows retrieving it much easier than trying to use needle-nose pliers. The motherboard can now be installed into the computer case. As a double-check, I count the number of mounting holes in the motherboard, and the number of brass studs in the case, to ensure they match. There's not much extra room; typical of horizontal desktop cases. Care must be taken to get all the motherboard's external connectors through the holes of the backplate correctly before installing any of the mounting screws. The bottom of the motherboard should not be dragged across the tops of the studs in the case while inserting it, but should be lifted off of them slightly when adjusting it. That can damage copper traces on the underside of the board. The screws are started using the tool above, and then tightened down firmly (again, taking care not to over-torque) using a standard #2 Phillips driver. More clues to the heat that will be generated by the motherboard are the three heatsinks on the 790FX Northbridge, SB600 Southbridge, and a third IC, all connected together with copper tubing. The power supply comes next. When power requirements hit exceed 650 Watts and higher, the power supplies get bigger. Because of standard mounting points and dimensions inside cases, they get longer, extending deeper into the case. This 850 Watt Antec is no exception, extending a couple inches farther toward the front of the case than the 500 Watt supply it replaced. Fortunately there's enough room between it and the drives mounted in the drive cage for the power supply and drive cables, but it's just barely enough room. The blank space just to the right of the motherboard in the photo above is now completely filled by the power supply. The processer heatsink/fan assembly and the RAM are now installed. Even though there are four card edge slots for the RAM, I used two 2GB "sticks" for a total of 4GB RAM. They go into the slots nearest the processor, and the heatsink crowds them a little. The fans AMD uses on the heatsinks included with a "retail boxed" processor are exceptionally reliable and quiet. The large fan size helps reduce noise as it can turn at lower RPM to move the same amount of air. The lifespan of every AMD fan I've had has exceeded every aftermarket processor cooling fan I've bought. Don't know what their service life is yet as they're all still running. It's a major reason I try to buy "retail box" versions of AMD's processors. In this case, it was less expensive than buying a "bulk pack" version of the processor (intended for manufacturers and system builders) and separate AMD heatsink/fan assembly rated for this processor (they can be found separately, but it requires some searching on-line for them). The RAM also has heatsinks; typical of low latency, high speed RAM as it must be run at a slightly higher voltage to work correctly. This generates more heat which must be dissipated. Could have used four 1GB "sticks" or one 4GB "stick" but there are reasons for using exactly two of the slots and putting identical RAM in them. An even number (versus 1 or 3) of the same size allows the motherboard to use "dual channel" RAM access which it cannot do with an odd number. This greatly speeds RAM reads and writes. In addition, if all four slots were populated, a couple RAM timing parameters would have to be slowed down slightly (at the least, the T1 in its timing specs would have to back off to T2). Using exactly two, with exactly the same RAM in each of them, allows optimal RAM access speed, to the timing limits specified for the RAM by its manufacturer. The first of the two graphics cards is inserted into its PCIe slot before adding the other cards as it's the innermost slot. The second one and CrossFireX cables will be added after the machine is up and running with the operating system, hardware drivers, firewall (I disable Windows' firewall and use one of my own), and Anti-Virus software. The size of its cooling fan and heatsinks are yet more clues about the amount of heat this system will generate. The remaining cards, a Soundblaster Audigy and a WiFi card, are installed in two PCI slots, and then all the internal interconnecting cables are run. Unless there's a piece of hardware that will prevent connecting a cable after all the hardware is installed, I save running all the internal cables until the end of the hardware build. Doing this keeps them from getting in the way of installing hardware, and it allows routing the cabling properly over and around the hardware as required. Doing it before this often requires re-routing cables again after additional hardware is added. After double checking that all the internal interconnect cables are properly installed and connected by systematically inspecting each component, it's time for the "Smoke Test" (apply power and see if anything starts emitting smoke; no smoke means at least there aren't anycatastrophic mistakes). It's the reason the CPU cover hasn't been installed yet. Keyboard, mouse, monitor and speakers (in this build, the monitor has built-in speakers) are connected, and then the power cords. Wiping the hard drive completely clean before gutting the old system prevents the machine from attempting to boot the old install of the operating system and makes accessing the motherboard's system BIOS easier. It's passed the "Smoke Test" and no errors were indicated when the BIOS performed its POST (Power-On Self-Test). The BIOS is accessed by depressing a specific key on the keyboard when the machine is powered on. The manual for the motherboard contains instructions on which key to use (typically the F1, F10 or DELETE key, it varies by motherboard manufacturer). Time to make some BIOS settings specific to the hardware in this build:
