# Zener diode modeling

Started by December 31, 2011
```Bill Sloman wrote:
> On Jan 1, 8:15 am, Robert Baer <robertb...@localnet.com> wrote:
>> BillSlomanwrote:
>>> On Dec 31, 5:49 am, Robert Baer <robertb...@localnet.com> wrote:
>>>>    Say one takes an E-B junction of a transistor or other diode and
>>>> measures the (reverse) breakdown voltage over a wide range of currents
>>>> (say 5nA to 5mA), and assume there is no point or range where one sees
>>>> characteristics of negative resistance or of oscillation.
>>>>    Also say that voltage is about 9 volts.
>>>>    Q: is that zener or avalanche breakdown?
>>>>    Q: how does one make a SPICE model of such a proposed beastie?
>>> At nine volts, the breakdown is by avalanche.
>>> It happens in very small areas, and the transit times is picoseconds,
>>> so at low currents the avalanche self-extinguishes from time to time
>>> due to the statistical fluctuations in avalanche multiplication
>>> process - sometimes a charge carrier gets through the avalanche region
>>> without generating any new charge-carrier pairs.
>>> This isn't easy to model in Spice - it probably needs a current-
>>> controlled random noise generator, but the real noise looks more like
>>> a series of Dirac spikes.
>>> For added extra complexity, you can throw in the fact that the
>>> avalanche process generates light, and in glass-packaged zener diodes
>>> this can trigger other avalanches.
>>> There was a thread on the subject here, many years ago - "Zener diode
>>> oscillation" from July 7, 1997.
>> 1) What about (say) 8.2 volts?
>
> The cross-over point is around 4.7V IIRR - the temperature coefficient
> of the break-down voltage is the clue, because it is negative when the
> Zener mechanism is dominant, and postive when avalanche breakdown is
> dominant
* Well, the voltage is well above that point, and the TC is positive, so
by both means it is avalanche mode.
Still..it behaves resistively well from zero to 5mA.
I have looked at a lot of transistor E-B junctions, and a lot of
zener diodes 6.2V to 15V and none behave nicely somewhere in that region.
Many start-up with a "backlash" or tunnel-diode type of bazzfazz
(very technical term meaning ugly).

>
> http://www.datasheetcatalog.org/datasheet/nationalsemiconductor/DS013023.PDF
>
>
>
>
> Searching on the exact phrase ""Zener diode oscillation" will do it.
>
> --
> Bill Sloman, Nijmegen
>
>
>
```
```Bill Sloman wrote:
> On Jan 1, 8:15 am, Robert Baer <robertb...@localnet.com> wrote:
>> BillSlomanwrote:
>>> On Dec 31, 5:49 am, Robert Baer <robertb...@localnet.com> wrote:
>>>>    Say one takes an E-B junction of a transistor or other diode and
>>>> measures the (reverse) breakdown voltage over a wide range of currents
>>>> (say 5nA to 5mA), and assume there is no point or range where one sees
>>>> characteristics of negative resistance or of oscillation.
>>>>    Also say that voltage is about 9 volts.
>>>>    Q: is that zener or avalanche breakdown?
>>>>    Q: how does one make a SPICE model of such a proposed beastie?
>>> At nine volts, the breakdown is by avalanche.
>>> It happens in very small areas, and the transit times is picoseconds,
>>> so at low currents the avalanche self-extinguishes from time to time
>>> due to the statistical fluctuations in avalanche multiplication
>>> process - sometimes a charge carrier gets through the avalanche region
>>> without generating any new charge-carrier pairs.
>>> This isn't easy to model in Spice - it probably needs a current-
>>> controlled random noise generator, but the real noise looks more like
>>> a series of Dirac spikes.
>>> For added extra complexity, you can throw in the fact that the
>>> avalanche process generates light, and in glass-packaged zener diodes
>>> this can trigger other avalanches.
>>> There was a thread on the subject here, many years ago - "Zener diode
>>> oscillation" from July 7, 1997.
>> 1) What about (say) 8.2 volts?
