low capacitance zener

down around there it's between the zener and impact mode. so I guess it should be some what stable as a voltage reference.

--
"I'm never wrong, once i thought i was, but was mistaken"
Real Programmers Do things like this.

Yeah but are you sure next month's shipment will be close enough to that

5.5V?

--
Regards, Joerg

http://www.analogconsultants.com

Doesn't matter these days with software calibration.

But does the junction capacitance stay at 1.7 pF when it goes into Zener mode? I thought that's what caused the huge increase in capacitance. I believe some typical ones may even reach 1nF.

There have been several discussions on using the base-emitter reverse breakdown. Here's some info from 1999:

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ some app notes say operating in avalanche mode, however briefly, will degrade the noise figure in low-level RF amplifiers.

The 2N2222 base-emitter junction has a very sharp breakdown - about 7.2V from < 1uA to > 10mA. Very repeatable on a number of units from the same purchase.

The 2N2369 (or 2N3227) is not so sharp but the knee is still under 10uA. About 6.4V.

The 2N5139 is sharp. About 7.8V.

The 2N4209 is sharp. About 6.0V.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Regards,

Mike Monett

Many NPN transistors will act like "reference zeners" of you use the c-e terminals. The b-e junction acts like a zener, in series with the forward-biased c-b junction, giving a net 6.2 volts with near zero TC.

Interestingly, the BFS17 doesn't do this. The b-e junction zeners at

5.5, but you only get about 3.5 volts in the "reference zener" mode. This is worth investigating some rainy day.

John

But then you might as well use a diode plus tempco adjust. Most uC have onboard temp sensors these days. Or take a BAV70 and compare the other half against a reference.

Yeah, it wouldn't be the ethical thing to sell them off via EBay afterwards ;-)

Interesting.

--
Regards, Joerg

http://www.analogconsultants.com

John just mentioned using a forward-biased b-c junction plus the b-e junction to give zero tempco. I guess that needs a transistor with a repeatable b-e breakdown voltage at the correct tempco. Which gets us back to the start of the thread:)

Scratch that. I just looked up the capacitance for the 1N751 series.

The maximum capacitance of around 100pf for a 5V Zener is at zero bias, as John said.

Increasing the bias reduces the capacitance, but there is no data for the capacitance at the Zener voltage.

I had the distinct impression it was quite high in Zener mode. Tony Williams might have much more info.

OTOH, the LM431 is neat. Adjustable, with 50 ppm/°C.

[...]

Regards,

Mike Monett

I guess it usually doesn't matter because the zener acts like a pretty conductive path up there. Maybe figure 5 in this datasheet helps a bit in estimating the trend:

It sure is. However, these days I find myself more and more in situations where those don't cut it anymore. Even when using them in a feedback path you normally have to make sure it sees the minimum cathode current when driven at the REF input. For the LMV431 that should be at least around 100uA. Too much for many apps. The old TL431 needs 1mA, with some species dropped to 400uA.

There are alternative versions but that gets expensive.

--
Regards, Joerg

http://www.analogconsultants.com

Yes, that's what I was looking at. The bias curves stop at 50% of VZ. Can we assume the capacitance keeps getting lower, or does it turn around and increase in Zener mode? I don't know the answer to that.

There was a lot of discussion about using Zener noise as a wideband noise source. Since the noise is negative-going this takes two zeners driving a differential input to get symmetrical noise on the output.

The bandwidth depends on the junction capacitance in Zener mode. It looks like the regular Zener might have quite high capacitance, so John's comment says the reverse biased b-e junction would be much better than a plain zener diode.

Regards,

Mike Monett

Probably does go down but once you are up the knee the impedance drops so far that it'll dominate for most applications except RF where zeners aren't used. If I (hopefully) find a quiet time some day I'll put one on the impedance analyzer and have it scan the capacitance from zero to the rated zener current. That should be easier than hooking up an inductor and measure resonances with a dip meter.

I was never too fond of using zeners as noise sources. Those little bulbs for watches are much nicer. It's just a matter of finding one that presents a nice 50ohm source while glowing hot enough. But not too bright, of course ;-)

John has the luxury of small production runs with high margins and really high per unit prices. So he could probably buy a reel or two, have that all measured, stash it somewhere and it'll last him until he retires.

