Shot noise and resistor geometry

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

featherpillow

Well-known member
Joined
May 14, 2005
Messages
214
Location
USA
Thermal noise appears to be affected by a resistor's physical size, but I can't find any information that discusses shot noise and resistor geometry. Is there a correlation?
 
I think the shot noise is more related to materials. Carbon composition being the noisiest, metal films the quietest. Or maybe tantalum is best.

Geometry might play a role, in that bulk films seem to do a better job in audio than the spiral wound types.

jh
 
Whoah, wait a minute, I lost my head there. Was thinking of "additive" noise sources (also thermal, voltage gradient, etc.).

Isn't "shot" noise purely a function of the number of electroncs flowing per unit time (current)? It's a counting phenomenon. This is an issue with CCD devices and photon counting or electron counting sensors. But in regular audio electronics, typically the count is so high that shot noise is not an issue.

Or maybe my memory is "shot". Should have looked this up in one of my texts...

jh
 
Shot noise per se is associated with discrete charges transiting a potential barrier, like across a vacuum from one electrode to another. In a conductor according to theory the electrons "see" each other and their fluctuations are due to thermal lattice vibrations ("phonons") jostling them about.

Resistors don't have shot noise per se. They can have excess noise though.

When a detector converts a photon flux to electricity there is at least full shot noise associated with the absorption-conversion process iirc.
 
Perhaps I'm calling the wrong thing. As I understand it, we try to calculate an average estimate of noise within a resistor, and it's a combination of three basic types--thermal, contact, and shot. Large value resistances have higher thermal noise than small value resistances. Thoughts that I've read suggest that a large geometry resistor of value x will have less voltage noise than a smaller sized resistor of the same value. I wondered if the same could be said for current-based noise and resistor geometry.
 
Contact noise and other anomalies typically have a 1/f^n spectrum, or in extreme cases a RTS (random telegraph signal) sort of noise, a random fluctuation of resistance among discrete levels. Again, "shot" noise should be reserved for carriers transiting a more-or-less well-defined potential barrier. These are not supposed to exist in resistors. I guess if you have a resistor with some reverse-biased diodes in the middle you could still call it a resistor and say that it exhibited shot noise----but that would be perverse.

Shot noise at very very low currents has a Poisson distribution, a point process sort of affair, which merges into a Gaussian at higher levels.

See Oliver's paper, Thermal and Quantum Noise, an oldie but goodie, for a comprehensive and magisterial account. I posted the particulars some time back---I'll dig it out in a bit. EDIT: Proc. IEEE 53 (5) May 1965, pp. 436-454.

Larger value resistances have higher thermal voltage noise than smaller value resistances. The thermal noise power at a given absolute temp is constant regardless of value.

For excess noise, yes, larger geometry resistors will typically have lower noise than smaller, given the same sort of constitution.

EDIT: Also, when you say "we try to calculate..."---I don't know of anyone who calculates excess noise, unless that consists in plugging in numbers to a formula based on experiments by the resistor vendor. Of course thermal noise is well understood and inescapable (if you get less than thermal noise in a physical resistor it is safe to say you are making a bad measurement of something, or messing up a calculation) so we calculate that all day long.
 
[quote author="featherpillow"]Thermal noise appears to be affected by a resistor's physical size, but I can't find any information that discusses shot noise and resistor geometry. Is there a correlation?[/quote]

There is strong correlation, but we must start on basic.

There is several factors, Noise indices, and other historical and
technological aspects, but simple view is from spectra.

Noise power spectra is

u**2(f) = 4kTR(Fl/f + 1)

where Fl is low frequency noise corner and f is frequency
It can be computed, that:

Fl = fi*U**2

where fi is technological parameter of resistor and U is DC voltage
on it.

That fi is greater if current going trough resistor can vote {...or choose?, what is beter in english?}.

If current have only straight way {wirewound} fi is small,
If current can vote in one dimension {metalised resistors} fi is
greater, than in wirewound.
If current vote in two dimensions {carbon mass resistor}, fi is greatest.