>
> The cross-over point is around 4.7V IIRR - the temperature coefficient
> of the break-down voltage is the clue, because it is negative when the
> Zener mechanism is dominant, and postive when avalanche breakdown is
> dominant
>
> http://www.datasheetcatalog.org/datasheet/nationalsemiconductor/DS013023.PDF
>
>
>
>
> Searching on the exact phrase ""Zener diode oscillation" will do it.
>
> --
> Bill Sloman, Nijmegen
>
>
>
VERY interesting!
A snip from the article by Win Hill:
" We know that each zener diode behaves differently.  In many cases the
magnitude of typical microplasma discharges are much smaller than the
examples I've shown,
e.g. each 20 - 40uA event lasts 0.3 to 20ns, removing 0.8pC at most,
yielding small spikes and low "noise."  A similar argument holds for a
typical zener as the average current is raised into the mA region.  Here
the average zener voltage increases a bit, which changes the
microplasma-discharge relaxation-oscillator environment, resulting in
"quiet" operation, even though the identical randomly-occuring
roughly-fixed-current discrete microplasma-discharge event mechanism  is
still fully responsible for all the current flow.
"

So, reversing that - that is to say, _decreasing_ the diode current
should result in more "noisy" operation.
And that seems to correlate with the observation of oscillations
and/or "backlash" at the knee (very low currents).

Hmmm...and hmmm.

And i can put hundreds in series and still see no problems.
```
```Jim Thompson wrote:
> On Sun, 1 Jan 2012 08:57:38 -0800 (PST), George Herold
> <gherold@teachspin.com> wrote:
>
>> On Dec 31 2011, 4:24 am, Bill Sloman <bill.slo...@ieee.org> wrote:
>>> On Dec 31, 5:49 am, Robert Baer <robertb...@localnet.com> wrote:
>>>
>>>>    Say one takes an E-B junction of a transistor or other diode and
>>>> measures the (reverse) breakdown voltage over a wide range of currents
>>>> (say 5nA to 5mA), and assume there is no point or range where one sees
>>>> characteristics of negative resistance or of oscillation.
>>>>    Also say that voltage is about 9 volts.
>>>>    Q: is that zener or avalanche breakdown?
>>>>    Q: how does one make a SPICE model of such a proposed beastie?
>>> At nine volts, the breakdown is by avalanche.
>>>
>>> It happens in very small areas, and the transit times is picoseconds,
>>> so at low currents the avalanche self-extinguishes from time to time
>>> due to the statistical fluctuations in avalanche multiplication
>>> process - sometimes a charge carrier gets through the avalanche region
>>> without generating any new charge-carrier pairs.
>>>
>>> This isn't easy to model in Spice - it probably needs a current-
>>> controlled random noise generator, but the real noise looks more like
>>> a series of Dirac spikes.
>> It's even 'worse' than that.  The noise changes 'character' as a
>> function of current.
>> Right near the 'knee' you go from random pulses to random steps.
>> (random telegraph noise).  With more noise at low frequency.... But
>> I'm not sure it's 1/f.
>
> "Popcorn" noise... often seen in bipolar OpAmp input stages.
>
>>
>>> For added extra complexity, you can throw in the fact that the
>>> avalanche process generates light, and in glass-packaged zener diodes
>>> this can trigger other avalanches.
>> And that room light can get in and 'modulate' the avalanche.
>>> There was a thread on the subject here, many years ago - "Zener diode
>>> oscillation" from July 7, 1997.
>>>
>>
>> I find that for use as noise sources it's best to keep the current
>> below the knee.
>>
>> George H.
>>> --
>>> Bill Sloman, Nijmegen
>>>
>>> --
>>> Bill Sloman, Nijmegen
>
>                                         ...Jim Thompson
"Popcorn" noise....
I remember "back in th good old daze" when every maker of analog op
amps was plagued with that and all sorts of ideas as to cause, etc was
floating around to no avail.
Someone finally found a cure, and if i remember correctly, it was
chemical PURITY and CLEANLINESS, especially related to the surface.
