I didn't understand everything you were saying about the hotswap bay, so I'll just reply to this:


If you haven't done this already, I suggest that you include an "incoming inspection" step for all drives, to check whether all TVSs are already shorted or not, before you apply any power to them. This will confirm that the overvoltage (to trigger the TVS) is occurring on these test stations.

In the case of the above drive, then IMHO the 2 most likely options (although I can think of a third, mentioned previously, but IMHO very unlikely) are either:

a) That TVS was already shorted when you received the drive, if it wasn't checked first; or
b) The usual cause applies i.e. that drive received an input voltage of >Vbr (the TVS breakdown voltage, typically something >6V for a 5V TVS, depending on the specific TVS component part number) on the +5V rail, and therefore the TVS conducted, as it is designed to do. Depending on the specific PCB, there may, or may not, be a deliberately fusible element in the power (or ground) rail(s), to limit the duration of the resulting high current.

Point (b) could be investigated by the testing with a DSO that I described before.

sorry if I wasn't clear. The hotswap bay was/is a Thermaltake BlacX eSATA unit. I tested several drives in it and for a while it worked well. Once I switched to the problem drives it started shutting off the unit. As soon as the drive was removed, the power came back. I took this to mean that it was working as intended. However, when I tested the drives that shut the unit down in the way we've been doing it 3 out of 4 ended up being fine. The 4th shut down the computer.

I'm not sure that I can get my hands on a DCO. I'm also not sure that if I were able to I'd know what I was doing. I'm more of a computer tech person than an electrical engineer. I've just been tasked with finding out whats going on here and determining if we are causing the issue or if the drives we're getting are just trash. I've removed serveral boards and found that the 12v TVS is cracked before testing. It's pretty easy to spot drives where this is the case as they smell very strong as soon as coming out of the ESD bags.

Just out of curiosity. Is there any way that when these drives were last used they were in some sort of lower power standby mode, then never woken from that state before being removed from whatever system they came from?

Thanks for the extra details, although there's clearly something odd about the events with that docking unit, which is probably why I couldn't grasp the details before.


OK, and as I understand it, that docking station comes with its own small external PSU, and I presume that is the PSU which you were using.


That is the classic, expected behaviour when a drive with a shorted TVS (due to a previous prolonged overvoltage event) is attached to something with a small PSU. That small PSU will shutdown due to what it sees as an excessive load, due to that shorted (or, to be more accurate, low resistance) TVS diode.


Without detailed resistance measurements of the TVS diodes from those drives, and preferably seeing the drives & test stations in front of me, then I could only guess at the possible reason for that. IMHO that's an odd result.


Understood. Unfortunately TVS diode behaviour is an electrical (perhaps more specifically, electronic) engineering area. That's where my primary qualifications are, but if that isn't your background then you may need to get someone else involved to assist with (and provide the equipment for) that measuring & troubleshooting - unless my comments below apply, because something seems to be coming clearer, due to your recent update. If the below comments do apply, then investigating the PSUs may not need to be the priority.


A-ha! This is vital new info. I realised that we'd been focussing on the testing stations and burning TVSs there, but if the drives come to you with no "history", then you wouldn't know whether the TVS were already damaged (shorted / low resistance) unless you pre-screen them - hence why I mentioned this in my posting yesterday.

So as I understand it, there was no pre-screening for shorted TVSs being done, you were using high-wattage PSUs, and seeing burning PCBs during your testing. Therefore as I mentioned yesterday, one possibility is that the TVS diodes had already been damaged before you received the drives (and your mention of the cracked 12V TVS above confirms this, at least for that drive), and due to the high-wattage PSUs, they are much more likely to continue to supply power to a drive with a "shorted" TVS, which will always be a bad result. Such a "shorted" TVS on the end of cables & connectors (which will have measurable resistance) is just like a heavy load, which is more likely to be within the capabilities of the power output from a high-wattage PSU. That is why your problems were reduced with lower-wattage PSUs, as you mentioned before.