  • Enabling the USB and Firewire ports
  • Disabling the Serial and Parallel ports (I'll never use them)
  • Disabling the on-board audio (I've installed my own Soundblaster card)
  • Enabling the SATA drive interfaces and disabling the IDE drive interfaces (I'm using a SATA hard drive and CD/DVD burner)
  • Disabling RAID (multiple hard drives can be configured to appear as a single one)
  • Setting RAM operating voltage and timing to its manufacturer's specs
  • Some additional timing parameters for the system bus (if you don't know what they are and what they do, and it isn't explained in the motherboard manual, leave these alone until you can look them up using Google).
The machine is then re-booted several times into BIOS to ensure it's at least stable enough to boot into BIOS. The cover is now put onto the CPU and the operating system is installed. The machine should not be connected to any network while installing the operating system! Firewall (if the one bundled with WinXP isn't used) and Anti-Virus software should be installed before connecting the machine to a network. The operating system is vulnerable to attack while it's being installed, and if it's allowed broadband internet access in the process, the operating system will be compromised. There are thousands of automated "bots" that scan the internet for vulnerable computers. A WinXP machine was compromised in less the 4 minutes in a test done by PC Magazine several years ago. A close friend forgot to disconnect his computer from the internet before installing WinXP. When he got done, he had to reformat the drive and start completely over. It had become a "Zombie" drafted into a Botnet Army. I'm putting Windows XP /x64 Professional on this platform. The "/x64" variant of Windows XP is the 64-bit version of WinXP. The common variants of WinXP (Home, Pro and Media Center) are all 32-bit. They'll run on a 64-bit hardware platform, but I prefer /x64 as it was created from 32-bit Windows Server 2003 (the server version of WinXP), which came several years after WinXP Home and Pro were released. Windows Server 2003, and WinXP /x64 are noticeably more stable than the 32-bit (aka /x86) WinXP Home or Pro. It does require having true 64-bit hardware drivers for all the hardware inside the machine, and for anything connected to it using a USB, Firewire, serial or parallel ports (e.g. printers, scanners, PDAs, cell phones, MP3 players, etc.). It's not a problem now, although it was one just after /x64 was released. Even iTunes can be installed for iPods (requires an easy to perform, minor hack to iTunes for Vista /x64). After intalling the operating system, hardware drivers are installed as needed for the processor, its supporting chipset, and other peripherals (e.g. graphics and audio). The general sequence I use:
  • Operating system (WinXP /x64 for this build)
  • Hardware drivers
  • Firewall and Anti-Virus software
  • Connect machine to internet and update A-V software
  • Update operating system with all critical updates (Windows Update for this build)
  • Customize the Windows user interface
  • Benchmark software (after which I run some benchmarks)
  • Application software (e.g. MS Office, OpenOffice, Adobe Reader, Nero, various games)
One of the benchmarks I use is Futuremark's 3D-Mark. The reason for running some benchmark software before installing other applications is to put the entire machine under full load and ensure its stable. If the benchmark software crashes while running its tests, this is investigated until the root cause is found (typically a timing paramete in BIOS must be tweaked). It prevents headaches later. I was very pleasantly surprised with the results of 3D-Mark05 after installing the 2nd graphics card and setting up CrossFireX. It's at the very top of reported benchmarks for similar machines in its class, and holds its own against the guys who build machines for nothing but sheer speed by overclocking everything and watercooling the processor, RAM and graphics cards. These guys don't care if the machine lasts only a day, or if its of any practical use for anything. They're after the highest benchmark score possible with a single run of the benchmark software (why the "Fastest System" score on the right is ridiculously high). Before installing other applications, I customized the Windows XP user interface on this machine. It looks a like Vista, including a Sidebar with Widgets, but it's not Vista, it's WinXP /x64. Who needs Vista? I don't, and I don't want it either. WinXP /x64 is smaller, more efficient and very stable. It doesn't have the continuing problematic stability issues Vista still has.