[...]

--
Regards, Joerg

http://www.analogconsultants.com

[...]

I've never heard of that before. What kind of light bulb makes a noise source?

Regards,

Mike Monett

Not me Mike. When using zeners to clamp AC signals I've only ever biassed-up the zener and used diodes to isolate any below-clamp AC voltage from the zener.

--
Tony Williams.

Check out the various signal line surge clamp products. These have been designed for low capacitance so as not to disrupt reception of signals in the tens of MHz region and work at logic compatible voltage clamping levels. The clamp level is probably not a precision specification.

Good Morning, Tony. Thanks for the reply. I was thinking of some of the marvellous posts you and Win supplied long ago. A little searching found some of them.

Here's one from Win's post of May 7, 2002:

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > from my post of 2Aug1997 > /| 1N5245B at 50uA > 200mV FS / | > /| / | /| > /| / | /| / | /| / | /| > / | /| / | / | / | /| / | / |/ | /| _____ > | / |/ | /| / |/| / |/ | / |/ | / |/ | > |/ | / |/ |/|/ |/ | /|/ | > |/ |/ | > / > scope trace, continuous events, 10us FS 14.41 V > average value

The trace above is from a zener diode that was a bit more noisy than most. A type of relaxation oscillation is clearly seen, with the diode's 285pF self capacitance charging from the 50uA current, until a discharge event is triggered, which happens randomly with an increasing probability as the average voltage is reached and exceeded. The discharge step looks nearly instantaneous, stopping at random voltages at or below the average voltage.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Win calcuates a diode capacitance of 285pF from the charging slope.

It was difficult to find a capacitance plot for the 1N5245B, which is a 15V zener. Vishay has one at

http://www.ortodoxism.ro/datasheets/vishay/85588.pdf

Figure 6 shows the diode capacitance vs Zener voltage at 2V reverse bias. Reading from the graph, the capacitance for the 1N5245B is about 25pF. From Jeorge's post, the capacitance decreases with increasing reverse bias, as one would expect.

However, Win's post seems to indicate the capacitance is much higher in Zener mode.

The thread is at

Very interesting and well worth reading.

Another thread that Win refers to is the 1997 thread on Zener Diode Oscillation. I can't seem to figure out how to get the start of the thread, but google provided this link that enters partway through:

Very well worth reading.

Regards,

Mike Monett

Good grief. You have a memory like an elephant Mike. I remember those zener noise threads, but completely forgot Win's post above, using dV/dT to measure capacitance.

Useful way of measuring the C from 0v to Vz. Have (say) a 50uA fixed pulldown, and a clocked 1mA constant current pullup. Look at the dV/dT over the voltage range and see how it varies.

--
Tony Williams.

A fluorescent lamp generates noise at 4 GHz. I used to repair C-band LNAs and LNBs in my shop at home. I didn't have a C-band signal generator, so I would wave the repaired unit past the overhead lamp. The noise was different than normal background noise, and the signal level meter would show that the unit was working. Then, it was put on the ten foot / three meter dish to see if it was OK. Most failures were in the power supply section, so the noise figure wasn't affected by the repairs.

--
Service to my country? Been there, Done that, and I\'ve got my DD214 to
prove it.

Any light bulb. A 100 ohm filament at 1500 C will generate about 3 nv/rthz wideband noise. Whereas a zener running at a mA will generate about 300.

John

At higher currents the noise becomes almost gaussian.

John

Yeah, that would be more repeatable. Somebody (Central?) has some multiple (quad?) esd zeners that are low pF. GE used to make a wonderful symmetric zener in a plastic transistor can, D13T1 or some funny number like that.

Sometimes a dual zener is just the thing to clamp an opamp. We're playing with ideas for a frequency counter front-end circuit, differential input over a huge voltage range, 1 MHz max maybe. We were thinking about running each side through a nonlinear function generator, vaguely a bidirectional log circuit (yes, I know that's impossible) and then run them into a comparator with hysteresis.

John

That one word "almost" is what drove us to using light bulbs. Those little watch bulbs are amazing. I thought they were designed to less than one hour MTBF because they are only normally used for a couple of seconds a day but they lasted forever.

--
Regards, Joerg

http://www.analogconsultants.com