Understood, but you can't rely on a damaged TVS being that obvious. If the drive was previously (before the drive as sent to you) attached to a PSU which shutdown quickly, then the visible damage to the TVS can be minimal or none. I've seen "shorted" TVSs which look perfectly normal - they need to be measured (e.g. with a multimeter) to see that they have been damaged already by a previous overvoltage event.

Based on your comments, it seems possible that the original problem is that you're receiving drives which already have shorted TVS diodes; this condition isn't being screened first using a multimeter; then when power is applied from a PC PSU (even 250W model), the TVSs burn (as expected) due to those particular drive PCBs not having a fusible element to disconnect the power. That's more likely to happen with even higher-wattage PSUs, and the earlier suggestion from SAjunky to use small (but good quality!) drive PSUs would be much less likely to be able to supply enough power for extensive PCB damage, compared with using PC PSUs.

However, IMHO your starting point must be to pre-screen (by measurement, not just visually) for any drives which would present an excessive load (this can be due to a shorted TVS diode or, rarer, another component which has failed and is now acting as a low resistance) on the main power rails.

Hope that helps.

That definitely helps. Unfortunately, there is no way my boss will go for removing every board from every drive one at a time to test the TVS'. So it would seem we'll have to continue dealing with the ones that are already broken burning up.

As for the ones that are presumably healthy and still burning up, would a custom power supply(say 35-40W) with voltage regulation be a safer way to go? PC Power & Cooling apparently makes custom PSU's with many different features that I don't believe our current ones have.

Also, would such a power supply continue to burn bad TVS diodes? Obviously, while dangerous to the techs who're testing these drives, it's necessary to weed them out somehow.

One of my techs recently noticed that the majority of the drives that are burning up, at least in the last couple days, have been labeled as refurbished. I'm wondering if whoever is "refurbishing" these drives is causing the damage, or if we're getting them because they were already beyond repair(in which case we'd really have to have a talk with that supplier).

I don't want to interrupt conversation, I have nothing to add what Vulcan wrote. It can be confirmed by any experienced electronic engineer.
I don't understand docking station issue, but I have a crazy idea. It applies when working with PC power supply. In normal PC-ATX operation power supply is not started by mechanical switch. There is a mechanical button, but it drives logic circuit located on the motherboard which eleiminates bouncing of the switch and power supply receives steady on/off signal. Such circuit is mandatory according to ATX specification.
If you use mechanical switch to power on/off power supply, it receives series of on/off bounces in the transition period. This can confuse electronics and generate overshots, as it was never tested for such conditions.
I think if you are going to use PC power supply I suggest to connect it to (even defunct) motherboard, it will create balanced load condition and generate clean power on signal.

@MrSpiffdifilous,


Testing the TVS does not require the drive's PCB to be removed from the drive. It involves just a 1 minute check with a multimeter, with the drive unpowered (using appropriate ESD precautions, as always).

Another member here, fzabkar, included details of that test (and more info about TVS) in this web page:

http://www.users.on.net/~fzabkar/HDD/TVS_diode_FAQ.html

I humbly suggest that not performing such a test could be considered negligent, as omitting it could lead (and, as you mentioned before, actually has led!) to personal injury, when applying power to such a drive (especially when the drive is not inside a case). Furthermore, if the PSU doesn't shutdown then applying power to such a drive can cause extensive damage to the PCB, as you have seen. IMHO it seems false economy to allow such potential damage (which would require lengthy repair), rather than perform a quick & simple screening test first, doesn't it?

If your boss won't perform such a screening test, then I'm not going to spend further time trying to find ways to "work around" that decision - among other reasons, I don't want to be associated with the potential consequences, sorry. I hope your boss reconsiders, especially if their decision was based on an erroneous assumption that the drive's PCB needed to be removed, to perform the test.

Good luck!

Thank you very much for the link. I will definitely check it out and pass it along to my boss. Knowing that the boards do not have to be removed to perform the check definitely changes things. My boss is much more likely to implement this procedure. I wholly agree that not doing these checks would be a terrible financial and would be negligent to his employees safety. we just have to figure out who will do these checks and when, since most of my power testing techs have little or no technical background. We'll figure something out though.