```
```Bill Sloman wrote:
> On Jan 1, 8:15 am, Robert Baer <robertb...@localnet.com> wrote:
>> BillSlomanwrote:
>>> On Dec 31, 5:49 am, Robert Baer <robertb...@localnet.com> wrote:
>>>>    Say one takes an E-B junction of a transistor or other diode and
>>>> measures the (reverse) breakdown voltage over a wide range of currents
>>>> (say 5nA to 5mA), and assume there is no point or range where one sees
>>>> characteristics of negative resistance or of oscillation.
>>>>    Also say that voltage is about 9 volts.
>>>>    Q: is that zener or avalanche breakdown?
>>>>    Q: how does one make a SPICE model of such a proposed beastie?
>>> At nine volts, the breakdown is by avalanche.
>>> It happens in very small areas, and the transit times is picoseconds,
>>> so at low currents the avalanche self-extinguishes from time to time
>>> due to the statistical fluctuations in avalanche multiplication
>>> process - sometimes a charge carrier gets through the avalanche region
>>> without generating any new charge-carrier pairs.
>>> This isn't easy to model in Spice - it probably needs a current-
>>> controlled random noise generator, but the real noise looks more like
>>> a series of Dirac spikes.
>>> For added extra complexity, you can throw in the fact that the
>>> avalanche process generates light, and in glass-packaged zener diodes
>>> this can trigger other avalanches.
>>> There was a thread on the subject here, many years ago - "Zener diode
>>> oscillation" from July 7, 1997.
>> 1) What about (say) 8.2 volts?
>
> The cross-over point is around 4.7V IIRR - the temperature coefficient
> of the break-down voltage is the clue, because it is negative when the
> Zener mechanism is dominant, and postive when avalanche breakdown is
> dominant
>
> http://www.datasheetcatalog.org/datasheet/nationalsemiconductor/DS013023.PDF
>
>
>
>
> Searching on the exact phrase ""Zener diode oscillation" will do it.
>
> --
> Bill Sloman, Nijmegen
>
>
>
To refute someone's statement about zener application (and theory to
boot), here is an exact quote from the first page of the Motorola Zener
Diode Handbook May 1967:

INTRODUCTION

the exotic components of only a few years ago into ubiquitous elements
capability, a host of zener-like devices have joined the designer's
arsenal. Temperature compensated diodes, reference diodes, current
regulators, and zener transient suppressors provide capabilities not
previously available.

Motorola's new Zener Diode Handbook is designed to provide the circuit
designer with all of the information necessary for efficient use of
zener components. Since the very diversity of zener application
precludes a comprehensive index of applications, the Handbook emphasis
is on designing with zeners. Proven basic circuits provide a springboard
for the designer's own requirements.

In addition to detailed analysis of zener characteristics and their
response to electrical and environmental stimuli, the Zener Diode
Handbook is a major source of device data. The cross reference,
specification, and selection information in Chapter X will greatly aid
the designer in his choice of zener components.

** end quote **

I got a little curious and decided to make more current measurements.
* 2N2219, Signetics circa 1980's: about 7.6V E-B, slightly negative
slope 0-200uA, sharp knee, no oscillations seen to a few mA. E-C (base
open) however, shows those (random) bursts that Win mentioned around 6V,
more frequent as the voltage is slowly increased.
* 2N3053, TI circa 1980's: about 7.3V E-B, similar only at lower current
0-100uA. E-C has definite negative slope 700uA to 5mA; no oscillations seen.
* 2N2369A, ST Micro, an RF type as in smallest can made: about 5.3V E-B,
0-200uA rather rounded knee. E-C 5-20uA region definitely negative slope
and noisy; no oscillation seen.

These were quick and dirty observations just out of curiosity.
However the results make it clear that when the B-C junction is used
as an added forward biased junction, that the characteristic of the
transistor takes on a whole new character which might be called beta
multiplication and enhancement.