Saturday, February 23, 2008

iPod Mini Lazarus Act

Resurrecting an iPod Mini From the Dead

A friend at work asked me if I could resurrect an iPod Mini that booted to the dreaded "Sad iPod Icon." Told her I'd give it a shot, thinking one of the "5 R's" would bring it back to life. My wife has always managed to cure her own iPods, so I knew there was a source of advice available on what things can be tried to get it working again. Although she has a whole family of various iPod models, none of them is a Mini. Little did I know at the outset that it would be a little more work than fiddling with some button combinations on the click wheel to reset it and restore the operating software, if necessary.

I got my first clue of this when the iPod Mini was delivered to me by my friend. She told me they had already been through Apple's "5 R's" several times without avail. Not a good sign. Nevertheless, I took it home and went through them a couple times myself, including trying to put the iPod into mass storage drive mode. Same results as she had. Nothing more could be prodded out of it beyond the "Sad iPod" screen. OK, off to the computer and the search engine of search engines: Google. What I didn't know before getting pages of hits about the iPod Mini and the "Sad iPod" icon was soon revealed.

The Bad News:
With an iPod Mini, it's almost always catastrophic drive failure. This isn't just the file structure hosed up on the drive that reformatting it can fix. That would produce a file folder icon with an exclamation point.

The meaning of a "Sad iPod" icon on a Mini:
[with apologies to Monty Python]
The drive isn't just restin' from playing the LP version of Iron Butterfly's "In-A-Gadda-Da-Vida," or pinin' for The Beatles' "Norwegian Woods." Even 4 million volts wouldn't make it "voom" again. It's passed on; the drive is no more; ceased to be; expired and gone to meet its maker; a stiff. Bereft of life, it rests in peace. If it were't captive inside the iPod, it'd be pushing up the daisies. Its electro-mechanical processes are now history; off the twig. It's kicked the bucket, shuffled off its mortal coil, run down the curtain and joined the choir invisible. It's an EX-DRIVE!

The Good News:
Powering up with an Apple icon and displaying the "Sad iPod" icon almost always means the logic board (main circuit board) and display are OK. Very, very likely the click wheel and daughter board with earbud jack are OK too. More Good News: the iPod Mini can be disassembled and the original Seagate ST-1 MicroDrive replaced with a CompactFlash card.

Better News:
While it's cracked open for major thoracic surgery, might as well replace the battery too. Age and repeated recharging has likely reduced its capacity (translates into playing time). The original battery Apple used was rated at 400 mAh capacity. Target offers 400 mAh replacements for $40. Surprisingly high priced as on-line (easily found) sellers of iPod parts offer 650 mAh significantly higher capacity (~1.5X) drop-in replacements for 2/5ths of Target's price, including shipping cost. Better yet, CompactFlash also uses less power than a MicroDrive. An iPod Mini with both CompactFlash and new 650 mAh battery won't run just 1.5X longer between recharges, it will run even longer, likely 2X or more.

