Question: EMI Shielding for IDE cables

(Long story) I am having problems with hard drive data corruption. I have recently rebuilt some of my computers in rackmount casings. After running them for a few weeks, I noticed that all of my files were becoming corrupted. I have 4 x 160GB + 4 x 120GB hard drives where this is becoming a problem. 2 of the drives lost their partition tables and became completely useless. I made valiant efforts to recover and preserve my data, but I have basically given up.

After much investigation, I suspect the cause for these problems to be related to my IDE cables.

1. With the chassis layout, the power supply is located at the front of the case, leaving an AC power cable running directly under the hard drive bays and the ide cables.
2. The rack itself houses 6 computers, 2 UPS, 2 power bars, and about 10-15 other devices (modems, switches, etc), so I expect that there is quite a bit of interference here.
3. I am using flat ribbon 24" and 36" IDE cables, because shorter cables will not reach from the drive bays to the ide controller connectors.
4. All drives are running ATA100

When doing data recovery, I removed the drives from the cages and ran short,

It is unlikely that shielding is the culprit, and more likely that grounding or other power supply issues is the problem.

By reducing the cable length, you have actually reduced the need for good grounding. The suggestion by tempus fugit that 80-pin cables is right-on, since those cables are required for proper ATA100 operation. They provide 40 additional ground wires in each cable (which is not quite the same as shielding). Standard ATA cables have only 5 or 6 ground conductors.

You should also check the quality of the cabling in the power connections. If you changed cases, I assume you also changed power supplies and likewise the distibution of power. Watch out for those "Y" cables in the power supply -- the standard ATA power connector is notorious for getting bent out of shape if they are plugged/unplugged a number of times.

Kevin

they are not "shielded", just the extra wires going between signaling ones are "grounded" to minimize crosstalk in signal wires ...

Thanks for the replies, I am using new aftermarket power supplies (thermaltake 480W). I've dealt with power supply problems with hard drives (long ago in the past), and the symptoms don't fit my current situation. Drives are always detected in bios, drives always spin up properly, drives are always accessible through os.

The reason why I think it is interference of some sort: I did some checking of a bunch of corrupt files in a hex editor. I actually had chunks of text from other text files. So data got scrambled up during read/writes to the drives. Switching to a shorter cable to do data recovery also meant unscrewing the drive from the bay, and just sitting it in the chassis next to the ide controller plug. This moved the drive away from the power cabling. Also, when doing recovery, I removed the rack chassis from the rack. With this done, I did some file copying tests to see if I was still getting corruption, and I did not. The same tests, with long ide cables, and the drive mounted in the bay, gave corruption errors. Test file was a large rar archive set, test was a sfv file check. I am using flat ribbon IDE cables, 80 conductor / 40 pin. I understand that the second connector is a ground, which should eliminate cable crosstalk. But I do believe that my problem is actual RF or EMI interference. I need to use longer cables because of the chassis layout - 24" - 36". It would seem to be quite easy to shield the rounded cables from EMI with some aluminum tape, I just don't really know what I need to consider. An EMI shield doesn't really make sense to me without a ground. And, is aluminum muffler tape the right stuff to use?

I am definitey paranoid of data corruption now - the amount of data I lost was crazy. I could have never imagined losing a full terabyte of data all at once..

Thanks, Keith creekchubbAThotmailDOTcom (replace AT with @, and DOT with .)

becoming

becoming a

completely

basically

the

bays

bit

short,

looking

aluminum

too

chassis

power

heats

It would certainly work, as long as you're sure that you make a good connection to it. You can't solder aluminum, you know - I'd seriously look into some kind of contact that's designed to connect aluminum to copper.

And only ground it at one end.

It would also probably help to see to it that it's not a spiral-wound inductor - be sure that there's contact from one strip to the next.

Or, get really wide aluminum tape and attach it lengthwise.

Good Luck! Rich

It would be better to ground it, of course, but you would still get some shielding from just wrapping it in the tape.

the

bios,

actually

rar

that

some

at

After

cables

not

layout -

to

ground.

Bad shielding could cause data transfer errors but could it really cause the loss of the partition table? Assuming you are using FAT32 or something similar I could see the FAT getting corrupted but the partition table? That would require writes to sectors that shouldn't normally be written to.

are you sure something isn't erasing the drives themselves? like very strong interference hitting the drives from the power supply?

Also, couldn't you put those round metal rings around the power cables to stop some of the interference? not sure how much those help???

:{

"dude" wrote in message news: snipped-for-privacy@corp.supernews.com...

look

the

That

strong

Thanks all for the replies. Thanks for the thought on soldering to aluminum tape. I might take a look for tinned copper shielding tape. I used some a long time ago when putting a booster inside a guitar. Hope it isn't too expensive! The power supply is right next to my boot drive, which never had any problems with corruption (short cable?). My data drives are relatively far away from the power supply. I think I will definitely look for some of the ferrite RF shields for the AC cable running through the case. Any idea where on the cable to put these?