```
```On Sat, 31 Dec 2011 10:33:05 -0700, Jim Thompson wrote:

> Accurately?  Not easily.  Spice has BV and IBV terms in the diode
> model.  Some simulators, such as PSpice, support the Mextram BJT model
> which models avalanche in the _collector_:

From the "Undocumented LTSpice" section of the LT Wiki:

"Bipolar CB avalanche breakdown is modeled in the LTspice Gummel-Poon device:

* BVcbo: C-B breakdown voltage.
* nBVcbo: breakdown emission coefficient ; default value = 1?
* TBVcbo1: linear temperature coefficient of breakdown voltage.
* TBVcbo2: quadratic temperature coefficient of breakdown voltage.

Bipolar BE breakdown is also in the LTspice Gummel-Poon device:

* BVbe: B-E breakdown voltage.
* IBVbe: breakdown current at breakdown voltage.
* nBVbe: breakdown emission coefficient. "

I haven't tried it (yet).

--
"For a successful technology, reality must take precedence
over public relations, for nature cannot be fooled."
(Richard Feynman)
```
```On Sun, 1 Jan 2012 08:57:38 -0800 (PST), George Herold
<gherold@teachspin.com> wrote:

>On Dec 31 2011, 4:24&#2013266080;am, Bill Sloman <bill.slo...@ieee.org> wrote:
>> On Dec 31, 5:49&#2013266080;am, Robert Baer <robertb...@localnet.com> wrote:
>>
>> > &#2013266080; &#2013266080;Say one takes an E-B junction of a transistor or other diode and
>> > measures the (reverse) breakdown voltage over a wide range of currents
>> > (say 5nA to 5mA), and assume there is no point or range where one sees
>> > characteristics of negative resistance or of oscillation.
>> > &#2013266080; &#2013266080;Also say that voltage is about 9 volts.
>> > &#2013266080; &#2013266080;Q: is that zener or avalanche breakdown?
>> > &#2013266080; &#2013266080;Q: how does one make a SPICE model of such a proposed beastie?
>>
>> At nine volts, the breakdown is by avalanche.
>>
>> It happens in very small areas, and the transit times is picoseconds,
>> so at low currents the avalanche self-extinguishes from time to time
>> due to the statistical fluctuations in avalanche multiplication
>> process - sometimes a charge carrier gets through the avalanche region
>> without generating any new charge-carrier pairs.
>>
>> This isn't easy to model in Spice - it probably needs a current-
>> controlled random noise generator, but the real noise looks more like
>> a series of Dirac spikes.
>
>It's even 'worse' than that.  The noise changes 'character' as a
>function of current.
>Right near the 'knee' you go from random pulses to random steps.
>(random telegraph noise).  With more noise at low frequency.... But
>I'm not sure it's 1/f.

A typical small 10-volt zener has a region where it will make noisy
sawtooth oscillations, 50 uA maybe from my memory. As current goes up,
you get the asymmetric erratic pulsing thing, decreasing in amplitude
as current increases. By the time you get to a few mA, it's nearly
symmetrical almost Gaussian wideband noise, a few hundred nV per root
Hz.

>
>
>>
>> For added extra complexity, you can throw in the fact that the
>> avalanche process generates light, and in glass-packaged zener diodes
>> this can trigger other avalanches.
>
>And that room light can get in and 'modulate' the avalanche.
>>
>> There was a thread on the subject here, many years ago - "Zener diode
>> oscillation" from July 7, 1997.
>>
>
>
>I find that for use as noise sources it's best to keep the current
>below the knee.

If you want predictable noise with decent statistics and wideband
quality, keep the current up, 5 or 10 mA for a typical 1N758 type.

John

```
```On Mon, 02 Jan 2012 09:52:41 -0800, Fred Abse
<excretatauris@invalid.invalid> wrote:

>On Sat, 31 Dec 2011 10:33:05 -0700, Jim Thompson wrote:
>
>> Accurately?  Not easily.  Spice has BV and IBV terms in the diode
>> model.  Some simulators, such as PSpice, support the Mextram BJT model
>> which models avalanche in the _collector_:
>
>From the "Undocumented LTSpice" section of the LT Wiki:
>
>"Bipolar CB avalanche breakdown is modeled in the LTspice Gummel-Poon device:
>
>    * BVcbo: C-B breakdown voltage.
