We need a FET section

[mockingbird]

So a customer brings in a computer today from a few years ago... Asus P5K-VM... It's a G33 motherboard that takes anywhere from a 775 P4 to Core2Quad...

The customer called and said it wasn't working... Now he has it in a pretty tame environment... He uses it to check email. It has a 350-watt Fortron with all Teapo caps (None of which were bulged) and a very mediocre Pentium 4 6xx CPU. This system is never going to go anywhere near the maximum PSU output. I'd be surprised if it was drawing even 150 watts.

No problem I thought... Most likely there are some UCC KZGs in the VRM high as is usual with Asus. Replace them with those famous 8mm Nichicon HMs which I keep some stock of and away we go.

Imagine my surprise when I find out the motherboard has all solid caps, mostly from UCC, a few from Enesol or Apaq or whatever Asus uses.

Odd, I thought... When I cleaned everything in the garage, the PSU looked fine... Teapo usually does well when not stressed, and I have other customers with Teapo FSP PSUs which haven't complained... OK, let's change the PSU with a personally re-capped FSP. NADA!

After fidgeting with this system for a little while, I decided to try a brand new Seasonic S12-II series, which are famed for their reliability and all-Japanese capacitors. Maybe my refurbed PSU wasn't doing it (Eventhough I had tried 4 different PSUs I re-capped).

Same problem. Memory switched in from different system...

No beeps, nothing. Fans are spinning but no signal to screen.

At this point I try a re-capped Asus P5VD2-MX SE. SAME EXACT SYMPTOM!

Now I know that the problem with these boards are the shoddy transistors Asus uses. The P5K-VM used the infamous Taiwanese NIKOS brand FETs, and I couldn't tell what was on the P5VD2-MX SE.

We need to make some sort of Database of FETs like that which we have with caps... Manufacturers know now that they can no longer get away with using inferior capacitors... Even Capxon branded polymers are still probably reliable. Sot hey skimp a few cents for off-brand grab bag junk transistors...

I read a little about FETs in Britannica, And from what I know they have to introduce some sort of impurity to affect the flow of electrons a certain way. I know it's vague, but with capacitors we pretty much know that it's either the inferior aluminium or the inferior electrolyte or both combined which defines its quality (Or lack thereof). Maybe the quality of the impurity is what is important here... Maybe it's the quality of the lithography (Do they use lithography for FETs - I'm not sure)...

If someone could start a section, explain the phenomenon of transistors taking out whole boards and also creating a databse of known good and off-brand transistors, it would be greatly appreciated... There's about a million different off-brand FETs I see even on 10 or more year old boards. It seems that today, with the more complex engineering of modern boards, their weakness shows more readily.

STMicroelectronics, Intersil (Sorry, this is the one with the "i" logo - I know it's a reliable brand but I don't know the name), NEC, etc - boards with these FETs always last forever when coupled with good caps.

Is anyone interested?

[Topcat]

Re: We need a FET section

We need to make some sort of Database of FETs like that which we have with caps... Manufacturers know now that they can no longer get away with using inferior capacitors... Even Capxon branded polymers are still probably reliable. Sot hey skimp a few cents for off-brand grab bag junk transistors...

Interesting idea......it opens a whole new realm of fixes, issues, and potential members...... Most FET's are SMT on anything recent....changing them for the average DIY'er is nearly impossible....but for us seasoned geeks its not too bad, I've changed more than a few over the years.... Ok, now issue # 2, unlike caps, are some of these throw-away boards really worth all the money and effort to save? That's relative I'm sure, just like I'd tell someone is their bad capacitor board worth fixing....more than not they are....but changing caps is something I can do blindfolded....haha

I do think you might have something with a section....maybe just not for FET's, but for semiconductors as a whole. The rest of the site is dedicated to caps and computers, but over the course of the last couple years, the forum has taken a turn into more general electronics outside the computer world.... Lets hear some more opinions.

[tom66]

Re: We need a FET section

Semiconductors are pretty much all or nothing devices. Usually a failure in manufacturing means the FET is thrown away, not sold on. It would probably just not work or be significantly out of specification. Adding impurities is necessary to create the N and P type semiconductor - without it you get a very insulating silicon crystal.

What kills these devices is heat. These motherboards run hot, and the FETs do too because they are driven near their limit to reduce the cost. What's really bad is the failure mode is short circuit in most cases. I imagine most CPUs aren't fond of 12V direct from the PSU.

[mockingbird]

Re: We need a FET section

So you're saying that it's the design of the motherboard (not those necessarily who specify which parts to use, but those who actually get to choose which parts to use) that's causing these failures and that equal spec p/n channel FETs from any two semiconductors will always be the same quality?

[tom66]

Re: We need a FET section

So you're saying that it's the design of the motherboard (not those necessarily who specify which parts to use, but those who actually get to choose which parts to use) that's causing these failures and that equal spec p/n channel FETs from any two semiconductors will always be the same quality?

More or less, yes. They probably tried NIKOS semiconductors because they had a good price for a certain part which was "just" about capable. Although you do still get counterfeit transistors, so they could be fakes. NIKOS is a big company, they are unlikely to manufacture fakes or poor-spec parts, so any fakes are coming from black/grey-market sources (and that's Asus's fault). For what it is worth, I have seen failed Hitachi, NEC and Renasis FETs in plasma TV repair -- it's all about squeezing that last cent.

If you check the RoHS declaration for many FETs and other devices from big name manufacturers (for example, Fairchild), you'll see that various subcontractors are used to manufacture them and rarely is the big name actually making them at their own fabs. It isn't worth it for the big players to manufacture $0.007 devices. For $0.50 MOSFETs, there's not much margin to fight over either.

FETs are very reliable devices, lifespans under good operating conditions can be millions of hours. But a MTBF of 1 million hours means that if you have one million units, one will fail every hour. (Sort of... it's a bit more complex...)

[Uranium-235]

Re: We need a FET section

tc you should stock the most common fet replacements for boards and PSU's.

[Toasty]

Re: We need a FET section

And how do you define common? What's hot today is gone tomorrow. Plus the cross-referencing alone is insanity. Last year's devices do not have this year's because of the technology leaps made every 18 months. Mouser and Digikey get it wrong sometimes too, and spend years getting rid of the awww-crap purchases.

I vote NAY.

[shovenose]

Re: We need a FET section

I don't think we should have a section devoted to it. However, it would be cool if we had a central database that members above 100 posts could contribute to.

[mockingbird]

Re: We need a FET section

Well, I'm stumped!

I put the system back together as it was, and tested it again just to see if it would post. Lo and behold, it posts!!! BIOS wscreen was all sluggish, so booted and flashed it to latest 1001 beta bios, everything works great, even after a power cycle...

According to tom66, this can't be the transistors... Transistors either work or don't work... Does anyone think it could be the solid APAQ capacitors?

It's got a few of those scattered around (Mine are purple, not like the blue in the picture, but the values are the same - 16v and what I assume is 100uF).

Everything else is UCC and Fuji polymers. Any ideas?