Input protection (again)

Den fredag den 31. marts 2017 kl. 17.10.30 UTC+2 skrev George Herold:

current limiting/regulating diodes

OK thanks. Those look to be unipolar devices. I guess I could hook 'em up back to back. Maximum voltage is not that great either.

George H.

Yikes, those CMJH parts are over $6 each, qty 1. A pair of LND150s is a fifth of that.

These LED regulator things are interesting:

Looking at the curves, they appear to be ohmic near zero volts. And they are cheap.

Yup, except the diode is in parallel. Same deal with CMOS muxes--in the ON state they look just like resistors from a noise POV.

Cheers

Phil Hobbs

Those are interesting parts.

With two back to back, you'd need Schottky diodes across them to stay within the 500 mV abs max reverse voltage, but even adding a BAT54A to the BOM, they're still cheap.

Nasty tempco though--I wouldn't use them with any very expensive LEDs, but maybe good for input protection. Their zero-bias resistance appears to be about 60 ohms at 25 C, which will help with the Johnson noise, and

Cheers

Phil Hobbs

Because in most cases it is just a JFET with the gate shorted to the source by the factory. Provided that, I prefer to buy a regular JFET in SOT23 and do it myself.

The depletion mode MOSFETs have much higher DS voltage ratings and they make a great resettable fuse (ok, current limitter) on the supply lines. If one of your modules fails, the remeining ones continue to operate correctly, because the power supply will need to source only the FET saturation current. Just don't forget to provide the necessary heat dissipation mechanism for the transistor or it could unsolder itself. :-)

Best regards, Piotr

Check out the Microchip (Supertex) high voltage protection switches (MSO100 etc.)

I've wondered if the depletion MOSFET are perhaps vulnerable with the added resistor in place. What if you modify your test fixture and experiment.

Yes, too much heating for prolonged fault protection.

Nice. I need something that will go to a higher voltage. (The LND150's will work fine, but something with a bit more current might be useful.)

George H.

Indeed, they are. I have recently made a zero cross detector out of such a pair. They drive a transformer connected in series with them. Since the current is kind of sigmoidal, the voltage on the secondary is 0 except of the AC sign changing time, which produces a clear, 50us-wide spike, zero-centered by physics itself. :-)

A funny circuit, BTW, apparently violates all the approximate transformer equations. You *decrease* the turns ratio and the output voltage *increases*. But only when you do it by adding turns at the primary side. }:->

Best regards, Piotr

OK I'm obviously confused again. So what happens when the drain is negative wrt to the source? I've been assuming the part gets turned off and the body diode conducts. But you are saying... (and JL's data sheet shows) that to be wrong.

OK I'm going to have to hook one up by itself and see what happens.

George H.

How about these?

Bets regards, Piotr

The body diode does conduct. That's why you need two of them back to back. If the diode were in series, it wouldn't be linear at low current.

Cheers

Phil Hobbs

You want me to add 1 k ohm (cross connected as in fig 5.80) plug it in and look for smoke? I could try 240 Vac also.

Oops right. I was (wrongly) multiplying 8V * 23 mA

George H.

That's the general problem with most depletion FETs without source resistors.

One approach is to use something like the BSS139 with a slow parallel RC in series with the source. That would prevent forest fires and avoid blowing out the gates without grossly degrading the noise, at least above the RC corner frequency. A high-value ceramic cap would probably be okay since in normal operation it wouldn't see any voltage to speak of.

One could go all rococo on it and use a couple of these LED current limiters to control a couple of those big 1 kV IXYS depletion FETs in the same sort of way as the source-source resistor with cross connected gates.

Of course that's a lot of iron to hang on a poor little input line, unless the transients are really horrible.

Cheers

Phil Hobbs

They may have reverse substrate diodes, or equivalent. Sometimes reverse voltage ratings are actually current limits. They might be OK back-to-back.

They may be binned jfets. I'll order some and play with them.

The Supertex depletion fets are still the best protection for high voltage overloads.

Supertex DN2530 is a 300 volt depletion fet in SOT89, Idss around 500 mA, which you can reduce with a source resistor.

This will dissipate 3 watts on a good copper pour.

Yes, go for it!

Hi Phil, OK I was having some confusion as to what was going on. Here's a nice 'scope shot. (Probably not needed for most of you.) I've got a 1 k ohm resistor in series with an LND150 GS shorted, and drain connected to ground.

With forward voltage I'm seeing the body diode conduct at ~0.6V (whatever) And then in the negative direction (as you've said) the lnd150, looks like a 1 k resistor so about equal voltage division with "real 1 k ohm in series). (Chan. 1 is input at 2V/div vs Chan. 2 looking at the GS junction is at 1V/div. ) At right about 2 V (DS) the fet becomes a constant current source, taking up the voltage drop from then on. (Exactly what the spec. sheet says.)

OK so now in the region where (in my pic) the voltage is positive. It looks like a forward biased diode.. but the turn on slope is more gentle. (I've got another 'scope shot with just a pn diode) So until the diode starts to conduct, it still looks like a 1 k ohm resistor.

Sorry I just had to beat it to death.

George H.