>    * nBVcbo: breakdown emission coefficient ; default value = 1?
>    * TBVcbo1: linear temperature coefficient of breakdown voltage.
>    * TBVcbo2: quadratic temperature coefficient of breakdown voltage.

I don't think that's truly avalanche.

>
>Bipolar BE breakdown is also in the LTspice Gummel-Poon device:
>
>    * BVbe: B-E breakdown voltage.
>    * IBVbe: breakdown current at breakdown voltage.
>    * nBVbe: breakdown emission coefficient. "
>
>I haven't tried it (yet).

And that ain't Gummel-Poon... it's LTspice... just like their "diode"
model.

...Jim Thompson
--
| James E.Thompson, CTO                            |    mens     |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

I love to cook with wine.     Sometimes I even put it in the food.
```
```On Jan 2, 1:15=A0pm, John Larkin
<jjlar...@highNOTlandTHIStechnologyPART.com> wrote:
> On Sun, 1 Jan 2012 08:57:38 -0800 (PST), George Herold
>
>
>
>
>
> <gher...@teachspin.com> wrote:
> >On Dec 31 2011, 4:24 am, Bill Sloman <bill.slo...@ieee.org> wrote:
> >> On Dec 31, 5:49 am, Robert Baer <robertb...@localnet.com> wrote:
>
> >> > Say one takes an E-B junction of a transistor or other diode and
> >> > measures the (reverse) breakdown voltage over a wide range of curren=
ts
> >> > (say 5nA to 5mA), and assume there is no point or range where one se=
es
> >> > characteristics of negative resistance or of oscillation.
> >> > Also say that voltage is about 9 volts.
> >> > Q: is that zener or avalanche breakdown?
> >> > Q: how does one make a SPICE model of such a proposed beastie?
>
> >> At nine volts, the breakdown is by avalanche.
>
> >> It happens in very small areas, and the transit times is picoseconds,
> >> so at low currents the avalanche self-extinguishes from time to time
> >> due to the statistical fluctuations in avalanche multiplication
> >> process - sometimes a charge carrier gets through the avalanche region
> >> without generating any new charge-carrier pairs.
>
> >> This isn't easy to model in Spice - it probably needs a current-
> >> controlled random noise generator, but the real noise looks more like
> >> a series of Dirac spikes.
>
> >It's even 'worse' than that. =A0The noise changes 'character' as a
> >function of current.
> >Right near the 'knee' you go from random pulses to random steps.
> >(random telegraph noise). =A0With more noise at low frequency.... But
> >I'm not sure it's 1/f.
>
> A typical small 10-volt zener has a region where it will make noisy
> sawtooth oscillations, 50 uA maybe from my memory. As current goes up,
> you get the asymmetric erratic pulsing thing, decreasing in amplitude
> as current increases. By the time you get to a few mA, it's nearly
> symmetrical almost Gaussian wideband noise, a few hundred nV per root
> Hz.
>
>
>
>
>
>
>
> >> For added extra complexity, you can throw in the fact that the
> >> avalanche process generates light, and in glass-packaged zener diodes
> >> this can trigger other avalanches.
>
> >And that room light can get in and 'modulate' the avalanche.
>
> >> There was a thread on the subject here, many years ago - "Zener diode
> >> oscillation" from July 7, 1997.
>
>
> >I find that for use as noise sources it's best to keep the current
> >below the knee.
>
> If you want predictable noise with decent statistics and wideband
> quality, keep the current up, 5 or 10 mA for a typical 1N758 type.
>
> John- Hide quoted text -
>
> - Show quoted text -- Hide quoted text -
>
> - Show quoted text -

Hi John,  Yeah I guess I was looking for repeatabliity.
(I don't want to have to sort diodes...)
I've got exactly the pulse into RC in the schematic I posted.