There are also formula for computing fi from NI {noise index} which
is catalogue parameter of device, It can be easily deducated, but
better is to measure fi directly from measured spectras.


xvlk
 
[quote author="NewYorkDave"]Oliver on noise
Oliver on noise

Damn... Only available to members :mad:[/quote]

Also reprinted in at least two books: Electrical Noise: Fundamentals & Sources, ed. Gupta, IEEE Press, 1977; The Selected Papers of Bernard Oliver, Hewlett-Packard Laboratories, 1997. Both books are excellent and highly recommended, the first for historical papers on noise, the second for the breadth and depth of the late Barney Oliver.

I see copies of both books are available at the moment and not too expensive except for a new paper copy of the first one---see the bookfinder.com site. Links to them expire rapidly so there is no point to posting specifics.
 
> Only available to members

And a real brain-buster for most folks.

Abstract- One purpose of this article is to develop the theory of black body- radiation, thermal noise, and quantum noise from a few basic physical principles. A second purpose is to show how these results apply to certain areas such as antenna theory and ideal receivers. It is hoped that having this related material collected and presented in the language of the electronics engineer will be of tutorial value.
. . . . .
2) In any closed system the entropy eventually maximizes. Since entropy is a measure of randomness, this, the second law, says that any closed system tends toward a state in which its total energy is randomly distributed among the various degrees of freedom present.
In general the total energy of a system can be written as the sum of a number of terms each involving the square of a system coordinate or its derivative. For example, the total energy of a system of n free particles is....
{ :sad: it goes uphill from here}

> Large value resistances have higher thermal noise than small value resistances.

No. Large value resistances have higher thermal voltage noise than small value resistances. Small value resistances have higher thermal current noise than large value resistances. All resistors have the same thermal noise power. {Ah, I see bcarso caught this already.}

This seems to imply a paradox: put two resistors together and you get the same noise power as one resistor. There is no paradox.

> Here's the link I've been reading:

Seems like a good tour of thermal noise. SPICE won't calculate "excess" (dirty carbon) noise because every resistor brand and series is different. I don't think SPICE considers shot noise, but I may be very wrong.
 
Seems like a good tour of thermal noise

Exactly--I couldn't seem to find anything on current related noises, and was a little confused.

So it's less of an issue of whether it's current or voltage noise, but more of an issue of noise potential (power)...?
 
For thermal noise, the noise power is always there---and expression as a noise current or noise voltage is a matter of convenience. A "pure" resistor can be modeled as having a noise voltage generator (zero impedance) in series, or with a noise current generator (infinite impedance) in parallel.

If you hook up a hot resistor to a cold one you will get a small cooling of the hot one and vice versa just due to the electrical energy exchange---Nature strives for equilibrium in this case.

The choice as to which noise variable to use is arbitrary. I remember a Review of Scientific Instruments paper a while back where they made a low temperature resistive thermometer by measuring both resistor voltage noise and current noise. If you do calculations on the two numbers you can extract the actual resistance and the absolute temperature, regardless of how much the resistance changes (and it changes a whole lot at very low temps).
 
[quote author="featherpillow"]
I couldn't seem to find anything on current related noises, and was a little confused.
[/quote]
Yes, there is lack of continuity from current-related (flicker) noises
and white noises.
In technological (or better resistor production) literature, there
are flicker noise and white noise computed still diversely.

Used catalogue value is "noise index" in somewhat
unusable units (which depends on "pure flicker noise" asymptotic computation)
NI = uV/V/dec.
This value of NI is dependent to resistor ohmic value in several orders.

It can be deduced, than computing formula (to "flicker noise susceptibility",I called it fi) is

fi = (10**(NI/10))/(R*3.69E-8) [Hz/V**2;uV/V/dec,Ohms]

That fi is less dependent to resistor value and can be directly used
for resistor noise analysis in programs like SPICE {if you can output
adjoint circuit matrices from it, it is not implemented in every SPICE}

xvlk
 
Part of the ambiguity may be terminological. There is thermal noise, which may be expressed as a parallel current generator or a series voltage generator, with the former being called in context "current noise."

