SparkFun Forums

[Russel]

Doesn't really appear to be a forum to discuss circuit problems, but here goes. Move it or delete it as you wish. I put it here cause its driving me nuts

I have a simple voltage regulator. 10K resistor feeds a 15V zener in the usual fashion. That voltage then controls the gate of a mosfet and the source of the mosfet is my output voltage. It works out to be about 12V. I've used this circuit many times and its worked fine.

This time, I ended up with what looked like a lot of noise in my circuit, till I traced it back to the mosfet. On closer inspection it looked like some kind of random(ish) oscillation. The level was about 500mV pk-pk. Fine I said, might be a mosfet thing, so lets put a cap across the zener.

So, now I have a 1uF cap across the zener. And initially that seemed to work. Until, I looked more closely. I still had a little bit of what looked like noise.

But then, I looked closer.

Across the zener and cap I get an odd little waveform. Its about 5-15mV peak to peak, and its got nice neat straight ramps upwards (the kind you get when charging a cap), and sudden drops downwards. A sawtooth, yes.. but a random one. The ramps terminate randomly. So do the drops.

I'm fairly sure this is the zener. For one thing if I supply power at <15V, then there is no oscillation. Reach that threshold and it starts to oscillate. Increase the supplied voltage and the ramps climb more steeply. (Typical period 10-100 us).

What it looks like is a zener that wants to suddenly conduct.. then stops.. then conducts.. over a random(ish) 5-15mV interval.

Has anyone seen this before? Is it some kind of thermal cycling effect? How do I make it go away (for sure)?

[leon_heller]

That is a well-known effect that sometimes occurs with zeners, they behave as oscillators under some conditions. The cause is unknown, but it might be something like quantum tunneling causing a negative resistance, as with tunnel diodes. That's just a guess, though.

I'd try adjusting the current through the zener, or use a regulator.

Leon

[FartingMonkey92]

What Leon said. I only use them for high voltage zener+mosfet combinations...

[bigglez]

I have a simple voltage regulator. 10K resistor feeds a 15V zener in the usual fashion. That voltage then controls the gate of a mosfet and the source of the mosfet is my output voltage. It works out to be about 12V.
....
Has anyone seen this before? Is it some kind of thermal cycling effect? How do I make it go away (for sure)? :)

Back when analogue music synthesizers were new
technology, specially selected zener diodes were
used for the noise generator. To cure this the
diode is bypassed with a capacitor as you have
already tried. You may have found a particularly
noisy diode specimen.

Why are you not using a LDO? Seems that a modern
single component voltage regulator would outperform
a zener and voltage follower circuit.

[Russel]

That is a well-known effect that sometimes occurs with zeners, they behave as oscillators under some conditions. The cause is unknown, but it might be something like quantum tunneling causing a negative resistance, as with tunnel diodes. That's just a guess, though.

I'd try adjusting the current through the zener, or use a regulator.

Leon

Do you have any references? What sorta amplitude of oscillation are we talking about? As I said, its around the 5-15mV scale with a 1uF cap there.. And around 500mV without one (except for the mosfet itself).

In reference to zener noise generators. All the sources I've found seem to suggest the avalanche noise you can expect to produce is in the order of tens of microvolts. They all seem to require a lot of amplification.

I still can't rule out the theory that it's thermal cycling. Goes like this. The zener breaks down. Capacitor discharges. Heating occurs. The avalanche effect has a positive temp coefficient so the zener shuts off. Capacitor starts ramping up. As it does, the zener cools off. When I checked out the data (this is for a MM3Z15VC) it would indicate I'd need to cycle the zener by about 0.6C.

The thought that its an aberrant device bothers me. Makes me want to try a different family of zeners. But then there's no way to tell who makes a good one and who doesn't. It's one thing to fix the problem. Its another thing to know that the next 100 you build don't need individual testing :/

Here's a bit more background. The supply has to generate 50mA or so from a 24V vehicle bus. Vehicles with BIG alternators. Thus I design around 120V (peak) load dumps. That kinda rules out a LDO. Also, to increase the zener current means going from a 1206 sized 10K resistor to a 2512 sized 1W resistor. My board is warm enough as it is :/

In any case, to test whether current makes a difference I've simply cranked up my supply to 40V. The oscillation persists. It just simply scales downward. So it starts at a zener current in the microamps and seems to persist all the way into milliamps. Urgh.

Thanks for the help so far

[Russel]

That is a well-known effect that sometimes occurs with zeners, they behave as oscillators under some conditions. The cause is unknown, but it might be something like quantum tunneling causing a negative resistance, as with tunnel diodes.

