The Official Dual Rail Thread

[Merlinb]

That's an old electronics industry myth. MOSFETs do have a failure mode similar to second breakdown, caused by the negative tempco of Vgs at low currents and high voltages.

Since it is used in a source follower configuration, why would it matter if Vgs falls with increasing temperature?

No MOSFET is specified for linear use at low current and high voltage, except for some from APT

Then why do they provide a DC region on the SOA?

[Steve Conner]

http://www.nalanda.nitc.ac.in/indust...APT/AN0002.pdf explains better than I could.

I've never seen a power MOSFET datasheet that gives a SOA curve with a DC region. It's a switching SOA, and the datasheet often says "Not for linear use".

High voltage power FETs can probably dissipate quite a lot of power in the linear region, to be sure, but it's not specified, and it's not the amount the datasheet would suggest from the FET's thermal ratings.

[imaradiostar]

http://www.nalanda.nitc.ac.in/indust...APT/AN0002.pdf explains better than I could.

I've never seen a power MOSFET datasheet that gives a SOA curve with a DC region. It's a switching SOA, and the datasheet often says "Not for linear use".

High voltage power FETs can probably dissipate quite a lot of power in the linear region, to be sure, but it's not specified, and it's not the amount the datasheet would suggest from the FET's thermal ratings.

It is somewhat reasonable to assume that if you use a sufficiently overspec'd part it'll continue to function in DC use in spite of not being designed for it. Any MOSFET in a screen supply that sinks big current for more than a few seconds at a shot will have bigger issues- like the screen grids they're connected to lighting up like a Christmas tree. The thermal issues are more a case of proper heatsinking and design than anything else. Parallel mosfets may go up like firecrackers but for this kind of use (single transistor) it should be a walk in the park.

For a regulator we don't need a particularly fast device- maybe we should be looking for IGBT's. Newer devices have rendered previous slower generations much cheaper and they surely can handle the linear load!

jamie

[Merlinb]

I've never seen a power MOSFET datasheet that gives a SOA curve with a DC region.

That surprises me- I've only seen one data sheet that doesn't give a DC operating area! Although it was interesting to learn that the SOA is usually calculated rather than tested.

And I've never thought of 600V+ MOSFETs as expensive or hard to find either?
http://www.rapidonline.com/netalogue/specs/47-0190.pdf

http://www.nalanda.nitc.ac.in/indust...APT/AN0002.pdf explains better than I could.

Looks like the source-follower configuration should not be at risk from this, thanks to the 100% inherent feedback.

[Steve Conner]

Well no, the point is that mosfets are made of a whole lot of little cells paralleled internally. If some get hotter than others they'll hog current, irrespective of whether the device is wired as a source follower.

[R.G.]

I've never seen a power MOSFET datasheet that gives a SOA curve with a DC region. It's a switching SOA, and the datasheet often says "Not for linear use".

Fairchild FQPF6N80C, page 5, figures 9-1 and 9.2; area under the "DC" curve.

Well no, the point is that mosfets are made of a whole lot of little cells paralleled internally. If some get hotter than others they'll hog current, irrespective of whether the device is wired as a source follower.

Same link, Figure 8. On resistance of the device goes uniformly up as junction temperature rises. This has to be made of the parallel resistance of the cells, so the only way this can happen is if all the cells are increasing resistance. If some cells got hotter, they'd go higher resistance than their fellows and with the higher resistance dissipate less. MOSFETs spread the hot spots out, unlike bipolars. This non-hogging is one of the primary advantages of MOSFETs over power bipolars.

Also figure 3: drain resistance goes up with drain current, the opposite of current hogging.

Interestingly, the breakdown voltage increases with temperature too: Figure 7. I didn't know that.

Still digging. I found this at International Rectifier's site: http://www.irf.com/technical-info/appnotes/an-1155.pdf. It seems to uphold the hogging theory. Got more looking to do.

==================
Still more:Power MOSFET Thermal Instability Characterization

Looks like making MOSFETs better for switching has made them worse for linear operation.

So yes, even more calculation is needed for low current high voltage linear operation of power MOSFETs, as Steve indicates.

I'll do some more digging and worry this point. More here later.

[mooreamps]

So what are you saying ? That maybe a power tube, or a pair of power tubes would make a better VVR than a FET ?