The Best News:
At the time the iPod Mini was originally designed and made, MicroDrives were expensive. That's why Apple only used 2GB and 4GB drives. CompactFlash was emerging as a drop-in direct replacement for MicroDrives using flash memory, but it was even more expensive. Hitachi and Seagate still make MicroDrives, offering models with as much as 8GB capacity, but they remain expensive. The price of CompactFlash has since plummeted while capacity and read/write speed has increased dramatically. A 133X CompactFlash with 8GB, twice the capacity of an iPod Mini 4GB MicroDrive, now costs 1/4th the price of a 4GB replacement Seagate ST-1 MicroDrive, and while the capacity of a MicroDrive now tops out at 8GB, the capacity of CompactFlash has risen to 32GB. Designed as an end-user removable memory, it's most often found in this kind of application with some PDAs and most professional grade digital SLR cameras. Devices like the iPod Mini have them embedded inside, and they weren't intended to be end-user removable. Consequently, not every make/model CompactFlash works well inside an iPod Mini. Scouring a number of reviews showed the Transcend 133X CF cards worked well with decent write speed (when uploading music into the iPod; read speed during playback is at a comparatively very slow rate).

Now to get the iPod apart and extract its MicroDrive. I just happen to have memory card readers on two desktop machines that have a slot for MicroDrive/CompactFlash memory cards.
[with apologies to Frank Baum]
Attempting to read it directly will verify it legally, to see if the MicroDrive, is morally, ethic'lly, spiritually, physically, positively, absolutely, undeniably and reliably dead.

Examining the outside, the mechanical design concept for its case is obvious. It is an extruded aluminum shell with plastic top and bottom caps, with two holes in the front for the display window and click wheel switch pad. What isn't obvious is how the plastic top and bottom caps are attached. There are no fasteners (screws, nuts, rivets, etc.) visible and there are no labels or bezels that could be concealing fasteners.  That means the top and bottom plastic caps must either snap on or be held in place with an adhesive or glue. Turns out, after a little more Google work, that they're held in place on metal plates just under them with adhesive. However, they're thin and can be easily broken if removal is attempted too aggressively. Past experience has taught me that getting an adhesive to "release" requires just enough tension applied and patience to allow it to slowly give way. The bottom plastic cover is more fragile than the top too, with its wide rectangular hole for the docking connector and thin strips of plastic above and below the hole.

A small and extremely thin knife blade edge can be slipped between the aluminum shell and the plastic top and bottom caps. The trick is not going too deep and damaging any components immediately under the cap, or break the guide pegs on the underside of the caps around their edges (aids installation of them during assembly and adds a little strength when completely assembled). Some gentle sideways prying, just under the cap edge (not on anything under it) and alternately working both the front and back long edges eventually releases the adhesive. Some folks suggest using a hair dryer to heat the caps (and the adhesive) to help it release. Yes, it could help as hot adhesive usually doesn't hold as well, but controlling the amount of heat and where it's directed is difficult. Even if the hair dryer is on its "low" setting it could warp the plastic, and too much heat on the click wheel could damage it too. I avoid using any form of hot air "heat gun" if at all possible. Room temperature and patience was all that was required to get this iPod's caps off.

Sure enough, under the caps are metal plates. The bottom one clips into four grooves at each end of the iPod's front and back. The curved arms that clip into the grooves, and the holes in the ends of them are there to aid in installing and removing it. This photo was done during reassembly. The adhesive that holds the bottom cap to the plate can be seen clearly. I added a little more adhesive of the same type to help hold the cap on securely as the plate and cap had been handled some while they were off.

An extremely small pair of needle nose pliers, taking great care not to insert the nose tips too far into the holes allowed squeezing the clip arms and removing each side. What's found under this metal plate shows why care must be taken to not insert the blade used to pry the plastic cap off or the tool is used to unclip the metal plate under it too far into the ipod. The very thin mylar cable with copper flex traces embedded in it that connects the click wheel to the logic board is just under one end of this clip. The "button" connector on the end of the cable has a mating connector on the edge of the logic board. These types of cables are easily damaged! Go too far with a tool under the cap or plate and the cable or its connector will get damaged. A damaged the click wheel cable or connector cannot be repaired; it requires removing and replacing the entire click wheel assembly with a new one. The cable in this photo has already been disconnected from its mate on the logic board by very gently prying alternately on each side of the connector until it came loose. It's a snap fit, so it should be obvious when it finally pops out. Do not pull on the cable or exert too much force on the tab on the end if it!