Leon

Just an after thought here. It reminds me of unijunction transistors. Except, those things tended to be a lot more predictable

[wiml]

The thermal cycling theory sounds pretty plausible to me, actually. In which case I would expect it to get worse with more capacitance across the diode. What's the timescale or approximate period of the oscillations, anyway? Does the behavior change much if you aim a heat gun or a cold air spray at the diode?

Lots of chipmakers sell linear regulators that'll take high input voltages, for automotive applications— but the input voltages seem to top out around 80v. There's the MAX6330, which looks like it acts like a precision zener up to 200V.

If you have access to USENET, you might also ask on sci.electronics.design… lots of cranks there, but also a few people who really know their stuff.

[Russel]

I found some references to oscillating zeners on sci.electronics.design .

There appears to be a real effect called microplasma discharges. These happen where there is some small defect or inhomogeneity in the silicon. Avalanche occurs in a very small spot. One reference I saw said that the discharge stops due to localised self-heating. The problem I have is one of scale...

I have a 1uF cap being discharged by 10mV in less than 1us. Hence a discharge in the order of milliamps. A lot of references to microplasma discharges talk in terms of tens of microamps (typical) for a given breakdown site. But then the experiments quoted in sci.electronics.* seem to suggest much higher currents, consistent with what I'm seeing.

Anyhow all of this is destroying my faith in zeners :/

[Russel]

The MAX6330 only has 3 preset voltages.. 3, 3.3 and 5V

However the old faithful ZTL431 will get me out of trouble in future. Shame about the extra board space is all.

Btw, in my travels I saw a series pass low dropout voltage regulator chip thats basically an op amp controlling an N MOSFET. It required a separate higher voltage (but low current) supply to feed the op amp. Now I cant find its number.

[JamesK]

Hi,

For my education...

If I understand the problem correctly:
a) Need to regulate from 24V to 12V at 50mA
b) Need to cope with 120V transients

I was wondering if you could use a standard 12V linear regulator (small, cheap), typically rated for 35V input, and protect it from the transients with say a 30V zener (also small & cheap) in parallel?

I'm guessing I'm missing some drawbacks to this overly simple approach, like perhaps it wouldn't work

Regards,
James.

[Blackfin]

Hi,

For my education...

If I understand the problem correctly:
a) Need to regulate from 24V to 12V at 50mA
b) Need to cope with 120V transients

I was wondering if you could use a standard 12V linear regulator (small, cheap), typically rated for 35V input, and protect it from the transients with say a 30V zener (also small & cheap) in parallel?

I'm guessing I'm missing some drawbacks to this overly simple approach, like perhaps it wouldn't work

Using a linear to drop 24V to 12V at 50mA will result in a fairly warm regulator, since it's dropping 12V at 50mA, about 600mW worth of heat will be generated. If efficiency isn't paramount and the heat can be sunk away and dispersed, this would probably be fine. I'd probably use a pole or two of LPF in concert with a TVS and rather than a zener for its speed and power handling ability. What's ultimately used to protect the input ultimately really depends on the nature of the supply surges etc.

It sounds like thermals are already an issue though so a linear adding 600mW of heat probably isn't viable. I'd probably look into something like a National LM2672-12 "simple switcher" buck converter with some of the conditioning circuit I mentioned above (LPF with TVS etc).

[bigglez]

Here's a bit more background. The supply has to generate 50mA or so from a 24V vehicle bus. Vehicles with BIG alternators. Thus I design around 120V (peak) load dumps. That kinda rules out a LDO.

Automotive circuits are protected from similar surges
(for similar reasons) and typically use an LDO after
a series diode and a surge suppressor (MOV).
For a 50mA load you could use a small footprint LDO,
and avoid some of its dissipation with a series resistor.
Simple usually wins in the end.

[Russel]

Just a follow up here.

Added a 10uF cap on top of the 1uF cap. Oscillation slowed down, but the scale didn't change (still around 10mV).

Took the zener off the board and tested it on the bench with a 22K resistor in series. Yep, back to the wild 500mV or so oscillation.

Tried 2 other zeners, same type MM3Z.. , different batches. Same results.

Tried a different brand of zener - 15V 3W. Same results.

Tried a leaded IN5xxx series zener - same results.

Yikes! No wonder the manufacturers never publish results under 1mA

Finally got the circuit working with 10uF and a 14V 3W surface mount part.

In future I'll use a ztl431 driving the base of a mosfet. It should be stable, handle the heat, handle the transients, and its still a cheaper solution than using most linear regs. I've also found the ztl431 is pretty well behaved at low currents.

Thanks for the ideas folks. Now, should I write this up for a journal?