-g

[Steve Conner]

Fairchild FQPF6N80C, page 5, figures 9-1 and 9.2; area under the "DC" curve.

OK, conceded :-P I can also see that the SOA curve shows 0.2A at 600V, the full 160W dissipation. If that really is true (and not just a calculated figure like the APT paper suggests) then this device would be a great choice for a VVR type circuit. I also see they're in stock at Farnell for about $1 each, so maybe I should get some and indulge in some Mythbusting.

Same link, Figure 8. On resistance of the device goes uniformly up as junction temperature rises.

True, but it's not relevant, and partly to blame for the perpetuation of the myth. The on-resistance applies when the device is turned fully on in a switched-mode application. The cells, and indeed multiple parallel MOSFETs, share just fine in switched mode, for the reasons you stated.

When a MOSFET is used in linear mode (what the literature confusingly calls "saturation" - the term here has the opposite meaning to what it does for BJTs) then the temperature coefficient of Vgs(th) has a much greater, and opposite, effect, and this is what causes the current hogging.

Tubes would indeed make a "better" VVR if issues like cost, space taken up, and heater current weren't considered. I used to have a tube regulated power supply whose outputs were protected by fuses. They were meant to be 200mA, but the previous owner had replaced them with 20 amp ones. I left it that way because I often needed considerably more than 200mA for short periods. A prolonged short would probably have done it in, but none of the abuse I gave it ever seemed to bother it.

[imaradiostar]

this one is readily available in the states:

http://www.fairchildsemi.com/ds/FQ%2FFQPF7N80C.pdf

Looks to have similar charts to the earlier posted part.

I still don't see why this is much of an issue- asking an 800 volt part to drop 200 or 250 volts for a screen supply at 50ma with occasional pulses above that seems like a really easy job for a mosfet. It seems like a 50-60 watt part should easily be able to take the 20 watts of heat this would require.

jamie

[Steve Conner]

RG and I are practicing EEs, so we immediately thought of the worst case scenario where the regulated rail shorts to ground.

I guess with a MOSFET regulated screen supply, you could choose screen resistors that would sacrifice themselves to save the FET.

[imaradiostar]

RG and I are practicing EEs, so we immediately thought of the worst case scenario where the regulated rail shorts to ground.

I guess with a MOSFET regulated screen supply, you could choose screen resistors that would sacrifice themselves to save the FET.

I'm not an EE- I just build stuff in a lab and blow it up and rebuild it till I can't blow it up anymore.

I do the math to get me in the ballpark of not blowing things up of course- it's part of the whole R&D thing.

Why would you want to save the fet? I'd much rather have a stiffer screen supply and possibly blow a $2 fet and a $1 fuse than have reduced tube power output as a result of excessively large screen resistors. Maybe I'm missing something. I generally prefer screen resistors sized to make decent power but protect the screens when things get ugly- regulating down the screen voltage allows us to reduce the screen resistor values, right?

jamie

[Steve Conner]

In a home-made amp it doesn't matter. But in a commercial product, the user can replace a fuse, but he can't replace the FET, so it means a service or warranty repair call.

Then again he can't replace the screen resistors either, so maybe it's OK to consider the FET a consumable item like screen resistors.

[imaradiostar]

Good call- I wasn't thinking production, only basement tinkering.

That said- should it ever really fail? If we're really worried about an individual fet failing we could have a fet per screen or fet per pair of output tubes or something. Even then it's probably overkill. I was thinking brute force approach- if it's built stout enough it won't fail. I've built similar circuits with fets that couldn't even handle the full plate voltage and been OK in the short term. I'd imagine a better fet with proper protection would be OK for the long term.

Sorry to beat a dead horse.

jamie

[Merlinb]

That said- should it ever really fail?

Not really- I think Steve is doing too much hand wringing. Hundreds of amps are on the market already, that use common-or-garden switching FETs for this sort of duty.

[Steve Conner]

The original poster wanted something really over-engineered. I felt duty bound to take the over-engineering to a whole 'nother level.

Hundreds? I've seen one Yorkville EL84 amp that used a MOSFET as a soft standby switch, and no production amp from a major company that used one as a regulator or variable voltage dropper.

(I discount the boutique stuff like Maven Peal and London Power, who can't be critiqued because they don't publish schematics, and the VVR kits, because I don't think they're engineered to commercial standards.)

Maybe they exist and I need to get out more.