Under the top cap is another metal plate. Turns out this one is attached to a daughter board with the audio ear bud jack and "Hold" switch on it. The daughter board is, in turn, attached to the main logic board. There are two #000 size Phillips screws that attach this plate to the top end of the iPod and hold the entire logic board assembly inside the case. As with the bottom plate, the adhesive used to hold the plastic top cap on can be seen in the photo. Removal of the logic board assembly from inside the aluminum case requires removing these two screws using a jewelers driver. Ensure it's a #000 Phillips or cross-tip, one of the smallest made, as anything bigger will not fit into the heads properly and strip out the slots. The screws have non-permanent thread locking adhesive on them and they're driven in firmly when the iPod is assembled in Apple's factory. This requires firm pressure on the screwdriver pushing it into the screw head while carefully turning it. The driver can start to back out of the screw head (in spite of some pressure on it) which necessitates stopping as soon as it does so, reseating it, and trying to turn it again. The thread locking adhesive will eventually "give" and the screws will come out, but it's very important not to damage the screw heads while removing them! They are all that hold the entire electronics assembly inside the case. It also helps to work over a towel or old T-Shirt, well away from the table or bench edges, to keep from losing them when they come out. I suppose replacement screws can be had, but have never seen anything suitable as tiny as these are. They can't be replaced by a simple trip to the local hardware store! I also use a very small Tupperware container with seal (aka lid) to keep parts like this in, so they don't get lost.

Felis Catus Interruptus:
Do you ever get that feeling someone might be watching? Ever try to do fine, detailed work requiring a steady hand with someone supervising it by watching closely over your shoulder? Just as I was about to remove the two #000 Phillips screws from the top plate, I sensed being closely watched. Suddenly I was being closely supervised, by our cat, Frisky, who jumped up on my shoulder and perched there, watching everything going on. She does when she wants attention, sometimes several times in a singe evening. What she really wants is a free ride around the house perched on my shoulders, trilling in my ear and rubbing her head against mine the entire time. At least she did it before I had the Phillips driver firmly engaged in the screw head. Picture was done with my cell phone held out at arm's length. After I deposited my "supervisor" onto a high perch (she climbs from my shoulder onto it), I got back to work on the iPod.

Once the screws are out, and after double-checking to ensure the click wheel cable is disconnected, the entire logic board assembly can be carefully slid out through the top by firmly pushing on the bottom docking connector. Do not pull on the top plate as this can break the daughter board off of the main logic board, and slide it out slowly. My fingers aren't that huge, but they still didn't fit far enough into the bottom of the case to push the board all the way out. My little finger was almost narrow enough. Used the eraser end of a common wooden pencil to push it the rest of the way out. The cap end of a narrow plastic pen can also be used. Push on the docking connector, not on the board or click wheel connector. The photo shows all the parts: top and bottom plastic caps, the metal plate from the bottom, and the Phillips screws from the top. All the parts except the logic board went into the Tupperware bin, sealed up and set aside until they were needed for reassembling it. Flipping the logic board over shows the drive and battery. This photo was done with the new CompactFlash drive installed, but the old battery still in place. The battery is opposite the display at the top and the drive opposite the click wheel at the bottom. The dull silver rectangle in a metal frame is actually a piece of stiff foam inside a fine wire mesh tube. It grounds the body of the drive to the logic board's ground plane and aluminum iPod case. The drive is given shock protection by wrapping its edges in a neoprene U-channel held in place with tape, along with the header connector for the drive that has a very short ribbon cable running to a tiny board connector (hidden by the battery). The small piece of copper conductive tape also wraps around the blue neoprene channel to provide a conductive path from the metal plate on the drive to the grounded wire mesh. The battery is attached with a very small square foam adhesive pad to the top of a small integrated circuit chip on the board.

