FET's Issue 3

[zacky]

Anyone has an idea on:
For high current applications which choice is more reliable:
1- Single FET with high current rating.
2- Several FET's with lower current rating connected in parallel.
Thanks in advance

 

[itsmoked]

I would always opt for a single...
Unless there is some unique condition.

Much simpler.
More reliable.
Less labor.
Less expensive for equivalent power.

 

[ScottyUK]

Depends somewhat on what size FET you are talking about:
A single FET has no redundancy, and probably higher R_DSon compared with parallel group. Multiple TO-218 plastic package devices can be easier to PCB-mount in a manner suitable for heatsinking than single TO-3 metal cans. Large single isolated base FET modules are easier to deal with than multiple smaller modules.

When the voltage / current / power starts rising to high levels, why not consider using IGBTs unless you are working with really high frequency?


----------------------------------

Your body might be a temple. Mine is an amusement park...

 

[itsmoked]

Scotty... are you serious about the redundancy??!?!?

What product have you ever seen that used multiple FETs for redundancy?

What about the failed shorted FET?

 

[ScottyUK]

ItSmoked,

Yes, definitely seeen it in professional amplifiers from a large-scale PA. I dallied working in professional sound while at college (too many holidays) and decided that it was a great way live like a hobo and make no money. The amp had individually fused FETs, and the amp was rated for full output with one of the three output FET's gone. Not sure how effective it was because I never knew of one that blew - but an interesting design none the less. The only things we ever replaced were fans and filters.


----------------------------------

Your body might be a temple. Mine is an amusement park...

 

[itsmoked]

I guess that would be a major league amp to rate individual fuses per FETs. Course I suppose the show-must-go-on!

 

[ScottyUK]

Two channels @ 1.5kW apiece, but they never worked flat out. Reliability took priority.

On the subject of reliability, losing a large LF enclosure is immensely irritating to the sound engineer (my boss at the time) but is usually masked to a large degree by its neighbours; losing an HF bank is much more awkward to deal with because HF sound is much more directional and you get noticeable 'dead' areas where there is a gap in the HF coverage. That said, most of the kids probably wouldn't notice the difference!

Have we dragged Zacky's thread sufficiently off-track yet?


----------------------------------

Your body might be a temple. Mine is an amusement park...

 

[fsmyth]

zacky, you are generally better off with a single FET, since
gate control is simpler. Multiple gates can be hard to drive
properly if the rise/fall times are short.
The only time I use more than one is when the power
requirements are greater than a single FET can handle.

smoked, scotty - I have always been under the impression
that an FET was constructed of lots of tiny internal
paralleled devices. But maybe that is only certain parts.
I have built a small induction heater using 6 in each leg
of an H-bridge; works great. After years of finding places
to mount emitter resistors, and trying to match beta on the
driver transistors, I /love/ FET's.
<als>

 

[fsmyth]

Well, crap. Should'a added the obvious use of paralleled
FET's - smaller (1-3 kVA) backup power systems. Got about
20 that use 4 banks of 5 ea. RFP45N06.
<als>

 

[MacGyverS2000]

The rest of the guys have already beat me to the punch... the true answer is "It depends on the application". Multiple parallel FETs will lower the Rds and provide redundancy, but require a more powerful base driver.

If you need reliability, multiple FETs would be the way to go, and give yourself some breathing room spec-wise if one or more of the FETs should fail. Don't forget to include some sort of indicator for an individual failure, else you could run with a failing unit for a long period of time, only to find out when another failure drops you below spec at the worst of times.


Dan
Owner

http://www.Hi-TecDesigns.com

 

[IRstuff]

Additionally, for reliability, you need to consider the percentage of maximum rating that you are attempting to use. Let's say that your analysis shows that reliability requires 50% margin on current rating. It would be easier to find smaller FETs that provide 50% margin than a single FET to provide that margin.

Ultimately, most reliability models consider temperature as the driving force. To lower operating temperature, you need to have less dissipation in the FET relative to its maximum ratings. Therefore, you need to have the operating margin.

TTFN

 

[zacky]

Regarding to power MOSFET Failure ,is it probably fail shorted or opened?
What will be the casa for IGBT ?

