High Voltage leaky diode (or any diode like device)

Maybe just a diode in reverse parallel with the current limiting resistor...

A normal rectifier diode will leak in much the same manner as you need.

Another option may be to use a depletion mode MOSFET could be used as a constant current source in place of the resistor. Normally I would suggest a Supertex product but the last time I checked their web site, it appeared that they no longer make them.

The data sheets and selector guides are showing up for me today. Are they really discontinued?

Cheers,

Phil Hobbs

Dont think so, they are still on the bwebsite just harder to find.

Here's a possibility for you: instead of using a high-power zener, use a lower power one and a high voltage transistor and a resistor. The zener goes from transistor collector to transistor base, and the resistor goes from base to emitter. Because the B-E voltage is relatively constant, and given enough transistor gain and low enough B- E resistor value, the current in the zener diode when the three parts are acting like a zener is relatively constant. The result can be a much sharper knee--slightly soft till the transistor turns on, and then very low impedance above that. Since you'd be using a lower power zener with smaller die area, presumably the leakage current at lower voltage would also be lower, perhaps enough lower that it's no longer a problem. This method also moves the lion's share of the power dissipation out of the zener and into the transistor, allowing the zener to stay closer to ambient temperature (assuming it's not thermally coupled to the transistor). You can also do the same sort of thing with a power MOSFET--probably a good idea to provide some gate-voltage clamping though! Beware the possibility of transients punching excessive base current into the transistor, too.

Cheers, Tom

There is no such thing as a "low TC 200V zener"; the "sweet spot" for zeners is near 5.6V.

Hmmm...could you use a shunt regulator optimized at 60uA with almost zero TC? Name the voltage. Leakage current less than 20uA at 185C; rather sharp knee. Not cheap, as it is designed as a drop-in replacement for the old Victoreen Corotron(TM) and i am the only source (Titan's regulator did not hold up and they finally quit making theirs).

??? I thought Mouser still carried them...

Can you be more specific about what you're trying to accomplish? The statement "Vcathode = Vin" makes no sense in engineering terms. How closely does it have to match over what range of source voltages and load currents and whatever else is in the circuit. The devil is in the DETAILS.

Trying to match leakage of one device with a different type of device with WIDELY different voltages over temperature is often futile. If you're building one test fixture, it might be worth the time to select something that works...but probably cheaper to use more parts and less time. If you're designing a production device, you won't (shouldn't) get past the first design review. NEVER, EVER design using a parameter that is not tested/guaranteed and expected to stay the same when purchasing second-sources the part...or the vendor second-sources the raw material...or

And I must not understand your reasoning. If you expect Vcathode = Vin, that defines the voltage across the resistor to be ~zero...which means that there can't be any leakage in a parallel connected diode??? Pesky details again...

If the leakage is your only problem and the current design is close to working, you may be able to tweak it by using a two-stage zener setup. The first zener doesn't have to be very precise, just heavy enough to burn the power required to get the source resistance down to what your leakage can tolerate. You can also do some interesting stuff with PTC resistors and/or incandescent light bulbs.

If diode leakage across 32K is a problem, you can't be delivering much current to the load. Maybe should just bite the bullet and build a voltage controlled voltage source to get what you need. All depends on what you're trying to accomplish.

I looked at the logwell site and saw the victoreen corotron regulator replacement. What kind of price range are we looking at for those guys?

How much do they leak below the zener voltage? Dynamic impedance above the zener voltage?

thanks

In my case the heating is caused by the environment and not power disipation so the zener still needs to be low tempco. Vbe also has a 2mV/C tempo that must be dealt with. Doesn't look like that will cause too much error in this case.

I will take a look at this configuration see how well it does for this application.

thanks..

I'm not quite sure how that helps you, but current regulator diodes (really just JFETs with G/S shorted together) can be stacked in series to reach any voltage limit. Siliconix (now Vishay) used to have a complete line. SST-502 seems to be still available.

If I wanted to follow Vin until it hit 200V then limit out, I'd use an op amp and HV transistor , sensing the Vin through a high value sense resistor with a current sink load. Or, I'd keep the Zener and use a transistor/resistor to amplify it (hot resistors are better than hot Zeners).

That would be the absolute best way, but you can still get passable performance with transistors: View in a fixed-width font such as Courier.