The battery must be at least pulled up and away from the board as most of the drive interconnect cable and its connection to the logic board are under the battery. The battery is held in place, primarily to keep it from shifting around inside the iPod with a very small square of thin adhesive foam pad. After the drive has been disconnected from the logic board, the tape, shock absorbing frame and interconnecting ribbon cable must be removed. The short interconnect cable was made by Molex, a major name brand of industrial, commercial and military electrical and electronics connectors. The double row of header pins mate with a matching double row of header holes on the top edge of the drive. The double row of pins can be seen inside the MicroDrive/CompactFlash slot of a memory card reader. Once I had everything removed from the MicroDrive, it could be tested in the MD/CF slot of the memory card reader in my desktop computer.

Slid the drive into the MD/CF slot in the memory card reader, and the card reader's LED lit up (meaning a memory card has been inserted into one of its four slots). Not loud enough to be heard while inside the iPod, a soft grinding sound lasting about two or three seconds was emitted by the MicroDrive after it was inserted into the memory card reader. This kept repeating continuously until I pulled the drive out of the card reader. The computer's O/S also did not acknowledge the presence of a removable drive. Even if it were completely unformatted, the O/S (Windows XP) should at least come up with a dialog box offering to format the drive. I have more than one desktop machine, and all of them have memory card readers. So, it was inserted into the MD/CF slot in a second destop computer which resulted in exactly the same behavior as when it was in the first one. The picture shows what the inside of a Seagate ST1 looks like. Most likely the repeating grinding sound was the read head arm trying to deploy or move to read the drive platter.

[more apologies to Frank Baum]
I must aver, I thoroughly examined her. And she's not only merely dead, she's really most sincerely dead.

With the new drive and battery installed, reassembly is simply the reverse of disassembly. I added some thin adhesive strips to the top and bottom plates to rejuvinate what was there as handling them, along with the plastic end caps had likely degraded the original adhesive some. It's a very similar, non-permanent adhesive, which means the caps can still be removed at some point in the future if necessary.

After the iPod was completely reassembled, it was time for the infamous "Smoke Test." Contrary to popular belief, promoted by electrical engineers world-wide, electronics do not run on electricity or electrons. It's one of the greatest hoaxes ever foisted upon mankind. They run on pressurized smoke. Ever see an electron? I daresay everyone has seen smoke though. The proof of this is seeing the smoke leak out when electronic devices fail. Disengaged the "Hold" switch on top and depressed the center click wheel button. The iPod sprang to life with the Apple logo followed by a file folder with exclamation point icon. It's what I had expected as the CompactFlash drive had been formatted, but didn't have the iPod software installed yet. Much, much, much better than the Sad iPod icon. Best of all, no smoke leaked out. Connected it to the USB port on my computer, reinstalled the iPod software, including updating to the most current version and was rewarded with a full iPod Mini menu. Loaded some music, started the iPod playing, and exercised all the click wheel functions. Works just like a new iPod Mini! It's now headed back to its owner, who I hope will be tickled pink that her iPod has been resurrected from the dead, with twice its original drive capacity, and just a hair over 1.5x its original battery capacity. The iPod seems to be with its new lease on life.

Suggestion for this iPod's new name: Lazarus!

Saturday, March 31, 2007

WiFi Security

WEP: Unsafe With Any Key

Recently did a "Site Survey" which looked for all the WiFi access points within range of my own WiFi access points. I do this periodically to find out what channels they're using so I can set my system for the least interference from other WiFi systems by using the least populated channel. To my great dismay I found a number of WiFi networks still using WEP authentication and encryption, instead of its replacements, WPA or WPA2. New WiFi access points are designed to prevent this, or worse yet, an access point with no authentication or encryption. WEP, which stands for Wireless Equivalent Privacy, was the original client (user) authentication and encryption system created for WiFi Access Points (wireless routers, aka APs) when IEEE 802.11, the international standard for WiFi and its operating protocols was originally written and released. The cryptography community shuddered with horror when they learned what had been approved by the IEEE 802.11 Working Group and Committee. It was known at the outset to be weak, and "crackable" without too much difficulty. Originally it took a little skill and patience, capturing encrypted data being transmitted between a client and the AP, but now that isn't even required. A 10 year-old or "Grandma" can crack a WEP "secured" AP in 5-10 minutes, sometimes in only 3 minutes. Within a few months, practical methods were devised to "crack" the WEP encryption for any AP using it without taking too much time. The weaknesses of WEP weren't just theoretical any more, they were very real.

