current limiter thingie

Nice, (thanks) too bad they are not cheaper. (I guess the 1000 price is OK)

George H.

I wonder whether they are just JFETs with the gate tied to source. Those used to be sold as current regulators. I guess depletion MOSFETs would be much the same.

I don't think that JFETs have the same reverse characteristics.

What does a jfet current limiter do in the reverse direction?

Looks like a diode with some resistance. ...Jim Thompson

I've used the NSI50010. You can stack them to get higher Vak. I've done 2 in series to support a 10-70v input requirement. Used this to drive a optocoupler. I tested this setup up to 90V with no issues.

Perhaps that's true of JFETs in an IC (with some messy substrate interaction) - but it's not true of discrete devices. These are, to a first order, symmetrical. You can reverse drain and source with little effect {I just verified this on a good old curve tracer with a MPF102}.

This is why these devices (and not MOSFETs) are used in gain control circuits.

I'm sure there are second-order effects - capacitances, breakdown voltages,... that aren't symmetrical. But clearly discrete JFETS do _not_ act like the lousy diodes that MOSFETs emulate in the reverse direction.

-Frank

In "current-limiter-mode" the gate would be tied _always_ to source, hence "reversed" would have a diode characteristic.

You're mixing apples and oranges in your applications analysis ;-) ...Jim Thompson

Its the potentials on the device that decides what is source, gate and drain.

Many devices in both discrete and integrated are physically symmetrical, i.e. source=drain. However, special high voltage, extended drain devices are available which are not symmetrical.

Once you wire up a device, say and tie a "drain/source" to a "gate" to make a two terminal device, and then run positive and negative potentials, it won't conduct symmetrical. However, for a normal symmetrical device, if you swap over the ones used as source and drain, the current will be "mirrored symmetrically". For large signals, one way, as Jim said, will be a diode (jfet). For small mv voltages, the currents will be similar, but not exactly the same as the potentials, are still not the same, due to the asymmetric connection to the gate.

-- Kevin Aylward

And one other important connection, with source and drain connected; many a JFET gate error current is VERY low indeed, they make great low-leakage diodes. Some of the old PFETs in metal cases were rather spectacular (the metal case being a fourth-wire shield). Any old JFET will have nanoamp max leakage (versus 1N4148 at 5000 nA).

What would the forward characteristic look like, surely not a regular diode? I believe someone said ohmic until it current limits.

Maybe I'm not getting this. Is it on in one polarity and off in the other, or ohmic in one direction with a current limit and similar to a diode forward characteristic in the other direction?

The PAD1 and similar are Jfet based low-leakage diodes; PAD1 is sold as a 1 pA diode.

They are terrible diodes. Series resistance and capacitance are both high, and they are expensive.

The c-b junction of a BFT25 is a vastly superior diode for a fraction of the price.

The 5 fA leakage was about my measurement resolution.

You _can_ force a JFET to have essentially symmetrical current limiting...

I do a similar gimmick with MOSFET polarity protection schemes all the time. ...Jim Thompson

And in a JFET current limiter, the wire connecting the gate to the source is an asymmetry, because there is not also a wire from the drain to gate.

The gate to channel P-N junction acts like a diode.

Of course symmetry can be restored by connecting the JFET's gate to both source and drain. ;)

Cheers

Phil Hobbs

Or to neither.

For small gate "forward" voltages I think many jfet gate-source resistance can actually drop lower as the device further enhances, until the gate diode is forward biased of course!

piglet

drain-source (typo - not gate-source resistance)

piglet

Mosfets or phemts, used in their ohmic regions, can do gain control just like jfets.

The bad thing about jfets, aside from the price, is their usual gigantic spread in parameters from part to part. A 5:1 or even 10:1 spread of Idss is common.