Adding two back to back gives this RC-interuptus signal.  Randomly
spaced RC's going in both directions.

George H.
```
```On Mon, 2 Jan 2012 16:33:45 -0800 (PST), George Herold
<gherold@teachspin.com> wrote:

>On Jan 2, 1:15&#2013266080;pm, John Larkin
><jjlar...@highNOTlandTHIStechnologyPART.com> wrote:
>> On Sun, 1 Jan 2012 08:57:38 -0800 (PST), George Herold
>>
>>
>>
>>
>>
>> <gher...@teachspin.com> wrote:
>> >On Dec 31 2011, 4:24 am, Bill Sloman <bill.slo...@ieee.org> wrote:
>> >> On Dec 31, 5:49 am, Robert Baer <robertb...@localnet.com> wrote:
>>
>> >> > Say one takes an E-B junction of a transistor or other diode and
>> >> > measures the (reverse) breakdown voltage over a wide range of currents
>> >> > (say 5nA to 5mA), and assume there is no point or range where one sees
>> >> > characteristics of negative resistance or of oscillation.
>> >> > Also say that voltage is about 9 volts.
>> >> > Q: is that zener or avalanche breakdown?
>> >> > Q: how does one make a SPICE model of such a proposed beastie?
>>
>> >> At nine volts, the breakdown is by avalanche.
>>
>> >> It happens in very small areas, and the transit times is picoseconds,
>> >> so at low currents the avalanche self-extinguishes from time to time
>> >> due to the statistical fluctuations in avalanche multiplication
>> >> process - sometimes a charge carrier gets through the avalanche region
>> >> without generating any new charge-carrier pairs.
>>
>> >> This isn't easy to model in Spice - it probably needs a current-
>> >> controlled random noise generator, but the real noise looks more like
>> >> a series of Dirac spikes.
>>
>> >It's even 'worse' than that. &#2013266080;The noise changes 'character' as a
>> >function of current.
>> >Right near the 'knee' you go from random pulses to random steps.
>> >(random telegraph noise). &#2013266080;With more noise at low frequency.... But
>> >I'm not sure it's 1/f.
>>
>> A typical small 10-volt zener has a region where it will make noisy
>> sawtooth oscillations, 50 uA maybe from my memory. As current goes up,
>> you get the asymmetric erratic pulsing thing, decreasing in amplitude
>> as current increases. By the time you get to a few mA, it's nearly
>> symmetrical almost Gaussian wideband noise, a few hundred nV per root
>> Hz.
>>
>>
>>
>>
>>
>>
>>
>> >> For added extra complexity, you can throw in the fact that the
>> >> avalanche process generates light, and in glass-packaged zener diodes
>> >> this can trigger other avalanches.
>>
>> >And that room light can get in and 'modulate' the avalanche.
>>
>> >> There was a thread on the subject here, many years ago - "Zener diode
>> >> oscillation" from July 7, 1997.
>>
>>
>> >I find that for use as noise sources it's best to keep the current
>> >below the knee.
>>
>> If you want predictable noise with decent statistics and wideband
>> quality, keep the current up, 5 or 10 mA for a typical 1N758 type.
>>
>> John- Hide quoted text -
>>
>> - Show quoted text -- Hide quoted text -
>>
>> - Show quoted text -
>
>Hi John,  Yeah I guess I was looking for repeatabliity.
>(I don't want to have to sort diodes...)
>I've got exactly the pulse into RC in the schematic I posted.
>
>Adding two back to back gives this RC-interuptus signal.  Randomly
>spaced RC's going in both directions.
>
>George H.

