Panic.

June 12, 1997

Hard drive gurus and A1200 hardware hackers, I got a question for ya.

I have an A1200 with a Seagate ST5850A IDE 3.5"x0.75" hard drive attached via a Redmond Cable 44-to-40 cable. Until last night this was connected the IDE passthrough on a Dataflyer SCSI+ - which you may remember the fun times I had installing - and the drive shared power with the floppy drive. Last night I disconnected the Dataflyer and found a separate power supply for the hard drive - but the problems I'm having persist.

Here's the scenario: I'll be working at the machine, suddenly go to access the hard drive, and it just says "Workbench has a read error on (some huge random block number). [Retry] [Cancel]" or whichever partition I was trying to read. Then I notice the hard drive light is on and staying on - and there is no sound of head movement.

Or I'll be working on something and suddenly hear the heart-stopping ker-THUNKuh-WEEEOOOOOOOOOOOOOoooooooooooooooh of a hard drive spinning down. The ker-thunk-uh sounds to me like the drive heads parking - which is good, I suppose, I'd rather hear ker-THUNK-uh-weeeeeooooooh than the THUD-screeeeecccchhh of heads slamming into the platter and recording the sound of my scream into a record groove on the disk surface.

Last night another wacky one: power the system up, and the drive never spins up, you just see the drive light go blink, blink-blink, blink, blink-blink.

All these are fixable by powering the machine down, letting it sit for awhile, and powering it back up.

Amazingly enough, I have never seen it have to revalidate the drive after powering back up - it's unaware anything ever happened, and DiskSalv cannot find anything wrong. There's an obvious answer for this - the computer hasn't yet had the hard drive "vanish" while it's trying to write. It's stopped in the middle of reading a file before, though.

I considered at first that it's heat-related - the 1200 has been caseless for some time, though, and it's never had THAT kind of heat-related problem before even with the case on, on the hottest of days, with stuff over the vents. I DO know the drive seems more reliable when it's cool to the touch. If it's heat-related, it would mean the heat coils are acting up - but I don't even know if the Decathlon-series drives have heat coils at all.

(I did have the joy once of standing beside a Web server with a Seagate 'Cuda drive when its heat coils triggered - the ker-THUNK-uh was more like a solid POP like a shotgun, followed by the drive spinning down. You'll notice hard drives take longer to spin down when they still have power than when you just simply power them down.)

Then after repeated attempts to cool an already well-ventilated machine, I considered physical problems: the hard drive is just sitting on the 1200's RF shield, the only insulation is directly over the clock header because that's the only place the drive has ever caused shorting. But with a piece of paper under the drive, the problem persists.

For whatever reason I decided to attach a small "muffin" fan - one of the 2" jobbies not unlike those found bolted to the tops of hefty CPUs - except this one came off a GVP hard drive chassis and has almost no torque at all. You can safely stick your fingers in it, that's how weak it is. Anyway, I tried figuring out where to mount it, and decided the best approach would be to take the floppy drive out (it hasn't worked since adding the Dataflyer) - doubly good because I can then suck power off the floppy port.

This was not very successful - Amigas do strange things when there is no floppy drive attached. Apparently Paula is waiting for an interrupt that will never come - and thus will just STOP for seconds on end at random places during the boot. So I abandoned the fan idea and reattached the drive - and was amazed to discover the floppy drive works again.

Well, that much is sorta good news...

Finally I considered something: I'm running an A1200, a 68030 and 68881 at 50MHz, 16MB RAM, a couple floppies, a Dataflyer SCSI+ card, and a 3.5-inch hard drive on an A500 power supply.

So I hunted for another power supply with enough juice and a spare power connector to run the hard drive - and found it:

A Tandy FD-501 floppy drive chassis.

Plenty of power there - people have run those units with more than the recommended number of floppy drives for years, and the way the Color Computer handles floppy drives, when one drive spins, they all spin - this particular chassis, I long ago ran an extra drive power lead outside the case, because I'd found a TRS-80 Model III floppy drive in someone's trash and wanted to use it. That dangling power lead came in MOST handy now - just set the drive chassis next to the 1200 and run the power lead over to the Seagate, which can sit exactly where it is under the 1200's keyboard.

But the problem persists.

So I'm down to a couple theories. One is, the heat coil is getting old, and is triggering at a far lower temperature than it should. The obvious question there is, first off, it doesn't sound quite like a heat coil - not to mention it doesn't explain the read errors as the system "forgets" the hard drive.

Another is, something's mechanically out of whack inside the drive, causing the unit to overheat, again triggering the heat coil.

Another is, the drive's circuitry is conking out. This drive may have an MTBF of 300,000 hours - that's 34 years - but that's the mechanism. Solid state electronics are a different story. Seagate builds some of the best mechanisms around - unlike Conner, whom they bought, who used to strap good boards onto pitiful drive mechanisms that lasted precisely two years from the date of installation. I don't like this possibility - it's fixable, but it would cost more than a new drive, data recovery places have found people will pay ANY price to recover a hard drive, and have adjusted their prices accordingly. In 1994 I saw someone pay $300 to recover a Bernoulli cartridge; in 1996 we priced recovering a Seagate 'Cuda 1.2 gig at a few thousand dollars. The only solution is to hook up a second drive and copy the contents to it, as much as I can before it loses itself again, power down and let it sit awhile, power back up and try again, until the drive is completely copied. I've STILL never made a backup of that drive - Amiga backup software on 3.5-inch disks are an absolute joke. DD takes too many disks and HD isn't as reliable as we'd like - and this is an 850 meg drive, which means about 700 disks uncompressed, considering some of that 850 megs is already compressed into LHAs and jpegs and GIFs, figure 400 disks if I'm lucky.

The last possibility is that the Amiga's motherboard has been damaged. The A1200's IDE port is what the industry calls "trivial" - it's two 25-cent 74xx logic chips, a 30-cent 44-pin header, and some ROM code. That's it. It's unbuffered. It's not particularly robust. The only good thing about it, aside from being really cheap, is that the ROM code supports drives up to 2 gig, unlike comparable PC BIOSes of 1992 that stopped at 500MB. On the A1200, your floppy drives are DMA but your hard drive isn't. Disk errors will halt the CPU and halt it so solid even keyboard resets don't affect it. A1200 IDE problems are starting to be the talk of the town - chaining two IDE devices off that IDE port is a sure way to cause motherboard damage. I've been running a decidedly nonstandard IDE setup for some time now - the Dataflyer - and while the Dataflyer is buffered, there's always the possibility of the Dataflyer itself overloading the Amiga motherboard. And since the IDE passthru on the Dataflyer is still unbuffered, there's always the possibility of damage THAT way. I've tried the system without the Dataflyer and the problem persists - so it's clearly not the Dataflyer hardware itself.

I'd actually prefer the last possibility - I have a friend who has an A1200 with no hard drive, I may be able to swap motherboards temporarily - if I promise not to mess up his too. A1200s are replaceable. Hideaway Dark and my e-mail archives are not.

So... any help anyone can offer would be appreciated.

And next time it'll be a REAL Rumor Mill. Really. With content.