Then there is noise associated with the flow of current, usually the major source of excitation for lower frequency noise generators of various sorts as indicated by xvlk. This could also, in context, be called current noise---I prefer the term excess noise.

Then, shot noise is associated with the flow of granular current over a potential barrier, and per se is poisson distributed (hail on the roof sort of stuff) merging into a traditional white noise at high levels.

To quote Peter Carroll Dunn, Gateways Into Electronics (forgive my copyright violation please Wiley): "Shot noise is associated with currents---but only certain currents, as we shall see---and arises because charge is quantized in units of e." Later, after showing how it works for a temperature-limited vacuum diode, the standard example, he reiterates: "Full shot noise is not associated with arbitrary currents. As an example, it is shown...that the noise associated with drift current in a powered resistor is many orders of magnitude less than the value predicted by the Schottky formula."


If it is any consolation, a number of smart people have gotten this latter part wrong, including at least one textbook on semiconductor devices, and some Princeton astronomers in vol. 12 of the prestigious Methods of Experimental Physics. Trusting them, I foolishly wasted about a year, circa 1974, trying to reconcile experiment and theory, only to finally turn to nuclear science instrumentation papers for accurate accounts of JFET noise and preamp optimization.
 
[quote author="bcarso"]
Then there is noise associated with the flow of current, usually the major source of excitation for lower frequency noise generators of various sorts as indicated by xvlk. This could also, in context, be called current noise---

I prefer the term excess noise.

Then, shot noise is associated with the flow of granular current over a , pp.potential barrier[/quote]
-
Yes, term excess noise is well applied to voltage/current / power,

but to separate current (Schottki) and termical (Johnson) noise
is somewhat old-fashioned.

See:
Schalpeshkar,Debrueck,Mead: White Noise in MOS Transistors and Resistors, IEEE Circuits and Devices, November, pp. 23-29.

xvlk
 
[quote author="xvlk"]but to separate current (Schottki) and termical (Johnson) noise
is somewhat old-fashioned.

See:
Schalpeshkar,Debrueck,Mead: White Noise in MOS Transistors and Resistors, IEEE Circuits and Devices, November, pp. 23-29.

xvlk[/quote]

November when??

And when did separating thermal and shot noise become old-fashioned?? Is Oliver now hopelessly out of date?? EDIT: Ah, found it. Would help if you had spelled the lead author's name correctly:

http://www.ini.unizh.ch/~tobi/anaprose/noise/noise.pdf

Second EDIT: It's a very interesting paper, and I thank you for the information.

However, I find it downright shocking that Oliver is not even referenced---particularly when the paper opens with the statement, clearly intended to be provocative and revelatory: "Shot noise and thermal noise have long been considered the results of two distinct mechanisms, but they aren't." :mad: Alas, all too typical of the Caltech folk, where Not Invented Here is practically tattooed on foreheads :razz:

Where were the referees? Sloppy sloppy.
 
[quote author="bcarso"] clearly intended to be provocative and revelatory[/quote]
Not all usable for do some computation must be
based on real things, i.e. computations in electromagmetic fields
can be simplier with using magnetic monopoles,
computations in audio frequency circuits can be simplier
with using only thermal noises. Trick how to do
it it was done by Russian man Dementjev in his book Osnovy ob\vs\vcej teorii \vsumja\vs\vcich linejnych cepej {I used backslash for diacritics,
but someone from russia here can correct it} via introducing "technological
noise factor", what is the some like frequency dependent temperature
of real {or imaginar, like transconductancy} device. this frequency dependence can handle low frequency noise like a simple function.

Dementjev used symbol small gamma for his factors to notate how much
* is noise of device higher, than noise of real resistor at real temperature.


xvlk



xvlk
 
Back
Top