The TJ Maxx debacle, with the compromise of millions of credit card numbers, was traced to WiFi APs used inside TJ Maxx stores that carried Point-of-Sale data to (ultimately) parent corporation servers. The original intrusion into their network was made from the parking lot of a TJ Maxx store through the store's WiFi system that was "secured" (a dubious characterization) using WEP. From there, they mapped the corporation's network, eventually working their way to the credit card database, and then installed a "Back Door" to access it using normal wired network means so they could quit using the WiFi for access.

WEP has been "deprecated" (entirely obsoleted and removed) in the current IEEE 802.11, replaced now by WPA and WPA2, which are infinitely stronger authentication and encryption methods. WPA/WPA2 strength in Pre-Shared Key systems rests entirely on the strength of the pass-phrase; one that isn't susceptible to a "Dictionary Attack" that guesses the pass-phrase using a large dictionary of commonly used pass-phrases. You'd be surprised at how many people use simple, common passwords. The encryption algorithm isn't breakable by statistical analysis methods; in other words, there's no "short cut" as there is with WEP. Sadly, in my opinion rising to just short of criminal culpability, WEP had been deprecated in 802.11 for a couple of years when the TJ Maxx incident occurred, but they made a decision to accept the risk because it would "cost too much" to replace their store WiFi APs with WPA or WPA2 capability. I only wonder in hindsight if TJ Maxx's parent corporation still thinks their "too costly" decision was the correct one.


The key - pun intended - to WiFi security:
  • Using WPA or WPA2 authentication and encryption.
  • Using a long, esoteric and obscure passphrase that cannot be guessed using a dictionary attack.
I'll give you a few clues:
  • password
  • drowssap
  • 123456
  • 654321
  • abcdef
  • uvwxyz
  • qwerty
  • zxcvbnm
  • ~!@#$%
  • qwerty
  • BR-549
  • NCC-1701, and
  • THX-1138
  • Shazam!
  • Open Sesame (or OpenSesame)
  • Xyzzy
  • Zzyzx
. . . or any of their backwards variants, are exceptionally weak passwords. They're quite early on in the common dictionaries used in a dictionary attack, which list well over 100,000 commonly used passwords in not only English, but German, French and Spanish. If you've thought of one of these, or anything similar taken from a movie, TV show, or computer game, millions of others have also. Your next door neighbor's 12 year old might not guess beyond the simplistic few forward and backward sequences, or keyboard letters, but all of these are within the first part of a commonly and readily available password dictionary. Using your name, or even with numbers substituted for letters, such as J0hn 5m1th is also worthless. We're all savvy to the number for letter substitutions and an automated system can perform them in a few nanoseconds. Someone who knows your name will try that before going to a common dictionary.

In a previous life a few decades ago, I was doing contract work on the side helping to maintain a digital, radio-telecommunications system for passing emergency written messages. It was mostly related to logistics, to get that text intensive traffic off the tactical voice channels. Needed to update staff computers with a software revision over a weekend. The agency communications director and I went in on Saturday morning and found password protected PCs (hooray for security - almost, but not quite). There wasn't a single one of over a dozen I couldn't get into withing five minutes using common passwords, looking at the pictures and refrigerator art from the kids at their desks using stuff the kids wrote on them, or lifting up writing pads and other stuff on the desktop, or looking under the middle desk drawer to find stuffed taped there. Esoteric and obscure means truly esoteric and obscure. It's gobbledy-gook gibberish nobody could guess without doing a brute force attack using all possible permutations and combinations of numbers and letters.

Welcome to JLind.net

Dedicated to the Practice of Polymathy Polymathy: The possession of learning in many fields. Your host is John Lind. Here you wi...