--- More of my story (for those interested in the drama!) The drives were big ntfs partitions, some were in raid 0 arrays (2x array of [2x120gb]), some as single drives (4x [160gb]) . I think the data was degrading due to cable loss over time, but i didn't realize it. Eventually, the index tables got messed up, and I started losing directories. This caused immediate panic, as I thought one of the drives was going in the array, so I started moving data to my 160gb drives. Not much thought to it - I've had tons of hard drive failures (dropped heads, bad spindle motors, all the common stuff) and my first impulse is to get all the data off. Windows chkdsk started coming up on bootup, and made an attempt at 'repairing' the faulty tables. (This is now the most destructive thing i've ever seen. I now believe chkdsk to be the ultimate 'virtual shredder' program.) What I didn't realize was that chkdsk was also destroying my second array. After chkdsk was done, the majority of the files on both arrays had disappeared, or were in found.### directories, and completely corrupted. That's ok, because I had just backed up to the 160gb drives, right? Not exactly. Every file that I had copied from the array was corrupted. And after checking things out, other files on the 160gb drives were also corrupted. I don't know if the original files were all corrupted (unlikely), or if the faulty index tables corrupted the copies (maybe), or if the data was corrupted during cable transfer (i believe), or all of the above (probable). I switched the 160gb drives to short cables, out of the bays, by accident really. I took the drive out, then I reconsidered and wanted to make a last ditch attempt at the data. Being lazy, I didn't feel like unscrewing the drive cage again, so i just grabbed a short ide cable and set the drive down in the case. Suddenly, some of the files that were coming up as corrupt earlier, were fine! (blank clueless stare - i must have sat there for an hour just staring, no mental activity, would you like fries with that?) Got about 30% of my data back, but none of the stuff from the drive arrays was recovered.

So, being paranoid now, I check out another computer - same case, cables, and a pair of 160gb drives. Sure enough, I've got data corruption here too. I pulled the drives, attached them with a short cable, and was able to get most of this data off.

And now, ultra paranoia. I have a terabyte of blank data drives, and I'm afraid to put anything on them. All drives were checked with manufacturer utility, and are fine (western digital, by the way, drives are 6-12 months old). I'm still testing files for data corruption - all files were copied to a pile of 40 & 60 gb drives - and burning off whatever I can to dvd. When I'm done with this, I'll be shielding these rounded cables, and figuring out how to test the drive setup.

Thanks for making it to the end! Keith creekchubbAThotmailDOTcom (replace AT with @, and DOT with .)

The fact that you are seeing chunks of text from *other* files indicates that the FAT is screwed up. If your problem was transmission errors which somehow made it past the error correction stuff (which I think is not too likely, but possible), then you would have erroneous data in the file you sent. Most likely it would just look like garbage, or random byte replacements. I can't see any way that it could be parts of other files, barring some otherworldly interaction between parallel cables on 2 drives during simultaneous writes. Not likely!

Of course, the FAT on the drive is also written by the same process that writes the rest of the file. So it's *possible* that some kind of noise pickup could corrupt the FAT data as it was updated. There are 2 copies of the FAT, so if they both are identical I'd say that was proof the trouble was somewhere else, in software. If it was noise in transmission, they would be different.

If your drives use something other than FAT, like NTFS, then the same reasoning applies. But I don't know if or where they maintain a 2nd copy.

Best regards,

Bob Masta dqatechATdaqartaDOTcom D A Q A R T A Data AcQuisition And Real-Time Analysis

actually

Actually seeing a chunk of recognizable data was a very rare occasion. Here's how i found the chunk of text. I renamed some large directories of archives, giving all the extension *.txt. I did a search on these files for a specific string (my email address). I went to sleep, and the next day, the search had found a file. I opened this in a hex editor, and sure enough, there was my email address with a small chunk of text from an ancient email in there. There is absolutely no reason why a solid data archive should have this string, but it did happen, in one file of many. This occured once, late in the game, after chkdsk had rewritten my index files, so i'm not sure if this had anything to do with it. I did make attempts to recover the ntfs index. I mirrored the entire drive with ghost, and ran a recovery program from a boot disk. I got nothing useful from this except more chkdsk madness on reboot after recovery. I think that a scrambled up file index would have the off chance probablilty of referencing incorrect data sectors for a file. I only did this search once, on files that had been verified as corrupt. It was out of curiosity (or madness) only, and didn't really have any purpose.

Keith creekchubbAThotmailDOTcom (replace AT with @, and DOT with .)

Duh. Copper tape. Of course! :-)

Just one per cable, near one of the connectors. That's called a common-mode choke; a google search should turn something up.

Everything is pointing at the long cables here. There's a chance that the length of the cables itself is the culprit, in which case shielding won't help, or could make it worse. If it's easer to try that than redesigning the mounting of the drives so you could use shorter cables, then try it and see how well it works - I'm sure you could come up with some data torture tests. ;-)

But if it were me, I'd make every possible effort to shorten the cables.

I know, I'm the one who suggested shielding, but that doesn't help with stuff like rise times and transmission line effects and stuff like that. At 133 MHz, 3 ft. is starting to approach a significant fraction of a wavelength, so impedance matching might be an issue.

I thought a RAID array was supposed to _increase_ data storage reliability? ;-)

Good Luck! Rich