That makes lots of signal, but it won't be repeatable and will have
rotten statistics. At higher currents, you get lower amplitude but
better noise. 300 nv/rthz is easy to amplify... it's hard to find an

John

```
```On Jan 2, 9:55=A0pm, John Larkin
<jjlar...@highNOTlandTHIStechnologyPART.com> wrote:
> On Mon, 2 Jan 2012 16:33:45 -0800 (PST), George Herold
>
>
>
>
>
> <gher...@teachspin.com> wrote:
> >On Jan 2, 1:15=A0pm, John Larkin
> ><jjlar...@highNOTlandTHIStechnologyPART.com> wrote:
> >> On Sun, 1 Jan 2012 08:57:38 -0800 (PST), George Herold
>
> >> <gher...@teachspin.com> wrote:
> >> >On Dec 31 2011, 4:24 am, Bill Sloman <bill.slo...@ieee.org> wrote:
> >> >> On Dec 31, 5:49 am, Robert Baer <robertb...@localnet.com> wrote:
>
> >> >> > Say one takes an E-B junction of a transistor or other diode and
> >> >> > measures the (reverse) breakdown voltage over a wide range of cur=
rents
> >> >> > (say 5nA to 5mA), and assume there is no point or range where one=
sees
> >> >> > characteristics of negative resistance or of oscillation.
> >> >> > Also say that voltage is about 9 volts.
> >> >> > Q: is that zener or avalanche breakdown?
> >> >> > Q: how does one make a SPICE model of such a proposed beastie?
>
> >> >> At nine volts, the breakdown is by avalanche.
>
> >> >> It happens in very small areas, and the transit times is picosecond=
s,
> >> >> so at low currents the avalanche self-extinguishes from time to tim=
e
> >> >> due to the statistical fluctuations in avalanche multiplication
> >> >> process - sometimes a charge carrier gets through the avalanche reg=
ion
> >> >> without generating any new charge-carrier pairs.
>
> >> >> This isn't easy to model in Spice - it probably needs a current-
> >> >> controlled random noise generator, but the real noise looks more li=
ke
> >> >> a series of Dirac spikes.
>
> >> >It's even 'worse' than that. =A0The noise changes 'character' as a
> >> >function of current.
> >> >Right near the 'knee' you go from random pulses to random steps.
> >> >(random telegraph noise). =A0With more noise at low frequency.... But
> >> >I'm not sure it's 1/f.
>
> >> A typical small 10-volt zener has a region where it will make noisy
> >> sawtooth oscillations, 50 uA maybe from my memory. As current goes up,
> >> you get the asymmetric erratic pulsing thing, decreasing in amplitude
> >> as current increases. By the time you get to a few mA, it's nearly
> >> symmetrical almost Gaussian wideband noise, a few hundred nV per root
> >> Hz.
>
> >> >> For added extra complexity, you can throw in the fact that the
> >> >> avalanche process generates light, and in glass-packaged zener diod=
es
> >> >> this can trigger other avalanches.
>
> >> >And that room light can get in and 'modulate' the avalanche.
>
> >> >> There was a thread on the subject here, many years ago - "Zener dio=
de
> >> >> oscillation" from July 7, 1997.
>
>
> >> >I find that for use as noise sources it's best to keep the current
> >> >below the knee.
>
> >> If you want predictable noise with decent statistics and wideband
> >> quality, keep the current up, 5 or 10 mA for a typical 1N758 type.
>
> >> John- Hide quoted text -
>
> >> - Show quoted text -- Hide quoted text -
>
> >> - Show quoted text -
>
> >Hi John, =A0Yeah I guess I was looking for repeatabliity.
> >(I don't want to have to sort diodes...)
> >I've got exactly the pulse into RC in the schematic I posted.
>
> >Adding two back to back gives this RC-interuptus signal. =A0Randomly
> >spaced RC's going in both directions.
>
> >George H.
>
> That makes lots of signal, but it won't be repeatable and will have
> rotten statistics. At higher currents, you get lower amplitude but
> better noise. 300 nv/rthz is easy to amplify... it's hard to find an
>
> John- Hide quoted text -
>
> - Show quoted text -

Hmm, I don't know why, but the circuit seems to be very repeatable.
We've sold ~150 units with this noise source in them and looking over
my numbers
I see rms voltages from 2.5 to 2.0 volts... with most of them at 2.2
Vrms.

I have no idea of the 'physics' behind the numbers... or why other
zeners vary so much.

George H.
```