High Voltage leaky diode (or any diode like device)

On Mar 29, 2:43 pm, Fred Bloggs wrote: [....]

Perhaps if he is reading along, he can fill us in. He may have to shoot us afterwards.

[.....]

I don't know the real application. If accurate measurement was the gaol neither of our ideas will work. If signalling is the goal, perhaps an even simpler circuit could do the job.

Dynamic impedance is similar to that of the original Corotrons and possibly a bit lower depending how it is measured. Leakage depends on temperature; it runs from nanoamps near 20C to a microamp or so at 185C. Look at the datasheets for current production; note a small positive TC built in to partly compensate for lower PMT gain at high temp. We have a design that has less that 1% voltage variation from 25C to

185C at a given current and (without looking at the data) seems to be within that spec from 30uA to 100uA. Back when the Corotrons were rare but available, their prices were in the $400 range so we pegged our unit at that value, and when the Corortrons became "unavailable", we lowered our price to $300. Then Titan lowered their price to $200 and we followed and have kept that price even tho they quit making them (and buy ours).

I just reread my orignal post a a few things gut mumbled up between the brain and the fingers. Zener bias resistor = 32K ohms. input voltage range = 80 - 200V (150V nominal) but must withstand 5 second transients up to 400V. Ramp rates in the transients are in the 1V/ms range (not very fast).

This circuit will be providing a voltage reference that must do two things.

1) follow the input voltage from ~80 to 200V (within 0V is ideal but up to 1V lower is acceptable) 2) When the input is 200V - 300V, output voltage must clamp at 200V (+/- 5V)

There will very little current drawn from the output at all times. (~100nA max)

Ambient operationg temperature will be 0 - 150C

Seems easy at first, just a zener with a bias resistor and you're set right. Not quite.

Garden variety 200V zeners have high temperature coefficent. With a 150C swing that present a significant error in the clamp point. Stacking (35)

5.6V zeneres is just silly and no space for it anyway. I found the 1N4085 and the tempco is very good but it has some leakage at higher temperature. It is also somewhat delicate so you can't surge 100mA through it or it will give up the magic smoke.

So I figure its a 200V 2.5W (200V @ 12mA) device so I derate the peak power rating to 50% (6mA peak).

400VIN - 200Vz = 200V 200V / 6mA = ~32K

The 4085 leakage is ~250uA 150C and ~1uA at room temp. This generates a error of ~8V (250uA @ 32K) when the input is below the output. too much.

The 4085 has a softish knee and the Vout starts to diverge from VIN at ~170Vdc. It still clamps at 200V but I need 220V to get there. Also the default 1n4085 is a 5% tolerance.

I can live with the knee and the 5% tolerance but I cant live with the leakage current induced error.

I figure I can find a leaky diode that leaks say 500uA at 150C and put two of them in parallel with the resistor and that would get it.

If there is a circuit that would also greatly sharpen the knee and reduce the initial tolerance to say 1% that would be great too.

I'm already looking at a transistor version offere by Win Hill on a different thread and the TL431 derivitives as alternates.

Keep the ideas rolling in.

Much appreciated.

Can it have a negative slope in the 200V-300V span? One of my ideas was to do something at the bottom end of the zener based on a sensed input voltage.

A MOSFET can be turned on by a photovolt-opto-isolator. This would let you bias a N-MOSFET all the way on when the voltage is under

200V. When the Zener concucts, a NPN could snap its gate down.

How much current can we draw from the input side or is there only a power limit set by practical considerations?

IIRC, its about +0.1V per degree C up at about 200V. That's about 50 silicon diode drops worth of correcting to do.

I think the zero tempco point moves up as you run at lower currents.

A Dep-MOS fet would be very nice here. A part like the LND150 runs 1 to 3mA or so of Idss. It has an on resistance of about 1K so it will make for a much smaller drop at the normal condition.

Maybe Interfet would select 2N6449s for you.

Do you have a low voltage supply we can use to make a faked Zener?

A micro power op-amp / reference and a high voltage transistor may leak less.

In the extreme, we could create a leakage at low temperatures to make the leakage constant.

You cannot use the Microsemi 1N4085 in this application. It is rated at

2.5W with 20mW/oC derating above 25oC meaning the allowable power dissipation at 150oC is 2.5-(150-25)*0.02=0 watts. Any clamping at 150oC will cause the diode to fail in short order. All of your component choices are probably inadequate. You need something else like this: Everything else should be a 200oC part minimum, and of course you need to pay attention to something called "derating" the power dissipation, the instructions for this are not banner headlines, usually contained in a footnote...

:)

By the book yes but in my business I design for 175C ambient temperature all the time using commercial parts. the 1N4085 does indeed work for extended time at 175C without failing as do several well selected transistors and various other parts. I work in wireline oilwell logging tools so testing comercial components to high temperature WAY beyond the datasheet (up to

220C) is the norm for me. 150C is not a problem with well selected components.

Yes in an extreme reliability sitution, I would specify the 200C parts but this is not one of those situations.

thanks.

From your description, you need something like this: View in a fixed-width font such as Courier.

. . . . . . . 10mA Vin . / \\ | . .------- -------| HV P-CH . | ---|+/ |- . | | |/ | . -/ | | . 20V ^ +---------[9R]---+---> out . | | . | [R] . | | . | | . ------+-----+------------------------- . . .

Our Codatron(TM) has an extremely sharp knee; to find the leakage at a given temperature, one must bias it below the knee. That leakage is not a part of the temperature coefficent up to 200C. The max (surge) current spec is 3mA, and the nominal max operating current is 500uA; and these values are good to 200C. However, one can use them to 6mA peak for short pulse periods. One percent voltage spec is easy. Now if you need low TC, the zener or reference must be low TC and at

200V, using an emitter follower (transistor to take the power) will not make any significant change in the reference TC. Every bandgap reference manufacture and type has a different temperature characteristic. At temperatures from roughly 150C and up, some oscillate, some have an "N" V/I characteristic, some leak like hell before the "knee", and some fail to work; most are useless above 165-170C. Ya gotta test, test and test!

I presume then that you know about the Yahoo wireline group hosted by Syd, who markets and sells our Codatron(TM). As far as parts working at temperature, i know whereof you speak. It is no accident that we guarantee our shunt regulator to 200C.

Perhaps i should have asked why dump so much current into the reference / zener, since you indicate such a small load current. With our Codatron(TM), you could set its current to the traditional 60uA. The Taylor brothers like the low noise, and call it "magic".