Using FETs in parallel

"Fred Bloggs"

Also, many commercial products have power devices operating in parallel quite successfully with little or no ballasting.

What outsiders *do not know* is the particular devices in each unit were selected for exact match in the factory.

....... Phil

The amps were the Jands J1000 and J700 (Australian designed & made).

The Motorola and later "Hi-Rel" output BJTs base pins were fed via 10 ohm

At high output currents (or under short cct tests), the drop across each 10 ohm was between 0.4 and 0.9 volts !!!

To add to the insanity - the 0.47 ohm, ballast resistors were on a separate PCB that attached to the one with the BJTs via multipin plugs and sockets - with the result that the extra copper track lengths altered the actual value of the ballast by up to 30 % for the ones at the end.

On test, under load - the Ic variation across the 10 BJTs had a ratio of typically of 2:1 !!!

In a desperate attempt to solve the high resulting *blow up rate* in the field - Jands went from Motorola MJ15024/25s to the more expensive Hi-Rel EB204 / ED 204 devices. The Hi-Rels had a much tighter spec for Hfe and Vbe, so gross mismatching was reduced.

All Jands really had to do was remove all those damn 10 ohm resistors and run a solder coating along the long thin tracks that fed the distant ballast resistors.

Done this many times when carrying out repairs:

Voila - near perfect Ic matching.

There were a whole BUNCH of other mindless stuff ups too - like speaker DC protection relays that didn't.

Cost Jands' reputation very dearly.

HF maybe ?

Fusible types are they ??

........ Phil

What is a bucker? I know a Buckler, but he lives in Santa Cruz.

No major heat sinking required. The over voltage is only a surge requirement, between 50 and 200 mSec at 100 volts down to 40 volts on a linear curve.

That's an idea, but I don't think we need that to handle surges.

Right, using parts from the same batch (tube?), with a carefully-determined selection test jig and setup.

I did that once, about 10 years ago, on my first NMR gradient amp (which is just now being retired.) Nightmare, even with moderate source resistors. Now I close a loop around each fet with an opamp per.

John

Not really. But it pays to use an opamp that has a low open-loop output impedance, so the fet gate capacitance doesn't make the local loop too twitchey. The nice thing is that the effective gate thresholds all go to 0 +-1 mV, and the fet Gm's become predictable and perfectly linear. (This with feedback taken from the source, using a small source resistor.) And quiescent current control is now exact.

You can push the fets a lot harder thermally if you are confident they're sharing the dissipation equally. You could even - just thought of this - apportion the dissipation according to the fet location on the heatsink.

John

If you knew how expensive that is you might not claim it.

Any interesting stability issues with that approach?

Best regards, Spehro Pefhany

Hah !

Does anyone recall RCA running an ad in the 70s about thermal cycling ? They reckoned their die attach ? was better than the competition's IIRC.

ISTR a figure of 10,000 thermal cycles being mentioned as a critical level.

Graham

My NMR gradient drivers push the fets to close to their rated power (as much as 300 watts dissipation per TO247) at, say, 10% duty cycle, one pulse per second. Pulse fatigue doesn't seem to be a problem with the parts I've used, in that we're not seeing any indication of a wearout mechanism.

John

Don't you hate it when the cool ideas come too late for production? Or everything is already in the field? Let's hear it for programmable load sharing, with the hardware's firmware updatable by Internet push from the factory. :-)

On Mon, 16 Jan 2006 20:19:08 -0800, John Larkin wrote: ...

I've seen this done, but open-loop, with different source resistors. It was in a "Battery Charger-Analyzer" that would determine the condition of Fibrous NiCd batteries by charging them, discharging them (which is what the banks of FETs were for), and charging them again. I didn't design it, I just did some mods to the 68HC11 firmware; but the guy who did had arranged it so that the hotter FETs were closer to the fan, and the unit had a definite air flow path.

It seems to have worked - the guy sold about 200 of them to the US Army.

Cheers! Rich

On my bigger amps, I digitize everything - heatsink temp, fet currents, fet voltages - and simulate junction temps in real time, with software-programmable Tj shutdowns.

John

Running Windows in your machines again? Didn't you learn?

Nice, but no software-programmable load-sharing constants, because you didn't think of it soon enough, right?

You need a cycle in the double-digit seconds to get enough temperature delta for optimium thermal fatigue. In the early days, SSRs were dying in two or three months from this cause (running 24/7).

Best regards, Spehro Pefhany

I'm just going through the security for something along those lines with a security expert... ugh.

Best regards, Spehro Pefhany

No, Linux on an ARM, but the real problem is that it's wireless.

Best regards, Spehro Pefhany

So an audio amp pumping music at 140BPM will last 1 minute 11 seconds.