. . . . .--------------. . | | . IN >----+ | . | |/ . '-[160k]-+---| Q1 KSC5603D . | |>

. - | . D1 v | . - [180] . | | . '-|>|-+---->OUT . D2 | . | . ---/ . // \\ 1N4085 . --- . | . --- . . .

I use a rule that a part must be:

(A) Found on the makers web site and (B) Easy to get such as from digikey

before I design it in.

The reason (A) is because many times I have seen parts still in the supply chain long after the maker has lots interest in that line. When that happens the part will go away some time soon for lack of interest by the maker.

The reason (B) is because if a part is hard to get, it doesn't get used so the sales are usually small and the maker gives it up at some point. I have also seen datasheets that are pure fiction about a part that the maker thinks that they may make if there are enough requests for it.

A side effect of (A) is that I don't use parts from NXP because the web site is so bad I can't find parts in it.

For the OP's need that won't really work because the circuit needs some voltage before it conducts at all. This circuit won't work either but is better:

Fixed font: Vin R1 -----+--------/\\/\\----- ! ! ! !/ e -->!-->!--+----! PNP D1 D1 ! !\\ / ! R2 \\ +---- Vout / ! ! /-/ Zener GND ^ ! GND

There is a resistance and a saturated transistor in the path up to the point where the zener conducts and pulls the transistor out of saturation. The current from R2 flowing in R1 scales the Vin down a bit but the EB drop of the PNP is mostly taken out of the picture by the fact that the voltages are so large. This makes the tempco small.

His minimum input is 100V so I don't think that is a consideration.

That's not bad but it does have its drawbacks compared to the NPN circuit: 1) low voltage saturation does not bode well for immunity to transients, and 2) the output impedance is about 2x. That's a big price to pay for a few less millivolts drop. Then are the HV pnp as readily available as the NPN, and same HFEs. He can dip into the ultra-cheap HV NPN deflection amps for the NPNs. That KSC5603D is $0.71 in unit quantity from Mouser.

Not yet but look at:

Notice they aren't listed on the "product families" page. This is usually a sure sign that the product is going away.

He was concerned about there being a tempco on the difference between the input and output voltages. I think the intent was to protect some sort of measurement circuit.

Yes, the could be a problem if the high voltage arrives quickly. The zener's current would be very large as the PNP came out of saturation.

and 2) the output impedance is about 2x. That's a big price

I don't see why the impedance would be lower. Your design has a non- saturated NPN in the current path during the "normal" operation. Mine has a saturated PNP which should be a much lower impedance.

Finding HV PNP is hard but not completely out of the question. It is, I admit, going to be trouble at 600V.

I had a couple of other interesting ideas:

Does anyone make "Static Induction Transistors". In theory, they would be a good part for this sort of circuit but nobody seems to make them.

For the OPs purposes a variation of your circuit may work better:

-------+--------------- ! ! ! !/ --/\\/\\--+----! NPN ! !\\ e ! ! ! +------- OP's output \\! ! NPN !----+ e /! ! ! \\ ! / ! \\ ! ! +------ ! /-/ ^ ! GND

I think it was more basic than that, he said " My problem is at 100V the zener has leakage that is proportional with temperature. This will cause an error when the sener is in vollower mode that gets worse with temperature." Here was talking about the 1mA leakage causing a 32V drop across his 32K current limiting resistor, that is a -32% error at minimum voltage and Tmax (which I figure is 85oC from his derating). I don't know for sure but it may have something to do with serial signaling of a network of remote devices through a common power supply feed.

It was just an observation, I don't think it's going to break the bank. The NPN is always in emitter follower mode.

Why don't you consider replacing the Zener? I'd use a TL431 in a cascode configuration: 32K

-----------/\\/\\--+----+-- out | | | | | | | | 1K |/ c \\ +--/\\/\\----| NPN / 22M

The low-voltage supply could be varied between 4 and 36V volts. The main leak is caused by the 22M resistor. and about 100=B5A at 200V. If a BF459 is used, its leak would be max. 5=B5A at 150 =B0C junction temperature. So total error is going to be below 3.5 volts. if that isn't accurate enough, an emitter follower could be used to drive the divider. But a circuit needing higher accuracy is not a good design anyway.