 

[Warpspeed]

The very large "block type" single FETs are actually multiple FETs if you pull one apart and have a look inside. The question is, what is better, a big single, or do it yourself and parallel your own ?

My biggest concern would be current sharing. With dc or low frequencies there is no real problem. But if extremely high di/dt currents are anticipated, the very compact low inductance package of a large single would be preferred.

If it was an audio amp, 60Hz inverter, or dc regulator I think I would opt for multiple small FETs, the heat into the heatsink would be better distributed, and the parts cost may be less.

If it was a monster >20Khz high frequency switching supply, the big single would be my choice. Cost of repair may come into this too. A big single is easy but expensive to replace. Multiple cheapies may be a real pain to repair, but the parts cost to do so may be lower.

Reliability with FETs mainly has to do with NEVER exceeding their ratings. Easier said than done, but with care and some inbuilt protection either approach should be equally reliable.

 

[IRstuff]

There are several possibilities:
> gate short
> gate open
> junction short -- metallization spikes

Of the three, the most likely is junction short, which is prevented by minimizing the current density, i.e., running at well below maximum rating. As stated above, even a single FET is composed of a paralleled array of smaller FETs, so junction shorts affect either configuration equally.

TTFN

 

[IRstuff]

One thing that's slightly more likely to happen with a larger FET is that the load current may not distribute uniformly within the die, thereby aggravating the current densities, precipitating earlier junction fails. On a smaller die, the current non-uniformity is more likely to have been accounted for in the device rating.

TTFN

 

[zeitghost]

According to legend, if you try really, really, really hard, you can turn on the parasitic bipolar transistor and then have all the joys of 2nd breakdown.

It's a dv/dt or di/dt thing from what I can recall from 15 - 20 years ago...

 

[zacky]

Warpspeed, I am fully agree with what you said, but I have seen some scaring articles such as "Turn-off Failure on Power MOSFETS" & "Dynamic stress might cause Power MOSFET Failure" &.......
So extra precautions my be required for selecting the suitable switching device specially for high frequecy applications.
Actually, I am working on DC/DC Flyback converter, 120V/250V-25Amps output at 50Khz switching frequecy, and I am looking for a heavy reliable single switching device, if you can suggest specific one ( FET or IGBT ) - cost is not the subject.

 

[Warpspeed]

Yes, I think that is probably the whole point. Static dc maximum ratings are one thing, you do your thermal heatsink calculations, and all is well.

But once you start really punishing the thing with extreme speed and violence, unequally distributed parasitic effects can become most serious. Localised voltage and current spikes can sometimes seriously overload individual devices in a large distributed array.

The manufacturers of the large singles are very careful to form all the internal connections and bonding wires into a compact "pyramid" formation with very short internal interconnections. But with dc or relatively low frequencies, hardly any of that really matters.

 

[sreid]

Take a look at the FET and IGBT offerings from IXYS.

http://www.ixys.com/

 

[Warpspeed]

zacky, For flyback, fast clean turn off is very important, which really rules out IGBTs. The tail current of an IGBT may create significant switching losses in a flyback circuit. Snubbing a flyback supply is counter productive and not really possible.

I would definitely go the big single mosfet route. Fastron and Semicron and possibly other sources should have exactly what you need.

Another approach to really high powered flyback systems is to run two or more flyback supplies in parallel, but out of phase. That can significantly reduce the peak currents both at the source and at the output capacitor. Ripple frequency will then be a multiple, and much easier to filter. It has a lot of other advantages too.

Another useful trick would be to use an IGBT and MOSFET in parallel to switch your flyback converter.

Turn on both together, the low conduction voltage of the IGBT will ensure it carries most of the rising current with fairly low conduction loss. Turn off the IGBT first so that all the load then transfers briefly to the MOSFET. The IGBT can then turn off relatively slowly while the MOSFET maintains the peak current. Then switch off the MOSFET to perform the actual flyback function. The IGBT will be well off, and tail current will not then be a problem.

THe IGBT carries most of the conduction current, The FET does the actual switching. With careful design the whole thing can be made far more efficient than by using either a MOSFET or IGBT by itself. The particular advantages of each type of device can be exploited to the full, and the disadvantages of each device minimised. They compliment each other so well when used that way.