Help! I've burned up 3 SATA drives today

root Inscribed thus:

That sort of confirms an open ground issue ! Though I don't think the PSU itself is the culprit.

On Thu, 16 Feb 2012 19:26:26 +0000 (UTC), root put finger to keyboard and composed:

Hopefully you really did mean "put" rather than "throw". :-)

- Franc Zabkar

On Thu, 16 Feb 2012 19:57:14 +0000 (UTC), root put finger to keyboard and composed:

It's not a fuse, it's a diode. A fuse goes open when it fails, whereas the TVS diode goes short circuit. A fuse is in series with the supply, whereas a TVS diode is in parallel with it, reverse biased, of course.

- Franc Zabkar

I haven't seen the D-Vine case mentioned, but various problems that have surfaced over the years have been related to numerous required ground points, primarily on large circuit boards (in various types of equipment) including motherboards.

Board designers sometimes don't connect critical/required ground points with circuit traces. All of the mounting screws on some motherboards were required for proper connection of the ground points.. and in some instances, case backplates may not have had threaded holes in the correct locations, so some of the screws for the metal mounting posts (standoffs) were casually omitted by DIY computer builders, and the boards would appear to be defective.

Another example is ground shields in other types of equipment (RCA), where the solder around the shield tabs would crack, and the equiment would fail to operate, due to the circuit board grounds needing to be tied together by shield mounting tabs.

In various types of equipment where several voltage potentials exist within certain sections of the circuitry, when a critical ground point is left open the likelyhood that a required ground/zero-volts-point potential can float (positive or negative).

On motherboards particularly, ground points may not be clearly marked as critical circuit connections, whereas required safety ground points typically will be marked in equipment with hazardous voltages.

-- Cheers, WB .............

Yes, the ground thingie(s). I would like to think this was caused more by something connected that shouldn't be rather than the other way around. However now we know that what fried was a protection diode. Could it have been across what was suppose dto be 3.3V but was 5V ? What would they have in there, a 3.9V zener ?

If I did the math right, feeding a 3.3V device with 5V would result in somewhere aroung 2.3 X the power dissipation is it's analog. Digital may be a bit different, I would think lower, by that I mean a slightly smaller increase in dissipation. This because for the most part heat in a digital device is from the saturation voltage of the semiconductors and the capacitive load on anything that switches.What this all means I think is thast a 3.3V device wil actually work on 5V for a while, how long depends on alot of things. If the components just barely meet spec, minutes or hours, but if you happen to have a really superior specimen, who knows how long it may work, years even.

If a similar diode exists for the 12V suply, it is probably 15V, which means one volt will not bother it. In fact driving a motor, it might just be fine with an extra volt, but the next standard zener value up from 3.3 is 3.9. One volt is too much.

So if somehow a negative 1V somehow got on the commons, this would explain it. But then how would that thappen ? Concievably if it were an open circuit and chassis ground made that connection, even though it's a bit far fetched as it assumes quite a bit of resistance in that ground. It could happen though, with the uPrecessor drawing alot of amps during bootup. Something like this may actually require a chain of partial failures, just to make that one volt.

I agree that this is an intersting subject because the actual cause seems to be elusive. If we don't know why it happened this time there is no way in hell of keeping it from happening again.

Jeff Urban Inscribed thus:

I can see where you are going, but consider this... A PC PSU has three major voltage outputs +12, +5, and +3.3. These are the Yellow, Red and Orange wires respectively. Black is common to all three. The Black wires, are normally common to the chassis. That is the negative pole of all three are bonded to the case. The case is also ground and would normally have an earth connected to it. At least in the UK.

Now if you place a volt meter between the case and any output you will read an appropriate voltage, 12, 5 or 3.3 volts.

Now do the same but connecting the voltmeter between the 12v and 3.3v outputs. Depending upon which way round the meter test leads are you will read + or - 8.7 volts. Do the same with any pair and you will see a voltage equal to the difference between the pair.

This is exactly the scenario if the ground was OC on the sata drive or if any other device had an OC ground.

The molex connectors have two ground pins. I have seen times that one of the pins is pushed out of the connector, but not both. In my case, neither of the ground pins were pushed out.

root Inscribed thus:

Yes I've seen that happen many times. I've always replaced bad pins, though I don't recall having to replace many female pins.

I've done the same thing with a Molex/SATA converter. To my shame, I put the Molex connector in upside down which lets the smoke out of a SATA drive quite effectively.

Cheapo Molex connectors are not foolproof in their keying. Might be useful if you see if you could have fitted yours upside down (with the power off :-) )

Chris K

I just tried the test using one of the molex-sata cables and one of the same power supplies I used. I couldn't get the two together upside down. Moreover, I almost always have trouble lining up make and female parts in a molex connector. The damned pins wiggle independently, so I have to look carefully.

It's good to see you posting in SER from time to time J.. always found your replies concise and explained wrt hands-on experience and knowledgeable background.

Yeah, when all the acceptable design tolerances of voltages and component values are taken into consideration (not even considering quality control issues) for almost any peculiar/odd scenario, it generally depends upon a lot of different issues.

Load testing used PSUs can involve many seemingly wasted hours, but used electronic gear is typically always iffy/suspect, at best, even if it seemed to be in perfect operating condition.. just yesterday.

And then there is always the "specifications subject to change" clause for nearly everything manufacured in recent decades.

-- Cheers, WB .............